Split (1)

train · 5.09k rows

pmcid stringlengths 6 6	title stringlengths 9 374	abstract stringlengths 2 4.62k ⌀	fulltext stringlengths 167 106k	file_path stringlengths 64 64
546006	General practitioners' reasoning when considering the diagnosis heart failure: a think-aloud study	Background Diagnosing chronic heart failure is difficult, especially in mild cases or early in the course of the disease, and guidelines are not easily implemented in everyday practice. The aim of this study was to investigate general practitioners' diagnostic reasoning about patients with suspected chronic heart failure in comparison with recommendations in European guidelines. Methods Think-aloud technique was used. Fifteen general practitioners reasoned about six case vignettes, representing authentic patients with suspected chronic heart failure. Information about each case was added successively in five steps. The general practitioners said their thoughts aloud while reasoning about the probability of the patient having chronic heart failure, and tried to decide about the diagnosis. Arguments for and against chronic heart failure were analysed and compared to recommendations in guidelines. Results Information about ejection fraction was the most frequent diagnostic argument, followed by information about cardiac enlargement or pulmonary congestion on chest X-ray. However, in a third of the judgement situations, no information about echocardiography was utilized in the general practitioners' diagnostic reasoning. Only three of the 15 doctors used information about a normal electrocardiography as an argument against chronic heart failure. Information about other cardio-vascular diseases was frequently used as a diagnostic argument. Conclusions The clinical information was not utilized to the extent recommended in guidelines. Some implications of our study are that 1) general practitioners need more information about how to utilize echocardiography when diagnosing chronic heart failure, 2) guidelines ought to give more importance to information about other cardio-vascular diseases in the diagnostic reasoning, and 3) guidelines ought to treat the topic of diastolic heart failure in a clearer way.	Background Chronic heart failure (CHF) is a major cause of morbidity and mortality and it has a considerable impact on the health care system [ 1 ]. In a recent study, the prevalence in Sweden was estimated at 1.3–2.5% [ 2 ]. Early detection of CHF has become increasingly important, as modern drug treatment has the potential to improve symptoms and quality of life, slow down the rate of disease progression, and improve survival. However, diagnosing CHF is known to be difficult, especially in mild cases, as many features of the condition are not organ specific, and there may be few clinical features in the early stages of the disease [ 3 - 5 ]. Most of the patients are old, which also makes the diagnosis difficult. Older patients may have atypical symptoms, they may suffer from other diseases, and they may be on treatment that modifies their symptoms [ 3 ]. Diagnosing CHF has been found to be especially difficult in women and in obese patients [ 4 ]. A large proportion of patients with CHF are managed by general practitioners (GPs), especially older patients and patients early in the course of disease, i.e. those patients for whom the diagnostic process is characterized by the greatest uncertainty [ 6 ]. The European Society of Cardiology adopted guidelines for diagnosing CHF in 1995, and these were revised in 2001 [ 7 - 9 ]. Swedish guidelines, based on the 1995 European guidelines, were published in 1996 by the Swedish Medical Products Agency [ 10 ]. However, guidelines are often not easily or accurately integrated into daily practice [ 11 , 12 ]. The full versions of the above-mentioned guidelines are comprehensive documents, covering epidemiology, aetiology, pathophysiology and diagnostic methods, but may be difficult to apply to specific diagnostic situations [ 13 ]. However, the recommendations are summarized in 1) a definition, 2) an algorithm for the diagnosis of CHF, and 3) a table of assessments to be performed routinely to establish the presence of CHF. The definition includes three criteria: a ) one or more typical symptoms (at rest or during exercise), b ) objective evidence of cardiac dysfunction (at rest), and c ) response to treatment directed towards CHF (in cases where diagnosis is in doubt). Criteria a and b should be fulfilled in all cases. Echocardiography (ECHO) is mentioned as the single most effective tool in widespread clinical use for objective assessment of cardiac dysfunction. In the algorithm for the diagnosis of CHF, a sequence of investigations is recommended: suspect CHF because of symptoms and signs; assess presence of cardiac disease by electrocardiography, X-ray or Natriuretic peptides (where available); imaging by echocardiography; assess aetiology, degree, precipitating factors and type of cardiac dysfunction; additional diagnostic tests where appropriate; choose therapy. Table 1 shows the assessments to be performed routinely [ 9 ]. In the present study, the list of assessments recommended in Table 1 was used for evaluation of the GPs' diagnostic reasoning. For most Swedish GPs, the main source of knowledge regarding CHF diagnostics is probably locally adapted protocols developed by cardiologists, or by cardiologists and GP representatives in collaboration. Table 1 Diagnostic assessments according to guidelines Assessments to be performed routinely to establish the presence and likely cause of heart failure (Eur Soc Cardiol 2001). Assessments The diagnosis of heart failure Suggests alterantive or additional diagnosis Necessary Supports Opposes Appropriate symptoms +++ +++ (if absent) Appropriate signs +++ + (if absent) Cardiac dysfunctioning on imaging (usually echocardiography) +++ +++ (if absent) Response of symptoms or signs to therapy +++ +++ (if absent) Electrocardiography +++ (if normal) Chest X-ray + (if pulmonary congestion or cardiomegaly) + (if normal) Pulmonary disease Full blood count* Anemia/secondary polycythemia Biochemistry and urinalysis* Renal or hepatic disease/diabetes Plasma concentration of natriuretic peptides in untreated patients (where available)* + (if elevated) +++ (if normal) + = of some importance; +++ = of great importance ) Recommended assessments which are not included in the corresponding table in the Medical Products Agency version of guidelines from 1996. Relatively few studies on how patients suspected of having CHF are diagnosed have been performed in primary health care settings, and most of them report over-diagnosis [ 3 , 4 , 14 - 16 ]. In the present study we used written case vignettes (case descriptions) and think-aloud technique to study how GPs' diagnostic reasoning and diagnostic judgements about patients with suspected CHF are related to the recommendations in the European guidelines [ 9 ]. What clinical information is considered important by the GPs in the sense that it is used as an argument for or against the diagnosis of CHF? What information that is considered important for diagnosing CHF in the guidelines is also considered important by the GPs? Methods Think-aloud method Process-tracing techniques are used to study the cognitive processes involved in decision-making such as, for example, how judgements change over time as new information is presented, and which decision rules are used [ 17 ]. A method that is often used to describe the sequence of thoughts behind decision-making is the think-aloud technique [ 18 ]. Subjects are instructed to say their thoughts aloud while performing a task, and the verbal reports are usually audio-taped, transcribed to a written form, and then analysed. The think-aloud technique has been used in a number of studies in the field of medical decision-making [ 13 , 19 ]. The value of conclusions reached in such studies depends on the validity of the think-aloud method, and on the reliability of the coding process. Thinking aloud while performing a task often lengthens the time for completing the task, but does not seem to change the accuracy of task fulfilment or the cognitive processes [ 18 ]. In a recent study we found that think-aloud data were at least as valid as ratings in describing a clinical decision process [ 20 ]. Participants All health care centres in northern Stockholm within a distance of 20–30 km from the city centre ( n = 61) were listed and contacted in a random order. The distance from central Stockholm was chosen for practical reasons. In each health care centre the GPs were contacted in a random order by one of the authors (YS). Only one GP at each centre was included in the study, and this person had to be a specialist in family medicine. We contacted the GPs during their regular telephone hour, during the period October 2001 to October 2002. Our goal was to include 15 GPs in the study. A total of 30 GPs were reached, and 15 agreed to participate. Those who declined to participate were not asked why they did so, but the majority of those who spontaneously gave a reason mentioned a heavy workload. The participants had been specialists in family medicine for an average of 14.8 (range 3–25) years, they were on average 52.7 (range 42–62) years of age, and six of them were men. The non-participating GPs were on average 52.7 (range 35–62) years of age, and seven of them were men. Case vignettes Six case vignettes (CV), based on authentic patients, were presented to the participants. The information presented in the case vignettes was obtained from the patient records and included information about relevant diseases (e.g. coronary heart disease, stroke, diabetes), lifestyle factors (e.g. smoking, alcohol consumption), symptoms, signs, electrocardiography (ECG), chest X-ray findings, and ECHO. Chest X-ray and ECHO results were presented in the same format as in the patient records. ECHO results could contain information about ejection fraction (EF), valvular disorders and ventricular wall motility. The diagnoses made by the attending cardiologists (based on all available clinical information, including ECHO) were used as a reference standard when assessing the participants' diagnostic accomplishments. The six cases were selected to represent patients with various types of potential diagnostic problems: A "prototypical" CHF patient (CV2), a patient with both CHF and chronic obstructive pulmonary disease (COPD) (CV6), a patient with CHF, tachycardia and mitral valve insufficiency (CV3), an obese non-CHF patient with normal ECG and EF (CV5), a non-CHF patient with COPD (CV4) and a non-CHF patient with alcohol abuse and a metabolic syndrome (CV1). Additional file 1 shows some of the characteristics of the six cases. For one of the cases (CV3) there was a disagreement between the diagnosis according to the cardiologists and the diagnosis that could be deduced from a simplistic interpretation of the guidelines. This patient had typical clinical findings including gallop rhythm, cardiomegaly, and pulmonary congestion, but normal left ventricular function according to ECHO. It could therefore be categorized as not CHF according to the definition given in the guidelines. However, this patient also had a mitral valve insufficiency, which can give a "false normal" ejection fraction value: the left ventricle is emptied both forward (cardiac output) and backward (leakage through the mitral valve). Procedure Before the sessions the GPs had received written information about the aim of the study (to study clinical judgements) and about the method (think aloud), but not about the kind of medical problems that would be presented to them. The study was conducted at the GPs' offices. All visits and recordings were made by one of the authors (YS). The participants were instructed that six authentic patients, suspected by GPs to have CHF, would be presented, and that their task was to say aloud their thoughts about the case, and to try to decide whether the patient had CHF or not. The order of the cases was the same for all participants. The order in which the information was presented was arranged to be as realistic as possible in relation to clinical practice (first history and symptoms, then findings, and then results of investigations). Each vignette was presented on a computer screen in five successive steps using QA software [ 21 ]. All previously shown information about a case was repeated at the top of the later screens in a different colour to reduce and control for memory effects. The participants could control the shift to a new screen by clicking with the mouse on a continue button at the bottom of the screen. After all the information had been presented, the participants were asked, on the sixth screen, to summarize their judgements about the case and to try to decide about the diagnosis. The doctors could express their diagnostic judgements freely, with their own words. The doctors first got a test case (not recorded) in order to get acquainted with the think-aloud method, and then continued with the six study cases. The only intervention from the researcher during the think-aloud session was that a participant who was silent for more than about 15 seconds was reminded to say his or her thoughts aloud about the information presented [ 18 ]. All sessions were recorded and transcribed by a secretary. Response measures and coding of data Coding of variables in the case vignettes All information in the case vignettes that was of relevance for the diagnosis and that could take on different values was considered to be variables. Fifty variables were defined: 19 of them were included in all six vignettes (e.g. symptoms, signs and investigations mentioned in the guidelines), six in five, one in four, six in three, one in two, and 17 in one vignette (e.g. alcohol abuse, history of a bypass operation, and panic disorder). For each case vignette, the presented variables were coded for content and value (Table 2 ). Table 2 Variabel codings Information as presented in the case vignette Content Value "Shortness of breath on level" Dyspnoea Positive (presence of finding) "Pathological R-progression on ECG" ECG Positive (pathology) "He has not had swollen legs" Oedema Negative (absence of finding) "Regular rhythm" Rhythm Negative (normality) "Relative heart volume 630 ml/m 2 " Heart volume 630 (numeric values as presented in the text) Examples of coding of variables in the case vignettes. Positive value = precense of finding, or pathological finding, negative value = absence of finding, or normal finding. Coding of think-aloud protocols For each participant, every mention of a variable was coded for how the GP seemed to use it: as an argument for the diagnosis of CHF, as an argument against CHF, or as not being of any explicit use for the diagnosis (mentioned only). "He has basal rales. This guy has CHF!" is an example of a participant using the variable "rales" (positive value) as an argument for CHF. "So I'm not really sure that he has got CHF. Just a moderate cardiac enlargement, no, I wouldn't think so" is an example of a participant using the variable "relative heart volume" (value 630 ml/m 2 ) as an argument against CHF. For each participant, a specific evaluation of each variable value was only counted once for each case vignette in order not to give more weight to thoughtful repetitions of an argument than to a single, firm statement. However, if a participant used the same variable value as an argument both for and against the diagnosis of CHF, both evaluations were coded. Ten percent of the 90 case vignette protocols were selected at random and coded independently by two of the authors (YS, LB) to estimate the interrater agreement of the coding process. The rest of the protocols were coded by one of the authors (YS). Comparing think-aloud protocols with guidelines The list of diagnostic assessments recommended in the guidelines (Table 1 ) was used for comparing GPs' diagnostic reasoning with the guidelines. Breathlessness, ankle swelling, and fatigue are mentioned in the guidelines as appropriate symptoms, and leg oedema, tachycardia, gallop rhythm, and pulmonary crepitations (rales) as appropriate signs. (Neck vein distension and liver enlargement are also mentioned, but these signs were not present in the case vignettes.) Use of the variables in relation to recommendations was analyzed for frequency among GPs and case vignettes. Classification of diagnostic judgements The participants were not forced to express their diagnostic judgements in a specific format, and their free verbal statements therefore had to be interpreted and coded. Two of the authors (YS, LB) independently classified all the diagnostic judgements ( n = 90) in three categories: CHF or probably CHF; uncertainty about diagnosis; probably not CHF or not CHF. Analyses Stata 8.0 was used for the statistics. Cohen's kappa test (κ) was used to determine interrater agreement regarding the coding of the think-aloud protocols and the classification of the diagnostic judgements. Kappa values are classified as follows: <0, worse than chance; 0 to 0.2, poor; 0.21 to 0.4, fair; 0.41 to 0.6, moderate; 0.61 to 0.8, good; and >0.8, very good [ 22 ]. The research ethics committee of Huddinge University Hospital approved the study. Results Reliability – interrater coding agreement Think-aloud protocols The randomly selected test protocols contained 322 segments of propositions, 36 of which were excluded since they contained variables that were not going to be investigated in this study (e.g. treatment suggestions). There was disagreement between the two coders about the content of variables in 14 of the remaining 286 segments (4.8%). The remaining 272 segments were then tested for interrater agreement on argument values (for CHF; against CHF; just a mention), which was 95% (κ 0.85). Diagnostic judgements The interrater agreement was 92% (κ 0.85). For the few diagnostic judgements where there was initial disagreement, it was possible to agree upon an interpretation. Diagnostic reasoning Assessments to be performed routinely according to guidelines The information that was used most frequently in diagnostic arguments was the ejection fraction value on ECHO, pulmonary congestion, and cardiac volume (Figure 1 ). The most frequent argument for CHF was pulmonary congestion on chest X-ray, and the most frequent argument against CHF was the ejection fraction value. Figure 1 The most frequently used diagnostic arguments The ten most frequently used arguments making use of different categories of clinical information. Number of arguments favouring the diagnosis CHF and number of arguments against CHF are given for each category. One variable (indicated by ) was only presented in five of the vignettes. Symptoms and signs were not often used as arguments in the GPs' diagnostic reasoning (Figure 1 ). Symptoms were most frequently used as diagnostic arguments when reasoning about CV2 ( Additional file 1 ), which represented the prototypical CHF case, with dyspnoea when walking on level ground and orthopnea ("in need of three pillows to be able to sleep"), and about CV5, which represented the prototypical non-CHF case (absence of dyspnoea). Signs were most frequently used for two of the case vignettes. In CV1, the presence of rales was used by nine of the GPs as an arguments for CHF (eight of them incorrectly ending up with this as the final diagnosis), and in CV3, tachycardia was used by nine of the GPs as an argument for CHF (eight of them correctly ending up with this as the final diagnosis). CV3 also had a gallop rhythm, which is reported to be fairly specific for CHF. However, only one GP used this as an argument for CHF. According to the guidelines, a normal ECG opposes the diagnosis of CHF (Table 1 ). Only one of the patients (CV5) had a normal ECG. Three of the GPs used this information as an argument against CHF when reasoning about this patient, and two of them diagnosed the patient as not CHF. Four GPs used a pathological ECG as an argument for CHF (CV1, CV6), and all four diagnosed those patients as CHF. According to the guidelines, information about cardiac enlargement or pulmonary congestion on chest X-ray gives some support for CHF if there are any pathological findings, and is of some importance as an argument against CHF if the findings are normal (Table 1 ). Chest X-ray findings were frequently used by the GPs as arguments in their diagnostic reasoning (Figure 1 ). When considered separately, information about cardiac volume was used as an argument 36 times (26 for, and 10 against CHF), and information about pulmonary congestion 38 times (32 for, and 6 against CHF). ECHO findings as arguments for or against CHF Each of the 15 GPs judged 6 case vignettes, which resulted in 90 judgement situations. In 48 of them, the EF value (or the information about left ventricular function in CV3) was used as an argument for or against CHF (Table 3 ). In some of the judgement situations in which the EF value was not used as an argument, other ECHO information was utilized, such as left ventricular hypertrophy or restricted motility of the ventricular wall. However, in 33 judgment situations, no ECHO information was used as an argument in the diagnostic reasoning. Table 3 shows the use of ECHO in all the judgement situations: Five GPs used information about EF in their diagnostic reasoning for five of the case vignettes, four GPs used it for four of the vignettes, one GP used it for three of the vignettes, one GP for two of the vignettes and two GPs for one of the vignettes (and in one case in the wrong direction). Two GPs never used information about EF in their diagnostic reasoning. Some of the GPs expressed uncertainty about the EF values (e.g. GP14, CV1: "I think... think I am not certain about the meaning of ejection fraction"). Table 3 GPs' use of ECHO information CV2 CV6 CV3 CV5 CV4 CV1 CHF CHF CHF Not CHF Not CHF Not CHF GP1 + other other 0 0 0 GP2 + + 0 0 0 other GP3 + + 0 - - - GP4 + + 0 - - 0 GP5 + + other - - - GP6 + + 0 - - - GP7 + + 0 - - other GP8 + + 0 - - 0 GP9 + + 0 - 0 - GP10 + + - - - 0 GP11 + + 0 - - - GP12 0 0 other 0 other other GP13 + + 0 0 0 - GP14 0 0 other 0 0 0 GP15 0 0 0 0 + 0 Use of information about ejection fraction (EF) value, or, in the case of CV3, about left ventricular function, as arguments for (+) or against (-) the diagnosis CHF. "Other" indicates that other ECHO information than the ejection fraction was used in the diagnostic reasoning and (0) that no ECHO informtion was used. (CHF = chronic heart failure, CV = case vignette, GP = general practitioner) In 17 judgement situations, there was a conflict between the GPs' evaluations of the chest X-ray information and their evaluations of the ECHO information. In seven of those situations, the final diagnosis was in the same direction as the ejection fraction argument (four CHF, three not CHF). In three judgement situations, the final diagnosis was in the same direction as the X-ray argument (three CHF). In seven judgement situations, the GP was uncertain about the diagnosis. Other relevant diseases In our study, the GPs used other diseases as an argument in a total of 70 judgement situations, mostly as arguments for CHF (91%). Atrial fibrillation, emphysema, history of myocardial infarction, and hypertension were the diagnoses most commonly used in this way. Information that GPs disagree about For certain variables, the same information value was used by some GPs as an argument for and by others as an argument against CHF. The presence of emphysema was sometimes seen as increasing the risk of CHF (e.g. GP 7, CV3: "...and then she has emphysema ... chronic obstructive pulmonary disease, which can also contribute to CHF."), or as an alternative explanation for symptoms (e.g. GP 13, CV3: "And then she also has emphysema, which could give her this severe breathlessness."). Diabetes could also be seen as increasing the risk of CHF (e.g. GP 9, CV5: "...if I think the patient has CHF? Well, there are some facts in particular, she's diabetic, and she has uncontrolled hypertension, well, too high, and stasis – so I couldn't rule out the idea of CHF after all."), or as an alternative explanation for symptoms (e.g. GP 1, CV5: "... we have to improve her diabetes, since her fatigue may be due to that."). Advanced age could be seen as increasing the probability of CHF (e.g. GP 2, CV6: "He's the age for it!"), or as an alternative explanation for symptoms (e.g. GP 1, CV6: "I'm not so sure that CHF alone can explain his symptoms. After all, he's 84 years old."). Age was used as a diagnostic argument only for the two patients over 80 years of age. For relative cardiac volume, the reasoning could be compatible with GPs using different threshold values in their reasoning. The two lowest values were only used as arguments against CHF, the two highest values only as arguments for it, and the two intermediate values were used as arguments in both directions. Diagnostic judgements There was total agreement among the GPs only for the prototypical CHF case; otherwise there was a large variation among GPs regarding diagnoses. Case vignettes representing CHF patients were more likely to be correctly diagnosed than those representing non-CHF patients (Table 4 ). Table 4 GPs' classifications of case vignettes CV2 CV6 CV3 CV5 CV4 CV1 CHF CHF CHF Not CHF Not CHF Not CHF Total number of arguments used by the GPs (proportion of arguments for CHF) 60 (98%) 52 (75%) 54 (85%) 57 (44%) 63 (70%) 63 (75%) Number of GPs classifying the patient as CHF 15 11 11 3 6 11 Uncertain, no classification 0 2 2 3 3 1 Correct diagnoses of CHF cases and not CHF cases (proportion of correct diagnoses) 37/45 (82%) 18/45 (40%) Correct diagnoses, all judgements (proportion of correct diagnoses) 55/90 (61%) The GPs' classification of the case vignettes and the number of arguments used by the group of GPs. Cells with bold numbers indicate correct judgements. (CV = case vignette, CHF = chronic heart failure) Discussion GPs' diagnostic reasoning compared with guidelines When comparing the GPs' diagnostic reasoning with guidelines, we found that the clinical information in the case vignettes was not used to the extent recommended in the guidelines. It is true that information about the ejection fraction value on ECHO was the single most frequent diagnostic argument, and it was the most common argument against CHF. This is in line with the guidelines, which emphasize the need for objective evidence of cardiac dysfunction. However, in more than one third of the judgement situations, the information about ECHO that was presented was not used as an argument. Over-diagnosis of CHF in primary health care has been demonstrated in a number of studies, with ECHO findings as the gold standard [ 3 , 4 , 15 ]. Limited access to ECHO has been suggested as an explanation for this finding. However, our data indicate that simply providing access to ECHO might not be enough. In the diagnostic algorithm, symptoms and signs are the entry criteria. However, the GPs did not seem to use them consistently in this way, except when diagnosing the prototypical CHF and non-CHF cases. One reason for this might be that most symptoms and signs considered typical for CHF are fairly non-specific as regards the diagnosis CHF. Information about other relevant diseases, which was important in the GPs' diagnostic reasoning, is not included in the list of assessments to be performed routinely (Table 1 ) [ 9 ]. However, information about a history of myocardial infarction, for example, increases the probability of CHF. In a study of CHF diagnostics in primary health care, it was shown that the combination of cardiac enlargement and a history of myocardial infarction had the best positive predictive value for CHF when systolic dysfunction measured by ECHO was used as gold standard [ 23 ]. This finding is compatible with the notion that experienced physicians structure their knowledge more according to enabling conditions than according to biomedical reasoning [ 24 - 26 ]. Enabling conditions are patient contextual factors such as sex, age, medical history, and occupation. In most routine diagnostic situations, biomedical details of a disease and its cause are not so important, and the physician's images of the diseases ('illness scripts') are rather characterized by these enabling conditions, which form a characteristic pattern. The GPs' frequent use of this kind of information may thus indicate that they are experienced physicians, with illness scripts for CHF which include other diseases. It might be valuable to include this kind of information in a clearer way in the guidelines, because it would reflect the higher probability of CHF in patients with those characteristics. Some methodological considerations The case vignettes represented authentic patients referred by GPs to a cardiology department for problems related to heart failure. This may have led to a selection of more complicated patients than the "typical" heart failure patients in primary health care. The reason we chose this group of patients was that we wanted to include patients who were thoroughly investigated, with a well-founded clinical diagnosis, and for whom information about all variables of interest could be found in the patient records. Selecting GPs only from health care centres in, or relatively close to, the city centre may have biased the results, since differences in catchment areas, working conditions, and access to echocardiography may influence GPs' diagnostic habits. This could make it difficult to generalize the results to other GPs. Only 50% of the GPs who were contacted agreed to take part in the study, which could bias the results. However, since the age distribution was the same in the two groups, it seems unlikely that the drop-out group would differ from the study group regarding clinical experience. Guidelines as decision support when diagnosing CHF The full version of the guidelines is difficult to apply to individual diagnostic situations and it is also difficult to use it for assessment of diagnostic behaviour [ 13 ]. In this study, we have used the table of routine assessments as a reference for evaluating the GPs' diagnostic reasoning (Table 1 ). This table includes a rough weighting of the importance of different types of information, which could serve as a guide for diagnostic judgements, even if it is not obvious how it should be used in individual cases. The two compulsory criteria in the definition are included in this table as necessary conditions. However, in some situations these judgment tools will not be satisfactory. One example is case vignette CV3, where the clinical picture was strongly indicative of CHF, with dyspnoea, rales, tachycardia, gallop rhythm, cardiomegaly and pulmonary congestion, while according to ECHO findings there was normal left ventricular function ( Additional file 1 ). The patient could therefore be classified as a non-heart failure patient according to the definition, while the clinical diagnosis, based on the attending cardiologist's judgement of all accessible information, was in fact heart failure. However, the ECHO in this case also included information about atrial dilatation, mitral insufficiency and pulmonary hypertension, i.e. a rather complex situation. A patient with clinical findings suggestive of CHF, but with a normal ejection fraction value, could be considered not to have CHF, i.e. not to have a systolic CHF, but could alternatively have a diastolic CHF [ 27 , 28 ]. This situation is not dealt with in the guidelines. Some implications of this study GPs' tendency to over-diagnose CHF has been explained by their relying on symptoms, signs and less specific investigations such as chest X-ray, and by limited access to ECHO in the primary health care. However, this study indicates that a substantial minority of GPs seem to be less familiar with the use of ECHO and EF. Thus, access to ECHO ought to be accompanied by education about how to integrate this information better in the diagnostic reasoning. Guidelines ought to include search of information about other cardio-vascular diseases in the list of assessments to be performed routinely (Table 1 ) and in the algorithm for diagnosis of heart failure. This would reflect the increased probability of CHF in presence of those diseases. The problem of diastolic heart failure should also be addressed in a clearer way in guidelines. Conclusions The information in the case vignettes was underused as arguments for and against the possibility of CHF as compared with the guidelines. Information about the EF value was the single most frequently used argument for or against CHF; nevertheless, in one third of the diagnostic judgements the GPs did not consider any information about the ECHO in their diagnostic reasoning. Information about symptoms and signs were not used to to the extent suggested in the guidelines. Information about other relevant diseases was frequently used in the GPs' diagnostic reasoning, indicating that they often relied on illness scripts. Some implications of our study are that 1) GPs should be taught how to use ECHO information better in their diagnostic reasoning, 2) guidelines ought to give more importance to information about other cardio-vascular diseases in the diagnostic reasoning, and 3) guidelines ought to treat the topic of diastolic heart failure in a clearer way. List of abbreviations used CHF Chronic heart failure COPD Chronic obstructive pulmonary disease ECHO Echocardiography ECG Electrocardiography EF Ejection fraction GP General practitioner Competing interests The author(s) declare that they have no competing interests. Authors' contributions All authors conceived of the study and participated in the design. YS carried out the data collection, performed the statistical analyses and drafted the manuscript. All authors participated in the interpretation of the results and the discussions of the drafts. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 Case vignette characteristics The additional file shows some important characteristics of the case vignettes. Format: Word-table. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546006.xml
368177	Conserved Genes Preferentially Duplicated in Evolution	null	Over the course of evolution, some organisms have gained many genes and become increasingly complex whereas other, simpler, organisms have survived with comparatively fewer genes. (Compare, for example, the 30,000 genes of humans to the 5,500 of brewer's yeast.) But where do these “new” genes come from? Evolutionary biologists have long known that duplication of existing genes is an important source of genetic novelty—it is easier to copy and modify an existing gene than to create a completely new one from scratch. Because gene duplication makes such a major contribution to evolution, researchers have attempted to understand the mechanisms of gene duplication, how genes evolve once they become duplicated, and what functional effect gene duplications have for the organism. Recent genomic studies, for example, appear to show that most duplicated genes go through a period of accelerated evolution and also that the presence of duplicated genes adds robustness to the functioning of genomes. In research published in this issue, however, Jerel Davis and Dmitri Petrov look at gene duplication from a different perspective. Rather than asking how genes are duplicated, they asked which genes tend to be “good” at duplicating over the course of evolution. The answer is important for our understanding of the forces underlying gene duplication and will also help us understand why genomes contain duplicates of some genes and not others. The authors began by identifying duplicated and nonduplicated gene pairs in the yeast Saccharomyces cerevisiae and the worm Caenorhabditis elegans , two model organisms whose genomes have been sequenced. They then looked for the corresponding genes in two distantly related species, the fruitfly and the mosquito, in order to obtain an independent measure of evolutionary rate. This independent measure is vital because of the likelihood that gene duplication itself influences the rate of evolution. After obtaining these rates, the researchers compared the evolutionary rates of duplicated and nonduplicated genes. Stated simply, the authors found that slowly evolving (that is, more conserved) genes are more successful at generating duplicates than faster evolving genes. This is no recent trend—more conserved genes have been better at generating duplicates of themselves consistently over hundreds of million of years. Phylogenetic studies show that slowly evolving genes are more likely to be duplicated than faster evolving genes This research challenges the assumption that genes are duplicated in an unbiased manner. In addition, it provides the essential background for other genomic studies of gene duplication. For example, the acceleration of protein evolution upon duplication is likely to be even more dramatic considering that it is the slowly evolving genes that duplicate preferentially. These findings also open up new questions in the study of gene duplication. The authors convincingly demonstrate the bias toward conserved genes in the process of duplication, but how and why does this happen? For a duplicated gene to be retained in a species, the duplicate must be fixed in the population and then must be preserved by natural selection. The preferential duplication of slowly evolving genes might come from a bias in either of these steps, and the authors outline several models for why this might be the case. Further analysis may enable researchers to test these and other models for gene duplication—especially as more sequence data become available—and learn more about this potent phenomenon in genome evolution.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC368177.xml
521076	Psychosocial risk factors for obesity among women in a family planning clinic	Background The epidemiology of obesity in primary care populations has not been thoroughly explored. This study contributes to filling this gap by investigating the relationship between obesity and different sources of personal stress, mental health, exercise, and demographic characteristics. Methods A cross-sectional survey using a convenience sample. Five hundred women who attended family planning clinics were surveyed and 274 provided completed answers to all of the questions analyzed in this study. Exercise, self-rated mental health, stress, social support, and demographic variables were included in the survey. Multiple logistic regression analysis was performed. Results After adjusting for mental health, exercise, and demographic characteristics of subjects, analysis of the data indicated that that being having a large family and receiving no support from parents were related to obesity in this relatively young low-income primary care sample, but self-reported stress and most types of social support were not significant. Conclusion Obesity control programs in primary care centers directed at low-income women should target women who have large families and who are not receiving support from their parents.	Background Many variables may influence eating behavior and therefore may also influence obesity, including depression, anxiety, stress, social support, race, ethnicity, education and income [ 1 - 8 ]. While the national media and federal websites emphasize the importance of physical activity in controlling body weight, exercise alone is not effective for this purpose. In fact, some research shows no relationship between exercise and body weight in community samples [ 1 ]. Therefore, the important questions are: what other risk factors in addition to exercise may affect obesity and how can they be changed? Studies of stress as a risk factor for obesity are limited. In fact, few epidemiological studies have been reported in primary care journals. An exception is found in a report by Sammel et al, who included a stress index in their study of weight gain among women in their late reproductive years (ages 35–47) [ 1 ]. Three hundred and thirty-six women were followed for four years. A 14-item Perceived Stress Scale was used to assess the degree to which situations were stressful to the subject. Women who gained more than ten pounds were not different in regard to this stress measure than other subjects. Stress can be measured in different ways. The personal circumstances experienced by individuals might be expected to have a direct effect on depression, anxiety, and general health status. After all, when conflict arises within a family, the psychological consequences can be dramatic. Even though the relevance of various sources of personal stress to obesity has not previously been examined in the community health literature, their potential importance is worthy of investigation. The purpose of the study reported here was to investigate the importance of personal stressors in determining obesity. The sample was drawn from a low-income female population: women using a family planning clinic for primary care. Exercise, social support, mental health and other personal characteristics were measured and held constant in order to determine the independent effect of different sources of personal stress on obesity. Methods A cross-sectional survey of primary care patients (adult non-pregnant women) attending one of five Planned Parenthood clinics in the Panhandle of Texas was conducted. The Amarillo Institutional Review Board granted exempt status to the study because no protected health information was collected. Planned Parenthood provide a valuable and unique sampling frame because it supplies basic primary care to low-income women in this area, including birth control, but not abortions. Eligibility for the study was limited to patients who were over age 18 and not pregnant. Questionnaires were placed on a table in waiting areas, with a poster inviting participation. Clinic staff also handed out survey forms. Subjects placed the completed forms in a sealed box. Sealed boxes containing survey forms were returned to Texas Tech for data entry. Microsoft Access was used for data entry. Five hundred forms were distributed. Twenty surveys were returned by persons ineligible for study participation and were excluded from the sample. The final data set was comprised of 345 subjects. Computing the response rate as completed returns divided by eligibles (345/(500-20) produces a participation rate of .719. Complete cases were available from 274 subjects for the multivariate analysis. Measures The dependent variable was obesity. Body mass index (BMI) was computed as weight in pounds times 703 divided by height in inches squared. BMI greater than 30 was classified as obese, and made up about 20 percent of the sample. Cases with missing heights and weights were dropped. Independent variables were stress, social support, age, race/ethnicity, education, income, number of persons in the home, exercise, marital status, anxiety, and depression. Key instruments are discussed below. Health items were taken from the Duke Health Profile [ 9 ]. Mental health was measured in terms of feeling depressed or sad and nervousness. Possible responses for all of the mental health items were "Yes, describes me exactly", "Somewhat describes me," or "No, doesn't describe me at all." Stress and social support items were taken from the Duke Social Support and Stress Scale (DUSOCS). The DUSOCS contains items addressing personal support from various sources. A person who stresses the respondent is defined as one who causes problems or makes life more difficult [ 10 ]. Respondents were asked how much they were stressed by spouses, parents and children. Possible responses were "None," "Some," "A Lot," and "There is No Such Person". Categories were combined into 'none' versus 'a lot or some.' Exercise was measured in terms of times per week (none, one day, two days, three days, four days, more than four). Age, number of persons in the home, race/ethnicity (non-Hispanic white, Hispanic, other), marital status, and educational level (less than high school, high school or GED, more than high school) were used to control for demographic differences among subjects. The median age was 25. Age was categorized as 18–21, 21–30 or 31 or over. Breaks in the age distribution were made at the first and third quartile. Statistical analysis Chi-square tests were used to test for the relationship between each independent variable and obesity. Variables that were significant at p < .10 in univariate tests were included in a multiple logistic regression analysis. EpiInfo 3.2.2 was used for data analysis. Results Nearly half of respondents were classified as being obese (47.8 percent). About half of the respondents reported 'a lot or some' stress from parents or a spouse. 'A lot or some' stress from a child was experienced by about 40 percent. More than three-fourths of respondents said they receive 'a lot or some' support from a spouse or parent. About sixty percent received 'a lot or some' support from a child. Over 90 percent reported 'a lot or some' support from a friend. Table 1 shows the results of the univariate chi-square tests. Of the three stress variables, only parent stress met the selection criterion for inclusion in the logistic regression model (p = .0988). Support from parents was marginally related to obesity (p = .0542) while support from a child was significantly to obesity (p = .0390). Table 1 Psychosocial Risk Factors and Percent Obese in Family Planning Clinics (Chi-square tests) Pct Obese Pct Not Obese p Overall 47.8 52.2 Nervous .6064 None 46.7 53.3 Some 45.2 54.8 A lot 57.1 42.9 Depression .1944 None or some 45.6 54.4 A lot 59.4 40.6 Stress from Parents .0988 None 52.8 47.2 A lot or some 42.7 57.3 Stress from Spouse .8084 None 49.1 50.9 A lot or some 46.8 53.2 Stress from Child .1285 None 44.6 55.4 A lot or some 54.2 45.8 Support from Spouse .1607 None 55.4 44.6 A lot or some 45.6 54.4 Support from Child .0390 None 39.5 60.5 A lot or some 42.2 47.8 Support from Parents .0542 None 61.7 38.3 A lot or some 45.2 54.8 Support from Friend 3472 A lot or some 45.9 54.1 None 57.1 42.9 The sample was comprised of relatively young women, with most being under 30 years of age. Most respondents lived with two or more other people and most had high school degrees. More than one-fourth of respondents were married. Over one-fourth were Hispanic and over half were non-Hispanic White. Most had incomes under $30,000 per year. Over 35 percent got no exercise at all. Table 2 shows the results of univariate chi-square tests for the demographic variables and for exercise. Obesity differed significantly by the number of persons in the home (p = .0047), level of education (.0060), income level (p = .0328), and marital status (p = .0183). Over 58 percent of married respondents were obese, compared to 42.5 percent of unmarried persons. Person who lived alone were much less likely to be obese than persons who lived with four or more people (32.5 percent vs 64.8 percent). Over sixty percent of those lacking high school education were obese, whereas only about 40 percent of those who had more than a high school education were obese. The $10–20,000 income category had the lowest percent obese (36.1). Table 2 Other Risk Factors and Percent Obese in Family Planning Clinics (Chi-square tests) Pct Obese Pct Not Obese p Days of Exercise per Week .3857 None 49.6 50.4 One 57.1 42.9 Two 53.4 46.6 Three 39.6 60.4 Four 31.8 68.2 Five or more 41.7 58.3 Missing 54.5 45.5 Number of persons in home .0047 None 32.5 67.5 One to three 44.7 55.3 Four or more 64.8 35.2 Education .0060 Less than high school 63.6 36.4 High school degree or equivalent 55.3 44.7 More than high school 39.7 60.3 Race/Ethnicity .3747 White, non-hispanic 44.6 55.4 Hispanic 51.1 48.9 Other 54.8 45.2 Income .0328 Less than $10,000 54.2 45.8 $10–20,000 36.1 63.9 $20–30,000 56.4 43.6 Over $30,000 46.3 53.7 Marital Status .0183 Married 58.3 41.7 Other 42.5 57.5 Age .3857 Less than 21 49.2 50.4 21–30 44.6 55.4 over 30 53.7 46.3 'Some' or 'a lot; of nervousness was reported by about one-third of respondents, while over one-third said they were depressed 'some' or 'a lot'. Neither depression nor anxiety was retained for use in the multivariate model, since significance levels were below .10. Variables that were significant at p < .10 were included in the multiple logistic regression model (see Table 3 ). Women who reported no support from parents had greater odds of being obese (adjusted odds ratio (AOR) = 2.17, p = .0420). Stress from parents and support from a child had no independent relationship with obesity. Persons who lived in homes of four or more were more likely to be obese (AOR = 4.05, p = .0089). Being in the $10,000 to $20,000 income category lowered the odds of obesity in comparison to the under $10,000 category (AOR=.4864, p = .0267). Table 3 Unconditional Logistic Regression of Obesity in Family Planning Clinics (N = 274) Variable Odds Ratio (Conf. Interval) P Stress from parents (none vs a lot or some) 1.2427 (.73–2.13) .4294 Support from child (none vs a lot or some) .8949 (.51–1.57) .6994 Support from parent (none vs a lot or some) 2.1710 (1.03–4.58) .0420 Number in home One to three vs none 1.8413 (.80–4.24) .1518 Four or more vs none 4.0503 (1.42–11.55) .0089 Income $10–20 vs less than 10 .4864 (.26–.92) .0267 $20–30 vs less than 10 1.1426 (.54–2.40) .7248 over $30 vs less than 10 .4945 (.20–1.19) .1176 Marital status Other vs married .6095 (.32–1.16) .1340 Education High school vs less 1.1435 (.43–3.07) .7900 More than HS vs less .7489 (.29–1.91) .5439 Comparisons of cases with missing obesity information to cases with complete obesity information revealed no significant differences in regard to age, marital status, or income. However, ethnicity and education were significantly different for persons who were missing obesity information. Missings were more likely to be Hispanic or other than non-Hispanic white. Missings also were less likely to have a high school degree or higher. Discussion According to our univariate analysis, the profile of an obese woman in this low-income population is having a large family, less than a high school education, and being married. They also were more likely to fall into income groups above or below the $10,000 to $20,000 range. Variables assessing stress from various sources were not significant at p < .05. Multivariate analysis revealed that receiving no support from parents was independently related to higher rates of obesity, while women in the $10,000 to $20,000 income category were less likely to be obese. The reasons for the income differences are not clear, though varying access to food assistance may offer a partial explanation. Additional investigation of the relationship between diet and income among low-income women is needed. The findings of this study should be treated with caution since it is based on a convenience sample and may not be representative of the population from which it was drawn. In addition, the response rate was not optimal and also a number of cases were dropped from the analysis due to missing information, which reduced statistical power and could have biased our conclusions. However, since the sample was not randomly selected we cannot be sure that dropping cases with missing data made the sample less representative of the low-income female primary care population. An additional limitation of the study was its cross-sectional in design which does not allow for proving causal relationships. Because of these limitations, our results must be considered suggestive rather than definitive. Nevertheless, the findings may be important to primary care physicians, epidemiologists and others who study the determinants of obesity in clinic populations. We could not demonstrate a significant relationship between self-reported stress and obesity in this relatively young, female population after adjustment for other variables. However, personal stress may have indirect effects on obesity, an issue not investigated in this study. Furthermore, the relationship between self-reported (perceived) stress with objective (psychological and physiological) measures of stress in this population group are unknown. Consistent with the findings of our study, Sammel et al did not find stress to be related to obesity in their study of women aged 35–47 [ 1 ]. Exercise was not significantly related to obesity in our data. Interestingly, Sammel et al also found no relationship between exercise and obesity 1]. Kaplan et al reported that physical activity was related to obesity in older Canadians, but their sample was quite large (N = 5,980) thus giving them more statistical power [ 3 ]. Obesity rates increased with age in our univarate analyses. This is consistent with what has been reported by other investigators [ 1 , 3 ]. We found no independent relationship with education or race, which agrees with Sammel but conflicts with other research [ 3 , 4 , 8 ]. This contradiction might be due to the fact that our sample was constrained to include primarily low-income women; obesity may be more strongly related to income than race, ethnicity or educational level. We could not show self-assessed depression to be predictive of obesity. Several other investigators have examined this issue, with some seeing obesity as a consequence of depression and others regarding depression to be a result of obesity. In a large study by Carpenter et al [ 11 ], increased BMI in women was associated with major depressive disorder as diagnosed in a structured interview using DSM-IV criteria. A study of the National Health and Nutrition Examination Survey (NHANES) also found that obesity was related to depression. Furthermore, Sammel et al reported that weight gain was related to baseline depression, providing some support that depression may be a cause rather than a consequence of obesity [ 2 ]. Goodman and Whitaker studied the development and persistence of adolescent obesity and found that depressed mood at baseline was an independent risk factor of persistent obesity [ 4 ]. Obesity may also increase the risk of depression in women due to stigma and social isolation related to obesity, particularly among women in western cultures. Since our depression variable was drawn from a single question, it may have contained too much measurement error to permit it to achieve statistical significance in our data set. The reasons why large families increase the risk of obesity are not entirely clear. One obvious mechanism is that, since women traditionally prepare meals, they may have more frequent opportunities to consume food and households with large families are more likely to have greater volumes of food available. Conclusion The research question for this paper was about the risk factors for obesity in a low-income female population in a single community. Our study differs from some other studies of obesity by its inclusion of several types of personal stress as well as social support and mental health measures. We were able to show that personal stress, as defined and measured in this study, was not an important risk factor for obesity in this population group. A limitation of this study is that it does not address occupational stress. Work-related stress has been shown to be related to the health of employees [ 12 - 14 ]. However, since many of the subjects in this sample were not employed, it was not feasible to test hypotheses regarding the relationship between work stress and obesity. Another limitation of the study is its cross-sectional design, which precludes making firm conclusions regarding causality. Despite these limitations, our findings have significance for public health practice related to weight control. Health promotion programs that seek to educate and encourage healthier eating behaviors in low-income female populations should focus on women who are not receiving support from their parents and have large families of their own. In addition, income and eligibility for food assistance may affect dietary practices in unexpected ways, so primary care providers should explore this issue. Competing interests None declared. Authors' contributions JR planned the study, organized the survey, and wrote the first draft of the results. BR wrote the section addressing mental health and obesity. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC521076.xml
550664	Assembly of a gene sequence tag microarray by reversible biotin-streptavidin capture for transcript analysis of Arabidopsis thaliana	Background Transcriptional profiling using microarrays has developed into a key molecular tool for the elucidation of gene function and gene regulation. Microarray platforms based on either oligonucleotides or purified amplification products have been utilised in parallel to produce large amounts of data. Irrespective of platform examined, the availability of genome sequence or a large number of representative expressed sequence tags (ESTs) is, however, a pre-requisite for the design and selection of specific and high-quality microarray probes. This is of great importance for organisms, such as Arabidopsis thaliana , with a high number of duplicated genes, as cross-hybridisation signals between evolutionary related genes cannot be distinguished from true signals unless the probes are carefully designed to be specific. Results We present an alternative solid-phase purification strategy suitable for efficient preparation of short, biotinylated and highly specific probes suitable for large-scale expression profiling. Twenty-one thousand Arabidopsis thaliana gene sequence tags were amplified and subsequently purified using the described technology. The use of the arrays is exemplified by analysis of gene expression changes caused by a four-hour indole-3-acetic (auxin) treatment. A total of 270 genes were identified as differentially expressed (120 up-regulated and 150 down-regulated), including several previously known auxin-affected genes, but also several previously uncharacterised genes. Conclusions The described solid-phase procedure can be used to prepare gene sequence tag microarrays based on short and specific amplified probes, facilitating the analysis of more than 21 000 Arabidopsis transcripts.	Background Extensive transcriptional profiling of the plant model system Arabidopsis thaliana has been limited when compared to other model organisms, such as human and mouse, mainly due to the lack of high-quality cDNA microarrays offering genome-wide coverage. However, during the recent years both academic and commercial alternatives to these cDNA arrays have emerged. The public initiative by the CATMA consortium has aimed at the production of high-quality probes for each of the 29 787 genes predicted in the Arabidopsis genome [ 1 , 2 ]. The design of the CATMA gene sequence tag (GST) probes is based on de novo gene prediction from the genome sequence [ 1 , 3 , 4 ], since only a relatively now number of ESTs are available for Arabidopsis (dbEST at NCBI contains only about 320 000 Arabidopsis ESTs compared with 6 million for the human species). Commercial alternatives for genome-wide monitoring of the Arabidopsis transcriptome have been developed by Affymetrix, Agilent Technologies, MWG Biotech, Operon and others. In a recent study the CATMA, Affymetrix and Agilent arrays were found to perform equally, but with a minor advantage for the CATMA arrays in terms of dynamic range [ 5 ]. In the first phase of the CATMA program 21 120 GSTs covering more than 70% of the predicted genes were designed. The length of the probes is kept low to ensure specificity and ranges from 150 bp to 500 bp, with the size distribution heavily shifted towards the shorter fragments. To further increase the specificity of the GSTs their distribution is shifted towards the 3'-end of the genes, with 60%, 16% and 24% representing 3'-, centre and 5'-regions, respectively [ 1 , 3 ]. Both 3' and 5' untranslated regions of the genes were included in the design. As a consequence of the GST fragment length, an efficient and robust high-throughput method for purification of short fragments is needed. Here we demonstrate that the purification can be accomplished by taking advantage of the recent finding that the streptavidin-biotin bond can be broken, in a fully reversible fashion, without denaturation of the protein [ 6 ]. The approach is based on incorporation of a biotin molecule during PCR amplification of the GSTs, binding of the products to streptavidin-coated paramagnetic beads using high ionic-strength conditions and elution through disruption of the streptavidin-biotin bond in a non-denaturing and fully reversible fashion with deionised water. We exemplify that this feature can be applied for generation of high-quality gene sequence tag microarrays in a cost-effective and high-throughput manner. We also demonstrate the use of these arrays by presenting results on the alteration in gene expression levels at different time points in Arabidopsis plants treated with physiological concentrations of the well-known plant hormone indole-3-acetic acid (auxin). Finally, we compare our results with those obtained in two previous studies [ 7 , 8 ] carried out on the Affymetrix 8 k Gene Chip platform to identify auxin regulated gene expression. Results and discussion In this study we present a method suitable for purification of gene sequence tags, which have recently been designed and successfully used for transcriptional profiling of the plant model system Arabidopsis thaliana [ 1 , 3 ]. The purification method is based on reversible biotin-streptavidin binding, utilises streptavidin-coated paramagnetic beads and can be automated on a robotic workstation dedicated for magnetic separation and equipped with a temperature control [ 6 ]. We exemplify the performance of the method by studying the purification of three representative biotinylated amplification products and subsequently show that arrays prepared using this method can successfully be used for large-scale transcriptional profiling. The amplification products we use to study the purification process are 500 bp, 1 kb and 1.3 kb in length, covering the size range typically used for probes on cDNA arrays. For successful purification of the probe both efficient capture by the beads and release is important. An example of the capture of a 50-μl PCR product and subsequent release is shown in Figure 1A . As shown, the initial capture and release is highly efficient (upper left panel), and with no product detectable in the eluate corresponding to the second release (lower left panel). Next we analyse the efficiency of the capture reaction using an increasing amount of beads while keeping the amount of PCR product and length of incubation time constant. The results indicate that for a standard 50-μl PCR product highly efficient capture is achieved already at approximately 100 μg of beads for all products up to 1.3 kb (Figure 1B ), but for a highly optimised amplification reaction a higher amount of beads may be necessary (data not shown). As expected, the binding of the biotinylated product to the streptavidin moiety is a rapid and efficient process with the majority of the binding taking place during the first minutes of incubation (Figure 1C ). Shorter products appear to have faster binding kinetics reaching saturation at earlier time points. Also important to note is that the molar amount of captured and eluted product is not equal for the different-sized products, indicating that other factors such as steric hindrance also contribute to capacity of the beads and should be considered for purification of longer products > 1000 bp. Repeated use of the magnetic beads after a single round of capture and elution is the key feature of the described strategy. To investigate the cross-contamination between iterative cycles of purification as well as the total number of bead purifications that can be used without significant loss in performance, we used agarose gel electrophoresis, DNA Lab-on-chip technology as well as a more sensitive approach based on printing of eluates onto glass slides and hybridisation using a fluorescently labelled oligonucleotide complementary to the purified and printed probes. A carry-over free purification requires that all captured product is released at the first elution so that no product is transferred to the next sample to be purified using the same set of beads. We analysed the presence of cross-contamination by analysing the eluates of two consecutive release reactions from a single immobilised product (Figure A, left panels). Furthermore, the cross-contamination issue was analysed by using the reversible beads in six sequential capture reactions containing either a PCR product or a water-only control in an alternating order (Figure A, right panel). A released product is detected in the first eluate, as expected, but not in the second, by all three methods including the sensitive fluorescence assay. As shown in Figure 1A , right panel, hybridisation with a labelled oligonucleotide complementary to the purified probes shows a signal in features originating from a PCR product of the three sizes, but not in features originating from the negative water-only control. We studied the capacity of the beads after multiple capture and release cycles by using a constant amount of the three PCR products as input for each iterative purification cycle. An extra washing step was carried out between the purification cycles. Data for nine consecutive binding, washing, elution and regeneration rounds of the PCR products is presented in Figure 1D . The yields of purified products are similar for six rounds of reuse with a minor decline during the subsequent cycles, which more likely correlates with loss of beads during the washing steps than with reduced capacity. We continued to analyse the efficiency of the bead regeneration and reuse by using amplification products of twelve additional clones (range 0.3 – 2 kb) and a hybridisation-based quantification approach. The clones were amplified, purified using beads reused up to nine times, printed onto glass slides and finally hybridised with a DNA-binding dye to determine the amount of purified product. A clone-wise scaling of the hybridisation signal of each subsequent reuse versus the signal corresponding to the first use was carried out, followed by a calculation of the overall average, which is shown as the solid black line in Figure 1D . The results from the quantification through hybridisation are in close agreement with the pattern observed using the probes discussed in more detail above. Assembly of the Arabidopsis gene sequence tag microarray We applied the described method for purification of 21 120 Arabidopsis biotinylated gene sequence tags (GSTs), with sizes ranging from 150 to 500 bp. The use of GSTs in transcript profiling offers improved specificity when compared to the more common EST or cDNA libraries since each GST has been designed to have minimal cross-hybridisation to other genes, including members of the same gene family. The investigated set of GSTs covers approximately 70% of the genes in the genome, as described in more detail by the CATMA consortium [ 1 ]. The consortia amplification strategy is based on a two-step PCR system. This facilitates, as shown in this study, the incorporation of a biotin label in the second PCR by generic handle sequences introduced at the initial amplification step. This circumvents the need to design individual gene-specific biotinylated primers. The products were purified in an automated fashion onto 200 micrograms of magnetic, streptavidin-coated beads that were reused up to six times. To compensate for the higher molar amount of the GST amplification products, an initially higher amount of beads was used for the purification of the GSTs than was used for the optimisation of the method. After elution with 12 μl water an equal amount of DMSO is added to eluted products, which are then printed onto the glass slides. Changes in gene expression caused by auxin treatment The arrays generated by large-scale purification of GSTs are used in a pilot time-point study where the plant hormone indole-3-acetic acid (IAA), also known as auxin, is used to cause transcriptional changes in Arabidopsis thaliana seedlings. Total RNA is collected at three post-treatment time points and compared, using a reference design, to RNA extracted from untreated plants. A general overview of the data is shown in Figure 2 . Using the filtered and normalised data (for details see Materials) genes which are differentially expressed upon auxin treatment are identified using a Bayesian approach [ 13 , 14 ]. Using a false discovery rate adjusted p-value of less than 0.001 as threshold level for differential expression, a total of 120 and 150 genes are found to be up- and down-regulated, respectively, at one or more of the three time points (see Additional data files 1 and 2 ). As expected, none of these genes are differentially regulated in the control self-to-self hybridisation experiment (Figure 2A ). It is previously known that auxin influences several key processes during plant growth and development and several lines of evidence indicate that auxin regulation of gene expression plays a key role in its mode of action. A particularly well-studied pathway of auxin-regulated gene expression is the auxin induction of the Aux/IAA genes. The Aux/IAA genes encode small short-lived nuclear proteins that interact with the ARF (auxin response factor) family of transcription factors and are thought to modulate the transcriptional activity of the ARFs in an auxin-dependent manner. These ARFs have been shown to bind to auxin-responsive elements (AuxREs) that are found in promotors of several auxin-regulated genes [ 16 ]. The ARFs function as both transcriptional activators and repressors [ 17 ], and the combination of ARF and Aux/IAA proteins is thought to mediate the tissue-specific effects of auxin [ 18 , 19 ]. Thirteen of the up-regulated genes identified in this study are previously shown to be auxin-regulated and include several members of the Aux/IAA family. These exhibit different induction patterns, with for example IAA5 and IAA19 being strongly up-regulated (>30- and 10-fold, respectively) already at 30 minutes, while a two-hour treatment is required for the IAA7 transcripts to reach a two-fold up-regulation. In a recent independent study where Arabidopsis seedlings were treated for only 15 min with 1 μM IAA, all of the Aux/IAA genes listed in Additional data file 1 , with the exception of IAA7 , were found to be up-regulated [ 7 ]. In addition to the Aux/IAA genes four members of the GH3 family that have also been shown to be induced by auxin exhibited a rapid and sustained 2- to 8-fold up-regulation in our study, again confirming the findings reported in a previous study on auxin regulation of gene expression using Affymetrix 8 k oligonucleotide arrays [ 8 ]. A key feature of auxin regulation of development is the polar transport of auxin that is mediated by auxin transporters. Our data indicate that the polar auxin transporters PIN1 [ 20 ] and PIN7 are up-regulated by auxin whereas in contrast the expression of several members of the Aquaporin gene family [ 21 ] are down-regulated. Expression of the PIN transporters is up-regulated already at 30 minutes and remains high throughout the studied time frame. These observations, of control of auxin transporters by auxin, are interesting since it is known that auxin regulates its own transport but to date there has been little data on this type of feedback. Other genes that are influenced by auxin in our study include transcription factors (8 up-regulated and 15 down-regulated), genes involved in signal transduction (7 and 6, respectively), metabolic enzymes (15 and 30, respectively), as well as several genes classified to other categories and also currently unknown genes. The most down-regulated gene at all time points (CATMA5a08790), for example, shows no sequence similarity to any known sequence and has no recorded expression in any of the sequenced tissue libraries deposited into the public domain. These 270 genes are interesting candidates for further research, but it is important that additional validations are carried out to identify and separate the immediate auxin target genes from the indirect. Conclusions We have described an efficient procedure for large-scale purification of gene sequence tags that can be used for several purposes including microarray fabrication. We demonstrate the utility of the technology by applying it to generate more than 21 000 short (150 – 500 bp) and highly specific Arabidopsis gene sequence tags for use as microarray probes in transcriptional profiling. Biotinylated amplification products are rapidly captured and eluted using a reusable streptavidin-coated solid-phase support in an automated high-throughput manner directly compatible with subsequent microarray printing. Our results demonstrate that the assembly and purification of gene-specific tags is an alternative to currently used purification methods, especially suitable for short amplification products such as gene sequence tags. In addition, the possibility to generate single-strand probes in the range of 150–500 nucleotides by a sodium hydroxide treatment of immobilised probes with subsequent elution of the remaining biotinylated strand, opens up for new microarray applications that would extend probe length beyond current oligonucleotide synthesis limits. Methods Optimisation of the purification procedure The performance of the described purification method was investigated by varying either the amount of beads, the length of the incubation time for binding of the biotinylated product to the streptavidin-coated beads and the number of times the beads were reused. We also investigated if multiple reuses of the beads did introduce a well-to-well cross-contamination. The section below describes the general aspects of the purification and is followed by a more detailed description of the experiments carried out to investigate the different above-mentioned aspects of the purification method. For all experiments three randomly chosen EST clones (0.5 kb, 1 kb and 1.3 kb) were amplified in 50-μl reactions containing 20 mM Tris-HCl, pH 8.4, 50 mM KCl, 1.5 mM MgCl 2 , 200 nM dNTPs (Amersham Biosciences Europe GmbH, Sweden), 5 pmole universal sequencing primer (USP, 5'-TAAGCTAGGCACTGGCCGTCGTTTTACAACG-3', MWG Biotech AG, Germany), 5 pmole biotinylated reverse sequencing primer (RSP, 5'-AGGCCTAATGGTCATAGCTGTTTCCTGTGTG-3', MWG Biotech AG) and 1.5 units Platinum Taq DNA Polymerase (Invitrogen AB, Sweden). The temperature cycling (5 min at 95°C, 30 × (30 s at 95°C, 30 s at 64°C, 2 min at 72°C), 10 min at 72°C) was carried out in a Hybaid thermal cycler (Thermo Electron Molecular Biology, MA, USA). Pooling and splitting into aliquots of 50 μl in a clone-wise manner was used to remove variances introduced by the amplification step. All purification steps, including bead dispensing, binding of the biotinylated product to the streptavidin moieties, washing, elution and regeneration of beads, were carried out in the Magnatrix 1200 automated workstation (Magnetic Biosolutions AB, Sweden). The biotinylated amplification products were bound to Dynabeads M-270 Streptavidin beads (Dynal Biotech ASA, Norway) during an incubation at room temperature using a high-salt binding buffer [1 M NaCl, 10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 5% PEG-6000 and 0.1% Tween-20] and when bound, washed with 1 × TE-buffer [10 mM Tris-HCl, pH 7.5, 1 mM EDTA]. During incubation the beads were kept in suspension by mixing through pipetting every third minute. Elution was achieved by breaking the streptavidin-biotin bond in a 20-μl volume using deionised H 2 O. By use of a peltier thermal element, the immobilised products kept in suspension were heated in deionised water to 80°C (1°C / 2 s) for 1 second and cooled to room temperature (1°C / 2 s). Efficient elution is achieved through a combination of elevated temperature, appropriate temperature ramping and incubation at the elevated temperature, as described in more detail elsewhere [ 6 ]. The beads were separated from released products by magnetic separation, reconditioned through a repeated wash procedure with 1 × TE-buffer and, finally, prepared for the next round of purification. Quantification of DNA was carried out using the Nanodrop ND-1000 spectrophotometer (NanoDrop Technologies Inc, DE, USA). For all the purifications described below an aliquot of the pooled PCR product corresponding to a 50-μl reaction was used. The binding capacity of the beads was studied using an increasing amount (10 μg, 50 μg, 100 μg and 150 μg) of fresh beads (first use) while keeping the incubation time constant at 20 minutes. To estimate the variability of the method eight independent purifications were carried out (n = 8). The effect of the length of the incubation time was studied using 150 μg of beads and eight different incubation times (1, 5, 10, 15, 20 or 30 min) with four replicates of each (n = 4). To study the effects of multiple reuses of the beads, the same beads were reconditioned and used up to nine times. The amount of beads in the first capture was 150 μg and the capture time 20 minutes. The variability was estimated using four independent replications. The multiple reuse of the beads was also analysed using a hybridisation based approach. Twelve clones ranging from 0.3 to 2 kb were amplified and purified multiple times using reconditioned beads. The purified products were subsequently printed in eight replicates onto glass slides and quantified using Syto61 (Molecular Probes Inc, OR, USA). The well-to-well carry-over of product was analysed by first purifying one of the PCR products, followed by a purification reaction with no PCR product added (water-only control). This pattern was repeated six times for all three products, while the same set of beads was used for all purifications. Eluates from all these purifications were printed on slides and hybridised with a Cy5-labelled oligonucleotide complementary to the common vector sequence present in all products. Hybridisation was carried out using 10 pmole of the labelled oligonucleotide for 1 hour at 35°C in a hybridisation solution containing 50% formamide, 5 × SSC and 0.1% SDS. Slides were washed with 2 × SSC containing 0.1% SDS (5 min at room temperature) and three times with 1 × SSC (1 min at room temperature). Scanning using the G2565BA DNA microarray scanner (Agilent Technologies) was carried out at the highest possible photo multiplier tube setting in order to reveal low-level signals. Preparation and arraying of gene sequence tags Initial amplification from BAC-clones or genomic DNA was carried out by the CATMA consortium at different nodes throughout Europe [ 1 , 2 ]. One percentage of the first amplification product, obtained using gene-specific primers with 5' handle sequences, was used as template for the second amplification. A total of 51 cycles [11 × (15 s at 94°C, 15 s at 55°C (-1°C / cycle), 30 s at 72°C), 40 × (15 s at 94°C, 30 s at 55°C, 30 s at 72°C)] were carried out in the presence of 20 mM Tris-HCl, pH 8.4, 50 mM KCl, 1.5 mM MgCl 2 , 200 nM dNTPs, 20 pmole of each primer (forward primer biotinylated, Thermo Hybaid GmbH, Germany) and 1.5 units Platinum Taq DNA Polymerase in a total volume of 50 μl. Biotinylated products from each amplification reaction were bound to 200 μg of Dynabeads M-270 Streptavidin beads (reused up to six times) during a 15-minute incubation at room temperature using a high ionic-strength buffer, washed with 1 × TE-buffer and, finally, eluted with 12 μl deionised water. After each capture round the beads were reconstituted, pooled plate-wise and randomly assigned to a new plate. An equal volume of 99.9% dimethyl sulfoxide (DMSO) (Sigma-Aldrich Sweden AB, Sweden) was added and the purified amplification products were arrayed into 22 by 22 patterns in 48 individual blocks with the QArray arrayer (Genetix Limited, UK) and SMP2.5 pins (TeleChem International Inc, CA, USA). When dried, printed products were immobilised to the reactive surface of the Ultra-GAPS slides (Corning B.V. Life Sciences, The Netherlands) using 250 mJ/cm 2 UV-light (Stratalinker, Stratagene Europe, The Netherlands). Auxin treatment and sample preparation 10-day-old Arabidopsis Col-0 seedlings were grown at 22°C in MS medium (Duchefa AB, The Netherlands) supplemented with 0.5% sucrose and using a 24-h photoperiod with 16 h of light at 75 mE m -2 sec -1 PAR. The samples were treated with 1 μM indole-acetic acid for a period of 0, 30, 120 and 240 minutes, washed once with an excess of MS medium with 0.5% sucrose for 5 minutes, frozen in liquid nitrogen and stored at -70°C. For each time point, frozen seedlings from three independent vials were pooled, grinded and total RNA extracted using the RNeasy kit (Qiagen GmbH, Germany). The quality of the RNA was determined using the RNA 6000 Nano kit and the Bioanalyzer instrument (Agilent Technologies, CA, USA). Target labelling, hybridisation, washing and scanning Ten μg anchored oligo dT primer (dT 20 VN, MWG Biotech AG) was annealed to 20 μg total RNA after a denaturating step (10 min at 70°C). The cDNA synthesis reaction was carried out at 42°C for 1 h 45 min in a 30-μl reaction containing 2 mM dNTPs (dTTP:aminoallyl-dUTP in 1:4, unmodified Amersham Biosciences, modified Sigma-Aldrich), first-strand buffer (50 mM Tris-HCl, pH 8.3, 75 mM KCl, 3 mM MgCl 2 ), 0.01 mM DTT and 400 units Superscript II (Invitrogen AB). The synthesis reaction was terminated by addition of EDTA, the RNA strand hydrolysed with NaOH (15 minutes at 70°C) and the reaction neutralised with HCl (final concentrations 20 mM, 150 mM and 150 mM, respectively). The cDNA strands were purified using the MinElute spin columns (Qiagen GmbH) with the provided wash and elution buffers replaced by 80% ethanol and 100 mM NaHCO 3 , pH 9.0, respectively. Monofunctional NHS-ester Cy3 or Cy5 fluorophores (Amersham Biosciences) were coupled to the amino-allyl groups during a 90-minute incubation at room temperature after which unincorporated ester groups were inactivated through a hydroxylamine treatment (final concentration of 730 mM). The pooled labelling reactions were purified using MinElute spin columns and hybridised to the arrays using a two-step protocol in the GeneTac hybridisation station (Genomic solutions Ltd, UK). The pre-hybridisation at 42°C for 45 min (5 × SSC, 1% BSA (Sigma-Aldrich), 0.1% SDS, 40 μg poly(dA) (Sigma-Aldrich) and 20 μg tRNA (Sigma-Aldrich)) was followed by a 16–18 h hybridisation at 42°C with the labelled material and a hybridisation buffer containing 5 × SSC, 25% formamide, 0.1% SDS, 40 μg poly(dA) and 20 μg tRNA. The slides were washed with 2 × SSC and 0.1% SDS at 42°C, followed by 0.1 × SSC + 0.1% SDS at room temperature and finally by three repeated washes with 0.1 × SSC at room temperature. Slides were scanned at 10-μm resolution using the G2565BA DNA microarray scanner for which the photo multiplier tube (pmt) setting was adjusted so that the images for the Cy3 and Cy5 channels were in balance as determined by visual observation. Each time point-to-reference sample comparison was carried out on two arrays, with the dye labels exchanged between the replicated hybridisations in order to avoid sequence-dependent labelling and hybridisation effects. A control self-to-self hybridisation was also carried out for the untreated sample in order to assess the level of noise in the experimental system. Image processing and data analysis The acquired tiff-images were processed using the GenePix 4.1 software (Axon instruments Inc, CA, USA) and the data with the R environment for statistical computing [ 9 ], Bioconductor [ 10 ] and the aroma package for microarray data analysis [ 11 ]. Expression values for each feature and dye channel were obtained by subtracting the median of the local background value from the median of the foreground value. Features for which the background subtracted value were zero or below in one of the channels, but not in the other, were given the expression value of 1. A feature was considered uncertain and removed from subsequent data analysis by setting its value to NA (not available) if a) it was flagged as Not Found by GenePix, b) it was manually flagged as bad (dust particles etc), c) the signals for both channels were saturated, d) the percentage of foreground pixels above the median background + 2 SD were below 60 for both channels or e) the feature diameter was <70 μm or >120 μm. Filtered data was normalised separately for each individual block on the slide using the intensity-dependent lowess method [ 12 ] and no between-slides scaling of the ratio values was deemed necessary. Differentially expressed genes were identified using a moderated t-test based on gene-wise standard errors estimated by an empirical Bayes method [ 13 , 14 ]. Genes with a false discovery rate adjusted p-value of less than 0.001 for any of the three time points were considered as potentially differentially expressed and are included in the Additional data files 1 and 2 . The MIAME compatible data set, including processed and unprocessed data, is made available to the research community through the ArrayExpress expression data repository at the EMBL using the accession number E-MEXP-140 [ 15 ]. List of abbreviations CATMA a complete Arabidopsis thaliana transcriptome microarray EST expressed sequence tag GST gene sequence tag IAA indole-3-acetic acid MIAME minimum information about a microarray experiment Authors' contributions VW carried out the laboratory work and the data analysis and participated in the design of the study and drafting of the manuscript. AH and ML participated in the solid-phase purification procedure. PN participated in the array production. PH designed and provided the CATMA probes. MU participated in the automation of the solid-phase procedure. RB carried out the auxin treatment and participated in the interpretation of the expression data. JL conceived of the study, participated in the drafting of the manuscript and coordinated the study. All authors read and approved the final manuscript. Supplementary Material Additional File 1 List of genes identified as up-regulated by the indole-3-acetic acid treatment. Group labels refer to (A) previously identified auxin regulated, (B) transcription related, (C) signal transduction related, (D) transport, (E) cell wall establishment related, (F) metabolic enyzmes, (G) light related, (H) disease repsonse related, (I) other and (J) unknown genes. Click here for file Additional File 2 List of genes identified as down-regulated by the indole-3-acetic acid treatment. Group labels refer to (A) previously identified auxin regulated, (B) transcription related, (C) signal transduction related, (D) transport, (E) cell wall establishment related, (F) metabolic enyzmes, (G) light related, (H) disease response related, (I) other and (J) unknown genes. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC550664.xml
544836	In vitro activity of antiamoebic drugs against clinical isolates of Entamoeba histolytica and Entamoeba dispar	Background Amoebiasis is a major public health problem in tropical and subtropical countries. Although a number of antiamoebic agents are used for its treatment, yet the susceptibility data on clinical isolates of Entamoeba histolytica and Entamoeba dispar are not available. Therefore, the present study was aimed to assess the in vitro susceptibility of clinical isolates of E. histolytica and E. dispar to metronidazole, chloroquine, emetine and tinidazole. Methods A total of 45 clinical isolates (15 E. histolytica and 30 E. dispar ) were maintained in polyxenic cultures followed by monoxenic cultures. In vitro drug sensitivity (IC 50 ) of clinical isolates and standard reference strain of E. histolytica (HM1: IMSS) was assessed by nitro blue tetrazolium (NBT) reduction assay after exposure to various concentrations of each drug. Results The results showed that all clinical isolates had a higher IC 50 compared to reference strain to all the four drugs. E. histolytica isolates appeared to be more susceptible [IC 50 (μm) 13.2,26.3,31.2 and 12.4] compared to E. dispar isolates [IC 50 (μm) 15.6,28.9,32.8 and 13.2] and the reference strain of E. histolytica [IC 50 (μm) 9.5, 15.5, 29.9 and 10.2] to the metronidazole, chloroquine, emetine and tinidazole respectively. Conclusions The results indicate that till date, Entamoeba isolates in India do not seem to be resistant to the commonly used antiamoebic drugs.	Background Entamoeba histolytica , is the etiological agent of amoebic dysentery and amoebic liver abscess (ALA). Worldwide, 40–50 million symptomatic cases of amoebiasis occur annually and 70,000 to 100,000 deaths due to this infection [ 1 ]. There are two distinct, but morphologically identical species of Entamoeba : Entamoeba histolytica , which is pathogenic and Entamoeba dispar , which is non-pathogenic. E. histolytica , has the capacity to invade intestinal mucosa resulting in intestinal amoebiasis and cause extra intestinal amoebiasis [amoebic liver abscess (ALA)] [ 2 ]. Infection is primarily treated by instituting antiamoebic therapy. Drugs of choice for invasive amoebiasis are tissue active agents, like metronidazole, tinidazole and chloroquine or the more toxic emetine derivatives, including dehydroemetine. Metronidazole and tinidazole are derived from 5-nitroimdazole which kill the trophozoites by alterations in the protoplasmic organelles of the amoeba, but are ineffective in the treatment of cyst passers. Chloroquine is derived from 4-aminoquinolines, which acts on the vegetative forms of the parasite and kills it by inhibiting DNA synthesis. Emetine, a plant alkaloid, kills the trophozoites of E. histolytica mainly by inhibiting protein synthesis. Indiscriminate use of drugs has led to an increase in the minimum inhibitory concentration (MIC) of these therapeutic agents [ 3 ]. Although, drug resistance to E. histolytica does not appear to be a serious problem, there are occasional reports of failure with metronidazole suggesting that this could probably be heralding the development of drug resistance clinically [ 4 ]. Recurrence of ALA even after treatment with metronidazole has been reported and parasites may survive in spite of adequate treatment [ 5 ]. However, differences in drug sensitivity between strains of E. histolytica have been reported, indicating that there may be a small percentage of amoebae which are either resistant to the drug or may even eventually become resistant due to abuse of antiamoebic agents [ 6 ]. Although, earlier studies have been focused on in vitro sensitivity of the only axenic strains of E. histolytica [ 7 - 9 ], yet to the best of our knowledge, studies on in vitro drug susceptibility studies on clinical isolates of E. histolytica and E. dispar have not been reported. Therefore, in the present study an attempt has been made to assess the in vitro activity of antiamoebic drugs (emetine, chloroquine, metronidazole and tinidazole) against clinical isolates of E. histolytica and E. dispar . Methods Clinical isolates Forty-five isolates from patients attending the Out Patient Departments of Nehru hospital, attached to the Post Graduate Institute of Medical Education & Research, Chandigarh, India, identified earlier [ 10 ] as either E. histolytica (15) or E. dispar (30) by hexokinase isoenzyme analysis and by Techlab ELISA were used in the present study. These have been cultured in modified Boeck and Drbohlav (NIH) medium [ 11 ] followed by Robinson's medium [ 12 ]. Standard reference strain (HM1: IMSS) Reference strain of E. histolytica (HM1: IMSS) maintained axenically in TYI-S-33 medium was included as control [ 13 ]. Preparation of antimicrobial agents The drugs (metronidazole, chloroquine, emetine dihydrochloride and tinidazole) used in the study were procured as pure salt from Sigma-Aldrich Co., St. Louis, MO., 63178 USA. The stock solutions of drugs (each 0.1 M) were prepared in dimethyl sulphoxide (DMSO) [ 14 ] and stored at -20°C till use. The stock solutions were diluted in medium to the required concentration. A starting concentration used was 200 μM, which yielded a maximum concentration in the assay of 17.1 μg/ml metronidazole, 51.59 μg/ml chloroquine, 55.3 μg/ml emetine, and 24.7 μg/ml tinidazole. In vitro drug sensitivity assay Drug sensitivity to all the compounds was carried out by nitroblue tetrazolium (NBT) reduction method [ 15 ]. Each clinical isolate was tested in duplicate along with the reference E. histolytica strain (HM1: IMSS). Amoebae were harvested from 24 hour old cultures and suspended in medium. The parasite count was adjusted to 3 × 10 5 parasites/ml in medium by haemocytometer [ 15 ]. The assay was carried out in microtiter plates (Grenier bio-one, Germany). Briefly, in row A 200 μl of drug and in all other rows (B-H) medium was added and doubling dilutions of the drug were performed down the plate. Final drug concentration in rows A-H was as follows: 100, 50, 25, 12.5, 6.25, 3.12, 1.6 and 0.8 (μM). Further 100 μl of parasite suspension (3 × 10 5 /well) was added to all the rows (A-H). Each test included control (without drug) and blank wells (medium only). The plates were incubated at 37°C for 4 hrs. The contents of the plates were discarded and washed with pre warmed Hank's balanced salt solution (HBSS pH 7.2). Thereafter, 100 μl of NBT/well in HBSS was added and the plates were incubated at 37°C for 45 min. followed by aspiration of the contents. Plates were then washed with HBSS twice and 200 μl/well of DMSO (100% v/v) was added. Following incubation at 37°C for 10 min, the optical density (OD) was measured in an ELISA recorder at 540 nm. The percentage of non-viable organisms, which failed to metabolize NBT and therefore did not produce the dark blue formazan product, was determined by applying the following formula: Percentage of non-viable organisms at each drug conc. = Statistical analysis The mean IC 50 values of all clinical isolates against the four drugs were compared with corresponding IC 50 values of the reference E. histolytica strain (HM1: IMSS). Standard deviation (SD) was used to indicate the extent of variation around group mean values. The p value was calculated using the student's-t test. Results The IC 50 values of emetine, chloroquine, metronidazole and tinidazole for the 45 clinical isolates [15 E. histolytica and 30 E. dispar ] and the reference strain HM1: IMSS were determined by the NBT reduction assay. The mean IC 50 values were significantly higher ( P < 0.001) in E. dispar isolates to all the four antiamoebic drugs as compared to the E. histolytica isolates and the reference E. histolytica strain (Table 1 & Figures 1–4 ). Figure 1 Percentage inhibition of E. histolytica (HM1: IMSS) and clinical isolates of E. histolytica and E. dispar by metronidazole Figure 2 Percentage inhibition of E. histolytica (HM1: IMSS) and clinical isolates of E. histolytica and E. dispar by chloroquine Figure 3 Percentage inhibition of E. histolytica (HM1: IMSS) and clinical isolates of E. histolytica and E. dispar by emetine Figure 4 Percentage inhibition of E. histolytica (HM1: IMSS) and clinical isolates of E. histolytica and E. dispar by tinidazole Discussion Treatment failure among amoebiasis patients often raises the possibility of drug resistance [ 16 ]. In the present study the 15 E. histolytica and 30 E . dispar clinical isolates maintained by in vitro cultivation in monoxenic medium were subjected to drug susceptibility tests against four antiamoebic drugs: metronidazole, chloroquine, emetine and tinidazole by NBT reduction assay. E. histolytica reference strain (HM1: IMSS) was also included in each set of experiments. Results showed a significant difference in drug sensitivity in clinical isolates as compared to the reference strain with all the four drugs. The mean IC 50 values (μm) of the E. histolytica / E . dispar isolates against metronidazole, chloroquine, emetine and tinidazole were 13.2/15.6, 26.3/28.9, 31.2/32.8 and 12.4/13.2 respectively. The IC 50 values (μm) of the reference strain against all the four respective drugs were 9.5, 15.5, 29.9 and 10.2. Recently Upcroft & Upcroft [ 14 ] have reported that the MIC values of metronidazole ranges from 12.5–25 μm for laboratory-passaged E. histolytica strains. Adagu, et.al. [ 9 ] have shown the mean metronidazole IC 50 value as 18.47 μm for the most susceptible isolates of E. histolytica with a > 30 μm value as the cut off for resistance. Burchard & Mirelman, studied in vitro sensitivity to metronidazole and emetine of non-pathogenic zymodemes and showed that all were similarly sensitive to both the drugs (1–10 μg/ml) [ 6 ]. In the present study, clinical isolates maintained in monoxenic culture were used to detect the in vitro sensitivity as earlier it has been concluded that bacterial flora associated with the amebae did not significantly interfere with the test performance and sensitivity values [ 6 ]. Although resistance to metronidazole has been reported against Trichomonas vaginalis [ 17 ], Giardia lamblia [ 18 ] and Leishmania donovani [ 19 ], yet to the best of our knowledge there is no documented resistance among clinical isolates of E. histolytica and E . dispar. Conclusion The results of the present study are in agreement with previous findings [ 6 , 9 , 14 ], except that there was a significantly higher IC 50 value of all four drugs to the clinical isolates as compared to the reference strain. E. dispar isolates showed higher IC 50 values when compared to E. histolytica or reference strain. This is the first report of in vitro drug sensitivity pattern to clinical isolates of E. histolytica and E. dispar . There is definitely a need to monitor the random drug susceptibility among clinical isolates especially in context to widespread use of metronidazole and tinidazole, which are available over the counter in many countries. Increased awareness and continued surveillance for the possible emergence of resistance among clinical isolates is necessary for the ultimate prevention and control of amoebiasis. Authors' contributions DB , carried the practical work mentioned in the manuscript. RS , was responsible for formulation of the project and provided guidance time to time. YC , he is clinician and carried clinical examination and proposed clinical diagnosis of the patient. RCM , he guided the proposed work related to differentiation of Entamoeba histolytica and Entamoeba dispar and in vitro drug sensitivity. NM , proposed the concept for this manuscript and guided the practical work and writing of the manuscript. Table 1 Comparison between 1C 50 value of clinical isolates ( E. histolytica and E. dispar ) vs reference strain (HM1: IMSS) COMPOUND MEAN IC 50 OF CLINICAL ISOLATES (μm ± SD) IC 50 VALUE OF REFERENCE STRAIN (μm ± SD) E. histolytica E. dispar Metronidazole 6.5 ± 0.81* a 15.6 ± 2.12* c 9.5 ± 1.53* b Chloroquine 18.9 ± 1.39* a 28.9 ± 2.45* c 21.5 ± 1.26* b Emetine 26.8 ± 1.27* a 32.8 ± 1.68* c 28.0 ± 2.62* b Tinidazole 8.2 ± 1.09* a 13.2 ± 1.43* c 10.2 ± 0.43* b Results expressed as mean ± SD from two experiments conducted in duplicate Student's t-test [*** P < 0.001] a = Eh Vs Ed Eh – E. histolytica b = Eh Vs C Ed – E. dispar c = Ed Vs C C – Reference strain (HM1: IMSS)	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544836.xml
523852	Characterization of a new full length TMPRSS3 isoform and identification of mutant alleles responsible for nonsyndromic recessive deafness in Newfoundland and Pakistan	Background Mutant alleles of TMPRSS3 are associated with nonsyndromic recessive deafness (DFNB8/B10). TMPRSS3 encodes a predicted secreted serine protease, although the deduced amino acid sequence has no signal peptide. In this study, we searched for mutant alleles of TMPRSS3 in families from Pakistan and Newfoundland with recessive deafness co-segregating with DFNB8/B10 linked haplotypes and also more thoroughly characterized the genomic structure of TMPRSS3 . Methods We enrolled families segregating recessive hearing loss from Pakistan and Newfoundland. Microsatellite markers flanking the TMPRSS3 locus were used for linkage analysis. DNA samples from participating individuals were sequenced for TMPRSS3 . The structure of TMPRSS3 was characterized bioinformatically and experimentally by sequencing novel cDNA clones of TMPRSS3 . Results We identified mutations in TMPRSS3 in four Pakistani families with recessive, nonsyndromic congenital deafness. We also identified two recessive mutations, one of which is novel, of TMPRSS3 segregating in a six-generation extended family from Newfoundland. The spectrum of TMPRSS3 mutations is reviewed in the context of a genotype-phenotype correlation. Our study also revealed a longer isoform of TMPRSS3 with a hitherto unidentified exon encoding a signal peptide, which is expressed in several tissues. Conclusion Mutations of TMPRSS3 contribute to hearing loss in many communities worldwide and account for 1.8% (8 of 449) of Pakistani families segregating congenital deafness as an autosomal recessive trait. The newly identified TMPRSS3 isoform e will be helpful in the functional characterization of the full length protein.	Background Genetic analysis of inherited deafness is a powerful tool for discovering the molecular mechanisms that control the development, function and maintenance of the auditory system. Linkage analysis of deafness segregating in large families has resulted in the mapping of more than 55 non-overlapping loci and the discovery of over 30 genes that are essential for hearing [ 1 ]. Given this extensive genetic heterogeneity, extended families from endogamous populations are ideally suited for identifying novel deafness genes and for genotype-phenotype studies. DFNB8/B10, an autosomal recessive deafness locus, was independently mapped in two consanguineous families from Palestine and Pakistan [ 2 , 3 ]. Haplotype and gene sequence analyses of individuals in these two families led to the identification of mutations in a gene encoding a serine protease, TMPRSS3 [ 4 , 5 ]. TMPRSS3 belongs to a subfamily of type II transmembrane serine proteases, which also includes TMPRSS1, TMPRSS2, TMPRSS4 and TMPRSS5 [ 6 ]. The TMPRSS3 gene, spanning approximately 24 kb on chromosome 21, contains thirteen reported exons [ 4 ]. In humans there are alternatively spliced transcripts ( TMPRSS3 a , b , c and d ), encoding predicted polypeptides of 454, 327, 327 and 344 amino acids, respectively [ 4 ]. Here, we report the identification of a fifth isoform, TMPRSS3e , which encodes 538 amino acid residues, including a signal peptide, and is expressed in many human tissues. We previously reported pathogenic TMPRSS3 mutations in four out of a total of 159 Pakistani families segregating profound congenital recessive deafness [ 5 ]. To determine the contribution of TMPRSS3 mutations to recessive deafness in Pakistan, we screened an additional 290 Pakistani families for linkage to the DFNB8/B10 locus. We also screened TMPRSS3 for mutations in a large kindred from Newfoundland, Canada segregating hearing loss linked to markers for DFNB8/B10 and surprisingly found two different mutant alleles. Methods Family ascertainment Institutional review board approvals (OH93-N-016 and OH95-DC-N-050) were obtained for this study from the National Institutes of Health, USA, the Centre of Excellence in Molecular Biology, Lahore, Pakistan and the Newfoundland and Labrador Medical Genetics Program, Health Sciences Centre, St. John's. Participating individuals gave written informed consent. Medical histories indicated that all four Pakistani families segregate congenital, profound, non-syndromic sensorineural hearing loss. Pure tone air and bone conduction audiometry was performed on affected and unaffected individuals of the family from Newfoundland. Pneumatic otoscopy was used to confirm or deny cases in which conductive hearing loss was suspect. All testing was done in non-sound-attenuated exam room. Prior to testing, a biologic calibration was done in the room used for testing with a listener having normal hearing thresholds (LS). Possible effects of ambient noise on thresholds were then taken into account and only those frequencies unaffected were included. DNA was extracted from venous blood samples from the participating individuals. Linkage and sequence analyses DNA samples were PCR amplified using fluorescently labeled primers surrounding microsatellite repeats at known DFNB loci and analyzed on an ABI 377 DNA sequencer. Genotypes were determined using Genescan and Genotyper software (PE Biosystems). To detect TMPRSS3 mutations in families segregating DFNB8/B10, the sequence of each exon of TMPRSS3 was evaluated in two affected subjects from each family as described previously [ 5 ]. 5' RACE and cloning of TMPRSS3e Sequence analysis of the subcloned PCR product amplified from human retina GETLarge full length cDNA (Genemed Biotechnologies) using primers 5'GGGTTGCTTCAAATGGCTTACTAGATCC3' and 5'CATTTTCCCCCATGGTGACTATTTCAG revealed an additional 385 bp of transcribed sequence downstream of TMPRSS3a exon 1. 5' RACE PCR using human retina Marathon-Ready cDNA as template (Clontech) with primers 5'CAGACCAATGGCCAGTGCTAATATC3' and the AP1 primer was performed using the following thermal cycling conditions: 94°C for 1 min, 25 cycles of 94°C for 30 s and 72°C for 5 min. Five microliters of amplified PCR product was used for the second round of amplification with nested primer 5'TTTTCAAATCATCAAGGCCAAAAAG3' and the AP2 primer. Five microliters of each sample were analyzed on a 1.2% agarose/ethidium bromide gel, extracted using a QIAquick gel extraction kit (Qiagen) and cloned into E. coli using pGEM ® -T Easy vector (Promega). DNA was purified from minipreps using QIAquick miniprep kit (Qiagen) and inserts were sequenced using T7 and SP6 primers. To amplify the full length TMPRSS3e isoform, 5'ATGGTGAGTAAAATGGGTGTGAGGA3' and 5'CTTGGAAGTAGAAAGGGTGGGTTTG3' primers and LA-Taq polymerase (Pan Vera) were used as recommended by the manufacturers. Multiple-tissue cDNA panel analysis Expression of TMPRSS3e was evaluated using a Human Tissue cDNA panel (Clontech). TMPRSS3e cDNA was amplified with primers 5'CCAGAAATGGTGAGTAAAATGG3' and 5'AGCAACAGCATCTGCATCTGGT3' using 20 pg of each cDNA as template following the manufacturer's recommended protocol (Clontech). PCR amplified products were sub-cloned into pGEM-T Easy vector (Promega) and sequenced. Results In four out of 290 newly investigated Pakistani families, nonsyndromic congenital deafness was found to co-segregate with DFNB8/B10 linked haplotypes (Figure 2 ). Mutational analysis of DNA samples from the affected individuals from these families (Table 1 ) revealed three previously reported mutations (207delC, C407R, C194F). In our families segregating 207delC there is a common disease associated haplotype [ 5 ]. The second mutation, C407R, was found previously in affected individuals of two Pakistani families and in a heterozygous state in one person of Indian descent. The same C407R ancestral haplotype was also found in family PKDF040 [ 5 ]. The third mutation, C194F, was reported in one Pakistani family [ 5 ], which has the same haplotype as family PKDF064 (Figure 2 ). Figure 2 Pedigrees of Pakistani families. Four families with nonsyndromic recessive deafness from Pakistan. Table 1 Mutant alleles of TMPRSS3 found in this study Family Mutation Domain Amino acid substitution First reference Family B IVS8+8insT SP Frameshift This study Family B 207delC LDLRA Frameshift [9] PKDF003 207delC LDLRA Frameshift [9] PKDF040 1219T>C SP C407R [5] PKDF064 581G>T SRCR C194F [5] PKDF311 207delC LDLRA Frameshift [9] SP; Serine Protease, LDLRA; Low Density Lipoprotein Receptor A, SRCR; Scavenger-Receptor Cysteine-Rich After completing a genome wide screen we found linkage of deafness to markers for DFNB8/B10 in one branch of a large six-generation pedigree (Figure 1 ) from Labrador [ 7 ], a Canadian island whose small population lives in villages isolated by geography and weather [ 8 ]. Medical histories and pure tone air and bone conduction audiometry revealed nonsyndromic, pre-lingual, severe to profound, sensorineural hearing impairment in ten participating affected individuals. Linkage analysis in this family demonstrated homozygosity for DFNB8/B10 linked microsatellite markers among seven affected individuals from five different sibships (Figure 1 ). An apparent founder DFNB8/B10 associated haplotype was constructed using the available genetic information (Figure 1 ). Three hearing impaired individuals (IV:10, V:2 and VI:1) carry only a single copy of the DFNB8/B10 haplotype, although their deafness is clinically indistinguishable from that of the five homozygotes. Figure 1 Pedigree of Newfoundland family. There are ten hearing impaired individuals in a six generation extended family structure. Drawn below the enrolled subjects is a haplotype for STR markers around the DFNB8/B10 locus on chromosome 21. The carrier status of each person for the mutant alleles of TMPRSS3 found in this family is shown. "C" represents 207delC, while "T" stands for IVS8+8insT. Individuals V:17 and V:20 are compound heterozygotes. Sequencing of the TMPRSS3 gene in affected individuals revealed two mutant alleles. The first mutation occurs in exon 4 of TMPRSS3 and is a deletion of a cytosine at position 207 (207delC) of the mRNA (Figure 1 , Table 1 ). The second mutation inserts a "T" residue at the eighth position after the splice donor site of exon 8 (IVS8+8insT). Although the precise effect of this mutant allele is not known, Genscan predicts this results in the skipping of exon nine. Individual IV:16 is a carrier of IVS8+8insT and her two affected children are compound heterozygotes (Figure 1 ). Neither of these two mutations was found in 100 random normal control DNA samples (200 chromosomes) from Newfoundland. We previously reported a Pakistani family (PKSR7) in which deaf subjects were homozygous for the markers spanning the DFNB8/B10 region with the simulated maximal lod score of 3.8 [ 5 ]. However, we did not find a mutation in the known coding and non-coding exons of TMPRSS3 [ 5 ]. One possible explanation is that family PKSR7 has a mutation in a regulatory element of TMPRSS3 . Alternatively, there might be additional uncharacterized exons of TMPRSS3 . Considering the later possibility, we searched for additional exons and alternatively spliced transcripts of TMPRSS3 . Previous studies have documented four isoforms of TMPRSS3 [ 4 ]. To probe for novel coding sequence of TMPRSS3 we designed primers from the known exons to amplify smaller overlapping cDNA product(s). The amplicon generated using a forward primer from the reported non-coding exon 1 and a reverse primer from the protein coding region of exon 2 was larger than the expected size product based on the sequence of TMPRSS3 isoform a [ 4 ]. After subcloning and sequencing this transcript, an additional 385 bp of sequence was found at the 3' end of exon 1 [ 4 ]. We further confirmed the results by 5'RACE using gene specific primers and human Marathon-Ready retina cDNA as template. Using specific primers we amplified a full-length transcript of this novel isoform, which we designated TMPRSS3e (Figure 3 ). Figure 3 Mutational spectrum, structure and expression of TMPRSS3 isoforms. (A) Coding and non-coding exons of all the known isoforms of TMPRSS3 are shown with black and gray rectangles, respectively. The newly identified isoform e has translation initiation codon (arrow) in exon 1, while the termination codon in exon 13 is marked with an asterisk. Shown also are predicted protein motifs encoded by the 3192 bp long mRNA of isoform e . All the known mutant alleles of TMPRSS3 causing hearing loss are shown above the protein motifs. Modified and updated from Ben-Yosef et al. 2001 [5] (B) Nucleotide sequence of the cDNA encoding the amino terminus of TMPRSS3e and its deduced amino acid sequence. The underlined nucleotide sequence represents the region predicted by SMART to encode a signal peptide. The last ATG shown is the reported translation initiation site for isoform a [4]. (C) RT-PCR specific to the TMPRSS3e transcript was performed on cDNA from seven human tissues, which include retina, lung, liver, heart, pancreas, placenta and kidney as indicated. All tissues, except heart, demonstrated expression of TMPRSS3e . G3PDH was used as a positive control. Although the ATG codon at position 705 (Accession # AY633572) does not meet Kozak consensus criteria, there are two reasons for believing that this is the translation initiation site. First, there are stop codons in all three reading frames upstream of this site. Secondly, the 252 bp open reading frame between ATG 705 and ATG 957 (Figure 3A ), the previously reported initiation codon, includes a predicted signal peptide, which is expected in a secreted serine protease. Considering the ATG at position 705 as a starting codon, analysis of the 1614 bp open reading frame indicates that this larger TMPRSS3e transcript encodes a putative polypeptide of 538 amino acids, including 84 previously unknown amino acids at the amino terminus (Figure 3B and 3C ). The SMART program predicts a cleavable signal sequence between residues 1 through 36 (Figure 3A and 3B ). Among all the known isoforms of TMPRSS3 , only isoform e reported here has a predicted signal peptide, consistent with the hypothesis that TMPRSS3 encodes a secreted serine protease. The remaining 48 new amino acids do not contain other predicted structural or functional motifs (Figure 3A and 3B ). Downstream of the newly identified 84 amino acids, isoform e shares a common protein sequence with isoform a, which includes a transmembrane (TM), low density lipoprotein receptor A (LDLRA), scavenger-receptor cysteine-rich (SRCR) and serine protease domains (Figure 3A ). Through non-quantitative RT-PCR analysis with TMPRSS3e specific primers and multiple human cDNAs, we found expression of this isoform in retina, lung, liver, pancreas, placenta and kidney (Figure 3C ). Mutational screening of exon 1 and exon 2 as well as other exons of TMPRSS3e , using DNA samples from two affected individuals of family PKSR7 did not reveal a pathogenic change. This may be an indication of as yet uncharacterized additional exons of TMPRSS3 , or the hearing loss phenotype in this family is due to a mutation in a regulatory element of TMPRSS3 . Alternatively, it remains possible that family PKSR7 has a mutation in a gene closely linked to TMPRSS3 . Discussion TMPRSS3 is the only protease reported thus far to be involved in nonsyndromic deafness [ 4 ]. In this study, we document the mutant alleles of TMPRSS3 segregating in a large family from Newfoundland and in additional Pakistani families with nonsyndromic, recessive deafness. The Newfoundland family has at least two different mutated genes associated with hearing impairment. The majority of the affected individuals in this family are homozygous for 207delC (Figure 1 ), while two of the hearing impaired subjects are compound heterozygotes for 207delC and a newly described mutation of intron 8 (IVS8+8insT). The 207delC mutation appears to be wide-spread; it has been found in affected individuals of Spanish (one homozygote), Greek (one heterozygote), Newfoundland (homozygous individuals) and Pakistani (homozygous individuals, PKDF003 and PKDF311) origins, and is the most common of the mutant alleles of TMPRSS3 . Although there is a predicted effect of IVS8+8insT mutation on splicing of exon 9, further studies are required to determine the precise role of IVS8+8insT on TMPRSS3 transcripts. The family from Pakistan that originally defined the DFNB8 locus was reported to have childhood onset hearing impairment [ 3 ], while all remaining families segregating TMPRSS3 mutations have pre-lingual deafness [ 4 , 5 , 9 , 10 ]. Mutational screening revealed a splice site mutation (IVS4-6G>A) in the original DFNB8 family, which may allow some normal splicing of TMPRSS3 transcripts [ 4 ]. The second mutant allele (IVS8+8insT) in the Newfoundland family may also be a leaky mutation, although the two individuals who are compound heterozygotes were described as having prelingual hearing impairment. The hearing impaired individuals IV:10, V:2 and VI:1 of our Newfoundland family are carriers of 207delC and we did not find any other TMPRSS3 mutation in these individuals. Individuals IV:10 (affected) and IV:12 (normal hearing sibling) show an identical chromosome 21 haplotype. Similarly the individuals VI:1 (affected) and VI:2 (normal hearing sibling) share identical haplotypes for the TMPRSS3 locus. Hence, it is unlikely that the deafness of individuals IV:10 and VI:1 is due to cryptic TMPRSS3 mutations (Figure 1 ). As connexin 26 mutations are the most common cause of nonsyndromic recessive deafness [ 11 ], we sequenced the coding exon of GJB2 in affected individuals of Newfoundland family and found no mutation. We have analyzed a total of 449 Pakistani families [[ 5 ] and this study] segregating severe to profound congenital recessive deafness and found a total of eight families in which the deafness phenotype is due to mutations of TMPRSS3 . Therefore, the relative contribution of TMPRSS3 mutations in the deaf Pakistani population is approximately 1.8%, a significant amount considering the extensive genetic heterogeneity of deafness in this population. Of ten TMPRSS3 mutations that have been reported world wide, five (IVS4-6G>A, 207delC, R109W, C194F and C407R) are segregating in the Pakistani population [ 4 , 5 , 9 , 10 ]. Among the ten known mutant alleles of TMPRSS3 , six are present in exons common to all isoforms, including isoform " e " reported herein. Of the isoforms of this serine protease, TMPRSS3e has the longest open reading frame and is the only isoform of this gene with a predicted signal sequence at the amino terminus. The identification of 84 additional amino acids, including a predicted signal sequence, may aid functional assays to determine the role of TMPRSS3 in cochlear development and function. TMPRSS3 isoform a was identified with a forward primer (TMa-F; see supplement table B [ 4 ]) in exon 2 and not in exon 1 as illustrated in figure 1 of Scott et al. 2001 [ 4 ]. Therefore, it is possible that TMPRSS3a is an incomplete version of isoform e and isoform a may not exist in vivo . TMPRSS3 message is expressed in supporting cells of the organ of Corti, in the stria vascularis and in the spiral ganglion cells of the cochlea [ 12 ]. Although the specific role of TMPRSS3 in the development and maintenance of the audiosensory apparatus is still unknown, the reported mutant alleles of TMPRSS3 abolish catalytic activity of the serine protease, implying a proteolytic function during the inner ear development [ 12 , 13 ]. Although the in vivo substrate(s) of TMPRSS3 have not been reported in the auditory system, TMPRSS3 is thought to regulate the activity of the epithelial amiloride sensitive sodium channel (ENaC) in vitro , which was suggested to control critical signaling pathway(s) in the inner ear and may have a role in the maintenance of the low sodium concentration of endolymph [ 12 ]. However, the absence of an abnormal auditory phenotype in individuals with Pseudohypoaldosteronism type I (PHA 1), which are homozygous for null alleles of ENaC subunits [ 14 ], suggests that the DFNB8/B10 deafness phenotype is due to aberrant proteolytic processing by TMPRSS3 of some other substrate in the inner ear. Conclusions TMPRSS3 mutations account for hearing loss in 1.8% (8 of 449) of Pakistani families segregating deafness as an autosomal recessive trait. We also identified two recessive mutations of TMPRSS3 segregating in a six-generation extended family from Newfoundland. Our study also revealed a longer isoform of TMPRSS3 with an exon encoding a signal peptide, which should help in the functional dissection of this secreted serine protease. Abbreviations TMPRSS3; Transmembrane Protease, Serine 3 [OMIM entry 605511]. DFNB8; Deafness, childhood-onset neurosensory autosomal recessive 8 [OMIM entry 601072]DFNB10; Deafness, congenital neurosensory, autosomal recessive 10 [OMIM entry 605316] Competing interests The authors declare that they have no competing interests. Authors' contributions ZMA performed DNA sequencing, characterized the novel TMPRSS3 isoform e and prepared the manuscript. XCL analyzed all linkage data from family B from Newfoundland and identified DFNB8/B10 segregation in this family. SDP performed DNA sequencing to determine mutant alleles of TMPRSS3 and confirmed linkage in Pakistani families. SR, KR and ZA enrolled families in Pakistan. TLY re-sampled some members of family B from Newfoundland, confirmed genotypes and sequenced normal controls from Newfoundland. SL participated in the clinical assessment of Newfoundland deaf family members. KD sequenced normal controls from Newfoundland. LM was the genetic counsellor involved in family B enrolment from Newfoundland. BP provided considerable technical assistance. LIS performed hearing tests on Newfoundland family members. EI identified Newfoundland families with inherited deafness. SR was responsible for the enrolment of 449 families from Pakistan. TBF was in charge of oversight, editing and analyses. RJM and ERW trained investigators and were involved in the experimental design and genetic analyses of all families. All authors contributed to and edited the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC523852.xml
538255	Mitochondrial inhibition of uracil-DNA glycosylase is not mutagenic	Background Uracil DNA glycosylase (UDG) plays a major role in repair of uracil formed due to deamination of cytosine. UDG in human cells is present in both the nucleus and mitochondrial compartments. Although, UDG's role in the nucleus is well established its role in mitochondria is less clear. Results In order to identify UDG's role in the mitochondria we expressed UGI (uracil glycosylase inhibitor) a natural inhibitor of UDG in the mitochondria. Our studies suggest that inhibition of UDG by UGI in the mitochondria does not lead to either spontaneous or induced mutations in mtDNA. Our studies also suggest that UGI expression has no affect on cellular growth or cytochrome c-oxidase activity. Conclusions These results suggest that human cell mitochondria contain alternatives glycosylase (s) that may function as back up DNA repair protein (s) that repair uracil in the mitochondria.	Introduction Mitochondrion plays an important role in various cellular functions ranging from synthesis of lipids to maintenance of ion homeostasis [ 1 , 2 ]. However, the singular function that defines this organelle is the production of energy by the electron transport chain. Mitochondrion is also a significant source of reactive oxygen species (ROS), known to be a potent DNA damaging agent [ 3 ]. The integrity of the mitochondrial genome is essential for effective cellular processes. The mitochondrion has various active and passive safe guard strategies to deal with the damaging effects of ROS on the mitochondrial DNA (mtDNA), one of them being the repair of the lesions caused by the ROS production [ 3 ]. Mitochondrial repair is not well studied. It is interesting to note that mtDNA experience more DNA damage than nuclear DNA [ 5 ]. Unlike the nuclear DNA that does not replicate in terminally differentiated cells mtDNA is continuously replicated in cells that have undergone differentiation. Hence lesions in the mtDNA can prove to be more deleterious [ 6 ]. Earlier it was believed that the mitochondria lack DNA repair mechanisms as thymidine dimers were not repaired in the mtDNA [ 4 ]. However, recent evidence indicates that DNA repair mechanism do function in the mitochondria [ 7 , 8 , 25 , 26 ]. Various enzymes that are involved in nuclear DNA repair have isoforms that are targeted to the mitochondria [ 9 , 10 ]. Whether these enzymes function in an identical fashion in the repair of both the nuclear and the mtDNA is not clear. The uracil DNA glycosylase (UDG) removes misincorporated uracil or deaminated cytosine from DNA. Human UDG gene encodes two alternative spliced isoforms, UNG1 and UNG2 [ 11 - 13 ]. Of these the UNG1 is translocated to the mitochondria [ 14 , 15 ]. UNG2 localizes to the nucleus [ 15 ]. Although UNG2's role in repairing nuclear DNA is well established, the role for mitochondrial UNG1 is not well studied. In this paper we inactivated mitochondrial UNG1 by expressing a natural uracil DNA glycosylase inhibitor (UGI) from PBS2 phage that binds to the active site of UDG in equimolar ratio and inhibits the UDG enzyme [ 16 ]. UGI has been successfully used as a tool to examine the role of nuclear UNG2 in base excision repair of misincorporated uracil or deaminated cytosine in the nuclear DNA [ 17 , 18 ]. In order to elucidate the role of UDG in in vivo mtDNA repair we targeted UGI to the mitochondria to inhibit UDG activity. Our studies suggest that mitochondrial inhibition of UDG is not mutagenic. This study indicates that alternative DNA glycosylase(s) may be operative in the mitochondria that might repair uracil in the mitochondrial genome. Materials and Methods Constructs The reading frame of uracil DNA glycosylase (UDG) that codes for functional UDG was amplified by PCR using forward primers (5'CCAGTGCCGCGCGCCAAGATCCATTCGTTGTTTGGAGAGAGCTGGAAGAAG) specific to human uracil DNA glycosylase that had a BssH II site at the 5' end and the reverse primers 5'TTGA TCTCGAGTCACAGCTCCTTCCAGTCAATGGG that had the Xho I site engineered at the 5' end. The template used for the amplification was pTUNGΔ84 [ 13 ]. The PCR fragment was cloned into pCMV/myc/mito (Invitrogen) treated with BssH II and Xho I. The vector has a mitochondial targeting signal of the subunit VIII of human cytochrome c oxidase that facilitates targeting of the cloned protein to the mitochondria. The construct was named as pCMV UNG. The complete reading frame of uracil DNA glycosylase inhibitor gene was amplified using pTZUgi (a gift from Dr. Umesh Varshney) as a template with forward primers (5'CCAGTGCCGCGCGCCAAGATCC ATTCGTTGATGACAAA TTTATCTG ACATC) specific to uracil DNA glycosylase inhibitor from phage PBS2 that had a BssH II site at the 5' end and the reverse primer(5'CGCCCGTTTGATCTCGAGTTATAAC ATTTTAATCCATTAC) which had the Xho I site engineered at the 5' end. The PCR fragment was cloned into pCMV/myc/mito (Invitrogen). The construct was named as pCMV UGI. Transfections Stable transfectants of the above constructs were made in immortalized normal breast epithelial MCF 12A cells using lipofectin as a transfecting agent. Briefly, MCF12A cells were plated to 70 % confluency in a 35 mm dish and transfected with 2 ug of pCMV UNG and pCMV Ugi. The cells were selected using G418 as a selection medium. The clones were selected after plating the cells in a 96 well plate to single cell dilution and the clones were screened for integration using PCR. A pCMV/myc/mito/GFP that has a GFP protein fused to the mitochondrial signal was used as a control to assay the efficiency of transfection and the expression of the protein using the vector. An empty vector was stably transfected and used as a control in all the experiments. PCR Screening of clones for stable integration of the constructs Each construct was assayed for stable integration after transfection using PCR. The primers were the same that were used for amplifying the gene for cloning namely UDG specific primers, forward primer: 5'CCAGTGC CGCGCGCCAAGATCCATTC GTTGTTTGGAGAGAGCTGGAAGAAG reverse primer 5'TTGATCTCGAGTCAC AGCTCCTTCCAGTCAATGGG, for screening UDG stable integrants and UGI specific primers, forward primer 5'CCAGTGCCGCGCGCCAAGATCCATTCGTTGATGACA AATTTATCTGACATC and reverse primer 5'CGCCCGTTTGATCTCGAGTTATAAC ATTTTAATCCATTAC for screening Ugi stable integrants. Briefly, the each clone was transferred from the 96 well plate to a 24 well plate and DNA was extracted when the wells were confluent using standard methods. 100 ng of the DNA was used to PCR amplify the DNA that was transfected. Clones that showed an intact gene were selected for further analysis. Isolation of mitochondria Stable clones and parental MCF12A cells were grown in T75 flask to seventy percent confluency. The cells were washed with 1X PBS and treated with 1.5 ml of 0.04% Digitonin solution (0.4 mg Digitonin /ml,2.5 mM EDTA,250 mM mannitol, 17 mM MOPS., pH 7.4). The cells were thoroughly resuspended and homogenized using ten strokes of a dounce homogenizer on ice. One ml of 2.5 X sucrose mannitol buffer (525 mM Mannitol, 175 mM Sucrose, 12.5 mM tris-HCl., pH 7.49) was added and homogenized further using 20 strokes of the homogenizer. Ten micro liter of the homogenate was visualized under the microscope to assess complete breakdown of the cells. The mitochondria were isolated by differential centrifugation [ 19 ]. The homogenate was centrifuged at 2500 rpm at 4°C to pellet the nuclei and the supernatant was further centrifuged at 2500 rpm till no pellet was visually observed. The supernatant was finally centrifuged at 14000 rpm at 4°C to pellet the mitochondria. Western Blotting Stable transfectants were assayed for production of the UDG protein in the mitochondria by western blotting. Twenty micrograms of the mitochondrial protein was electrophoresd on a 12% SDS polyacrylamide gel and transferred on a nitrocellulose membrane. The membrane was blocked overnight in a blocking solution containing 5% non-fat milk and probed with the primary antibody (1:1000 dilution) against UDG (a gift from Dr. Hans Krokan, Norway). The membrane was washed twice with TBST and probed with a secondary antibody linked to horseradish peroxidase. The bands were visualized using ECL (Amersham Pharmacia) kit. The membrane was then probed for the house keeper protein beta actin to assess for equal loading. RT-PCR RNA from Ugi stably transfected MCF 12A cells was extracted using TRIZOL reagent following the manufacturers instruction. One and a half micrograms of total RNA was used for reverse transcription using Superscript II Rnase H - reverse transcriptase (Invitrogen). Two microlitres of the reverse transcribed products was used in the subsequent PCR reactions. Twenty-five microlitres of the PCR reactions contained 20 mM Tris-HCL, pH 8.4, 50 mM KCl, 1.5 mM MgCl 2 , 200 μM dNTP and 10 picomoles of each primer (forward primer: 5'CCAGTGCCGCGCGCCAAGATCCATTCGTTGATGACAAATTTATCTGACATC and reverse primer 5'CGCCCG TTTGATCTCGAGT TATAACATTTTAATCCATTAC and one unit of Taq DNA polymerase (Invitrogen). The PCR profile consisted of an initial denaturation at 94°C for 5 minutes and 32 cycles of denaturation at 94°C for 45 sec, annealing at 58°C for 1 min and extension for 2 min at 72°C with a final extension at 72°C for 10 min. The PCR products were electrophoresed on a 1% agarose gel stained with ethidium bromide (0.5 μg/ml) and visualized under UV. Flow Cytometric Analysis Proliferation assay was done using a flourescent lipophilic molecule, 5-(and-6)-carboxyfluorescein diacetate succinimidyl ester (CFSE) that gets incorporated into live cells and gets diluted into daughter cells with every cell division. The dilution in the intensity of the dye as estimated by flow cytometry with respect to a "0" hour time point gives an indication of the proliferation of the cells. Cells were plated at a density of 1 × 10 5 in a 60 mm dish and stained for 15 min using the fluorescent dye CFSE (Molecular Probes). Cells were fixed in 70% alcohol just after staining to have a 0 hour time point and after a period of 72 hours. Proliferation was then estimated using flow cytometry using a FACSvantage™, Becton Dickinson [ 20 , 21 ]. SIN1-1 and SNAP treatment and of mitochondrial damage MCF12A parental cells were used for dose optimization of the SIN1 and SNAP. An optimal dose was used for further experiments. The parental and the transfected cells were plated on a 60 mm dish to 70% confluency. Each of the cell lines were treated with 4 mM 3-morpholinosydnonimine (SIN-1) and 2 mM S -nitroso-N-acetylpenicillamine (SNAP), NO donors for a period of 1 hour after which the medium was changed and cells were harvested after 0, 2, 4, 6 hour period intervals. DNA was extracted from these cell lines and Cox I was PCR amplified and sequenced using an automated sequencer (ABI PRISM) for mutation analysis. Uracil DNA Repair Assay Uracil DNA repair assay was conducted as described by Radany et al., [ 17 ]. Oligonucleotides used for the assay were and T-34-mer 5'AGCTTGGCTGCAGGTXGACGGATCCC CGGGAATT-3' containing a uracil or thymine residue at position 16 (X=U or T, respectively) and (A-34-mer and G-34-mer) 5'-AATTCCCGGGGATCCGTCXACCTGC AGCCAAGCT-3' containing an adenine or guanine residue at position 19 (X = A or G, respectively). Twenty picomoles of oligonucleotide substrates were labeled with 32 P using T4 polynucleotide kinase. The labeled products were precipitated and then resuspended in a lower volume of distilled water. These were directly used as single stranded substrates in the enzyme assay. To prepare double stranded substrates twenty picomoles of the labeled products were annealed to 10 pmoles of the unlabelled complementary or mismatch oligos by heating at 70°C and slowly cooling it down to room temperature for an hour. UDG assay was performed using 50 μg of mitochondrial extract in 1X UNG buffer (20 mM Tris-HCl pH8.0,1 mM EDTA,1 mM DTT) and 4 pmoles of labeled oligos. The reaction was carried out at 30°C for 45 min. The assay using commercially available Ugi (NEB) was performed using similar conditions. Ten units of Ugi per reaction was used. Apyrimidinic sites (AP-sites) generated by uracil removal from DNA substrates were hydrolyzed by the addition 0.1 N NaOH and incubating for 10 min at room temperature and terminated using a formamide buffer (80% formamide in 1XTBE) to generate single stranded products. Half of the reaction was electrophoresed using a 15% acrylamide gel containing 8.3 M urea and 1X TBE buffer. The gel was autoradiographed after electrophoresis to visualize the bands. Results Generation of stable transfectants expressing UGI and UDG in the mitochondria Previous studies have shown that uracil DNA glycolyase can be inhibited by PBS2 phage protein UGI in a stoichiometric fashion [ 17 , 18 ]. This protein has been used to inactivate nuclear UDG by targeting it specifically to the nucleus by attaching a nuclear localization signal [ 17 ]. We have used the pCMV/myc/mito (Invitrogen) vector to target UGI protein in the mitochondria to inhibit UDG activity. Expression of UDG (UNGΔ84), that retains the wild type function of the enzyme, was also targeted to the mitochondria and was used as control. The pCMV/myc/mito vector contains a mitochondrial localization signal (MLS) of subunit VIII of human cytochrome c oxidase that specifically targets a protein of choice to the mitochondria. Clones containing stable integration were isolated and were confirmed by PCR upon transfection with UGI gene and the UDG after G418 selection. To confirm that the UGI gene was expressed in transfected cell lines we did RT-PCR analysis (Figure 1 ). Our results show that UGI was expressed (Figure 2 ). Western blot analysis on extracts isolated from mitochondria using antibody against UDG protein demonstrates that cells containing UNG stable integration express higher level of UDG protein in the mitochondria (Figure 3 , lane 3). It is important to note that the UDG band was absent in cells expressing UGI because UDG epitope was not available for binding with antibody. Figure 1 PCR screening for stable integrants of pTZUGI in MCF12A cells. PCR using pTZUGI primers were used to screen for stable integrants. Lane 1 is a positive control (pTZUgi plasmid DNA), lane 2, 3, 4, 5, 7, 8, 9 and 10 show the presence of stable integrants. Figure 2 RT PCR to verify expression of Ugi gene transfected in MCF12A cells using primers specic to the UGI gene: RT PCR products electrophoresed on a 1% agarose gel. Lane 1 shows RT PCR product from MCF12A cell line, lane 2 shows RT PCR product from MCF12A cells transfected with pCMV UNG, lane 3 shows RT PCR product from MCF12A transfected with empty pCMV/myc/mito control vector, lane 4 shows RT PCR product from MCF12A transfected with pCMV UGI vector. Figure 3 Western blot analysis of mitochondrial UDG expression in transfected cell lines: Upper panel shows western blotting of mitochondrial extracts with UDG antibody the lower panel shows the same blot probed with Cox II antibody to assess for equal loading of the samples. Lane 1 is mitochondrial extract from parental MCF12A cells, lane 2 is mitochondrial extract from MCF12A cells transfected with empty pCMV/myc/mito vector, lane 3 is mitochondrial extract from MCF12A cells transfected with pCMV UNG vector, lane 4 is mitochondrial extract from MCF12A cells transfected with pCMV UGI vector. A band of lower molecular weight was seen in some extracts. Expression of UDG and UGI in the mitochondria does not affect cell growth It is possible that inhibition of UDG in the mitochondria may affect cell growth. To determine if UGI expression in the MCF12A cells resulted in a difference in cellular growth, cell cycle analysis was conducted using flow cytometry. The cell cycle distribution of parental MCF12A cells, wild type UNG and UGI transfected cell line and the cell line containing the control vector is shown in figure 4 . Interestingly, a very similar growth pattern was observed between all the cell lines examined. We conclude that expression of UGI in the mitochondria does not affect cell growth. Figure 4 FACS analysis of growth rate using fluorescent dye CFDA-SE: The first graph (green) in each panel shows fluorescent cells at 0 hour time point and the second (black) shows a decrease in fluorescence at 72 hr after the cells proliferate. There is no difference in the growth rate between the parental cell line and the transfected one. Lack of mutations in COXI, COXII, and COXIII gene encoded by mtDNA Our previous studies suggest that inactivation of UDG in yeast Saccharomyces cerevisae leads to mutations in mtDNA [ 25 ]. We therefore asked whether UGI transfected cells showed spontaneous increase in level of mutation in mtDNA. We isolated mtDNA from cell expressing wild type UNG, UGI and the control MCF12 A cells containing vector. We amplified mtDNA encoding COXI, COXII and COXIII by PCR. PCR fragments were sequenced. Sequencing revealed no differences in mtDNA sequence between the cell lines expressing UGI, wild type UNG1 and the cell line containing the vector (data not shown). We also treated the transfected cell lines with two agents SIN1 and SNAP. Both SIN1 and SNAP are known to deaminate mtDNA [ 22 ]. The transfected cells were treated for one hour with the agent and were harvested at different time intervals to accumulate mutations. The DNA from these cell lines was isolated and analyzed by sequencing for mutations in the COXI, COX II and the COX III genes encoded by the mtDNA. Our analysis showed no increase in mutation in mtDNA in the treated cell lines (data not shown). We conclude that UGI expression in the mitochondria does not lead to mutations in mtDNA. Uracil repair is unaffected by inhibition of UDG in the mitochondria It has been previously reported that the UGI protein when targeted to the nucleus lowers the activity of the nuclear UDG enzyme [ 16 ]. To analyze the effect of UGI expression on the mitochondrial UDG activity in the transfected cell line, we carried out UDG activity measurements in mitochondrial extracts with and without commercially available UGI as a control. The commercially available UGI was found to inhibit mitochondrial UDG. However, constitutively expression of UGI in the mitochondria in cell line transfected with UGI was not observed (Figure 5 ). These results suggest two possibilities i) that an alternative uracil glycosyalase activity is present in the mitochondria and/or ii) mitochondrially expressed UGI is incapable of inhibiting UDG present in the mitochondria. Since commercially available UGI does inhibit mitochondrial UDG activity, it is likely that alternative uracil glycosylase(s) are present in the mitochondria. We conclude that uracil repair is unaffected by inhibition of UDG in the mitochondria. Figure 5 UDG activity in mitochondrial extracts of parental MCF12A cells and transfected cell lines: Lanes 1, 3, 5 and 7 show UDG activity in mitochondrial extracts from MCF12A parental cell line, cells transfected with pCMV UNG, cells transfected with pCMV UGI, cells transfected with pCMV/myc/mito control vector and commercially available UDG enzyme, that acted as a positive control, respectively. Lanes 2, 4, 6, 8 and 10 shows an inhibition of UDG activity when commercially available Ugi was added in mitochondrial extracts from MCF12A parental cell line, cells transfected with pCMV UNG, cells transfected with pCMV UGI, cells transfected with pCMV/myc/mito control vector and commercially available UDG enzyme, that acted as a positive control, respectively. Discussion Cells are exposed to DNA damaging agents generated both as a process of normal physiology as well through extrinsic mutagens. Cells repair damage done to the DNA by a variety of repair mechanisms each specific for the type of DNA damage [ 23 ]. Many proteins involved in the repair mechanism are conserved in prokaryotes and eukaryotes. One of the repair mechanisms is the base excision repair pathway that repairs lesions of DNA that involve base modification as well as damage by reactive oxygen species. The enzymes involved in the base excision repair pathway for the repair of the nuclear DNA are well studied [ 23 , 24 ]. Base excision repair involves a DNA glycosylase that cleaves the damaged base by hydrolysis of the glycosidic bond, producing an abasic site. The abasic site generated is then removed by AP endonuclease and the gap is filled by DNA polymerase and then ligated by DNA ligase [ 23 , 24 ]. The first enzyme involved in the base excision repair pathway differs depending upon the lesion introduced in the DNA. Thus uracil DNA glycosylase is specific for misincorporated uracil or deaminated cytosine and would only act on these lesions [ 11 - 13 ]. Oxoguanine DNA glycosylase is specific for 8-oxoguanine and other oxidative species, and 3-methyl adenine glycosylase is specific for alkylated residues [ 10 ]. The mitochondrial DNA is subjected to a greater risk of DNA damage due to reactive oxygen species generated as a result of normal physiology of this organelle. The proximity of the mitochondrial DNA to the electron transport chain makes it more vulnerable to the DNA damaging effects of the reactive oxygen species. Therefore, many of the base excision enzymes including UDG have isoforms that are targeted to the mitochondria [ 9 , 10 ]. UDG's role in the nucleus is well established [ 17 ]. It is also established that UGI, a PBS2 phage encoded protein when expressed inhibits UDG activity in the nucleus [ 17 , 18 ]. In this paper we investigated whether UDG is the major protein that plays an important role in repairing uracil residues in the mitochondria. In order to address this question, we cloned UGI gene in frame with the mitochondrial localization signal present in the pCMV/myc/mito vector. We isolated stably transfected MCF12A cell lines and measured uracil-DNA repair activity in the mitochondria. We found no difference in DNA repair activity of uracil in mitochondrial extracts. These results were further substantiated by lack of spontaneous mutations in mtDNA in the COXI, COXII and COXIII genes. Similar results were obtained after treating the cells with SIN1 and SNAP that deaminate DNA [ 22 ]. Cells expressing UGI also showed no difference in the growth rate suggesting a lack of mitochondrial defect due to UGI inhibition of mitochondrial UDG. Our results of UGI expression in the mitochondria are different when compared with UGI expression in the nucleus. A previous study has shown that expression of UGI results in inhibition of uracil DNA repair in the nucleus and subsequently mutation in the nuclear DNA [ 17 ]. Our results are intriguing and hints to the presence of alternative DNA repair proteins that may repair uracil in mtDNA. Indeed, cells contain several classes of enzymes that can remove uracil residues from DNA and maintain genomic integrity [ 27 ]. These include the thymine-DNA glycosylase (TDG), mismatch specific uracil-DNA glycosylase (MUG) and the single-stranded monofunctional uracil-DNA glycosyalse (SMUG1) [ 27 , 28 ]. It is not clear whether any of these proteins are present in the mitochondria and may function as a back up enzyme when UDG is inactivated by UGI. It is also possible that an extremely low level of mutant mtDNA may be present in the cells expressing UGI in the mitochondria and PCR technique used to identify mutant copies among a heterogeneous population of mtDNA was unable to detect mutant mtDNA molecules. It is also conceivable that targeted UGI is present in a subset of mitochondria and at any given time there is always enough active UDG in vivo and in the extract from untargeted mitochondria to carry out the uracil repair activity in vitro . However, these possibilities are ruled out because UGI expression did not result in lower level cytochrome C oxidase activity (data not shown).	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538255.xml
550658	Nonrandom distribution and frequencies of genomic and EST-derived microsatellite markers in rice, wheat, and barley	Background Earlier comparative maps between the genomes of rice ( Oryza sativa L.), barley ( Hordeum vulgare L.) and wheat ( Triticum aestivum L.) were linkage maps based on cDNA-RFLP markers. The low number of polymorphic RFLP markers has limited the development of dense genetic maps in wheat and the number of available anchor points in comparative maps. Higher density comparative maps using PCR-based anchor markers are necessary to better estimate the conservation of colinearity among cereal genomes. The purposes of this study were to characterize the proportion of transcribed DNA sequences containing simple sequence repeats (SSR or microsatellites) by length and motif for wheat, barley and rice and to determine in-silico rice genome locations for primer sets developed for wheat and barley Expressed Sequence Tags. Results The proportions of SSR types (di-, tri-, tetra-, and penta-nucleotide repeats) and motifs varied with the length of the SSRs within and among the three species, with trinucleotide SSRs being the most frequent. Distributions of genomic microsatellites (gSSRs), EST-derived microsatellites (EST-SSRs), and transcribed regions in the contiguous sequence of rice chromosome 1 were highly correlated. More than 13,000 primer pairs were developed for use by the cereal research community as potential markers in wheat, barley and rice. Conclusion Trinucleotide SSRs were the most common type in each of the species; however, the relative proportions of SSR types and motifs differed among rice, wheat, and barley. Genomic microsatellites were found to be primarily located in gene-rich regions of the rice genome. Microsatellite markers derived from the use of non-redundant EST-SSRs are an economic and efficient alternative to RFLP for comparative mapping in cereals.	Background The genetic maps of grass species have been constructed using a variety of marker types. Most of the older species-specific molecular maps were constructed with RFLP markers, but in recent times there has been increased utilization of PCR-based markers because of accessibility and higher throughput. Conservation of gene content and order has been detected among grass genomes through the use of comparative maps [ 1 , 2 ]. The applications of comparative maps have been discussed many times in the past (See for example: [ 3 ]), however genetic maps are not always designed with a comparative study in mind, thus, current maps from different grass species (and in many cases, within the same species) seldom share an adequate number of common (anchor) markers to allow researchers to bridge across maps with an adequate resolution. This is especially true when comparing the genome maps of the Triticeae tribe with the maps of rice or maize, which on average share 3 to 4 markers per wheat homoeologous chromosome group. The lack of anchor markers for bridging across species is exacerbated as new maps are constructed using PCR-based markers such as AFLP, genomic microsatellites and single nucleotide polymorphisms (SNPs) rather than the transferable but laborious cDNA-based RFLP markers. Genomic SSR (gSSR) markers are biased towards genome specificity [ 4 , 5 ] and generally do not transfer to other species, making them less useful for the generation of comparative maps. For comparative mapping, markers must identify orthologous loci and be polymorphic in two or more species [ 6 ]. Recently, several researchers [ 6 - 17 ] have addressed the lack of transferability of gSSRs to other genomes by limiting primer design to transcribed regions, that are expected to have higher levels of conservation across related organisms. Public EST sequence databases from the Poaceae family can be scanned for the presence of SSRs both in protein-coding regions and in untranslated regions of genes (5' or 3' UTRs). When compared to gSSRs, EST derived SSRs (EST-SSRs) were less polymorphic in a study in hexaploid wheat [ 9 ] with only 25% polymorphism, but the successful markers were of high quality and were also polymorphic in durum wheat. Thiel [ 13 ] reported a higher level of polymorphism in barley (42%). Lower polymorphism requires more effort to design primers for testing a larger set of candidate markers, but the ease and speed of finding SSRs among freely available EST sequence data offsets this extra effort. This approach is only feasible in species for which there have been EST sequencing projects. We used the rice genome sequence generated by the International Rice Genome Sequence Project (IRGSP) [ 18 ] to identify gSSRs that can potentially serve as sources of markers for mapping. In an equivalent experiment, non-redundant sets of transcript sequences from rice, wheat and barley were scanned (dataset was obtained from the TIGR gene-index databases [ 19 ]) and those transcripts containing SSRs were collected and mapped in-silico to the rice genome. SSR-containing transcripts derived from different species, sharing a pre-determined threshold of similarity and matching the same location in rice were considered putative orthologs that may be used as anchors in comparative mapping studies. This paper describes a methodology for developing EST-SSR markers from wheat, barley and rice as markers for developing independent species maps as well as for homologous anchor markers for comparative maps. Over 13,000 untested PCR primer pairs for EST-SSRs were generated from the three gene indices and made available to the research community interested in grass genomes. Researchers are encouraged to evaluate a subset of primer pairs and send feedback regarding their utility to the GrainGenes [ 20 ] database for posting. The list (along with all other materials: scripts, programs source code and database schemas) is available from the additional files as well as from the Triticeae EST-SSR Coordination webpage in GrainGenes [ 21 ]. Results Frequency of microsatellite types and motifs Based on combinations of all four nucleotides, the canonical set of SSR motifs is represented by four different duplets (AC, AG, AT, CG), 10 different triplets, 33 different quadruplets and 102 different quintuplet motifs. In the source sequences, all these basic nucleotide motifs can be represented in variant forms of the same basic set or by their reverse complements but to keep a consistency in the database for estimating frequencies, they were transformed into the canonical motifs. Reverse complements and variants would include, for example, CT for AG and GAG for AGG. Sets of unigene sequences such as the TIGR gene indices have the advantage of built-in elimination of redundant SSR counts allowing for more precise estimates of EST-SSR frequency. The rice genomic SSR (gSSR) counts were processed a-posteriori to eliminate redundancy due to BAC/PAC clone overlaps (see methods). Mononucleotide repeats are common in genomic DNA and some are known to be polymorphic but these were deliberately avoided in the unigene database, because they are usually added by the RNA polymerase and are not present in the template DNA (e.g. poly A tails). Table 1 summarizes the frequencies of SSRs in the TIGR gene indices and the rice genome, grouped by SSR type (di-, tri-, tetra-, and penta-nucleotides) and by several minimum acceptable microsatellite lengths starting at 12 bp or longer. Counts were cumulative, meaning that the totals for SSRs with 12 or more nucleotides included the number of longer SSRs displayed in columns to the right in this table (Table 1 ). If microsatellite expansion were motif-sequence and location independent, under the null hypothesis one would expect that all types of SSRs would expand at the same rate, and thus, the proportions of SSR types would remain equal from short to long SSRs. However, the proportions of SSR types changed with the length of the SSRs (as did the proportions of motifs, Table 2 ) and the different SSR types (and motifs) seemed to expand or contract at different rates in the genome of rice, and at different rates when compared to the unigenes of the other two cereals (Tables 1 and 2 ). SSRs of the trinucleotide type were the most frequent overall. The relatively higher proportion of short trinucleotide-based SSRs in rice unigenes was apparent when compared to the same length categories in wheat or barley unigenes and when compared to the rice gSSRs. The proportion of dinucleotide repeats was greater among genomic microsatellites than among EST-SSRs, and this proportion increased with longer SSRs, overtaking trinucleotide repeats in all datasets when the minimum length was set to 20 bp in gSSRs and 30 bp in EST-SSRs. For instance, in the rice genome 72% of the SSRs longer than 30 bp were of the dinucleotide type; thus, it appears that the rice genomic sequence is relatively richer in dinucleotide SSRs than the gene indices from rice (a subset of the genome) and the other two species. Table 1 Frequency table of perfect and imperfect microsatellites grouped by type of repeat in the non-redundant rice genome (IRGSPnr), rice gene index (Osgi), barley gene index (Hvgi) and wheat's gene index (Tagi) under nine different constraints for minimum SSR length, starting at a minimum of 12 bp. 'n' is SSR count, 'f' is relative proportion (fraction). 'U.f' is unigene fraction: the percentage of unigenes from that species containing at least one SSR. Dataset SSRLength > = 12 > = 14 > = 15 > = 16 > = 18 > = 20 > = 24 > = 30 > = 40 SSR type n f n f n f n f n f n f n f n f n f Osgi dinucleotide 1751 0.07 1206 0.09 950 0.07 950 0.14 673 0.12 507 0.16 366 0.25 237 0.45 136 0.67 trinucleotide 17342 0.67 8921 0.63 8921 0.65 4272 0.62 4272 0.75 1993 0.62 912 0.61 242 0.46 56 0.28 tetranucleotide 4039 0.16 1242 0.09 1242 0.09 1242 0.18 275 0.05 275 0.09 84 0.06 18 0.03 9 0.04 pentanucleotide 2706 0.10 2706 0.19 2706 0.20 458 0.07 458 0.08 458 0.14 124 0.08 33 0.06 2 0.01 Total SSR 25838 14075 13819 6922 5678 3233 1486 530 203 U.f 50% 27% 27% 13% 11% 6% 3% 1% 0.4% Hvgi dinucleotide 1304 0.07 916 0.10 706 0.08 706 0.15 508 0.15 387 0.17 237 0.25 162 0.48 96 0.67 trinucleotide 9176 0.52 4326 0.45 4326 0.46 1990 0.43 1990 0.57 971 0.42 464 0.48 115 0.34 34 0.24 tetranucleotide 4122 0.23 1359 0.14 1359 0.15 1359 0.29 379 0.11 379 0.16 103 0.11 10 0.03 4 0.03 pentanucleotide 2976 0.17 2976 0.31 2976 0.32 594 0.13 594 0.17 594 0.25 157 0.16 53 0.16 10 0.07 Total SSR 17578 9577 9367 4649 3471 2331 961 340 144 U.f 36% 20% 19% 10% 7% 5% 2% 1% 0.3% Tagi dinucleotide 2606 0.08 1890 0.11 1526 0.09 1526 0.18 1122 0.18 890 0.22 668 0.32 478 0.54 317 0.70 trinucleotide 18650 0.55 8671 0.50 8671 0.51 3847 0.46 3847 0.61 1840 0.45 989 0.48 343 0.39 124 0.27 tetranucleotide 8189 0.24 2253 0.13 2253 0.13 2253 0.27 572 0.09 572 0.14 181 0.09 26 0.03 8 0.02 pentanucleotide 4570 0.13 4570 0.26 4570 0.27 773 0.09 773 0.12 773 0.19 243 0.12 33 0.04 5 0.01 Total SSR 34015 17384 17020 8399 6314 4075 2081 880 454 U.f 31% 16% 16% 8% 6% 4% 2% 1% 0.4% IRGSPnr dinucleotide 32202 0.17 23056 0.21 18541 0.17 18541 0.33 14081 0.34 11286 0.39 8516 0.53 6307 0.72 4284 0.81 trinucleotide 78832 0.41 38759 0.35 38759 0.36 17615 0.31 17615 0.43 8392 0.29 4431 0.28 1400 0.16 582 0.11 tetranucleotide 48836 0.25 15030 0.14 15030 0.14 15030 0.26 3789 0.09 3789 0.13 1636 0.10 625 0.07 378 0.07 pentanucleotide 34102 0.18 34102 0.31 34102 0.32 5591 0.10 5591 0.14 5591 0.19 1401 0.09 385 0.04 54 0.01 Total SSR 193972 110947 106432 56777 41076 29058 15984 8717 5298 Table 2 Frequency table (counts and relative proportions) of the ten most common motifs from perfect and imperfect microsatellites under the same minimum length constraints used in Table 1 found in the non-redundant rice genome (IRGSPnr), rice gene index (Osgi), barley gene index (Hvgi) and wheat's gene index (Tagi). IRGSPnr Osgi Hvgi Tagi SSRLength motif prop. count motif prop. count motif prop. count motif prop. count > = 12 CCG 0.18 34238 CCG 0.32 8309 CCG 0.19 3415 CCG 0.20 6643 AG 0.08 14603 AGG 0.10 2501 AGG 0.09 1536 AGG 0.08 2720 AT 0.06 11594 ACG 0.06 1677 AGC 0.07 1250 AGC 0.07 2506 AGG 0.05 10534 AGC 0.06 1583 AAG 0.04 730 AAC 0.05 1805 ACG 0.04 7151 AG 0.04 1074 AG 0.04 714 AG 0.04 1463 AGC 0.03 6576 ACC 0.04 1031 ACG 0.04 668 AAG 0.04 1305 AAG 0.03 4889 AAG 0.04 994 ACC 0.03 604 ACC 0.03 1184 AAAAG 0.02 4809 ATC 0.02 604 ATC 0.03 446 ACG 0.03 1162 AAAT 0.02 4756 ATCG 0.01 341 AGGGG 0.02 426 ATC 0.02 758 ACC 0.02 4517 AAAG 0.01 270 AGGG 0.02 391 AC 0.02 749 > = 14 CCG 0.16 17881 CCG 0.33 4622 CCG 0.18 1717 CCG 0.19 3270 AG 0.09 10293 AGG 0.09 1277 AGG 0.07 716 AGG 0.07 1256 AT 0.08 9385 AG 0.06 822 AGC 0.06 607 AGC 0.07 1190 AGG 0.05 5343 ACG 0.06 776 AG 0.06 543 AG 0.07 1140 AAAAG 0.04 4809 AGC 0.05 774 AGGGG 0.04 426 AAC 0.05 907 AAAAT 0.03 3221 AAG 0.03 482 AAG 0.04 359 AAG 0.03 559 ACG 0.03 3178 ACC 0.03 478 ACG 0.03 287 AC 0.03 530 AGC 0.03 2995 ATC 0.02 273 ACC 0.02 236 ACG 0.03 479 AGAGG 0.02 2627 CCGCG 0.02 249 AC 0.02 231 ACC 0.03 478 AAG 0.02 2448 AGAGG 0.02 248 CCCCG 0.02 224 ATC 0.02 322 > = 16 AT 0.15 8334 CCG 0.32 2238 CCG 0.17 784 CCG 0.17 1452 AG 0.14 7931 AG 0.10 674 AG 0.09 424 AG 0.11 951 CCG 0.14 7820 AGG 0.09 643 AGC 0.07 317 AGC 0.07 560 AGG 0.04 2438 AGC 0.05 354 AGG 0.07 316 AGG 0.06 542 AC 0.03 1714 ACG 0.05 339 AAG 0.04 186 AC 0.05 413 AAAG 0.03 1457 AAG 0.04 249 AC 0.04 175 AAC 0.04 344 AGC 0.03 1421 ACC 0.03 218 AGGG 0.03 154 AAG 0.03 291 AAT 0.02 1403 AT 0.02 127 AGGGG 0.03 150 ACG 0.03 221 AAAT 0.02 1390 ATC 0.02 114 ACC 0.02 115 AGGG 0.02 203 ACG 0.02 1345 ATCG 0.02 112 ACG 0.02 110 ACC 0.02 202 > = 18 CCG 0.19 7820 CCG 0.39 2238 CCG 0.23 784 CCG 0.23 1452 AT 0.17 7103 AGG 0.11 643 AG 0.09 318 AG 0.12 753 AG 0.14 5580 AG 0.09 503 AGC 0.09 317 AGC 0.09 560 AGG 0.06 2438 AGC 0.06 354 AGG 0.09 316 AGG 0.09 542 AGC 0.03 1421 ACG 0.06 339 AAG 0.05 186 AAC 0.05 344 AAT 0.03 1403 AAG 0.04 249 AGGGG 0.04 150 AAG 0.05 291 ACG 0.03 1345 ACC 0.04 218 AC 0.04 124 AC 0.04 267 AAG 0.03 1228 ATC 0.02 114 ACC 0.03 115 ACG 0.04 221 AC 0.03 1133 AT 0.02 100 ACG 0.03 110 ACC 0.03 202 AAAAG 0.03 1131 AGAGG 0.01 66 ATC 0.02 83 ATC 0.02 145 > = 20 AT 0.22 6360 CCG 0.32 1032 CCG 0.16 383 CCG 0.15 625 AG 0.14 4055 AG 0.12 385 AG 0.11 251 AG 0.15 618 CCG 0.12 3431 AGG 0.10 317 AGC 0.07 170 AGG 0.06 262 AGG 0.04 1190 AGC 0.05 156 AGGGG 0.06 150 AGC 0.06 250 AAAAG 0.04 1131 ACG 0.04 143 AGG 0.06 139 AAC 0.06 249 AAT 0.04 1121 AAG 0.04 134 AAG 0.04 101 AC 0.05 195 AC 0.03 806 ACC 0.03 97 AC 0.04 85 AAG 0.04 176 AGAT 0.02 693 AT 0.03 84 ACC 0.03 64 ACC 0.02 92 AAG 0.02 685 AGAGG 0.02 66 CCCCG 0.03 59 ACGAT 0.02 91 AGC 0.02 573 ATC 0.02 57 AGAGG 0.02 52 ACG 0.02 77 > = 24 AT 0.34 5501 CCG 0.29 436 CCG 0.19 180 AG 0.23 484 AG 0.16 2504 AG 0.19 278 AG 0.18 171 CCG 0.13 275 CCG 0.10 1574 AGG 0.10 152 AGC 0.10 93 AAC 0.10 203 AAT 0.05 866 AAG 0.06 83 AGGGG 0.06 56 AGG 0.06 131 AGG 0.04 615 AGC 0.05 69 AAG 0.05 48 AC 0.06 127 AC 0.03 487 ACG 0.04 65 AGG 0.05 47 AAG 0.06 122 AGAT 0.03 486 AT 0.04 65 AC 0.04 41 AGC 0.06 118 AAG 0.03 447 ACC 0.03 48 ACC 0.04 37 ACGAT 0.04 86 AAAAG 0.02 345 ATC 0.02 26 AT 0.03 25 AT 0.03 56 ACAT 0.02 302 AC 0.01 21 ATC 0.02 21 ATC 0.02 39 > = 30 AT 0.54 4680 AG 0.35 184 AG 0.39 133 AG 0.42 368 AG 0.15 1343 CCG 0.17 90 AGGGG 0.09 29 AAC 0.15 132 AAT 0.05 420 AT 0.09 46 CCG 0.07 24 AC 0.08 68 CCG 0.04 310 AGG 0.08 40 AAG 0.07 23 AAG 0.07 59 AGAT 0.03 295 AAG 0.06 32 AGC 0.07 23 AT 0.05 42 AC 0.03 280 AGC 0.05 24 AC 0.05 16 CCG 0.05 42 AAG 0.03 236 ACG 0.03 16 AT 0.04 13 AGC 0.04 37 ACAT 0.02 198 ACC 0.02 13 AAC 0.03 10 AGG 0.03 27 AGG 0.02 176 AAC 0.02 10 ACC 0.03 9 AAT 0.01 13 AAAAG 0.01 119 AGAGG 0.02 9 CCCCG 0.02 8 ATC 0.01 12 > = 40 AT 0.69 3631 AG 0.51 104 AG 0.60 86 AG 0.57 260 AG 0.09 494 AT 0.14 29 AGC 0.07 10 AAC 0.14 62 AAT 0.07 353 AAG 0.08 17 AAG 0.06 9 AC 0.07 31 AGAT 0.04 192 AGC 0.04 8 AT 0.06 8 AAG 0.07 31 AC 0.03 157 CCG 0.03 7 AGGGG 0.05 7 AT 0.06 26 ACAT 0.03 153 ACG 0.03 6 AAC 0.03 4 AGC 0.03 14 AAG 0.02 109 AGG 0.03 6 ACC 0.01 2 ACAT 0.01 5 CCG 0.01 27 AAC 0.02 5 ACAT 0.01 2 AGG 0.01 3 AAC 0.00 24 ATC 0.02 4 ATC 0.01 2 CCG 0.01 3 AGG 0.00 19 ACAT 0.02 4 ACT 0.01 2 AAT 0.01 3 ESTs are a rich source of SSRs The abundance of SSRs (perfect and imperfect) in the unigenes can range from one in every 100 to one in every two unigenes depending on the minimum length (Table 1 ). When all SSRs with a minimum length of 12 bp are tabulated, 50%, 36% and 31% of rice, barley and wheat unigenes have at least one SSR. When the minimum length was raised to 16 bp the proportion was reduced to 13, 10 and 8%, respectively. Rice unigenes had a higher frequency of SSRs than did barley and wheat for most minimum lengths, but not for SSRs longer than 20 bp, where the relative abundance was similar in all three gene indices. The nearly two-fold difference at a minimum length of 18 bp was mostly due to the high abundance of trinucleotide SSRs in rice unigenes relative to wheat and barley. The abundance of trinucleotide repeats decreased by about one half for each repeat unit added to the series. The decline in abundance was steeper for tetranucleotide and pentanucleotide repeats but was less than one half for dinucleotide repeats, which at lengths greater than or equal to 30 bp, became the predominant type in all datasets. At 30 bp or longer, the AT motif was most common among gSSRs while AG was more numerous among EST-SSRs motifs (Table 2 ). Wheat unigenes contained a larger number of SSRs for all repeat length categories, followed by rice and barley unigenes. This is probably because wheat has more than twice the number of unigenes than rice or barley with 109,782 for wheat, 51,569 for rice and 48,159 for barley. The larger number of unigenes in hexaploid wheat may result from divergence of the genes in the three genomes, but also from a relatively larger EST dataset, i.e., more ESTs have been sequenced for wheat, with a sequence redundancy of 3.8×, versus 6× and 2.7× for barley and rice, respectively (see methods). The number of the ten most frequent motifs was tabulated for different minimal SSR lengths (Table 2 ). The relative proportions of motifs fluctuated with different length constraints as well as source species. At a minimum SSR length of 12 bp, CCG was predominant in all datasets, but AT and AG were more frequent in the higher range of minimal SSR lengths. Among dinucleotides in the rice EST-SSRs, AG and AT were the most common, but AG and AC were more common in wheat and barley EST-SSRs. Besides CCG, other frequent trinucleotide motifs were AGG and AGC. The trinucleotide (CCG)n microsatellite was present in both coding regions and UTRs. In coding regions, this triplet has the potential to code for the amino acids proline (CCG), arginine (CGG), alanine (GCC), glycine (GGC), but among these, expansion of the motif leading to additions of the amino acid proline could have the strongest effects on protein structure while alanine and glycine would have relatively small effects. The longest SSRs were genomic microsatellites (as long as 726 bp). Unigenes had few SSRs longer than 40 bp, up to 333 bp in wheat ESTs. Often these were not useful for developing SSR markers because no flanking sequence was available to design primers. The overall mean length for rice gSSRs equal to or longer than 12 bp was 16.5 (s.d = 12.7), with no significant differences in mean gSSR lengths among the chromosomes in rice, while the mean length for EST-SSRs was 15.3 (s.d = 6) with no significant differences among the three species gene indices (t-test 2-sample with unequal variances, p > 0.01). Density of gSSRs and comparison to EST-SSRs mapped in silico in rice chromosome 1 The contiguity of the pseudomolecule sequence of rice chromosome 1 (R1) (a virtual sequence composed of the assembly of tiling path clones), with only eight gaps for the whole chromosome, provided a convenient framework of coordinates for calculating density estimates of gSSRs features (by in-silico scanning with the Sputnik program), and for anchoring rice, wheat and barley unigenes associated with microsatellites (by sequence similarity). Best similarity matches between the rice genome and 8,259 barley, 16,917 rice and 13,565 wheat EST-SSR unigenes were identified using BLASTN. From these, a total of 6,373 EST-SSRs mapped to R1 pseudomolecules including 1,104 from barley, 3,568 from rice and 1,701 from wheat with 88.8%, 98.4% and 89.1% average sequence similarity respectively. Density of gSSRs equal to or longer than 12 bp in R1 ranged from 1 gSSR in 2.8 kbp near the centromere to 1 gSSR per 1.1 kbp in the distal regions (Figure 1A ). For a more stringent subset of gSSRs (≥ 16 bp tetranucleotides, ≥ 18 bp dinucleotides and trinucleotides, and ≥ 20 bp pentanucleotides), the density ranged from 1 gSSR in 10 kbp around the centromere to 1 gSSR in 3.8 kbp in the densest region of the short arm (Figure 1A ). Figure 1 A) Features in the contiguous 42.5 Mb of rice chromosome 1: Comparison between the density (counts per 500 kbp) of the stringent subset of gSSRs (gSSR Stringent: ≥ 16 bp tetranucleotides, ≥ 18 bp dinucleotides and trinucleotides, and ≥ 20 bp pentanucleotides; in orange), density of the best matches to rice unigenes associated with microsatellites (OsgiSSR_vs_chr1; in green) and density of genomic microsatellites (gSSR; in blue) with length ≥ 12 bp. The centromere location is indicated by "CEN". B) Features in the contiguous 42.5 Mb of rice chromosome 1: Comparison between the density of gSSRs ≥ 12 bp (in blue), the density of best sequence similarity matches to rice unigenes associated with microsatellites (OsgiSSR_vs_chr1; in green), and the density of best sequence similarity matches between rice unigenes (Osgi_vs_chr1) and rice chromosome 1 (gray line and red points). C) Rice chromosome 1 density plots of the stringent subset of gSSRs and its component types (di-, tri-, tetra- and pentanucleotides) The comparison of the density of gSSRs to, a) the density of all rice unigenes mapped to R1, and b) the density of EST-SSRs (a subset from the unigenes) mapped to R1 provided an estimate of the relationship between gSSRs and gene regions in the rice genome. There was a striking resemblance in the patterns of the plots for the density of gSSRs and the density of unigene-derived EST-SSRs in R1 pseudomolecules (Figure 1A ). The similarity in density patterns was less apparent but still present between gSSRs and R1 matches to all rice unigenes (Figure 1B ), and these densities were significantly correlated ( r = 0.45, p ≤ 1E-5; Figure 2 ). Figure 2 Linear regression of the density of genes in rice chromosome 1 (roughly estimated by the matches of OsGI sequences to this chromosome: Osgi_vs_chr1) on the density of gSSRs (≥ 12 bp) in rice chr1. Pearson correlation is 0.45. Regression coefficients are highly significant ( P -value = 1.2E-05). Density is expressed in counts per 500 kbp. A decomposition of the set of stringent gSSRs (see the methods section for criteria defining the "stringent gSSRs") by types in R1 (Figure 1C ) showed that the relative proportions of the pentanucleotide gSSRs ≥ 20 bp were consistently lower over the majority of the chromosome, while the proportion of the other three types of microsatellites was higher, indicating that pentanucleotides are only a small component in the non-homogeneous distribution of the stringent subset. Development of primers for cereal EST-SSRs We designed primer pairs for 5,425 wheat, 3,036 barley and 4,726 rice EST-SSRs conforming to the stringent restrictions described in the methods. The average product size expected from the set of designed primers was 217 bp for rice EST-SSRs, 213 for wheat and 218.9 for barley. Of those EST-SSRs, 42% of the wheat and 56% of the barley were mapped in-silico to the rice genome. The additional file 1 contains the list of primer pairs that can be downloaded for testing. Discussion Are microsatellites preferentially associated with gene-rich DNA in rice? Morgante and colleagues [ 22 ] reported that in plants, gSSRs were preferentially associated with non-repetitive DNA such as the gene-rich regions. They found a highly significant, positive, linear relationship ( r 2 = 0.94, p < 0.006) between genomic microsatellite frequency and the percentage of single copy DNA in several plant species with a wide range of genome sizes. Estimates of repetitive and non-repetitive single-copy DNA fractions were based on reviews of the literature describing renaturation kinetics experiments for each of the species. Plant species that have gone through genome expansion due to retrotransposon amplification, such as maize and wheat, had a lower genomic microsatellite frequency indicating that SSR frequency is not a function of overall genome size but rather the relative proportion of single-copy DNA. In this study, the best similarity matches between rice unigene sequences and the genomic sequence of rice chromosome 1 were used to estimate the density of transcribed regions along R1. This estimate was compared to both the density of gSSRs and the density of EST-SSRs (the latter group being the intersection of the set of gSSRs and the set of transcribed regions, or unigenes). The density pattern of transcribed regions (unigenes in R1) and of SSRs within transcribed regions (unigenes in R1 with SSRs or EST-SSRs) followed closely the density pattern of gSSRs in rice chromosome 1 ( r = 0.45, p < 1 × 10 -5 and r = 0.62, p < 1 × 10 -10 , respectively) (Figures 1A, 1B and 2 ). The density (counts per 500 Kbp) of gSSRs along R1 was higher than both the density of transcribed regions and the density of EST-SSRs. A large number of gSSRs that are not already included in the set of EST-SSRs could still be associated with genes because, as Figure 1B suggests, they are preferentially found in genic regions near promoter regions or inside introns and away from the highly repetitive and gene-poor DNA in heterochromatin. In rice, (AT)n SSRs are rare among ESTs, but are the most common gSSRs among the long group (= 20 bp). The (AT)n SSR motif is frequently found along with sequences of the Micropon family of MITEs [ 23 , 24 ] which are associated with gene-rich regions. Other reports have documented a role for SSRs that are associated with genes in the control of gene expression. For example, several human diseases have been linked with events of triplet expansions in the past [ 25 ]. Chromatin remodeling and gene silencing via histone-deacetylation/cytosine-methylation are among the putative functions of SSRs in the vicinity of genes, especially if GC rich. Coffee [ 26 ] showed that histone deacetylation (and methylation of CpG bases) leading to lower expression at the FMR1 locus in fragile X was a consequence of CCG repeat expansion. In another example, the expansion of a (AGC)n SSR in the 3' UTR of the myotonic dystrophy (DM) protein kinase gene could potentially affect the expression due to changes in local chromatin structure [ 27 ]. It has been found that DM patients have a reduced or complete loss of a nuclease-hypersensitive site in the region of the gene. Further analysis showed that the majority of DM protein kinase transcripts from cells carrying the repeat expansion also lacked the last two exons of a normal transcript, showing that the repeat expansions affected the splicing at the 3' end. In rice, although the presence of a single-base mutation breaking an intron splice site is more directly responsible for the difference in phenotypes of the waxy gene, polymorphism due to SSR expansion has been associated with variation of expression levels in different japonica and indica varieties [ 28 , 29 ]. The effect that microsatellites might have in gene expression in plants may be observed as natural phenotypic variation. A strategy to exploit the EST database for microsatellite markers One strategy to better exploit a database of EST-SSRs in order to find polymorphic markers is to first sample the longest SSRs (≥ 30 bp), favoring dinucleotide repeats, then follow with trinucleotide, tetranucleotide and pentanucleotide repeats [ 23 ]. After exhausting the longest SSRs, one would then proceed with another cycle to select shorter SSRs. Short trinucleotide-based microsatellites such as (CCG)n, the most abundant group overall (Table 2 ), are more likely to derive from coding regions, thus reducing the chances for finding polymorphism [ 13 ]. This strategy is based on the following observations from our results and the literature: 1) Dinucleotides are a better source of polymorphic markers than the other types [ 30 ]. 2) Longer SSRs generally have a higher tendency to be polymorphic [ 23 ]. 3) SSRs deriving from UTRs have the potential for a higher polymorphism than those derived from coding regions, which are constrained by purifying selection [ 30 ]. One percent of unigenes from the three species examined in this study have SSRs starting with a minimum of 30 bp (Table 1 ). Overall, 880 wheat unigenes, 530 rice unigenes and 340 barley unigenes contain at least one of these long microsatellites and primer pairs were successfully designed for 451, 276 and 148 of these long EST-SSRs in wheat, rice and barley, respectively. At this minimum length, trinucleotide repeats were not the most frequent. Nearly 50% of the EST-SSRs longer than 30 bp were based on dinucleotide repeats (Table 1 ), with (AG)n being the most common motif (Table 2 ). Yet, the frequency of SSR types among those for which primers could be designed did not follow this pattern. Trinucleotide repeats were still the most common type in this group, followed by dinucleotides. This was due to the fact that dinucleotides are found preferentially in the UTRs of transcripts, and their sequences had fewer surrounding bases to anchor acceptable primers. After relaxing the microsatellite length constraint to a minimum of 20 bp, the overall number of SSRs in the unigenes increased to around 5%. An additional 3,195, 2,703 and 1,991 EST-SSRs in wheat, rice and barley become available for primer design. Acceptable primer pairs were designed for 2,622 wheat, 2,183 rice and 1,476 barley EST-SSRs in this category (which included the set mentioned previously). The set of EST-SSRs with acceptable primer pairs ( Additional file 1 ) were selected from among all dinucleotide and trinucleotide EST-SSRs with a minimum of 18 bp, tetranucleotide EST-SSRs longer than 16 bp and pentanucleotide EST-SSRs 20 bp or longer. However, from 5,424 wheat unigenes associated with microsatellites and having a set of PCR primers in our database, only 2,323 had a best match in the rice BAC/PACs with our stringency settings. The rest (57%) are not anchored to the rice genome but still have potential to provide polymorphic wheat microsatellite markers. The same applies to 44% of the barley EST-SSRs with acceptable primers. The reasons for a large number of wheat/barley unigenes without matches to rice genomic sequence include not having the complete sequence of the rice genome available (the majority of clones were still in sequencing phase 2, with gaps) and having a relatively high stringency setting for filtering wheat and barley sequence comparisons to rice. In previous comparisons between wheat EST unigenes and the same version of the rice genome sequence draft [ 31 , 32 ], we found that 40% of the unigenes did not significantly match a sequence in the rice genome. Conclusion The relative proportions of di-, tri-, tetra-, and penta nucleotide repeats and motifs varied widely depending on length and were not consistent among the species examined. We have shown that ESTs are a good source of SSRs that can be exploited to develop microsatellite markers for wheat, barley and rice. The advantage to this approach is that the sequences are already available resulting in a lower cost than designing and testing microsatellites from anonymous genomic libraries, even if the polymorphism rate for EST-derived markers is lower. EST-SSRs are useful for enhancing individual species maps, but can be used as anchor probes for creating links between maps in comparative studies when designed from sets of orthologous genes, as demonstrated by Yu et al [ 33 ]. The annotation and/or the sequence similarity between putative orthologous genes from two related species can provide the basis for their use in comparative maps. More than 13,000 primer pairs were designed to amplify fragments from a stringent subset of EST-SSRs in wheat, rice and barley and are available to the public for testing. Using a different methodology, our results substantiated the report by Morgante et al [ 22 ] suggesting that microsatellites are predominantly found in the vicinity of genes. In some instances, their presence in the vicinity of genes may implicate a regulating function by mechanisms involving chromatin remodelling and DNA methylation. Methods Source of unigene sequences TIGR's non-redundant gene indices [ 34 ] from wheat, barley and rice were downloaded in January of 2003. The databases were: OsGi rel.11 (January 2003) for rice transcripts, with 51,569 non-redundant sequences after processing 139,918 ESTs; HvGi rel. 5 (December 2002) for barley transcripts, with 48,159 non-redundant sequences from 304,061 initial sequences and TaGi rel. 6 (January 2003) for wheat transcripts with 109,782 non-redundant sequences (415,125 initial sequences). Source of rice genomic sequences All analyses of the rice genome used the version released in December 2002 by the International Rice Genome Sequencing Consortium [ 18 ] and consisted of a minimum tiling path of 3,280 BAC or PAC clones for the 12 rice chromosomes. There were nine pseudomolecules of assembled, contiguous sequence available for rice chromosome 1 that replaced the overlapping clones in that chromosome [ 35 ]. For the rest of the genome, accession numbers for the individual BAC/PAC clones in the tiling path were used to download the corresponding sequence from NCBI GenBank [ 36 ]. The tiling path for chromosomes 2 to 12 was used to facilitate the posterior ordering of clones. Scanning of the rice genome and the non-redundant EST-datasets for SSRs The TIGR gene indices and the genome of rice were scanned with a modified version of Sputnik [ 22 ] available from the University of Delaware [ 37 ] to find all perfect and imperfect SSRs having 2 to 5 nucleotides in the basic repeat unit and at least 12 bp in total length. For imperfect SSRs, up to 10% sequence deviation from a perfect SSR was included. We modified the way the program handles input sequences in the NCBI FASTA format and the format of the program's output, making it easier to export to relational databases. No changes were made to the underlying algorithms written by C. Abajian and modified by Morgante's group. The version of Sputnik used to generate the microsatellite data for this report can be obtained from the GrainGenes EST-SSR coordination webpage [ 21 ], or by downloading the additional file 2 . In order to eliminate the problem of counting the same microsatellites several times in the rice genome due to the redundancy created by overlapping regions between contiguous BAC/PACs in chromosomes 2 to 12, the gSSRs were annotated as redundant or not, according to their location in the tiling path. When located to a region in the BAC/PAC that overlapped with a neighbor clone (based on the tiling path information as well as MegaBLAST [ 38 ] pairwise alignments) only the SSRs belonging to the overlapped region of the top (northern) clone were counted while those present in the overlapped region of the bottom (south) clone were ignored. Of course, all SSRs found in unique, non-overlapping regions of rice clones were counted. A perl script that performed queries and updates to the SQL database (via the DBI perl module) scanned the tables of genomic microsatellites and flagged them according to the procedure explained above. Thus 24% of the genomic microsatellites (of length ≥ 12 bp) found in the rice BAC/PAC clones were ignored, as they were duplicates due to clone overlaps. Table 1 shows the counts and relative proportions of SSRs found in the four datasets (rice, wheat and barley gene indices as well as in the non-redundant rice genomic) for dinucleotides, trinucleotides, tetranucleotides and pentanucleotides when having different minimum microsatellite lengths (greater than or equal to 12 bp) as the starting point. Table 2 , on the other hand, shows the relative proportions of the ten most common motifs for each dataset when subject to different constraints for minimum microsatellite lengths. In-silico mapping of grass Non-redundant EST-SSRs The set of EST unigenes associated with SSRs from wheat and barley was matched against the sequence of the rice genome to provide a putative map location in rice. Only the best hits were recorded for any given EST unigene. The similarity threshold was set at an E-value < 1 × 10 -10 and at least 80% similarity over 100 bp of minimum alignment. Rice EST unigenes were matched with the same criteria except for a higher similarity threshold of 95%. The inferred location in the rice genome for rice EST unigenes was used to estimate the proportion of rice gSSRs that were associated with regions containing genes. PCR primer design We developed a perl script (see Additional file 3 ) that automatically queries the database of EST-SSRs to design primers in batch based on what was learned in previous experiments and on recommendations found in the literature to maximize the chance of selecting polymorphic microsatellite markers. The script used the BioPerl module [ 39 ] to control the Primer3 core program [ 40 , 41 ], feeding each of the SSR source sequences and specifying the target regions to be amplified via PCR. EST-SSRs were selected for primer design when conforming to the following more stringent restrictions (referred to as the set of stringent SSRs or gSSR stringent): a) The SSRs are dinucleotides or trinucleotides of length equal or larger than 18 bp, tetranucleotides equal or larger than 16 bp or pentanucleotides equal or larger than 20 bp. b) The imperfect SSRs have less than 10% mismatches or gaps relative to a perfect SSR of the same length and motif. c) There is a minimum of 50 bp surrounding the SSR edges in the source sequence to allow for possible primer design. The parameters used for the Primer3 program specified an optimal Tm of 60°C with a minimum and maximum of 57°C and 65°C, respectively, and a 30% to 70% GC content with a low chance of dimer or hair-loop formation. The range for PCR product length was set to be between 100 and 300 bp. Abbreviations EST: Expressed Sequence Tag. SSR: Simple Sequence Repeat. gSSR: genomic SSR. EST-SSR: EST-derived SSR. UTR: Untranslated region flanking a coding region in DNA and messenger RNA. IRGSP: International Rice Genome Sequence Project. R1: rice chromosome 1. BAC/PAC: Bacterial artificial chromosomes or bacteriophage P1 artificial chromosomes (for cloning of large DNA fragments). MITE: Miniature Inverted-repeat Transposable Element. Authors' contributions ML did all programming and design of computational experiments and databases. RVK contributed in the first database design. JY did wet-lab testing of a subset of primer pairs. ML and MES drafted the manuscript. All authors read and approved the final manuscript. Supplementary Material Additional File 1 EST-SSR designed primers Table listing the stringent subset of SSRs (≥ 16 bp tetranucleotides, ≥ 18 bp dinucleotides and trinucleotides, and ≥ 20 bp pentanucleotides) found in rice, barley and wheat gene indices for which primer sequences were designed, the source sequences, the primers, their in-silico mapping in the rice genome (to BAC/PAC clones or pseudo-molecule) and relevant metadata. File is a spreadsheet table, compressed with the zip program. The file (14 Mb uncompressed) is also available at the GrainGenes Triticeae EST-SSR Coordination page Click here for file Additional File 2 Modified Sputnik source code and executable This is the source code with modifications, to the microsatellite searching program "Sputnik", originally written by Chris Abajian from the University of Washington at Seattle. The set of files are compressed using the zip program. File is also available at Click here for file Additional File 3 Perl script to design primers in batch The script uses the Bioperl perl modules to control the Primer3 program in order to design primers from a microsatellite database stored in a MySQL database. The script can be modified to accommodate similar schemas on any database engine supported by the perl DBI module. File is also available at Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC550658.xml
551591	Chronic pneumonia with Pseudomonas aeruginosa and impaired alveolar fluid clearance	Background While the functional consequences of acute pulmonary infections are widely documented, few studies focused on chronic pneumonia. We evaluated the consequences of chronic Pseudomonas lung infection on alveolar function. Methods P. aeruginosa , included in agar beads, was instilled intratracheally in Sprague Dawley rats. Analysis was performed from day 2 to 21, a control group received only sterile agar beads. Alveolar-capillary barrier permeability, lung liquid clearance (LLC) and distal alveolar fluid clearance (DAFC) were measured using a vascular ( 131 I-Albumin) and an alveolar tracer ( 125 I-Albumin). Results The increase in permeability and LLC peaked on the second day, to return to baseline on the fifth. DAFC increased independently of TNF-α or endogenous catecholamine production. Despite the persistence of the pathogen within the alveoli, DAFC returned to baseline on the 5 th day. Stimulation with terbutaline failed to increase DAFC. Eradication of the pathogen with ceftazidime did not restore DAFC response. Conclusions From these results, we observe an adequate initial alveolar response to increased permeability with an increase of DAFC. However, DAFC increase does not persist after the 5 th day and remains unresponsive to stimulation. This impairment of DAFC may partly explain the higher susceptibility of chronically infected patients to subsequent lung injury.	Introduction Pseudomonas aeruginosa is a Gram negative bacteria producing a wide array of virulence factors frequently responsible for chronic airway infections in cystic fibrosis (CF) or chronic obstructive pneumonia disease (COPD) patients, as well as acute nosocomial airway infections in intensive care units [ 1 - 3 ]. In acute P. aeruginosa pneumonia, the functional consequences, and particularly lung fluid movements, have been studied extensively. Lung fluid balance is the result of fluid movements following active ion transport by functional alveolar cells, and permeability of the alveolar capillary barrier. In P. aeruginosa -induced acute lung injury (ALI), distal airspace fluid clearance (DAFC) is typically increased at 24 hours through a TNF-α pathway [ 4 ]. Studies have also shown that the capacity of maintaining alveolar active fluid transport is correlated with patient outcome in ALI [ 5 , 6 ]. Lung liquid clearance (LLC) is another functional marker reflecting the capacity of the lung to evacuate fluid instilled in the alveoli outside the lung, LLC involves DAFC, epithelial and endothelial permeabilities [ 7 ]. We previously showed that, even though DAFC is upregulated, LLC is decreased at both 4 and 24 hours in ALI [ 7 ] reflecting a major endothelial injury overwhelming the alveolar response. In chronic infection, these functional consequences on lung fluid balance are less clear. In the 70's, Cash developed an experimental model of chronic pneumonia by intra tracheal injection of P. aeruginosa embedded in agar beads [ 8 ]. Most of the work performed with this model has focused on immunological, inflammatory, or nutritional aspects [ 9 - 12 ]. To the best of our knowledge, no previous work has tried to evaluate alveolar permeability and lung fluid transport in P. aeruginosa chronic lung infection. In order to elucidate these functional aspects we studied lung fluid transport in an experimental model of chronic P. aeruginosa lung infection in the rat. After the validation of the experimental model, we studied alveolar function: alveolar-capillary barrier permeability, lung liquid clearance, distal airspace fluid clearance and its pharmacologic stimulation. Materials and Methods Animals Specific pathogen-free Sprague Dawley rats (n = 280) (230–270 g), (Depre, St Doulchard, France) were housed in the Lille University Animal Care Facility and allowed food and water ad lib . All experiments were performed with approval of the Lille Institutional Animal Care and Use Committee. Preparation of the bacterial inoculum The methodology was adapted from Cash et al [ 8 ]. Briefly, P. aeruginosa (PAO1 strain) was incubated in 125 ml of tryptic soy broth at 37°C in a rotating shaking water bath for 8 hours. The culture was then washed twice, and resuspended in phosphate-buffered saline. The resulting bacterial suspension was 1 × 10 9 CFU/ml. A sample of 1 mL of this suspension was mixed in agarose and mineral oil (Sigma Diagnoses, St Louis, USA) at 56°C. The resulting oil-agar emulsion was cooled to obtain agar beads. Dilutions of the final suspension were cultured to determine the size of the final inoculum. Experimental infection Under a short general anesthesia with ether (Mallinkrodt, Paris, France), with sterile surgical conditions, a small midline incision was made on the neck ventral surface after swabbing it with ethanol. The trachea was exposed by blunt dissection. Using a 28-gauge needle, 0.1 mL of agar beads followed by 0.5 mL of air were inoculated intra-tracheally. Quantitative bacteriological analysis After exsanguination of the animal, the lungs were isolated and homogenized in 2 mL of sterile isotonic saline. Bacterial culture after serial dilutions was performed and bacterial colonies counted after 12 h at 37°C. Antimicrobial therapy In a subgroup of animals, ceftazidime (GlaxoSmithKline, Marly-le-Roi, France), 100 mg/kg, was administered in the peritoneal cavity every 8 hours during 72 hours. Lungs were harvested, homogenized and cultures were performed to confirm bacterial eradication. Serum ceftazidime levels were measured in HPLC. Broncho-alveolar lavage (BAL) Broncho-alveolar lavage (BAL) was performed by cannulating the trachea. Lungs from each experimental group were lavaged with a total of 20 ml in 5-ml aliquots of PBS with EDTA (3 mM). BAL fluid samples were filtered and immediately frozen at -80°C. A cell count was performed directly. Cellular monolayers were prepared with a cytocentrifuge and stained with Wright-Giemsa stain. Cellular morphotype differential was obtained by counting 200 cells/sample and expressing each type of cell as a percentage of the total number counted. Protein concentration in the BAL was measured with an automated analyzer (Hitachi 917, Japan). Histological study After a vascular flushing with sterile isotonic saline through the pulmonary artery, the lungs were removed. Samples were fixed by intratracheal instillation of paraformaldehyde 10 %. Samples were included in paraffin and sections of 5 μm were realized. Analysis was performed after coloration with Hematoxyline-Eosine-Safran (Zeiss, LEO 906). Serum and BAL TNF-α measurement Levels of tumor necrosis factor α (TNF-α), in the serum, and the BAL fluid, were determined by use of commercial immunoassay kits (ELISA) specific for rat cytokines (Quantikine Murine rat TNFα, R&D Systems, Abingdon OX, UK). The reading was performed with a microplate reader Digiscan (Spectracount Packard Instrument Company; Meriden CT USA). BAL and serum measurement of epinephrine and nor-epinephrine Blood and broncho-alveolar lavage fluid were collected on heparin/Na-metabisulfite coated tubes. The samples were centrifuged (2500 g, 4°C), supernatants were frozen (-80°C). Catecholamines are specifically fixed on alumina (pH = 8.7), the eluent is analyzed with an inversed phase H.P.L.C (Coulochem II ESA). The results are expressed in μg/L. Functional study Surgical preparation Sprague-Dawley male rats were anesthetized with pentobarbital (Sanofi, Libourne, France). A catheter (PE-50) was inserted into the left carotid artery in order to monitor systemic arterial pressure (Acqknowledge Software v 3.7.1, Biopac systems, Santa Barbara, CA, USA) and obtain blood samples. An endotracheal tube (PE-220) was inserted through a tracheostomy. The rats were ventilated with a constant volume pump (Harvard Apparatus, South Natick, MA) with an inspired O 2 fraction of 1.0, a peak airway pressure of 8–12 cmH 2 O, and a positive end expiratory pressure of 2 cmH 2 O. The animals were placed in left decubitus position until the end of the protocol. The body temperature was maintained at 37°C. Preparation of the instillate The test solution, used for alveolar instillation, was prepared as follows : briefly, a 5% bovine albumin solution was prepared using Ringer lactate and was adjusted with NaCl to be isoosmolar with the rat circulating plasma [ 13 , 14 ]. A sample of the instilled solution was saved for total protein measurement, and water to dry weight ratio measurements. In different experimental groups, terbutaline (10 -4 M) (Sigma Aldrich, St Quentin Fallavier, France) was added to the instillate or injected intra-peritoneally to the animals. General Protocol For all ventilated rats experiments, the following general protocol was used. After the surgical preparation, heart rate and blood pressure were allowed to stabilize for 1 hour. To calculate the flux of plasma protein into the lung interstitium, a vascular tracer, 1 μCi of 131 I-labeled human albumin, was injected into the bloodstream [ 14 , 15 ]. 131 I-HSA was prepared in our institution according to a standardized technique. Administration of the instillate (3 ml/kg) was performed into the left lung over a 2-min period, using a 1-ml syringe and polypropylene tube (PE 50, Intramedic, Becton Dickinson, Sparks, MD, USA)[ 13 ]. One hour after the beginning of the alveolar instillation, the rat was exanguinated. The lungs were removed, and fluid from the distal airspaces was obtained (aspirate). The total protein concentration and the radioactivity of the liquid sampled were measured. Right and left lungs were homogenized separately for water to dry weight ratio measurements and radioactivity counts. Measurements • Hemodynamics, pulmonary gas exchange, and protein concentration Systemic arterial pressure and airway pressures were measured continuously. Arterial blood gases were measured at one hour intervals. The arterial PO 2 was used to quantify the oxygenation deficit [ 13 , 14 ]. Samples from instillated protein solution, final distal airspace fluid, and from initial and final blood were collected to measure total protein concentration with an automated analyzer (Hitachi 917, Japan). • Albumin flux across endothelial and epithelial barriers The flux of albumin across the lung endothelial and epithelial barriers was used to evaluate the permeability. This method requires measurement of the vascular protein tracer, 131 I-albumin, in the alveolar and extravascular spaces of the lungs. Endothelial permeability was assessed by measuring the ratio of 131 -iodine radioactivity in the aspirate to the radioactivity obtained in the plasma (Asp/plasma), it reflects the leak of the vascular tracer in the alveolar compartment. We estimated the quantity of plasma that entered the instilled lungs by measuring the transfer of the vascular protein tracer, 131 I-albumin, into the extravascular spaces of the instilled lung using the equation of plasma equivalents previously described [ 7 , 13 , 14 ]. • Extravascular lung water (EVLW) and lung liquid clearance (LLC) The EVLW was estimated by gravimetry: 300 μL of the lung homogenate were weighed, to determine the wet weight, and dessicated at 45°C during 7 days, to obtain the dry weight. The blood fraction was calculated from the homogenate hemoglobin supernatant content. The wet to dry weight ratio (W/D) was estimated using the values of the right lung which was not instilled [ 7 , 14 , 16 ]. Lung liquid clearance was calculated as previously described [ 7 ]. • Distal Airspace Fluid Clearance (DAFC): A change of native bovine albumin concentration over the study period (1 h) was used to measure alveolar fluid movement. DAFC was calculated from the ratio of the final unlabeled alveolar protein concentration, compared to the initial instilled alveolar protein concentration. Experimental groups 15 experimental groups were constituted for the study: - A control group (Ctr), which received an intratracheal instillation of sterile saline at the beginning of the protocol - 7 Sterile groups (St) received an intratracheal instillation of sterile beads and were studied at different days after inoculation: St 1, St 2, St 5, St 8, St 15, St 21 and St 28. - 7 Pneumonic groups (Pn) received an intratracheal instillation of Pseudomonas containing beads and were studied at different days after inoculation: Pn 1, Pn 2, Pn 5, Pn 8, Pn 15, Pn 21, Pn 28. Statistical analysis Comparisons between two groups were made using an unpaired, two tailed Student's t -test. Comparisons between more than two groups were made using a one way analysis of variance with post hoc test for multiple comparisons. A value of p < 0.05 was considered as significant. The data are expressed as means ± SD. Results Pseudomonas beads instillation is associated with the development of a chronic infection Clinically, a major weight loss was observed from the second day in P. aeruginosa beads infected animals compared to the sterile beads groups (Figure 1 ). 5% of the infected animals died within the first 48 hours after inoculation, none did in the sterile groups. Figure 1 Evolution of animals' weight during the four weeks of the analysis. An initial weight loss is observed for the infected animals compared to the sterile beads group. Footnote : Data are mean (± SD). Comparisons between groups were made using analysis of variance. p < 0.05 vs the Pn group. Pn : Pneumonic animals, St : animals which received only sterile beads Prior to the instillation, the size of the inoculum was 7.9 10 5 ± 1.5 10 5 CFU/mL. Lung bacterial load reached a peak on the second day of the infection; from the 5 th day, a progressive decrease occurred to finally remain steady between the 15 th day (8.25 ± 5.2 10 4 CFU/mL) and the 3 rd week (1.67 ± 1.63 10 5 CFU/mL). Total broncho-alveolar lavage (BAL) cells slightly increased in the sterile beads group, the difference was however not statistically significant compared to the control group, the analysis showed that the number of cells peaked on the second day and was constituted, at that time, of 25% polymorphonuclear cells and 75% macrophages. The results were not statistically different over time and therefore pooled in Table 1 . In the infected groups, alveolar cellularity was maximum on the 2 nd day mostly polymorphonuclear's neutrophils (PMN). From the 8 th day, the relative number of PMN progressively decreased as alveolar macrophages increased. All the results are summarized in Table 1 . Table 1 Analysis of the bronchoalveolar lavage All the animals who received sterile beads were included in the sterile group and compared to the control and pneumonic groups at respectively 2, 5, 8, 15 and 21 days post instillation. Total cells (× 10 6 )/mL PMNs (%) Macrophages (%) Ctr 0.4 ± 0.1 0.5 ± 0.4 98.5 ± 0.5 St 3.2 ± 0.6 5.6 ± 4.4 92.9 ± 4.4 Pn 2 10.5 ± 2.9 79.8 ± 5.2* 19.0 ± 4.6* Pn 5 7.9 ± 1.7* 19.0 ± 8.0 79.3 ± 8.4 Pn 8 4.9 ± 1.0 3.8 ± 0.7 95.5 ± 1.0 Pn 15 4.0 ± 0.8 1.2 ± 0.6 98.8 ± 0.6 Pn 21 4.9 ± 1.8 2.0 ± 0.6 97.2 ± 0.6 Footnote : Data are mean (± SD). Comparisons between groups were made using analysis of variance with post hoc for multiple comparisons. p < 0.05 vs the other groups. PMNs : Polymorphonuclear neutrophils, Ctr : Control group, St : Sterile group, Pn : Pneumonic group Histologically, in the infected groups, from the 2 nd day, large numbers of PMNs were observed, mostly centered on the alveoli (Figure 2C–D ). Agar beads were clearly observed in the Pn2 group (Figure 2D ). With time, increased extracellular material became more prominent (Figure 2G–L ). The lung architecture of animals inoculated with sterile beads remained strictly normal (Figure 2A–B ). Figure 2 Histological analysis of the different groups, controls and sterile beads instilled animals are compared to pneumonic rats from the second to the 21 st day post instillation. Coloration was performed with Hematoxyline-Eosine-Safran. A : Control group; B : Sterile beads; C-D : Pneumonia on the 2 nd day (the arrow on panel D underlines infected beads); E-F : Pneumonia on the 5 th day; G-H : Pneumonia on the 8 th day; I-J: Pneumonia on the 15 th day; K-L : Pneumonia on the 21 st day. A transient increase of alveolar-capillary barrier permeability is observed on the second day post infection No variation in permeability or clearance was observed between St groups, so all the results were included in a single group (St) for the analysis (at least 5 animals were included in each time point). Alveolar-capillary barrier permeability, evaluated by the leakage of the vascular marker into the alveoli (Asp/plasma ratio), was increased in infected animals on the second day compared to the control group (0.59 ± 0.08 vs 0.11 ± 0.02). This ratio came back to control values from the fifth to the 28 th day. In the St group a moderate but significant increase of the Asp/plasma ratio was observed throughout the study (0.31 ± 0.04). Both lung liquid clearance and DAFC increased on the 2 nd day post infection; DAFC increase is not related to a TNF-α or catecholamine dependent mechanism • Extra-vascular lung water and Lung liquid clearance (LLC) As shown in Table 2 , no difference in wet to dry lung weight ratio was observed between the groups. LLC increased in the pneumonic group on the second day after the infection (p = 0.02) to return to baseline on the 5 th day. A moderate but not statistically significant increase was observed in the Pn15 group (p = 0.13). Table 2 Lung liquid clearance (LLC) and lung wet to dry weight ratio (W/D). LLC increases on the second day post instillation and returns to baseline on the fifth day. W/D remains constant over time. W/D LLC (%) Ctr 4.33 ± 0.87 22.24 ± 3.65 St 4.29 ± 0.24 36.53 ± 4.95 Pn 2 4.66 ± 0.51 45.51 ± 4.26 Pn 5 4.03 ± 0.27 20.99 ± 5.94 Pn 8 3.47 ± 0.81 23.01 ± 2.80 Pn 15 3.92 ± 0.29 36.21 ± 8.23 Pn 21 4.31 ± 0.07 22.37 ± 2.56 Footnote : Data are mean (± SD). Comparisons between groups were made using analysis of variance with post hoc for multiple comparisons. p < 0,05 vs the other groups. Ctr : Control group, St : Sterile group, Pn : Pneumonic groups from the 2 nd to the 21 st days. • Distal alveolar fluid clearance Distal alveolar fluid clearance increased in the Pn2 group (Figure 3 ). This ratio decreased back to baseline on the 5 th day and remained comparable to both the St and the Ctr groups. Figure 3 Evolution of the DAFC over time in sterile and infected beads injected groups. We observe an increase on the 2 nd day post infection, the clearance returns to a basal level on the 5 th day. Footnote : Data are mean (± SD). Comparisons between groups were made using analysis of variance with post hoc for multiple comparisons. p < 0.05 vs the other groups. DAFC : distal alveolar fluid clearance, Ctr : Control group, St : sterile beads injected group, Pn : pneumonic groups from the 2 nd to the 21 st day. We tested whether the increase in DAFC observed at 48 hours was related to a TNF-α or a catecholamine dependent mechanism. No TNF-α was detected between the 2 nd and 21 st days in the serum or the alveolar compartment. Similarly, neither epinephrine nor nor-epinephrine could be detected in the alveolar compartment at 48 hours. The levels recovered in the plasma were comparable between control and pneumonic animals on the 2 nd and the 5 th days (Table 3 ). Table 3 Plasma catecholamines measurement Plasma catecholamines were measured in pneumonic animals on the 2 nd and the 5 th day post instillation compared to the control group. No statistically significant difference could be observed. Ctr Pn2 Pn5 Epinephrine (μg/L) 8.5 ± 2.1 11.2 ± 4.9 14.2 ± 3.5 Norepinephrine (μg/L) 5.8 ± 0.8 7.0 ± 2.2 8.2 ± 1.9 Footnote : Data are mean (± SD). Comparisons between groups were made using analysis of variance with post hoc for multiple comparisons. Ctr : Control group, St : Sterile group, Pn: Pneumonic groups. Distal airspace fluid clearance cannot be stimulated on the 5 th day post infection even after bacterial eradication Even though DAFC returned to baseline values on the fifth day post infection, alveolar function was not normal in these chronically infected animals. First of all, since bacterial load persisted in the alveoli at least a modest increase of DAFC would have been expected in response to this stimulus. This absence of the expected response led us to test the DAFC response, in each group, to well known pharmacological stimuli. • Terbutaline The administration of terbutaline is associated with an increase in DAFC in controls. Stimulation with terbutaline intratracheally could not increase DAFC on the 5 th day post infection, the intraperitoneal injection also failed to increase DAFC (Figure 4 ). Figure 4 Evaluation of DAFC in the control compared to the pneumonic groups on the fifth day post instillation at baseline and after stimulation with terbutaline. A last group received terbutaline after bacterial eradication with ceftazidime administered intraperitoneally. None of the pneumonic groups could increase DAFC after either stimulation or bacterial eradication. Footnote : Data are mean (± SD). Comparisons between groups were made using analysis of variance with post hoc for multiple comparisons. *p < 0.05, statistically different from the control group. DAFC : distal alveolar fluid clearance. Ctr : Control group, Terbut : Control instilled with terbutaline (10 -4 M), St + Terbut : Sterile beads instilled with terbutaline (10 -4 M), Cefta + Terbut : Control group treated with ceftazidime (100 mg/kg/8 h for 72 h) and instilled with terbutaline (10 -4 M), Pn5 : Pneumonic group on the 5 th day, Pn5 + Terbut : Pneumonic group on the 5 th day instilled with terbutaline (10 -4 M), Pn5 + Terbut IP : Pneumonic group on the 5 th day with an intraperitoneal injection of terbutaline, Pn5 + Cefta+ Terbut : Pneumonic group on the 5 th day treated with ceftazidime (100 mg/kg/8 h for 72 h) instilled with terbutaline (10 -4 M). • Terbutaline after bacterial eradication In order to eliminate the possibility of a direct bacterial effect inhibiting the expected response in the chronically infected animals, we performed a comparable stimulation with terbutaline on 10 animals treated with ceftazidime initiated 24 hours after the infection. On the 5 th day, all lungs were sterilized and measurement of ceftazidime levels showed a steady state level at 46.3 ± 4.8 μg/mL. However, even after the eradication of the pathogen, DAFC remained unresponsive to beta-adrenergic stimulation (Figure 4 ). Discussion In our study we validated an experimental model allowing us to explore alveolar function in chronic P. aeruginosa lung infection through measurements of lung liquid movements. In this model of chronic P. aeruginosa lung infection, after observing an initial increase of both alveolar permeability and lung fluid movements, we characterized an impairment of DAFC where, even though DAFC returned to baseline, it remained unresponsive to pharmacological stimuli. In the first part of our work, we validated, on several parameters, the chronic infection model previously described by Cash et al [ 8 ]. After reaching a peak on the second day of the infection and decreasing from the 5 th to the 15 th day, lung bacterial load persisted for 3 weeks. These results, as well as the analysis of the BAL and the histological features, are consistent with the literature [ 8 , 11 , 17 , 18 ]. Since, in this model, P. aeruginosa is associated with agar beads, we performed, as control groups, instillation of sterile agar beads. Sterile agar bead instilled rats did not show any evidence of weight loss and although they did present an increase in BAL cellularity, there were no PMN's except a slight increase on the second day which failed to reach a statistical significance (data not shown). This result is consistent with the literature, Nacucchio et al showed that agar beads alone could not reproduce the same level of injury than P. aeruginosa in agar beads [ 19 ]. From this first part of our work, we concluded that the model of chronic infection with P. aeruginosa is adequate, based on clinical, bacteriological, cytological and histological data. Although a clinical study has reported increased lung permeability in COPD patients infected by P. aeruginosa [ 20 ], few studies have focused on the consequences of chronic lung infection on alveolar function and particularly fluid movements. In our study, lung fluid movements were maximal on the 2 nd day post infection. We observed an increase of alveolar-capillary barrier permeability, DAFC and overall lung liquid clearance. A normal lung wet to dry weight ratio was a consequence of this adequate alveolar response. This contrasts sharply with the data we obtained in an acute lung injury model where LLC dramatically decreased and W/D weight ratio increased at 4 and 24 hours after Pseudomonas instillation [ 7 ]. In our chronic model, following the increase in both permeability and lung liquid clearance, we observed an improvement in permeability with a return to baseline of these 2 parameters on the 5 th day. The St group presented a moderate increase in permeability (Asp/plasma ratio: 0.31 ± 0.04), it has previously been reported that agar beads could alone be responsible for a moderate increase in permeability [ 19 ]. However, taking into account the association of the other parameters validating the model (clinical, bacteriological, cytological and histological), this effect does not challenge the model. Our results showed an increase of the DAFC at 48 hours post infection. In acute lung injury, the initial alveolar response is usually towards an increase of DAFC which many authors have documented in septic shock [ 21 ], or after endotoxin administration [ 22 ]. In septic shock, this increase was related to the release of endogenous catecholamines. In acute P. aeruginosa pneumonia, increased DAFC can be related to either Pseudomonas exoproducts [ 15 ] or to a TNF-α dependent mechanism during the first 24 hours of the infection [ 4 ]. We tested in our model whether TNF-α or catecholamines could explain our results. TNF-α was not detectable and systemic endogenous epinephrine or nor-epinephrine not different from controls on the 2 nd or the 5 th day. TNF-α is produced during the early phase of pneumonia, and its short half life probably explains the absence of detectable levels at 48 hours. A dynamic evaluation of TNF-α production with serial samples or antibody neutralization experiments would be helpful to precisely study the role of TNF-α. We therefore did not rule out that TNF-α may have triggered an inflammatory response which could be responsible for the increased DAFC. Other potential mechanisms such as Transforming Growth Factor β remain to be investigated [ 23 ]. Surprisingly, on the fifth day, DAFC returned to baseline along with the improvement in permeability. Although it is logical to see an improvement in permeability, consistent with a decrease of the bacterial burden and an adequate host response, DAFC was expected to remain increased. The persisting presence of the pathogen in the alveoli and many factors only related to its presence would normally lead to a persistent increase of DAFC [ 15 ]. We therefore decided to evaluate if a normal increase in DAFC could be elicited on the 5 th day post infection in response to known pharmacological stimuli [ 24 , 25 ]. In the normal lung, intra-alveolar administration of terbutaline generates a DAFC increase of approximately 30% [ 26 ]. We observed comparable results in our study in control animals as well as animals which received only sterile beads. In our model, on the 5 th day, terbutaline intratrachéal administration did not change DAFC. However the lack of effect may be due to airway inflammation and an inability to adequately deliver the drug, we therefore decided to use intraperitoneal administration with the same agent. Our results also show the absence of DAFC increase. We then hypothesized that the absence of response to the stimulation might be related to the persistence of the pathogen in the alveoli. To test this hypothesis, we injected the animals with ceftazidime to sterilize the lungs on the 5 th day. Sterilization was achieved but failed to restore DAFC stimulation with terbutaline. To explain this impairment of DAFC, different hypotheses still remain to be investigated concerning these agonist's receptors and their regulation. Other authors have shown in different situations that either an internalization or a decrease of affinity of the receptors [ 27 ] could be observed. Another hypothesis could be a lost of sensitization through a decrease of the AMPc dependent signal transmission. It was shown, in vitro, on tracheal cells that a continuous or repeated exposure to isoproterenol could lead to a lost of sensitization [ 28 ]. If this unresponsiveness exists in patients, the absence of an adapted DAFC response in chronic lung infection could lead to major damage in the presence of any new lung injury. Although chronic lung infection has not been isolated, per se, as an aggravating factor associated to mortality in COPD patients admitted in an intensive care unit, a pre-existing underlying pathology is associated with a worsening of the prognosis in community and nosocomial pneumonia [ 29 , 30 ]. DAFC impairment might be part of the answer to this effect of underlying disease. In conclusion, chronic P. aeruginosa pneumonia is characterized initially at 48 hours by an increased alveolar-capillary barrier permeability and an adapted host response with an increased DAFC and LLC preserving a normal lung wet to dry weight ratio. On the 5 th day, DAFC remains non responsive to pharmacological stimulation even after bacterial elimination. This impairment of DAFC could represent one of the factors responsible for the increased susceptibility of chronically infected patients to other respiratory insults. Authors' contributions SB and FA were responsible for the acquisition of the data. KF and MOH made substantial contributions to the drafting of the manuscript and the analysis of the data. TP performed the radioactive labelling of the albumin (I 131 ). EK was involved in the revision of the manuscript and the English editing. XL performed all the histological analysis. BG was involved in the acquisition of the data, the design and the conception of the study as well as the drafting of the article. All the authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC551591.xml
544188	Haplotype frequency estimation error analysis in the presence of missing genotype data	Background Increasingly researchers are turning to the use of haplotype analysis as a tool in population studies, the investigation of linkage disequilibrium, and candidate gene analysis. When the phase of the data is unknown, computational methods, in particular those employing the Expectation-Maximisation (EM) algorithm, are frequently used for estimating the phase and frequency of the underlying haplotypes. These methods have proved very successful, predicting the phase-known frequencies from data for which the phase is unknown with a high degree of accuracy. Recently there has been much speculation as to the effect of unknown, or missing allelic data – a common phenomenon even with modern automated DNA analysis techniques – on the performance of EM-based methods. To this end an EM-based program, modified to accommodate missing data, has been developed, incorporating non-parametric bootstrapping for the calculation of accurate confidence intervals. Results Here we present the results of the analyses of various data sets in which randomly selected known alleles have been relabelled as missing. Remarkably, we find that the absence of up to 30% of the data in both biallelic and multiallelic data sets with moderate to strong levels of linkage disequilibrium can be tolerated. Additionally, the frequencies of haplotypes which predominate in the complete data analysis remain essentially the same after the addition of the random noise caused by missing data. Conclusions These findings have important implications for the area of data gathering. It may be concluded that small levels of drop out in the data do not affect the overall accuracy of haplotype analysis perceptibly, and that, given recent findings on the effect of inaccurate data, ambiguous data points are best treated as unknown.	Background Haplotype analysis has become a valuable tool for researchers in population genetics. In particular, the value attached to the prediction of the constituent haplotypes of a given sample and their frequency of occurrence is such that a variety of methods have been developed for this purpose. Many of these methods, however, depend on knowledge of the phase of the data supplied. In general, genotypic data from polymorphic loci are ascertained phase-unknown. Various methods for determining the gametic phase exist. With sufficient data from the genotyping of family members, definitive haplotypes may be inferred. However, in particular for late-onset disorders, these data may be difficult or even impossible to obtain. At the laboratory level, techniques such as chromosomal isolation or long-range PCR [ 1 ] may be utilised in the prediction of haplotypes, but they suffer the dual drawbacks of being both technologically demanding and in many cases prohibitively expensive in practice. Thus researchers have moved towards computational solutions to this problem. Prominent among the techniques employed for the estimation of the true haplotype frequencies of a phase-unknown sample are those based on the Expectation-Maximisation (EM) algorithm. Hill [ 2 ] originally proposed the use of the EM algorithm in genetics, and three years later the term was first coined by Dempster et al. [ 3 ] and the method put on a more formal footing. A number of EM-based methods for haplotype frequency estimation (HFE) have been produced [ 4 , 5 ]. Excoffier and Slatkin [ 6 ] provide a thorough outline of the implementation of the EM algorithm as applied to the problem of HFE. Reliable computational techniques for the estimation of haplotype frequencies have been around for some time, and extensive studies of the accuracy of the EM-based methods have been carried out [ 7 , 8 ], but until recently there has been little investigation of the effect of missing data on these techniques. This is surprising considering that, even with modern automated DNA analysis methods, the problem of missing data is not uncommon, whether due to the failure of amplification or insufficient DNA. Zhao et al. [ 9 ] have developed the GENECOUNTING software specifically to take into account missing data in a sample, but have not produced any validation of the method. The HAPLO [ 5 ] program is also capable of analysing multiallelic data with missing alleles, using jackknife techniques for error analysis. The SNPHAP [ 10 ] algorithm can handle large numbers of loci and unknown alleles, but is restricted to the analysis of biallelic loci. In order to carry out an investigation of the effect of missing data on HFE, a program, based on the algorithm outlined in [ 6 ], has been developed which can accommodate multiallelic loci and a significant percentage of unknown alleles. The necessary alterations to the existing implementation of the EM algorithm are outlined in the Methods section. Following this, biallelic and multiallelic data sets were analysed with varying quantities of unknown alleles randomly substituted. The analysis is similar to previous work by Kirk and Cardon [ 11 ], which described the effect of genotyping error on HFE. Here we investigate the effect of missing data on the sizes of the confidence intervals (CIs) about the haplotype frequency point estimates (or simply "point estimates"). Surprisingly, the loss of as much as 30% of the allelic data did not have a significantly detrimental effect on the quality of the results. The frequencies of haplotypes which predominate in the complete data analysis remain essentially the same after randomly selected data have been relabelled as missing. The error estimates associated with the predicted frequencies, which are generated via a bootstrap method, are also quite stable, but increase as the proportion of missing data increases. Results Source of data Two sources of data were used for the principal part of this study. The first is real single nucleotide polymorphism (SNP) data; the second is multiallelic data generated via population generation software. Three additional sets of data containing 10%, 20% and 30% missing alleles respectively were generated from each of the two original sets. The process of generation is described in the Methods section. HFE was carried out on the eight data sets listed above. In each case 1,000 bootstrap iterations were performed for each HFE analysis and the 95% CIs about the point results were selected. For the sake of clarity the results from analyses of the 20% unknown alleles data sets have been omitted from the displayed graphs. Further tests were performed to investigate the effect of sample size upon the quality of the results. To this end two sets of progressively smaller data sets, with and without missing alleles, were generated from the SNP and multiallelic data sets, and HFE was carried out. The method of selecting these data is outlined in the Methods section. An additional data set, unrelated to those previously described, consisting of data from five SNP loci was generated for the purposes of performing tests on data with weak LD between the loci. A further data set with 10% missing alleles was generated from these additional data. Seven loci biallelic data sets Figure 1 displays point estimate results from the analyses of the seven loci SNP data sets with 536 sample points. Figure 1 is a comparison of the frequencies of the 26 haplotypes present in the phase-known data, and their predicted frequencies when the phase is assumed unknown and data are missing. The percentage of missing alleles varies from zero (labelled "complete data") to 30. The haplotypes derived from the phase-known data were labelled from 1 to 26 in non-increasing order of the magnitude of their frequency, hence the "haplotype label" of the x-axis. For a quantitative measure of the discrepancies in the frequencies between the phase-known and phase-unknown frequency predictions we use the measure D ( h , ) [ 6 , 11 ] given by where h i and are the haplotype frequencies derived from the phase-known and phase-unknown data respectively, and N is the number of possible haplotypes in the sample. As these data are from seven biallelic loci, N = 2 7 = 128 in this case. The results are displayed in Table 1 . Also recorded in Table 1 is the percentage increase in D ( h , ) as the percentage of unknown alleles in the sample increases. In each case it is the percentage increase relative to the complete data value that is measured. Three haplotypes absent from the phase-known data set appear in the results of the HFE analysis of the complete data. Their frequencies are 2.3 × 10 -3 , 1.4 × 10 -3 , and 1.1 × 10 -3 . Of the haplotypes present in the phase-known data, only one haplotype appears with a frequency less than these, the given frequency being 9.3 × 10 -4 . Figure 1 and Table 1 offer complementary illustrations of the effect of missing data. Table 1 provides a good overall picture of how the accuracy of the HFE method deteriorates with inferior data quality. The effect is most marked in the initial jump from complete to 10% missing data, where a 35% increase in D ( h , ) is recorded. The subsequent percentage increases going from 10% to 20% and 20% to 30% unknown alleles are 22% and 16%, respectively, of the value of D ( h , ) for the complete data. Figure 1 allows us to view specifically where this deterioration is most evident, in the mid-range frequency haplotypes. Figures 2 and 3 display the effect of increasing quantities of missing data on the 95% CIs of the haplotype frequencies estimated from the phase-unknown data. In an attempt to quantify this effect, the spread of the CIs for each haplotype (the difference between the two bootstrap haplotype frequencies which give the limits of the 95% CI) was summed for each data set. The sum for each data set containing missing data was compared with the sum for the complete data set (no missing data). The ratio of the two values (the ratio of the extent of the CIs or RCI) for each comparison is displayed in Table 2 . Here we see a superlinear increase in the RCI with increasing proportions of missing data. Despite this, we note from Figures 2 and 3 that, even for the 30% missing data case, the CIs for the complete data are not entirely contained within the CIs for the data with unknown alleles for many of the haplotypes. Multiallelic data sets Similar computations to those carried out for the SNP data sets were carried out for the four multiallelic data sets. Figure 4 is a comparison of the frequencies of the most prominent haplotypes in the phase-known data, and their predicted frequencies when the phase is assumed unknown and data are missing. As with the seven loci SNP data sets, the percentage of missing data varies from zero to 30. The haplotypes are labelled as before. However, as 118 distinct haplotypes appear in the phase-known data, only the frequencies for the 40 most common are illustrated in Figures 4 to 6 for reasons of clarity. No haplotype with a frequency greater than 0.005, as given by the phase-known data, was excluded from the graphs by this trimming. As with the biallelic data, the discrepancy between the phase-known and phase-unknown frequency predictions, D ( h , ), was measured. As the allele counts at each of the seven loci are 8, 2, 2, 9, 2, 5, and 2 respectively, the sum in Equation 1 is over the N = 5760 possible haplotypes in the sample. The results are displayed in Table 3 . As in Table 1 , the percentage increase in D ( h , ) as the percentage of unknown alleles in the sample increases is also recorded. 129 distinct haplotypes were estimated to have a frequency of greater than 10 -6 as a result of the HFE analysis. 29 of these do not appear in the phase-known data, with the most common of these having a frequency of 2.187 × 10 -3 . 68 haplotypes in the phase-known data display a frequency greater than this. As with the SNP case, Figure 4 and Table 3 together provide a good overall picture of the effect of missing data on the accuracy of the HFE method. Table 3 displays similar percentage increases in D ( h , ) with the 10% and 20% missing data cases to those of Table 1 (42% and 18% respectively), though there the similarity ends, as the jump in D ( h , ) going from 20% to 30% unknown alleles comes to 40% of the value of D ( h , ) for the complete data. In Figure 4 we see how the phase-unknown frequency predictions match well the observed phase-known frequencies for the more prominent haplotypes, but less well for the less common haplotypes, particularly for the 30% missing data case. Similarly to the SNP case, Figures 5 and 6 display the effect of increasing quantities of missing data on the 95% CIs of the haplotype frequencies estimated from the phase-unknown data. As before, measurement of this effect was made by observing the relative increase in the sizes of the CIs. The results are displayed in Table 4 . In contrast to the SNP case, we see a linear increase in the RCI with increasing proportions of missing data. This contrast is further marked by Figures 5 and 6 where we note that the CIs for the complete data are, in the case of most haplotypes, entirely contained within the CIs for the data with unknown alleles. Sample sizes Investigations were made into the effect of the sample size on the performance of the HFE method when 10% of the data was missing. Three further data sets of sizes 300, 100 and 50 individuals were generated by random selection from the original seven loci SNP and multiallelic sets. From these data, six additional sets with 10% missing alleles were created. HFE was performed upon these additional data, and the D ( h , ) results for each were displayed in Table 5 . In each case the phase-known haplotype frequencies used in the computation of D ( h , ) were those derived from the respective smaller samples (e.g. the accuracy of the HFE method as applied to the SNP sample with 300 individuals was calculated relative to the haplotype frequencies observed in the phase-known sample with 300 individuals, and not those observed in the original data). As may be expected, in all cases we see an increase in D ( h , ) as we move from the complete data to the data sets with missing alleles. D ( h , ) also is seen to increase as the sample size decreases. However, what is of note is the pattern involved. For the seven loci SNP case, the percentage increase in D ( h , ) from complete to missing data itself increases monotonically as the sample size is reduced. A similar pattern is not observed in the multiallelic data. Performance at low LD levels Fallin and Schork [ 7 ] illustrate how the performance of the EM-based HFE method diminishes with falling LD strength. Here we investigated how the accuracy of our implementation behaves on a data set exhibiting weak LD when 10% of the alleles are missing. A population of 500 individuals with data at five SNP loci was generated specifically for this part of the study. Lewontin's D ' [ 12 ] was found to range between 0.117 and 0.014 for all adjacent loci. Table 6 displays D ( h , ) readings for this particular case. Here we see a large percentage increase of 60% in D ( h , ) as we move from the complete data to 10% missing data. Discussion The results displayed here show the impact of the addition of increasing quantities of missing alleles on the quality of haplotype frequency estimates. Studying Figure 1 in tandem with Table 1 , and Figure 4 in tandem with Table 3 , we see a loss of accuracy of the HFE method as the quality of the data degrades. This is particularly true for the multiallelic data set with 30% missing alleles. Here the loss of accuracy is most apparent with the rarer haplotypes as may be seen in Figure 4 , whereas for the seven loci SNP case, Figure 1 illustrates that the low frequency haplotypes are dealt with remarkably well, even at high missing data proportions. For both sets of data the ability of the method to predict the frequencies of the most prominent haplotypes in the samples holds up well as the percentage of unknown alleles increases. Figures 2 and 3 and Figures 5 and 6 display a similar behaviour in the bootstrap generated CIs. To summarise, there are two significant aspects of the analysis of genotypic data containing incompletely typed individuals evident here. Firstly, that the HFE algorithm, given phase-unknown data with moderate to high levels of LD, predicts the frequencies of the underlying haplotypes with a high degree of accuracy, as is evident from the point estimate graphs, Figures 1 and 4 . Tables 1 and 3 quantify how the quality of the frequency predictions behave with increasing percentages of missing data. For the multiallelic case where 30% of the alleles are unknown, Table 3 shows that the discrepancy between the phase-known and phase-unknown predicted frequencies has doubled when compared with the complete data case, though from the study of Figure 4 the bulk of this discrepancy would appear to originate from the lower frequency haplotypes. The second aspect is the extent of the 95% CIs. We see a steady increase in the spread of the CIs with the addition of missing alleles, reflecting the growing uncertainty in the data. However, the most prominent haplotypes in both the SNP and multiallelic data sets maintain their distinctiveness, even at the 30% unknown alleles level. These data show that, in particular for the SNP data set, the effect of relabelling significant proportions of the data as unknown on the performance of the HFE algorithm is minor. Although study of the illustrated graphs suggests that the impact of missing data is more pronounced with the more complex multiallelic data sets, Tables 2 and 4 demonstrate that the relative increase in the size of the CIs is similar across the biallelic and multiallelic data sets, and is almost identical for the 30% missing data sets. There appears to be a discrepancy between the two measures, namely D ( h , ) and the RCI, used here to quantify the degradation in the quality of the results with increasing percentages of unknown alleles. Tables 1 and 3 imply that the HFE method works significantly better for biallelic data than for multiallelic data, whereas this phenomenon is much less evident in Tables 2 and 4 . This may be explained by the fact that D ( h , ) is an absolute measure of the performance of the algorithm, as the phase-known data are available for each data set and thus the exact sample haplotype frequencies are known. This discrepancy is to be expected; D ( h , ) is a sum over all possible haplotypes and there exist only 128 (2 7 ) possible haplotypes for the seven loci SNP data, whereas the multiallelic data, as noted in the Results section, have 5760 possible haplotypes. Also, it is not surprising that haplotype frequencies estimated from the multiallelic data set are found to be less accurate than those estimated from SNPs, given the more complex nature of the data. The RCI is a relative measure, and illustrates not so much the accuracy of the algorithm, rather the effect of additional missing data. The results displayed in Tables 2 and 4 show that the algorithm handles the increase in the proportion of unknown alleles equally well for both SNPs and multiallelic data, although it should be pointed out that the RCI measure gives no indication of the accuracy of the point estimates, and should generally be considered in tandem with a measure such as D ( h , ). Interestingly, the results for the multiallelic data set were achieved despite departure from Hardy-Weinberg equilibrium (HWE) at two of the seven loci (see Methods section). Although this technique relies on the assumption of HWE, Niu et al. [ 13 ] have demonstrated it to be reliable and robust even when the HWE assumption has been violated. Fallin and Schork [ 7 ] have shown that HWE violation which results in an excess of heterozygosity leads to an increase in HFE error, though their results are based on a five-locus system, and the observed error increase when two of the five loci were found to be in disequilibrium was minimal. As we are dealing here with a seven-locus system, the effect on the error was likely to have been even less apparent. The investigation into the effect of smaller sample sizes has produced some surprising results. Comparing Table 1 with Table 5 , we see that the relative increase in D ( h , ) observed when 10% of the seven loci SNP data is relabelled as missing does not change substantially as the size of the sample reduces. For the full sample of 536 individuals, the percentage jump in D ( h , ) moving from the complete data to 10% missing data is approximately 35%. For the sample of size 300, this increase is 37%. Likewise for the samples of size 100 and 50, the increases are 41% and 47% respectively. However, for the multiallelic data, we see a contrasting trend. The percentage jump in D ( h , ) decreases rather than increases with increasing missing data proportions. Inspection of Tables 3 and 5 shows us that the percentage increase in D ( h , ) when moving from the complete data to 10% missing data for the full sample of 500 individuals is approximately 42%, whereas for the sample of size 300 this drops to 32%. The recorded increase for the sample of size 100, 5%, is even more striking. (The sample of size 50 is not considered here, as the matching observed between the phase-known and phase-unknown frequencies was of poor quality (figure not shown), and any conclusions drawn from analysis of this case would be highly suspect). Thus no definitive conclusions may be made as to the effect of missing data as the sample size is reduced, other that to say that the matching between the phase-known and phase-unknown frequencies deteriorates with falling sample size, as would be expected. Table 6 underlines the relationship between strong LD and superior performance of the EM method [ 7 ]. For the weak LD data set, we see that D ( h , ) for the complete data is comparable to that of the seven loci SNP data with 30% missing alleles. It should also be borne in mind that, as the weak LD data set features only five SNP loci, the sum for D ( h , ) is over a mere 32 possible haplotypes, as compared to 128 for the seven loci SNP data, emphasising the fall-off in accuracy. Also of note is the similarity in the sample sizes -500 in the weak LD case, and 536 in the moderate to strong LD case. Moving to the 10% missing allele case, we witness a further 60% drop in accuracy, a considerably greater percentage that was observed for the medium to high LD data sets, a result which again calls into question the reliability of the method in the presence of weak LD. Conclusions Here we show that the EM method, with the modifications to the implementation for complete data detailed here, can generate accurate estimates of haplotype frequencies even when large amounts of data are missing, in this case up to 30%. Moreover, using this method, the degree of accuracy can easily be estimated using conventional bootstrapping approaches. This is of considerable importance in the design of experiments, as it is therefore obvious that small levels of drop out in the data for whatever reason do not affect the overall accuracy of the approach perceptibly. Furthermore, considering the strongly deleterious effects of even small amounts of inaccurate data [ 11 ], this analysis shows that large amounts of missing data are much less detrimental to the overall quality of the results than incorrectly typed sites. Thus from a practical standpoint it is clearly preferable that if any doubt exists as to a genotype's identity, it should be excluded rather than included using a "best guess". Methods Seven loci biallelic data The data used in this part of the study are derived from a genetic investigation of cystic fibrosis sufferers [ 14 ]. The haplotypes used here are actual haplotypes composed of a subset of the markers typed in the vicinity of the CFTR gene locus. The haplotypes comprise seven biallelic loci. From these haplotypes 536 phase-known genotypes were constructed via random resampling. Thus the data set comprised of 536 individuals each with seven SNP loci. In common with Kirk and Cardon [ 11 ] a linkage disequilibrium (LD) analysis was carried out on the data. For adjacent loci, D ' was found to be ≥ 0.9 for all intervals but the third and fifth, where D ' ≤ 0.25. As HWE is assumed for HFE, each locus was tested and found to be in HWE. Multiallelic data An initial population of fifty individuals with data from seven loci spaced 1 cM apart was generated in silico . The number of distinct alleles at each locus ranged from two to nine. A trait marker was introduced between the 3 rd and 4 th loci for 10 of the 50 founders. The population was evolved for thirty generations as an isolated group with random mating. The birth rate per couple was binomially distributed, with a range of zero to ten offspring and a mean of 2.5. 500 individuals bearing the trait were randomly selected from the final generation for analysis. As with the SNP data, the level of LD across the interval was measured. D ' was found to lie between 0.5 and 0.8 for all adjacent loci except between the second and third loci where D ' = 1.0 and the fifth and sixth where D ' = 0.24. A test for HWE [ 15 ] was performed, and it was found that the fourth and fifth loci were not in HWE (P-values < 0.001 and 0.003 respectively). In both cases an excess of heterozygosity was evident (observed heterozygosities of 0.848 and 0.232, respectively compared with expected heterozygosities of 0.729 and 0.205, respectively). Smaller sample sizes The data sets of reduced size used in this analysis were generated from the original seven loci SNP and multiallelic data sets via a random sampling process. The process was identical for both. Initially 300 individuals were chosen from the original data. Following this, 100 individuals were chosen from the newly created set of size 300. Finally 50 individuals were chosen from the set of size 100. In each case the selection process was random and done without replacement. From each of these six smaller data sets, six additional sets of data with 10% missing alleles were generated by the process outlined below. Low LD data A population of 500 individuals with data from five SNP loci was generated in silico specifically for the testing of the performance of the HFE algorithm in low LD circumstances. D ' was found to range between 0.117 and 0.014 for all adjacent loci. The data were also tested for HWE. The first locus was found to be marginally not in HWE (P = 0.0465), with excess homozygosity in evidence. All other loci were found to be in HWE. Phase-unknown data The HFE algorithm assumes that the input data are phase-unknown, and thus no alteration was necessary to the sample data sets which were phase-known before input. Comparison tests on the phase-known data, and phase-unknown data generated from the phase-known data via a process of phase-randomisation have confirmed that no bias is introduced by the use of phase-known data (results not shown). Generation of missing data Data sets containing unknown alleles were generated from the original data via the following procedure: 1. Each individual is selected in turn. 2. For each locus a random number between 0 and 100 is generated. 3. If this random number falls below the desired percentage of unknowns, both of the individual's alleles at the locus in question are redefined as unknown. This ensures that all unknowns appear in homologous pairs. 4. The process is repeated until all loci for all individuals are exhausted. Thus the desired percentage of unknown alleles is achieved globally, and the percentage of missing data at each locus may vary. Three additional sets of data were generated from each of the two original sets in this way, with 10%, 20% and 30% missing data respectively, giving eight data sets in all for the principal component of the study. Expectation-Maximisation algorithm For known gametic phase, HFE is a straightforward process of counting the constituent haplotypes in the sample. For the case where the gametic phase is unknown, maximum-likelihood haplotype frequencies are computed using the EM algorithm. The particular implementation used here for the finding of the haplotype frequencies is similar to that outlined by Excoffier and Slatkin [ 6 ]. The operation of the algorithm is based on the assumption of HWE, though as mentioned above, the method has been found to be quite robust in the presence of deviations from HWE [ 13 ]. Implementation of the EM algorithm Missing data in a sample necessitate alterations to the implementation for complete data of the EM-based algorithm. When all alleles in an individual are known, there exist c j possible genotypes consistent with this phenotype where and s j is the number of heterozygous loci in phenotype j . However, when unknown alleles appear at a locus, the situation is considerably more complex. In this case each unknown allele may take on the identity of any of the alleles observed at that locus. We require that unknown alleles always appear in pairs – the amplification of one allele only would result in the appearance of a homozygote which may bias results. Thus if there are N i distinct alleles (forms) observed at locus i in the entire sample, the number of possible complete phenotypes consistent with the observed phenotype is increased by a factor of N i ( N i + 1)/2 by the presence of an unknown site. This factor is the number of ways of selecting two alleles from a pool of N i distinct alleles when repetition is allowed. Thus the number of possible complete phenotypes given by phenotype j is given by where M is the number of loci in the sample and where N i is the number of distinct alleles observed in the sample at locus i . For each possible complete phenotype i of the κ j complete phenotypes possible for individual j , there exist c i possible genotypes, as given by Equation 2. Thus the number possible complete genotypes for phenotype j is given by Then, following [ 6 ], the probability P j of the j th phenotype, assuming random mating, is given by: where P i ( h k h l )is the probability of the i th genotype made up of haplotypes k and l , and where p k and p l are the population frequencies of the k th and l th haplotypes. Expectation step At the t th step of the EM iterative process, the probability of resolving each phenotype into the different possible genotypes is given by: where n j is the number of individuals with phenotype j , and n is the total number of individuals in the sample. Thus n j / n is the proportion of the total sample that has phenotype j , and P j ( h k h l )/ P j is the conditional probability of the particular genotype given the phenotype. Maximisation step The haplotype frequencies are then computed using a form of gene-counting [ 16 , 17 ] : where N is the number of globally distinct haplotypes (the number of different possible haplotypes in the sample), is the frequency of haplotype v, m is the number of distinct phenotypes in the sample, and ε iv is equal to the number of times haplotype v appears in genotype i . Generation of confidence intervals The technique of bootstrapping [ 18 ] was used to generate CIs about the point haplotype frequencies estimated from the phase-unknown data. Specifically, the percentile bootstrap approach was used. Authors' contributions EDK carried out the main programming work, performed the tests and drafted the manuscript. FS designed the population generation tool and assisted in the programming effort. RM assisted in the drafting of the manuscript and provided the SNP data. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544188.xml
406387	Neural Induction in Xenopus: Requirement for Ectodermal and Endomesodermal Signals via Chordin, Noggin, β-Catenin, and Cerberus	The origin of the signals that induce the differentiation of the central nervous system (CNS) is a long-standing question in vertebrate embryology. Here we show that Xenopus neural induction starts earlier than previously thought, at the blastula stage, and requires the combined activity of two distinct signaling centers. One is the well-known Nieuwkoop center, located in dorsal-vegetal cells, which expresses Nodal-related endomesodermal inducers. The other is a blastula Chordin- and Noggin- expressing (BCNE) center located in dorsal animal cells that contains both prospective neuroectoderm and Spemann organizer precursor cells. Both centers are downstream of the early β-Catenin signal. Molecular analyses demonstrated that the BCNE center was distinct from the Nieuwkoop center, and that the Nieuwkoop center expressed the secreted protein Cerberus (Cer). We found that explanted blastula dorsal animal cap cells that have not yet contacted a mesodermal substratum can, when cultured in saline solution, express definitive neural markers and differentiate histologically into CNS tissue. Transplantation experiments showed that the BCNE region was required for brain formation, even though it lacked CNS-inducing activity when transplanted ventrally. Cell-lineage studies demonstrated that BCNE cells give rise to a large part of the brain and retina and, in more posterior regions of the embryo, to floor plate and notochord. Loss-of-function experiments with antisense morpholino oligos (MO) showed that the CNS that forms in mesoderm-less Xenopus embryos (generated by injection with Cerberus-Short [ CerS ] mRNA) required Chordin (Chd), Noggin (Nog), and their upstream regulator β-Catenin. When mesoderm involution was prevented in dorsal marginal-zone explants, the anterior neural tissue formed in ectoderm was derived from BCNE cells and had a complete requirement for Chd. By injecting Chd morpholino oligos ( Chd- MO) into prospective neuroectoderm and Cerberus morpholino oligos ( Cer- MO) into prospective endomesoderm at the 8-cell stage, we showed that both layers cooperate in CNS formation. The results suggest a model for neural induction in Xenopus in which an early blastula β-Catenin signal predisposes the prospective neuroectoderm to neural induction by endomesodermal signals emanating from Spemann's organizer.	Introduction Vertebrate development results from a series of cell–cell interactions in which groups of cells induce their neighbors to acquire new cell differentiation fates. This process, known as embryonic induction, was first reported for the induction of the lens in surface ectoderm by the optic vesicles originating from the brain ( Spemann 1901 ; Lewis 1904 ). Subsequent work showed that the surface ectoderm itself also plays an important role (reviewed by Grainger 1992 ). From the analysis of lens induction, Spemann (1938 ) proposed that a double assurance mechanism ( doppelte Sicherung ) could provide a way of explaining the robustness of vertebrate development via reciprocal interactions between two layers of cells. Lens induction is an example of a secondary embryonic induction. Most experimental embryologists concentrated their research on the induction of the neural plate, which is considered the primary embryonic induction ( Spemann 1938 ; Saxén and Toivonen 1962 ; Harland 2000 ; Gilbert 2001 ; Stern 2002 ). In the classical organizer transplantation experiment, Spemann and Mangold (1924 ) demonstrated that dorsal lip mesoderm is sufficient to induce the differentiation of a complete central nervous system (CNS) in responding ectoderm. Spemann devoted an entire chapter of his book to the discussion of whether a double assurance mechanism existed in the case of neural plate induction (Chapter 8 in Spemann 1938 ) and concluded that the evidence supported a role for the underlying mesoderm, but not for the prospective neuroectoderm. A role for the gastrula ectoderm in neural plate formation had been proposed on the basis of experiments in which the mesoderm or the ectoderm had been damaged ( Goerttler 1925 ) and received some subsequent support ( Lehmann 1928 ). However, further consideration of the possible role of ectoderm in neural plate formation was hampered by a highly influential exogastrulation experiment performed in axolotl embryos ( Holtfreter 1933 ), in which endomesoderm involution was prevented and the entire ectoderm differentiated into epidermis. Since there was no trace of CNS tissue in these embryos, this experiment was interpreted as a demonstration that the underlying endomesoderm had the essential role in neural plate induction and that the prospective neuroectoderm had none ( Holtfreter 1933 ; Spemann 1938 ). The debate concerning whether the ectoderm itself has a role in neural plate formation has continued to this day. In dorsal marginal zone explants ( Keller and Danilchik 1988 ; Keller 1991 ), CNS differentiation can take place in the absence of underlying mesoderm. It has been proposed that in these Keller explants neural tissue induction results from a “planar” signal that diffuses in the plane of the ectoderm from the mesodermal organizer at gastrula ( Ruiz i Altaba 1992 , 1993 ; Doniach et al. 1992 ; Poznanski and Keller 1997 ) (see Figure 6 A). However, the existence of planar neural induction signals has been disputed, with neural induction in Keller explants proposed to result from “vertical” signals resulting from a brief contact between ectoderm and mesoderm at early gastrula ( Nieuwkoop and Koster 1995 ). Therefore, a central question remains unanswered despite many decades of research in amphibian neural induction: What is the differentiation potential of the presumptive neural plate material in the absence of a mesodermal substratum? This is the problem addressed here. Figure 6 Anterior Neural Induction in Keller Explants Requires Chd (A) Proposed vertical and planar signals in neural induction (following Ruiz i Altaba 1993). (B) Diagram of Keller explant preparation and subsequent elongation of the endomesoderm by convergent extension ( Keller 1991 ). (C) The neural and mesodermal regions of Keller explants contain descendants of BCNE cells (in blue) marked by blastomere injection at the 64-cell stage. (D) Expression of Otx2 and Krox20 in Keller explants ( n = 7). (E) Injection of 17 ng Chd- MO completely blocked Otx2 and Krox20 expression in neural regions, while expression of Otx2 in anterior endoderm was not affected ( n = 10). (F) The differentiated neuron marker N-tubulin is expressed in Keller explants ( n = 8). (G) Partial inhibition of N-tubulin by injection of Chd- MO ( n = 7). (H and I) Summary of the effects of Chd- MO in Keller explants. Abbreviations: SC, spinal cord; CG, cement gland; Epi, epidermis. (J) RT-PCR analyses of the effect of Chd- MO in Keller explants; samples injected with (plus) or without (minus) Chd- MO are indicated. Lane 1, whole embryos; lanes 2–7, Keller sandwiches. Note that expression of the neural markers NCAM and N-tubulin in Keller sandwiches was abolished by co-injection with 200 pg of dnFGF receptor 4a ( dnFGF4a ) mRNA and 17 ng of Chd- MO (lane 5). Injection with 600 pg of CerS mRNA, which eliminates mesoderm but not BCNE formation, does not affect neural induction in this assay (lane 6). Two recent technical advances led us to reinvestigate neural induction in Xenopus . First, it is now possible to completely inhibit mesoderm formation by microinjecting Cerberus-short ( CerS ) mRNA, a secreted antagonist specific for Nodal-related mesoderm inducers ( Agius et al. 2000 ). Interestingly, Xenopus embryos lacking mesoderm still developed a CNS, including a cyclopic eye ( Wessely et al. 2001 ). This was surprising, because such mesoderm-less embryos did not express multiple Spemann organizer markers such as Chordin ( Chd ), Noggin ( Nog ), and Goosecoid in dorsal endomesoderm at the gastrula stage. Second, a technical revolution has taken place with the availability of antisense morpholino oligos (MO) that permit loss-of-function studies in Xenopus ( Heasman et al. 2000 ). It is now possible to combine the tools of amphibian experimental embryology with investigations on the role of individual genes, such as the secreted bone morphogenetic protein (BMP) antagonist Chd ( Oelgeschläger et al. 2003 ) or its upstream regulator β-Catenin ( Heasman et al. 2000 ), in experimentally manipulated embryos. In whole embryos injected with Chd- MO, a CNS, although of reduced size, still develops. However, Spemann organizers depleted for Chd lose all neural-inducing activity when grafted to the ventral side of a host embryo ( Oelgeschläger et al. 2003 ). Surprisingly, when similar Chd-depleted grafts are placed on the dorsal side, ectodermal cells lose the ability to contribute to neural plate ( Oelgeschläger et al. 2003 ). This suggested that a cell-autonomous requirement of Chd for neural plate formation might exist in the ectoderm itself. At the blastula stage, the BMP antagonists Chd and Nog are expressed in the dorsal animal cap and marginal zone, in a region we had originally designated as the “preorganizer center” ( Wessely et al. 2001 ). This group of cells constitutes a blastula Chordin - and Noggin -expressing (BCNE) region that contains both prospective neuroectoderm cells and Spemann organizer precursors. The BCNE region also expresses Xenopus Nodal-related 3 (Xnr3) , a secreted factor with neural-inducing properties that is expressed at high levels in early Xenopus embryos ( Haramoto et al. 2004 ; Wessely et al. 2004 ). The early phase of expression of Chd and Nog in BCNE cells is regulated by the dorsal accumulation of β-Catenin, whereas later expression of the same genes in Spemann organizer endomesoderm requires in addition Nodal-related signals that can be blocked by CerS ( Wessely et al. 2001 ). In this study we analyze the mechanism of neural induction in Xenopus by means of embryological cut-and-paste and molecular loss-of-function experiments. We find that the BCNE center contains much of the presumptive anterior CNS. Loss-of-function studies show that gene products expressed at blastula—such as Chd, Nog, and β-Catenin—are required for neural induction in the absence of underlying endomesoderm. Cell-lineage studies show that the BCNE center itself gives rise to brain, notochord, and floor plate. Transplantation experiments show that the BCNE center is required for brain formation in Xenopus embryos. Microinjection experiments at the 8-cell stage, in which Chd- MO was injected into dorsal-animal and Cer- MO into dorsal-vegetal blastomeres, confirmed that secreted signals from both prospective neuroectoderm and underlying endomesoderm are required for anterior CNS development. The results support a double assurance mechanism for brain formation of the type proposed by Spemann (1938 ) for lens induction. Results The BCNE Center Is Distinct from the Nieuwkoop Center The initial asymmetry in Xenopus development is caused by a cortical rotation triggered by sperm entry, thought to redistribute “dorsal determinants” that in turn stabilize β-Catenin protein on the dorsal side of the embryo ( Figure 1 A) (reviewed in Gerhart et al. 1991 ; Harland and Gerhart 1997 ; De Robertis et al. 2000 ). At the blastula stage, the Nieuwkoop center located in dorsal-vegetal cells secretes mesoderm-inducing signals such as Xnr1, Xnr2, Xnr4, Xnr5, and Xnr6 that induce formation of the gastrula Spemann organizer in overlying mesoderm cells ( Agius et al. 2000 ; Takahashi et al. 2000 ). The Nieuwkoop center has also been called the “blastula organizer” in the early literature ( Gerhart et al. 1991 ). The BCNE region develops in the dorsal animal and marginal region. In situ hybridization analyses at the blastula stage (7 h after fertilization) showed that the neural-inducing secreted factors Chd, Nog, and Xnr3 are expressed in the animal cap, in a region that includes about 45 o of arc above the floor of the blastocoel, as well as in the dorsal marginal zone ( Figure 1 B– 1 D). At the gastrula stage, the same genes are expressed in more vegetal regions, in the Spemann organizer located in the dorsal endomesoderm of the marginal zone (e.g., Figure 2 E). Figure 1 Two Signaling Centers Coexist in the Xenopus Blastula (A) Diagram of early events between 1-cell stage and early blastula. (B–D) Expression of Chd, Nog, and Xnr3 transcripts just after midblastula transition (7 h postfertilization). Embryos were hybridized as whole mounts, stored in methanol for 1 mo at room temperature to improve contrast, and sectioned with a razor blade. (E) RT-PCR analysis of gene markers at midblastula, early stage 9. Six samples were prepared by dissections of blastula regions as shown in the diagram. (F) Summary of gene expression at blastula. The BCNE center expresses Chd, Nog, Siamois, and Xnr3, while the Nieuwkoop center expresses Xnr2, Xnr6, and Cer. Figure 2 The BCNE Center Contributes to Forebrain and Midline Structures (A) Method used for lineage tracing of the BCNE center with biotin-dextran amine (BDA) labeled grafts. (B) Sagittal section of a recently grafted BCNE at stage 9. (C) Chd mRNA expression at stage 9. (D) BCNE descendants at stage 10.5. (E) Chd mRNA expression at stage 10.5. (F) BCNE center descendants at stage 11. (G) Dorsal view of BCNE descendants at neural plate stage 14. (H) Double staining of transplanted BCNE region with nuclear lacZ mRNA and epidermal ectoderm of the host with epidermal cytokeratin (epi) probe in light red at stage 14. (I) Transverse section at the level of the trunk at stage 16. Abbreviations: fp, floor plate; no, notochord. (J–L) Transverse sections at stage 40. Abbreviations: fp, floor plate; hb, hindbrain; he, heart; le, lens; mb, midbrain; no, notochord; ov, otic vesicle; re, retina. (M) Dorsal view of 6-d embryo transplanted with a BCNE graft from CMV-GFP transgenic embryos. Abbreviations: br, brain; fp, floor plate; on, optic nerve; op, olfactory placode. (N) Side view at 4 d showing labeled retina and brain. Abbreviation: br, brain. The question arises as to whether two distinct signaling centers coexist in the Xenopus blastula. To address this, early blastulae with strong dorsoventral polarity ( Klein 1987 ) were dissected into six fragments, as shown in Figure 1 E. The results showed that, although some overlap existed, the region expressing Chd and Nog included the animal cap, whereas the Nieuwkoop center region that expresses Xnr2 and Xnr6 had a more vegetal location (see Figure 1 E). Xnr3 expression was observed in fragments 3 and 4, indicating a higher degree of overlap (see Figure 1 E). In addition, the results showed that the homeobox gene Siamois was expressed in the BCNE region, even though its expression has been reported to be more vegetal at later stages of development ( Lemaire et al. 1995 ). We also found that Cerberus (Cer), a gene expressed in anterior endoderm at gastrula ( Bouwmeester et al. 1996 ), was a component of the Nieuwkoop center. We conclude that two distinct signaling centers are present at blastula (see Figure 1 F). The more animal BCNE center expresses Chd, Nog, Xnr3, and Siamois, whereas the Nieuwkoop center expresses Xnr2, Xnr6, and Cer. Cell Lineage of the BCNE Region To map the fate of the blastula Chd- and Nog- expressing cells during normal development, we transplanted lineage-labeled BCNE regions isotopically into host blastulae at early stage 9 ( Figure 2 A). These grafts containing the lineage tracer biotin-dextran amine (BDA) marked the Chd- expressing region at blastula (compare Figure 2 B and 2 C). A few hours later, at early gastrula (stage 10.5), dorsal animal cap descendants were found both in organizer endomesoderm and in prospective neuroectoderm ( Figure 2 D). We note that by early gastrula stage Chd mRNA was expressed in organizer endomesoderm, but was no longer detectable in prospective neuroectoderm ( Figure 2 E). At midgastrula, stage 11, the transplanted tissue elongated in organizer endomesoderm and prospective neuroectoderm, with both layers remaining in close apposition ( Figure 2 F). At neural plate stages, stage 14, BCNE center descendants were found in a wide region in the anterior neural plate and, more posteriorly, in a narrow stripe in the midline ( Figure 2 G). Double staining using nuclear lacZ mRNA as lineage tracer in combination with an epidermal cytokeratin marker confirmed that BCNE cells give rise to anterior neural plate ( Figure 2 H). The midline staining in the trunk region corresponded to floor plate and notochord in histological sections ( Figure 2 I). At tadpole stage (stage 40), BCNE descendants contributed to a large part of the brain and retina (but not lens and otic vesicles) and to dorsal midline structures of the trunk-tail region ( Figure 2 J– 2 L). This lineage could be traced up to feeding tadpole stages ( Figure 2 M and 2 N) using dorsal animal cap grafts from cytomegalovirus–green fluorescent protein transgenic embryos ( Marsh-Armstrong et al. 1999 ). These results indicate that blastula Chd- expressing cells give rise to much of the brain and to the floor plate and notochord in the trunk region of the Xenopus embryo. The Dorsal Animal Cap Is Specified to Form CNS In embryology, the test of whether cells are specified to form a particular tissue is to culture them in isolation from the rest of the embryo. Dorsal animal cap explants from embryos injected with CerS mRNA expressed multiple neural molecular markers at stage 26, whereas animal or ventral explants did not ( Figure 3 A and 3 B). CerS was required to inhibit mesoderm formation; when identical explants were prepared without CerS mRNA injection, mesoderm contamination from the marginal zone was detected (data not shown). Neural differentiation could also be obtained in the absence of CerS mRNA when additional care was taken to avoid mesodermal contamination. As shown in Figure 3 C, small explants from the top half of the BCNE region were excised, sandwiched together, and cultured in saline solution for 3 d. The sandwich procedure allows such small explants to survive in culture for long periods of time. Dorsal BCNE explants differentiated into histotypic CNS, including gray and white matter ( Figure 3 D), whereas similar explants from ventral ectoderm differentiated into epidermis ( Figure 3 E). These results demonstrate that dorsal animal cap cells are already specified to form CNS at the blastula stage. Figure 3 The Blastula Dorsal Animal Cap Is Specified to Form CNS (A) Experimental diagram showing embryos injected with CerS mRNA from which three regions of the animal cap were dissected at blastula, cultured until stage 26, and processed for RT-PCR. The size of the explants was 0.3 mm by 0.3 mm in these samples. Abbreviations: A, animal pole; D, dorsal region; V, ventral animal cap. (B) RT-PCR analysis of animal cap fragments; note that anterior brain markers were expressed in the dorsal fragments in the absence of mesoderm ( α-actin ) and endoderm ( endodermin, Edd ) differentiation. Abbreviations: A, animal pole; D, dorsal region; V, ventral animal cap. (C) Experimental diagram of the small animal cap sandwich experiments; these embryos were not injected with CerS. In this case, the size of the explants was 0.15 mm by 0.15 mm leaving a 0.15-mm gap from the floor of the blastocoel to avoid contamination from mesoderm-forming cells. Fragments from two explants were sandwiched together (explants are too small to heal by curling up) and cultured in 1× Steinberg's solution until stage 40. Abbreviations: VSW, ventral sandwich; DSW, dorsal sandwich. (D) Histological section of dorsal animal cap explant (dorsal sandwich). These sandwiches differentiated into histotypic forebrain tissue including white and gray matter (4/17). Abbrevations: DSW, dorsal sandwich; gm, gray matter; wm, white matter. (E) Histological section of a ventral animal cap sandwich. All sandwiches differentiated into atypical epidermis ( n = 20). Abbreviations: ae, atypical epidermis; VSW, ventral sandwich. BCNE Tissue Is Required for Brain Formation To test whether the BCNE center is required for brain formation, we first deleted ventral or dorsal regions of the animal cap. Deletion of the dorsal region, but not of the ventral animal cap, resulted in headless embryos ( Figure 4 A and 4 B). Since Xenopus is one of the best-studied vertebrate embryos, it was surprising that this requirement of a region of the blastula for CNS formation had not been reported previously. To investigate this further, we replaced the deleted fragment with various ectodermal grafts. The brain defects could be rescued by transplantation of dorsal, but not ventral, animal cap grafts ( Figure 4 C and 4 D). Ectoderm from the animal pole was unable to rescue the ablated dorsal animal cap ( Figure 4 F). However, similar animal poles from lithium chloride (LiCl)–treated embryos, in which β-Catenin is stabilized and transcription of BCNE genes activated, were able to rescue head formation ( Figure 4 E). Figure 4 The Dorsal Animal Cap Is Required for Brain Formation (A) Ventral animal cap deletion (ΔV) produces a normal embryo. (B–F) Dorsal animal cap deletion (ΔD) results in loss of anterior brain structure. The headless phenotype of dorsal animal cap deletions was rescued by dorsal animal cap grafts (C) and animal pole grafts obtained from LiCl-treated embryos (E), but not by ventral animal cap transplants (D) or animal pole transplants (F). The average dorso-anterior indices (DAI) were 4.89 ([A] n = 28), 3.52 ([B] n = 25), 4.90 ([C] n = 10), 3.63 ([D] n = 19), 4.90 ([E] n = 12), and 3.50 ([F] n = 10). (G) Transplantation of the dorsal animal cap into the ventral animal cap region of a host embryo induced weak secondary axes (65.4%, n = 26). The embryo shown here was one of the strongest axes obtained. (H) Activity of BCNE transplanted ventrally was blocked by Chd- MO ( n = 15). Despite this requirement for brain development, blastula dorsal animal caps grafted into the ventral side of a host blastula were only able to form weak secondary axes ( Figure 4 G). Chd- MO, which blocks the activity of Spemann grafts ( Oelgeschläger et al. 2003 ), also inactivated BCNE grafts ( Figure 4 H). However, an important difference with the mature Spemann organizer was that BCNE cell transplants self-differentiated into spinal cord and muscle in these weak axes and were unable to induce CNS in neighboring cells as the Spemann organizer does (data not shown). We conclude that the dorsal animal cap BCNE center is required for brain formation. However, when transplanted into ectopic sites, BCNE tissue has only weak effects and does not induce neural tissue. Anterior CNS Formation in the Absence of Mesoderm Requires Chd and Nog We next investigated whether BCNE center signals are required for the anterior CNS that forms in embryos lacking mesoderm and Spemann organizer. CerS mRNA was injected at the 4-cell stage and the BCNE region marked with BDA at the 64-cell stage ( Figure 5 A; see also Figure S1 ). These mesoderm-less embryos developed forebrain tissue and prominent cyclopic eyes, which were derived from the lineage-labeled BCNE cells ( Figure 5 B and 5 C). To test whether there was a requirement for Chd in these embryos, we injected Chd- MO at the 2-cell stage. When Chd was knocked down, BCNE cells developed into epidermis instead of CNS ( Figure 5 D and 5 E). Brain and eye formation could be rescued by overexpression of Chd mRNA lacking the region targeted by Chd- MO ( Figure 5 F and 5 G). Figure 5 The CNS of Mesodermless Embryos Derives from BCNE Cells and Requires Chd, Nog, and β-Catenin (A) Experimental design. Embryos in which mesoderm induction was inhibited (by injection of 600 pg of CerS mRNA into the vegetal pole) were sectioned at stage 38 and stained with hematoxylin-eosin or for microinjected BDA lineage tracer marking the BCNE region. (B and C) Embryos injected with CerS mRNA alone ( n = 40). Abbreviation: br, brain. (D and E) Embryos injected with 17 ng of Chd- MO in addition to CerS ( n = 21). Abbreviation: epi, epidermis. (F and G) Coinjection of 17 ng of Chd- MO and CerS, followed by 100 pg of Chd mRNA together with the lineage tracer ( n = 19). Abbreviation: br, brain. (H) Expression of anterior CNS markers in mesodermless embryos requires Chd and Nog. RT-PCR analysis of CerS mRNA–injected embryos at tailbud stage 26. Markers of anterior brain ( Otx2 ), eye ( Rx2a ), midhindbrain boundary ( En2 ), hindbrain ( Krox20 ), and cement gland ( XAG ) were inhibited by injection of Chd- MO, Nog- MO, or both. A pan-neural marker (NCAM) and a neuronal marker (N-tubulin) were partially inhibited, and the posterior neural marker HoxB9 was not affected. α-actin serves as a mesoderm marker to show that CerS blocked mesoderm in these embryos and ODC as mRNA loading control. The effects of the Nog- MO described here can be rescued by full-length Nog mRNA lacking the 5′ leader sequence targeted by the antisense morpholino (data not shown). (I and J) β-cat- MO (13.6 ng) together with CerS mRNA ( n = 15). Abbreviation: epi, epidermis. (K and L) Rescue of β-cat- MO by 800 pg of β-catenin mRNA. Abbreviation: br, brain. (M and N) Rescue of the β-cat- MO phenotype by 100 pg of Chd mRNA ( n = 8). (O) Chd is required for the anterior neural induction caused by β-Catenin. Neural and cement gland markers were induced in animal cap explants by activation of β-Catenin signal by the injection of 600 pg β-catenin mRNA, dnGSK3 mRNA, or LiCl treatment (lanes 3–5). Markers of anterior brain ( Six3, Otx2 ), eye ( Rx2a ), midhindbrain boundary ( En2 ), hindbrain ( Krox20 ), and cement gland ( XAG ) were inhibited by Chd- MO (lanes 6–8). Although inhibition was not detected for the posterior neural marker HoxB9 and the pan-neural marker NCAM , the neuronal marker N-tubulin was inhibited. α-actin and α-globin are dorsal and ventral mesoderm markers, respectively, used to show the absence of mesoderm formation, and ODC serves as loading control. Molecular analyses confirmed that mesoderm-less embryos injected with Chd- MO did not express anterior neural tissue markers such as Otx2, Rx2a, En2, and Krox20 ( Figure 5 H, compare lanes 3 and 4). However, spinal cord ( HoxB9 ) or pan-neural markers (N-tubulin, neural cell adhesion molecule [NCAM]) were still expressed, indicating that only anterior neural differentiation was eliminated by Chd- MO and that posterior neural induction continues to take place. We also generated a Noggin antisense morpholino oligo ( Nog- MO) reagent, which, like Chd- MO, inhibited brain markers ( Figure 5 H, lane 5). Nog- MO was slightly weaker than Chd- MO, but even a combination of both morpholinos did not eliminate posterior neural markers ( Figure 5 H, lane 6). These results show that the brain tissue formed in embryos lacking mesoderm and Spemann organizer derive from BCNE cells. The formation of anterior CNS in mesoderm-less embryos requires the expression of Chd and Nog in prospective neuroectoderm. Neural Induction by β-Catenin Requires Chd It has recently been discovered that microinjection of β-catenin mRNA is able to induce neural tissue in Xenopus animal caps ( Baker et al. 1999 ). Stabilization of β-Catenin has a dual effect, inhibiting the transcription of BMPs ( Baker et al. 1999 ; Leung et al. 2003 ) and increasing expression of the BMP antagonists Chd and Nog in the blastula animal cap ( Wessely et al. 2001 ). We next tested the effect of β-Catenin knockdown on CNS differentiation. As shown in Figure 5 I and 5 J, β-cat- MO oligos ( Heasman et al. 2000 ) blocked formation of histological anterior brain and eye structures in CerS mesoderm-less embryos. Importantly, anterior CNS formation could be restored by overexpression of either β-catenin or Chd mRNA in these embryos ( Figure 5 K– 5 N). We conclude that brain formation in the absence of mesoderm requires the early β-Catenin signal. To investigate whether neural induction by β-Catenin in animal cap explants required Chd, the β-Catenin pathway was activated by β-catenin mRNA, dominant negative glycogen synthase kinase-3 ( dnGSK3 ) mRNA, or LiCl. These treatments induced multiple neural markers in animal caps ( Figure 5 O, lanes 3–5). Microinjection of Chd- MO inhibited the expression of anterior neural markers ( Six3, Otx2, Rx2a, En2 ), but not of posterior or pan-neural ones ( HoxB9, NCAM) ( Figure 5 O, lanes 6–8). The results indicate that neural induction by the β-Catenin signal requires expression of its downstream target gene Chordin. Anterior Neural Induction in Keller Explants Requires Chd Is the expression of Chd in prospective neuroectoderm at blastula responsible for the “planar” neural induction signals ( Figure 6 A) described by earlier workers? To investigate this, we used Keller sandwiches (Keller and Danilchick 1988; Doniach et al. 1992 ; Ruiz i Altaba 1992), in which neural tissue develops without contacting underlying mesoderm ( Figure 6 B). Marking of the BCNE region with lineage tracer indicated that Keller sandwiches contain cells that expressed Chd in prospective neuroectoderm at blastula ( Figure 6 C). Keller sandwiches expressed anterior CNS gene markers ( Figure 6 D and 6 F). However, explants prepared from embryos injected with Chd- MO failed to express Otx2 or Krox20 in anterior neuroectoderm, while retaining Otx2 expression in endoderm (compare Figure 6 D and 6 E). N-tubulin expression, which marks differentiated neurons, was inhibited by Chd- MO in the anterior CNS, but persisted in prospective spinal cord ( Figure 6 F and 6 G; results summarized in Figure 6 H and 6 I). These results show that the anterior CNS formation observed in Keller explants lacking underlying mesoderm requires Chd. Molecular analyses of Keller explants confirmed that brain markers were inhibited by Chd- MO, while pan-neural and spinal cord markers were less affected ( Figure 6 J, compare lanes 2 and 3). As before, posterior neural induction did not require Chd. The origin of this posterior neural differentiation is due to fibroblast growth factor (FGF) signaling ( Hongo et al. 1999 ; Pera et al. 2003 ), since it could be blocked in explants injected with dominant negative FGF receptor 4a (dnFGFR4a) mRNA ( Figure 6 J, lanes 4 and 5). Importantly, anterior CNS markers were still expressed in Keller sandwiches when mesoderm induction was blocked by CerS mRNA ( Figure 6 J, lane 6) and could be blocked by Chd- MO ( Figure 6 J, lane 7). Since mesoderm-less CerS Keller explants lack an endomesodermal Spemann organizer, their sole source of Chd is the BCNE center. Taken together, these experiments indicate that the anterior neural induction observed in Keller explants, known as “planar” induction, results from the activity of Chd- expressing cells located in the presumptive neuroectoderm at the blastula stage. Chordin and Cerberus Cooperate in Brain Induction Do vertical signals from endomesoderm cooperate with the BCNE center in brain differentiation? The endomesoderm secretes growth-factor antagonists with head-patterning activity, such as Cer, Frzb-1, Crescent, Dickkopf-1, Chd, and Nog ( Harland 2000 ; De Robertis et al. 2000 ). Several of these secreted antagonists are expressed in the anterior endoderm, which is homologous to the mouse anterior visceral endoderm ( Beddington and Robertson 2000 ). We chose to study one of these antagonists, the head-inducer Cer, because it is expressed in the anterior endoderm of the Spemann organizer ( Bouwmeester et al. 1996 ) and in the Nieuwkoop center, but not in the BCNE center (see Figure 1 E). Two recent studies have described morpholino antisense oligos targeting Cer . In both, Cer did not appear to be required for head development on its own, but cooperated when coinjected with other factors ( Hino et al. 2003 ; Silva et al. 2003 ). Xenopus laevis genes frequently have pseudoalleles thought to have originated from hybridization between two different Xenopus species in the course of evolution ( Kobel and Du Pasquier 1986 ). Examination of the EST database showed that a second Cer allele existed, and that the published morpholinos had three and four mismatches with it, respectively ( Figure 7 A) ( Silva et al. 2003 ; Hino et al. 2003 ). We therefore designed a new morpholino oligo, Cer- MO, targeting both X. laevis pseudoalleles ( Figure 7 A). Cer- MO inhibited head formation in Xenopus embryos, which could be rescued by Cer mRNA lacking the targeted 5′-leader sequence (data not shown). Figure 7 A Double-Assurance Mechanism in Xenopus Neural Induction That Requires Chordin and Cerberus (A) A new Cer- MO is complementary to both Cer pseudoalleles, while two MOs reported by other authors ( Hino et al. 2003 ; Silva et al. 2003 ) match only one allele, having three or four mismatches, respectively, with the other allele. The Cer- MO used in the present study inhibits head formation in intact embryos (data not shown), while the other two do not ( Hino et al. 2003 ; Silva et al. 2003 ). (B) Experimental procedure and cell lineages at 32-cell and early gastrula (stage 10.5) for dorsal-animal (FDA, green) and dorsal-vegetal (TRDA, red) blastomeres microinjected at the 8-cell stage. (C and D) Uninjected embryos. (E and F) Dorsal-animal injection with 8.5 ng of Chd- MO alone partially inhibited head formation; green fluorescence was seen in anterior CNS. (G and H) Dorsal-vegetal injection with 17 ng of Cer- MO also inhibited brain formation partially; red fluorescence may be seen in anterior endomesoderm. (I and J) Injection with 8.5 ng Chd- MO dorsal-animally and 17 ng Cer- MO dorsal-vegetally blocked brain formation, but not spinal cord and somites (histological sections not shown). To test whether Cer and Chd cooperated, we targeted Cer- MO to dorsal endomesoderm and Chd- MO to dorsal neuroectoderm at the 8-cell stage ( Figure 7 B). At the doses used, microinjection of Chd- MO or Cer- MO alone resulted in partial reductions of the anterior CNS ( Figure 7 C– 7 H). However, when neuroectodermal and endomesodermal progenitor cells were injected with Chd- MO and Cer- MO, respectively, brain differentiation did not occur ( Figure 7 I and 7 J). In histological sections, these embryos lacked brain structures but still developed spinal cord, somites, and notochord (data not shown). The results suggest that formation of the brain requires partly overlapping distinct factors in different cell layers. Chd is required in prospective neuroectoderm to predispose cells to form anterior CNS and cooperates with vertical signals from the underlying endomesoderm that include Cer ( Figure 8 ). Figure 8 Double-Assurance Model for Brain Formation by the BCNE and Nieuwkoop Centers Blastula Chd- and Nog- expressing cells are located in the dorsal animal region, while the Nieuwkoop center is found in the dorsal-vegetal region. At gastrula, the anterior endoderm derived from the Nieuwkoop center is found in close apposition to the prospective anterior CNS. See text for discussion. Discussion The results presented here are consistent with the following sequence of events during CNS development in Xenopus . A dorsal β-Catenin signal triggered by the early cortical rotation of the egg ( Gerhart et al. 1991 ; De Robertis et al. 2000 ) induces the expression of anti-BMP molecules such as Chd and Nog in a group of cells located in the dorsal animal region at the blastula stage (see Figure 8 ). The dorsal prospective neuroectoderm is already specified to form CNS at blastula (see Figure 3 ). Remarkably, transplantation studies showed that this BCNE center was required for brain formation in vivo (see Figure 4 ). The normal fate of these dorsal animal cap cells during development is to give rise to anterior CNS, floor plate, and notochord (see Figure 2 ). We note that our previous term “preorganizer” ( Wessely et al. 2001 ) was somewhat inadequate. Since BCNE grafts are unable to induce CNS in neighboring cells after transplantation, they lack organizer activity. The Nieuwkoop center arises at the same stage as the BCNE center, but in more vegetal cells (see Figure 8 ). The Nieuwkoop center expresses secreted factors such as mesoderm-inducing Xnrs and Cer and in later development gives rise to the endoderm that underlies the anterior CNS (see Figure 8 ). By inhibiting Chd in presumptive neuroectoderm and Cer in the endomesodermal substratum (see Figure 7 ), we were able to provide evidence that partly overlapping functions of distinct growth-factor antagonists secreted by different germ layers cooperate in CNS formation. Neural Induction Starts at Blastula At the blastula stage, gene expression in the BCNE region causes a neural predisposition in the prospective brain tissue itself. When prospective neuroectoderm is explanted at blastula and cultured in the absence of mesoderm, it can develop into histotypic neural tissue (see Figure 3 D). Once the Nieuwkoop center induces a Spemann organizer in dorsal mesoderm, a cocktail of growth factor antagonists is secreted by the endomesoderm. These Spemann organizer molecules require Nodal-related signals in order to be produced at gastrula ( Agius et al. 2000 ; Wessely et al. 2001 ). These endomesodermal factors include Cer (an inhibitor of Nodal, Wnt, and BMP signals); anti-Wnts such as Frzb-1, Crescent, and Dickkopf, and anti-BMPs such as Follistatin, Chd, and Nog ( De Robertis et al. 2000 ). Some molecules, like Chd and Nog, are expressed both in the dorsal animal cap at blastula and in the Spemann organizer at gastrula, whereas others, like Cer, are expressed only in the Nieuwkoop center and its endomesodermal descendants. In amphibians, the ability of a gastrula dorsal lip to induce a complete CNS when transplanted into a host gastrula had been taken as an indication that neural induction occurs at gastrula. We now show that brief expression of Chd and Nog during blastula stages, triggered by the maternal β-Catenin signal, is required for brain formation. This requirement becomes apparent when additional signals from underlying endomesoderm are blocked by inhibition of Nodal signaling or by preventing involution in Keller dorsal explants. We conclude that Xenopus neural induction starts in presumptive neuroectoderm shortly after midblastula. This is in agreement with other recent work in Xenopus ( Baker et al. 1999 ; Gamse and Sive 2001 ; Wessely et al. 2001 ). Redundant Signals in Neural Induction Multiple secreted growth factors antagonists participate in CNS induction ( Harland 2000 ; De Robertis et al. 2000 ), and their activities can be redundant. In the mouse, Chd and Nog mutants have normal neural plates, but in Chd -/- ; Nog -/- embryos, development of the forebrain fails ( Bachiller et al. 2000 ). In Xenopus and zebrafish, loss of Chd in the whole embryo results in animals that still are able to form anterior CNS, although its size is reduced ( Schulte-Merker et al. 1997 ; Oelgeschläger et al. 2003 ). This contrasts with the strong requirement for Chd revealed here when neural induction is driven by a single signaling center in Xenopus . In mesoderm-less embryos, ( CerS- injected) blastula dorsal animal cells are the sole source of Chd, and anterior CNS differentiation can be completely inhibited by Chd- MO (see Figure 5 ). Chd is also required for the neural inducing activity of Spemann organizer grafts ( Oelgeschläger et al. 2003 ). Multiple neural-inducing molecules, such as the above-mentioned growth factor antagonists, have been identified and may compensate for the loss of Chd in intact embryos. In addition other signals located outside of dorsal signaling centers, such as FGFs and insulin-like growth factors (IGFs) may participate in neural induction (see below). Could redundant signals from prospective neuroectoderm and endomesoderm also function in neural induction in other vertebrates? This seems possible in the case of the chick embryo. Chick Chd is initially expressed in the unincubated egg in epiblast central cells just anterior to Koller's sickle ( Streit et al. 1998 ), a region that may correspond to the Xenopus blastula Chd -expressing region. The progeny of this region of the chick epiblast contributes to the prospective forebrain and moves anteriorly during development. The descendants of the chick early Chd -expressing brain progenitors migrate at all times in front of the organizer, which is located at the tip of the primitive streak and has a caudalizing influence ( Foley et al. 2000 ). In zebrafish, mutant embryos lacking Nodal signaling still form brain tissue in the absence of a mesodermal organizer and express Chd ( Gritsman et al. 1999 ). In the mouse embryo, transplantation experiments at the gastrula stage support a role for different germ layers in brain induction ( Tam and Steiner 1999 ). However, expression of mouse Chd and Nog has only been analyzed from early primitive streak stage on ( Bachiller et al. 2000 ). Studies on the expression of these BMP antagonists in prestreak or peri-implantation mouse embryos, or on the earliest nuclear localization of β-Catenin protein, will be required to determine whether a region homologous to that of the Xenopus blastula Chd- expressing region exists in mammalian embryos. Neural-Inducing Signals in Chordates In amphibians the default model of neural induction proposes that BMPs expressed in ectoderm cause epidermal induction. When animal cap cells are dissociated, they become neuralized (reviewed by Weinstein and Hemmati-Brivanlou 1999 ). When exogenous BMP is added to dissociated animal cap cells, epidermal differentiation is restored. The present work with morpholinos that inhibit Chd, Nog, and Cer highlights the importance of BMP signaling regulation in Xenopus. The BCNE center appears shortly after midblastula and is required for anterior CNS formation when endomesodermal signals are inhibited. Induction of posterior neural tissue can still take in the absence of Chd and Nog (e.g., Figure 5 H). Formation of this posterior neural tissue can be blocked by dominant-negative FGF receptor 4a (see Figure 6 J). In Xenopus, FGF and IGF signaling are able to induce neural differentiation in animal caps ( Hardcastle et al. 2000 ; Pera et al. 2001; Richard-Parpaillon et al. 2002 ), and late canonical Wnt signals are known to inhibit anterior brain formation ( Kiecker and Niehrs 2001 ). In chick and ascidian embryos, current models of neural induction highlight the role of FGF and Wnt in neural induction and de-emphasize a role for BMP regulation ( Wilson and Edlund 2001 ; Stern 2002 ; Bertrand et al. 2003 ). We are unable to discuss in depth here the relative importance of the different signaling pathways in various organisms (reviewed in Wilson and Edlund 2001 ). It is clear, however, that multiple pathways cooperate in neural development. For example, in the chick embryo, Wnt or BMP antagonists applied to cells at the border region between epidermis and CNS expand the neural plate, and FGF signaling represses BMP4 expression in the neuroectoderm. In addition, the anti-neural effects of intermediate levels of an FGF antagonist can be reversed by the addition of chick Chd ( Wilson and Edlund 2001 ). One of the difficulties in comparing neural induction between Xenopus and other chordates concerned the different timing of events. We now find a requirement for critical signals triggered by β-Catenin in the prospective neuroectoderm just after midblastula. Thus, the neural induction process seems to start at blastula in all chordates ( Wessely et al. 2001 ; Wilson and Edlund 2001 ; Stern 2002 ; Bertrand et al. 2003 ). In addition, new molecular mechanisms are being discovered that help explain how disparate signaling pathways—such as those of FGF, IGF, and anti-BMPs—can be integrated during development. Tyrosine kinase receptors such as those for FGF and IGF have recently been found to inhibit the BMP pathway effector protein Smad1 by phosphorylation via mitogen-activated protein kinase (MAPK) ( Pera et al. 2003 ; Sater et al. 2003 ). Neural induction by the BMP antagonist Chd requires the extra boost in Smad1 inhibition provided by FGF and IGF signaling ( Pera et al. 2003 ). This molecular mechanism exemplifies one way in which signaling pathways hitherto considered entirely independent might be integrated in embryonic cells ( Massagué 2003 ). Primary neural induction in the chordate embryo has been an area of active investigation for many years and we can expect this to continue for the foreseeable future. A Role for Ectoderm in Amphibian Neural Induction The role of the ectoderm in amphibian neural induction has been the subject of much debate ( Spemann 1938 ; Holtfreter and Hamburger 1955 ; Nieuwkoop and Koster 1995 ). Gene marker studies in Xenopus had noticed a predisposition of dorsal ectoderm for neural induction by mesoderm ( Sharpe et al. 1987 ; London et al. 1988 ), but a requirement for any specific genes had not been addressed. In addition, it was known that the dorsal animal cap responds much better to the mesoderm-inducer Activin ( Sokol and Melton 1991 ). These earlier findings can now be reinterpreted as reflecting the effects of the early β-Catenin signal that induces expression of genes such as Chd, Nog, Xnr3, and Siamois. Chd is not only expressed in the organizer region during gastrulation, but also in the dorsal animal cap region during blastula, and this is required for neural specification. In this study we have provided evidence that presumptive neural plate material can differentiate into CNS in the absence of a mesodermal substratum. The BCNE center is required for brain formation in the embryo, but requires the cooperation of endomesodermal signals such as Cerberus. A requirement of gastrula prospective neuroectoderm for neural plate formation had been proposed by earlier workers on the basis of defect experiments ( Goerttler 1925 ; Lehmann 1928 ). However, these results were disputed ( Holtfreter 1933 ; Spemann 1938 ; Holtfreter and Hamburger 1955 ; Hamburger 1988 ; Nieuwkoop and Koster 1995 ), and vertical induction by the endomesodermal Spemann organizer was attributed the preeminent role in amphibian neural induction. A possible explanation for why the role of the prospective neuroectoderm remained unrecognized for so many years of research on the experimental embryology of neural induction is that, unlike Spemann's organizer, the BCNE center lacks inducing activity when transplanted to ectopic sites. The availability of new tools to investigate the function of individual genes—such as β-catenin, Chd, and Cer —has now provided evidence that both ectodermal and endomesodermal signals are required for primary embryonic induction in Xenopus. Materials and Methods Embryo manipulations Xenopus embryos obtained by in vitro fertilization were cultured in 0.1× modified Barth's medium ( Sive et al. 2000 ). For BCNE transplantation and deletion experiments, dissections were performed in 1× Steinberg's solution ( Sive et al. 2000 ). BCNE grafts were 0.3 mm squares isolated from the dorsal animal cap just above the floor of the blastocoel. They were excised at early stage 9 (6.75–7.25 h after fertilization at room temperature), before extensive epiboly movements begin, just one division after the large-cell blastula stage (stage 8), and could be monitored by the thickness of the animal cap and cell size. Embryos were cultured in 1× Steinberg's solution until healing (0.5–1 h) and then changed into 0.1× Barth's solution. Embryo stages were according to Nieuwkoop and Faber (1994 ). Keller sandwiches were prepared at early stage 10. The dorsal sector of the gastrula was excised at an angle of 30° from the dorsal midline, from the dorsal lip up to the animal pole, using stainless steel forceps. Two explants were sandwiched and cultured in 1 x Steinberg solution for 12 h for in situ hybridization, 1 d for RT-PCR analysis, and 2 d for morphological analysis. For RT-PCR analyses, RNA was pooled from five Keller explants, five animal caps, or single embryos. The RT-PCR conditions and primers, as well the protocol for whole-mount in situ hybridization, are described in http://www.hhmi.ucla.edu/derobertis/index.html . Lineage tracing To fate map BCNE descendants, an improved lineage tracing method was developed. Embryos were injected with 1–4 nl of 1% BDA (Molecular Probes, Eugene, Oregon, United States) in H 2 O and cultured explants or embryos were fixed for at least 1 h in MEMFA ( Sive et al. 2000 ). Subsequently, embryos were placed for 24 h in 70% ethanol, 1 h in 100% ethanol, 1 h in 100% isopropanol, 12–16 h in 100% xylene, and 1 h in paraffin at 65°C before embedding. We found that overnight incubation in xylene improved sectioning of early embryos, which are rich in yolk. Sections were cut at 8–10 μm and dewaxed in 100% xylene, 100% ethanol, and 70% ethanol for 2 min each. Next, sections were washed twice in binding buffer (100 mM Tris–HCl, 150 mM NaCl [pH 7.5]) for 5 min and incubated in binding buffer containing streptavidin-coupled alkaline phospatase (Roche, Basel, Switzerland) at a dilution of 1:5,000 overnight at room temperature. Afterwards, slides were washed twice with binding buffer, once with reaction buffer (100 mM Tris–HCl, 100 mM NaCl, 50 mM MgCl 2 , pH 9.5) for 5 min and incubated overnight with reaction buffer containing 10% BM purple solution (Roche) in a Coplin jar in the dark at 4°C. Staining was stopped by incubation in Stop solution (100 mM Tris-HCl, 1 mM EDTA, pH 7.4), and sections dehydrated in 100% methanol and completely air-dried before mounting in Vectashield medium (Vector Laboratories Inc., Burlingame, California, United States). In some experiments nuclear lacZ mRNA (kind gift of R. Harland), fluorescein dextran amine or Texas red dextran amine were used as lineage tracers. RNA injections To generate synthetic mRNAs, the plasmids pCS2- CerS, pCS2- Chd, pCS2- β-catenin, and pCS2- dnGSK3 were linearized with NotI and transcribed with SP6 RNA polymerase as described previously ( Piccolo et al. 1999 ). The following amounts of mRNA were used for microinjections: 600 pg (150 pg four times into the vegetal region at 4-cell stage) for CerS, 100 pg (50 pg twice into dorsal-animal region at 4-cell stage) for Chd, 800 pg (400 pg twice into the dorsal-animal region at 8-cell stage) for β-catenin, and 600 pg (300 pg twice into the dorsal-animal region at 8-cell stage) for dnGSK3 mRNA. Morpholino oligos Morpholino oligos were as follows: Chd- MO1 (5′-ACG TTC TGT CTC GTA TAG TGA GCG T-3′) and Chd- MO2 (5′-ACA GCA TTT TTG TGG TTG TCC CGA A-3′) ( Oelgeschläger et al. 2003 ); Nog- MO (5′-TCA CAA GGC ACT GGG AAT GAT CCA T-3′) (this work); β-cat- MO (5′-TTT CAA CCG TTT CCA AAG AAC CAG G-3′) ( Heasman et al. 2000 ); Cer- MO (5′-ACT TGC TGT TCC TGC ACT GTG C-3′) (this work); and a control-MO (5′-CCT CTT ACC TCA GTT ACA ATT TAT A-3′) ( Oelgeschläger et al. 2003 ). The morpholino oligos were resuspended to prepare a 1 mM stock solution (SS) that was then further diluted in sterile water to give a working solution: Chd- MO solution ( Chd- MO1-SS: Chd- MO2-SS:H 2 O = 1:1:6), β-cat- MO solution ( β-cat- MO-SS:H 2 O = 1:4), Nog- MO solution ( Nog- MO-SS:H 2 O = 1:1), Cer- MO solution ( Cer- MO-SS:H 2 O =1:1), control-MO solution (control-MO-SS:H 2 O = 1:1), and Chd- MO/ Nog- MO solution ( Chd- MO1-SS: Chd- MO2-SS: Nog- MO-SS:H 2 O = 1:1:4:10). A total of 8 nl (two times 4 nl or four times 2 nl) morpholino solutions were injected at the 2-cell stage or 4 nl (two times 2 nl) of morpholino solution injected at the 8-cell stage. Supporting Information Figure S1 The BCNE Region Can Be Reliably Marked at the 64-Cell Stage Using an Improved BDA Lineage-Tracing Method (A–H) Microinjection of individual 32-cell blastomeres does not faithfully recapitulate the lineage of BCNE grafts at gastrula (compare with Figure 2 B and 2 D). Note that the lineage of D1 includes part of the Nieuwkoop center and contributes to anterior endoderm at gastrula. This 32-cell map is in general agreement with previously published fate maps ( Dale and Slack 1987 ; Bauer et al. 1994 ); the minor differences observed are explained by our choice of batches of regularly cleaving embryos ( Klein 1987 ) with tightly adhering small animal blastomeres. Abbreviation: AE, anterior endoderm. (I–L) Diagram indicating the injection of the lower daughter of B1 and the upper daughter of C1 at the 64-cell stage (I), which reliably identify BCNE descendants at stage 9 (J), stage 11 (K), and stage 36 (L). Arrowheads indicate the blastopore. Abbreviations: fp, floor plate; no, notochord. (4.39 MB TIF). Click here for additional data file. Accession Numbers The GenBank ( http://www.ncbi.nlm.nih.gov/Genbank/index.html ) accession numbers discussed in this paper are for β-catenin (M77013), Cer (U64831), Cer pseudoallele (BG160114), Chd (L35764), Goosecoid (M81481), and Nog (M98807).	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC406387.xml
517705	Adaptation of an amphibian mucociliary clearance model to evaluate early effects of tobacco smoke exposure	Rationale Inhaled side-stream tobacco smoke brings in all of its harmful components impairing mechanisms that protect the airways and lungs. Chronic respiratory health consequences are a complex multi-step silent process. By the time clinical manifestations require medical attention, several structural and functional changes have already occurred. The respiratory system has to undergo an iterative process of injury, healing and remodeling with every exposure. Methods To have a better understanding of the initial changes that take place when first exposed to environmental tobacco smoke, we have developed an exposure model, using the frog palate that closely represents the features of obstructive airways where ciliary dysfunction and mucus hypersecretion occur. Results Mucus transport was significantly reduced, even after exposure to the smoke of one cigarette (p < 0.05) and even further with 4-cigarettes exposure (p < 0.001). Morphometric and ultrastructural studies by SEM show extensive areas of tissue disruption. Gelatinase zymography shows activation of MMP9 in mucus from palates exposed to tobacco smoke. Conclusions The clearance of mucus on the frog palate is significantly reduced after exposure to environmental tobacco smoke. Cilia and the extracellular matrix are anatomically disrupted. Tobacco smoke triggers an increased activity of matrix metalloproteinases associated with a substantial defoliation of ciliated epithelium. These studies enhance the knowledge of the changes in the mucociliary apparatus that occur initially after exposure to environmental tobacco smoke, with the goal of understanding how these changes relate to the genesis of chronic airway pathologies in humans.	Background Respiratory diseases, infectious and non-infectious, are a prime cause of morbidity, mortality and health system utilization in many countries. Exposure to cigarette smoke is an important factor in causing as well as increasing complications in several pulmonary disorders. The mucociliary clearance constitutes the first line of defense to maintain the airways as free as possible of foreign bodies [ 1 ]. Impairment of mucociliary function may be the result of epithelial airway damage, ciliary dysfunction, inflammation, and change in mucus viscosity and/or elasticity. In laboratory studies we have shown that the physical properties of mucus in nonsmokers are not altered by age or by restrictive pulmonary pathology. However, mucus properties show alterations when exposed to tobacco smoke and these alterations are noticeable in the very early stages of smoke exposure, even at an exposure level in the range of 1 to 5 cigarettes per day [ 2 ]. The majority of studies on the outcome of tobacco use have been performed in the late phases of the process [ 3 - 7 ]; the acute and early effects on cilia, mucus and mucociliary clearance after active smoking or side stream tobacco smoke exposure have not been well studied. We hypothesized that studying changes occurring in the initial stages will lead to a better understanding of the multifaceted problems of tobacco exposure. Therefore our laboratory has been developing and using in vivo and ex vivo epithelial injury models that replicate the features of airway diseases [ 8 - 12 ]. The mucus blanket in the frog palate is cleared by coordinated ciliary activity in an almost identical fashion to that observed in human airways. We anticipated that a modified frog palate exposure model could better serve our purposes. We were specifically interested in developing an exposure model that would allow us to study the initial effects and mechanisms occurring in an epithelial tissue after exposure to environmental tobacco smoke. We conceived this model to be free of interferences from other agents or systemic physiological responses resulting from other internal or external influences. We also aim to have a better understanding of the mechanism by which ciliated epithelial cells are exfoliated after being exposed to tobacco smoke, as this may relate directly to impaired mucus clearance in several human airway diseases including chronic bronchitis and chronic obstructive pulmonary disease (COPD). Hence we decided first to test our novel ex vivo model, and in later studies use in vivo models. Methodology Frog palate preparation From a bullfrog, Rana catesbiana , the upper portion of the head is removed following the procedures described in previous work [ 10 , 11 ], by cutting with scissors through from the junction of the posterior pharynx and esophagus out to the skin of the back. This procedure was carried out after lowering the body temperature of the frog for 30 – 60 minutes inside a refrigerator to abolish pain sensations. The palate was examined for macroscopic lesions, such as ulcers, petechia or redness as evidence of inflammation. Only palates free of inflammatory indicators were included in this study. Any blood remaining in the epithelial surface was carefully washed away, then the excised head was placed palate side facing upwards on a piece of gauze saturated with frog Ringer solution (FRS) in a Petri dish. The experimental procedures involving animals were approved by the Health Sciences Animal Policy and Welfare Committee, University of Alberta. The FRS was prepared by mixing standard Ringer injection with sterile water (2:1). The composition of standard frog Ringer (in mmol/l) is 90 NaCl, 3 KCl, 2 CaCl 2 , and 15 NaHCO 3 (220 mosm/l). The palate was placed inside the frog chamber, a wooden box with a glass top and fitted glass front, and manipulated trough glove openings. The humidity inside the box is maintained at 100% using an ultrasonic Pari nebulizer and the temperature is kept between 22° to 24°C by a rheostat-controlled, externally mounted light source. Before carrying out any measurement, the palate was allowed to stabilize inside the box for 15 minutes before testing. Exposure chamber The exposure chamber (Figure 1 ) with a volume ~10 liters had two inlets: one connected to an ultrasonic Pari jet nebulizer system set at 8 L/min measured by a Puritan flow meter to maintain the chamber near 100 % humidity. The other inlet was linked to a burning chamber. The latter, which contained a burning cigarette, was slightly pressurized with air flowing into the chamber at a rate of 2 L/min to promote cigarette combustion. Positive ventilation inside the burning chamber pushed the side stream smoke into the covered but not sealed exposure chamber, which was exhausted into a fume hood. Temperature inside the chamber was maintained to 22°C. and monitored through a thermocouple and digital-display thermometer. The palate was placed inside with the palate side upwards on a piece of gauze saturated with FRS in a glass dish at about five centimeters above the bottom of the chamber. Figure 1 Box designed to expose the frog palate to cigarette smoke. The exposure chamber was designed to allow a steady flow of side stream tobacco smoke to reach the epithelial tissue. The concentration of cigarette smoke inside the exposure chamber was maintained constant throughout the experiment by the presence of a mixing baffle on the outflow of the chamber. The concentration of tobacco smoke inside the chamber was not directly measured, but is estimated that the half palate was exposed to the smoke of each cigarette delivered, diluted in 180 liters of fresh air. As much as possible, the conditions (humidity, type of solution, temperature, physical manipulation and airflow in and out) in the frog chamber and in the exposure chamber were maintained similar, with cigarette smoke being the independent variable. Frog palate exposure model preparation The palate was divided longitudinally in two halves along the midline as shown in Figure 2 , cutting the epithelium with a scalpel to minimize damage. After five minutes the mucus transport was measured in both halves, left and right, to confirm that both half palates were functioning normally. One side was used as control and the other half was exposed to tobacco smoke. The control half of the palate was left in the frog chamber while the other half was placed in the exposure chamber maintained at similar conditions of 100 % humidity and room temperature. If any or both halves showed a variation in mucus transport greater than 30 % from the baseline, the experiment was aborted. Figure 2 The frog palate was further divided longitudinally in two halves following the midline. Mucus transport velocity (MTV) determination The palate was placed under a dissecting stereomicroscope provided with a reticulated eyepiece. Mucociliary clearance was determined by observing the movement of particles of charcoal powder gently deposited on a sample of mucus on the palate surface; its clearance was visually monitored and MTV determined. The displacement of 3 – 5 μL of endogenous frog mucus sample was calculated by dividing the distance traveled by the transit time across the 0.3 inch (7.62 mm) segment marked between 0.1 and 0.4 inches in the graduated eyepiece. At least five measurements of the time required for the mucus sample to travel the defined distance were made every time to obtain control and smoke exposure mucus transport velocity. We used cigarettes regularly available at commercial outlets of a representative brand in terms of customer preference and toxic emissions: nicotine 0.5 – 2.1 mg, tar 4 – 24 mg, carbon monoxide 5 – 25 mg, hydrogen cyanide 0.04 – 0.21 mg, benzene 0.025 – 0.069 mg, formaldehyde 0.018 – 0.1 mg. One cigarette took 17 minutes on average to burn completely in this preparation. Measurements and mucus samples collection After being exposed to one cigarette, the half palate was brought back to the frog chamber and MTV was measured in both control and exposed halves. A sample of mucus was collected from each half, and placed in separate containers and frozen at -80°C in liquid nitrogen followed by storage in a -80° freezer until analysis. The exposed half, brought back to the exposure chamber was further exposed to 3 more cigarettes (51 additional minutes of exposure). Following tobacco exposure, the half palate was again brought back to the frog chamber to measure mucus transport and collect tissue and mucus sample in both halves, control and exposed. One set of half palates (control and exposed to four cigarettes) was stored at 4°C overnight, and transport of mucus was reassessed the next day. Tissue and mucus samples collection Sample of mucus from both halves were immersed in glutaraldehyde 2.5%, and placed in properly identified separate containers for storage at 4°C for later Scanning Electron Microscope (SEM) studies. The epithelial tissue was carefully dissected and separated from the palate musculature. Samples of the palate tissue from each half were sectioned, frozen in liquid nitrogen and stored at -80°C for gelatinase zymography studies. Zymography Samples of tissue, as well as mucus, were taken out of the freezer and ground to a powder in a mortar and pestal. Liquid nitrogen was added to keep the samples frozen. Homogenizing buffer (KCl 100 mM, ZnCl 2 0.5 mM, EDTA 10 mM, Tris-HCl 1 M, pH 6.8) was added to the ground samples (approximately 500 ul buffer per 10 mg tissue sample, 200 buffer per 10 mg mucus sample) that were sonicated for 30 seconds on ice and then centrifuged at 8000 rpm for 5 minutes at 4°C. The supernatant was collected for protein assay (BCA protein assay kit, Pierce). Normalizing the protein content as 5 μg, different amount of samples were loaded into the 7.5% separating zymography gel (30% acrylamide/ 0.8% bisacrylamide 3.75 ml, 4 × Tris-Cl/ SDS, pH 8.8 3.75 ml, H 2 O 6 ml, 2% gelatin A 1.5 ml, 10% ammonium persulfate 0.1 ml, TEMED 0.01 ml, for 4 gels) and were run at 100 volts for 20 minutes, 150 volts for 40 minutes. After electrophoresis, the gel was washed 3 times (20 minutes per a time) in 2.5% Triton X-100 at room temperature followed by incubation for 96 hours in zymography development buffer (0.15 M NaCl, 5 mM CaCl 2 , 0.05% Azide NaN 3 , 50 mM Tris-HCl pH 7.6). The gel was then stained for 2 hours with stain solution (Coomassie brilliant blue R-250 1 g/L, methanol: acetic acid: H 2 O= 2.5: 1: 6.5) followed by de-staining (ethanol: acetic acid: H 2 O= 1: 2: 22) overnight. The expression of the gelatinases was shown on the gel as clear bands against the dark background stained with Coomassie blue. The bands were compared with the gelatinase standards for MMP 2 and 9 running in the first lane of each gel. The density of the gels was measured in the Bio-Rad scanning densitometer. Scanning electron microscopy Samples of mucus and tissue were placed in 2.5 % glutaraldehyde solution immediately after collection and stored at 4°C until processing. The samples were post-fixed in 1 % osmium tetraoxide in Milonig's buffer at room temperature for one hour. They were then washed in a series of ethanol (50 – 100 %), ten minutes at each step, followed by two additional periods of absolute ethanol (10 minutes each). The samples were further dehydrated by critical point drying at 31°C for 5 – 10 minutes, then mounted on a specimen holder for SEM and dried overnight in vacuum desiccators. In the final stage of preparation for viewing, the samples were sputter coated with gold (Edwards, model S150B Sputter Coater). Samples were viewed using SEM (Hitachi S-2500). Images were scanned directly to a computer and stored as image files for subsequent viewing and analysis. Statistical analysis Data are expressed as mean ± standard deviation unless otherwise stated. A paired Student-T test was used for simple comparison. The level of significance was set at 5 %. Results The modified fresh frog palate exposure model was relatively easy to prepare and practical to handle. On gross examination, the surface of the palate exposed to side stream tobacco smoke did not show any macroscopic change in appearance after cigarette smoke exposure (CSE) compared to the control halves maintained in the normal chamber. One half of the palate was used as a control and the other half was exposed to cigarette smoke. Baseline mucus transport velocity( MTV ) was measured in both half palates prior to exposure of one half palate to cigarette smoke, and were identical (19.5 ± .03 mm/sec). Two additional MTV determinations were carried out in the control half after one and four cigarettes and after 24 h period of recovery during which time the palate was kept at 4°C. A paired T-test showed no statistical difference in mucus transport velocity among the control measurements during the entire experiment (Figure 3 ). Figure 3 Three MTV controls are shown in the graph, no differences were observed among them. After immediate exposure to four cigarettes, mucus transport on the palates was drastically reduced. One of the four cigarette-exposed palate and its respective half palate control were maintained overnight in a refrigerator, and it was not possible to measure a transport time the next day in the exposed palate. A paired T-test of MTV on the control half compared to the exposed half showed that mucus transport velocity in the exposed half palate was reduced (p < 0.03) immediately after exposure to the side stream smoke of one cigarette compared with the non-exposed half palate. Further exposure to the side stream smoke of three more cigarettes (4 in total) significantly reduced MTV (p < 0.001), with no signs of recovery after 24 hours. We obtained a sample of mucus from the half palate exposed to four cigarettes and used it on the control half palate to measure MTV. Clearance was within the normal range. Immediately after, the same sample of mucus was tested on the exposed half palate and mucus clearance was again seen to be very slow. Mucus transport had basically ceased in some areas of the tobacco-smoke exposed palate. Scanning electron microscopy (SEM) studies to assess the integrity of the epithelium after exposure to side stream smoke of one cigarette showed areas where the layer of cilia looked disordered. However, we did not see loss of cilia or exfoliation of ciliated epithelial cells after careful examination under the SEM in lower and high power of the entire surface of the sample after this level of exposure. In Figure 4 , SEM images from palates exposed to the smoke of four cigarettes showed greater epithelial tissue disruption (panels 2 and 3) compared to a control palate (panel 1). Large areas of deciliated cells were observed, as well as exfoliation of intact ciliated cells. Examples of exfoliated cells found on the surface of the epithelium in the representative SEMs are indicated with black arrows. Morphometric analysis of the area of cilia loss from 3 paired palates exposed to four cigarettes, evaluating 12 different areas randomly selected in each palate, showed cilia loss of 51 ± 14 % compared to < 2 % on control palates. Figure 4 Magnification is ×400. On the left, the surface of the normal non-exposed palate is shown with a continuous ciliary layer, punctuated with secretory gland openings. The middle and right micrographs show the surface of palates exposed to the smoke of four cigarettes. Gelatinase zymography showed increased activity of MMP-9 in mucus collected from the palates exposed to tobacco smoke of four cigarettes compared to mucus from control palates as shown in Figure 5 . MMP2 activity was not different in mucus samples obtained from palates exposed, or not to cigarette smoke, but these results are inconclusive. Figure 5 A representative gelatinase zymogram on mucus collected from control and cigarette-exposed palates. In the left lane, standards for MMP 2 and 9 are shown. In the next two lanes MMP activity is shown in control mucus (two samples). The next two lanes show MMP activity in two aliquots of the mucus collected from palates exposed to the smoke of four cigarettes. Discussion Major findings in this study include: a) our amphibian mucociliary clearance model appears to be appropriate to study the early effects of tobacco smoke exposure; b) the clearance of mucus is significantly reduced after exposure to side stream tobacco smoke associated to dose response; c) ciliated cells are anatomically and physiologically acutely affected therefore drastically impairing mucus clearance. Acute irritation and inflammation of cilia due to exposure to tobacco smoke could explain this alteration; d) mucus seems to not be physiologically affected at this stage; and e) tobacco smoke triggers an increased activity of matrix metalloproteinases (in particular MMP9) associated with disruption of the extracellular matrix, most likely affecting cellular attachments to the basal membrane generating a substantial exfoliation of the ciliated cells. These findings might assist us in enhancing the knowledge of the changes in the mucociliary apparatus that occur early after exposure to environmental tobacco products, with the goal to understand how these changes relate to the development of chronic airway pathologies in humans. The frog palate has been used for several decades as a model to assess mucociliary clearance [ 13 - 17 ]. Different species of frogs have also been used, as well as a variety of study designs. Researchers had to deal with several sources of variability, making it difficult to standardize a widely acceptable model. There are disadvantages in our model such that it pertains to a non-mammalian species, in addition that the epithelial tissue is non-respiratory. However as a model it has advantages over some mammalian models like rodents in that the ciliated epithelium has a well developed mucus blanket that work in coordination with cilia similar to the human situation. This exposure model is an isolated system, theoretically free of interferences from other agents or systemic physiological responses resulting from other internal or external influences. In this injury model we randomly exposed either half palate to side stream tobacco smoke and used the opposite half palate as the control (internal control). Since there was no statistical difference in mucus transport between them, previously established in different sets of pairs of palates, one side was arbitrarily selected for control and the opposite half for exposure to tobacco smoke. This study follows an approach utilized previously by Zayas et al [ 7 ] that compared samples of mucus obtained from both mainstem bronchi in smokers and in nonsmokers. In assessing the mucus transport rate in both half palates before any experimental procedure and to insure that they functioned identically, we established baseline data for our experiment. This helped to reduce error variability due to sex, weight, age, as well as control for any seasonal effect on the mucus transport rate in the frog palates as demonstrated by Rubin et al [ 18 ]. Since the palate is excised from the frog we may assume that any observed effect or response is a local and direct effect of exposure to side stream tobacco smoke, which is an advantage of our model. From the data obtained, we would conclude that our frog palate exposure model is suitable to study the acute effects of second hand tobacco smoke. We may also conclude that clearability of mucus seems not to be altered at this stage, at this level of exposure and in this particular exposure preparation or design. Therefore, at this stage mucus transport seems not to be functionally affected for being cleared in a normal fashion by non-exposed cilia. However, after more prolonged tobacco exposure there will be modifications in mucus properties, as seen and measured in smoking dogs where after two months of tobacco exposure, the galactose content and the viscoelastic properties of mucus were observed to present alterations. After further exposure the viscosity and elasticity properties of mucus were reversed to quasi-normal levels, but galactose content did not [ 12 ]. Hence the model makes it very useful for differentiation of cilia-related effects versus mucus-related effects on mucociliary clearance after acute tobacco exposure. Mucus clearance is a function of cilia beat frequency and mucus viscoelastic properties. A change in mucus clearance may be due to one or the other component or both. In future studies, incorporating high speed digital imaging of the palate surface to determine cilia beat frequency will allow us to further differentiate between cilia related effects versus mucus related effects. Exposure to tobacco smoke with all its noxious agents and components will possibly allow us to separate and study in detail, the timing and appearance of different phases of the organism response. Our exposure model may allow us to individualize and study the inflammation phase resulting from the tobacco exposure. We can then focus on the injury phase occurring in the mucociliary system. Later we will explore mechanisms involved in the healing process. The remodeling of the ciliated epithelium following acute injury will be an important component of our studies and particularly how the remodeling is mediated. The time frame of the injury and recovery phases needs to be determined. Cilia do not work alone, but in association with other cilia to produce metachronal waves. Several metachronal waves may contribute to the propulsion of the mucus layer to flow over irregularities or non-ciliated areas. Our data indicates that acute tobacco exposure may have an initial and early irritation effect, possibly mediated by an unknown mechanism on the exposed palate that leads to inhibition of ciliary beat frequency or discoordination of metachronal waves. Such factors adversely affect clearance by imposing stasis or local eddies resulting in erratic clearance that may be the reason why scanning electron micrographs of palates after one cigarette showed cilia somewhat disordered, but not visibly disrupted. We have shown that tobacco smoke exposure interferes with mucociliary clearance. Sustained exposure may lead to loss of ciliated epithelium associated with activation of matrix metalloproteinases. Significant loss of cilia or ciliated epithelial cells results in disruption or even cessation of mucociliary clearance. Matrix metalloproteinases may be implicated in this injury through disruption of epithelial cell-to-cell or cell-to-basement membrane connections. Further clarification of the mechanisms involved will be undertaken in subsequent studies. Our exposure model can assess physiological, ultrastructural and molecular parameters in response to the initial deleterious effects of acute exposure to side stream tobacco smoke in an epithelial model homologous to the human airways. In these preliminary studies we did not attempt to mimic "human smoking conditions". However, our results show that MTV is affected even after exposure to one cigarette. Although concentration of smoke used in the present study is higher than likely encountered in typical environmental tobacco exposures, they are within an order of magnitude of those computed for exposure in poorly ventilated cars or homes. Future studies will try to replicate real conditions faced by non-smokers exposed to environmental tobacco smoke and to characterize the mediation and effectors of this acute injury.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517705.xml
548956	A New Paradigm in Eukaryotic Biology: HIV Tat and the Control of Transcriptional Elongation	Studies of the transcriptional transactivator (Tat), a key regulatory protein of HIV, have yielded insight into the control of eukaryotic transcription	Viruses are intracellular pathogens that are subject to intense selective pressures during their ongoing battles within the host. To propagate successfully, they must exploit numerous machineries of the infected cell. Thus, studies of their replicative cycles have yielded fundamental insights into eukaryotic biology. A prime example is the human immunodeficiency virus (HIV), which is a lentivirus that causes the acquired immunodeficiency syndrome (AIDS). Unlike simpler oncoviruses that rely exclusively on host cell machinery, lentiviruses code for additional accessory and regulatory proteins that act as molecular switches at different stages of viral entry and exit from the infected cell. Studying the actions of these viral proteins has yielded understanding of diverse cellular functions such as the innate immunity against retroviruses, control of transcriptional elongation, export of macromolecules from the nucleus to the cytoplasm, and intracellular trafficking of proteins (reviewed in [ 1 ]). The transcriptional transactivator (Tat) is a key regulatory protein of HIV. It is expressed early after the virus integrates into the cell, and stimulates the elongation of RNA polymerase II (RNAPII). This type of transcriptional control had not been previously appreciated; thus, work on Tat established a new paradigm in the field of eukaryotic biology. Moreover, these findings impacted greatly studies of cotranscriptional processing of nascent mRNA. To understand these processes better, we need to start with the basics of transcriptional control. RNAPII is the enzyme that transcribes protein-coding genes in eukaryotic cells. Elegant studies in vitro first suggested that the simple recruitment of RNAPII to transcription units was not sufficient for the copying of genes and cotranscriptional processing of their transcripts. Rather, distinct steps could be defined, which began with the assembly of the preinitiation complex (PIC), promoter clearance, pausing, and arrest, and ended with efficient elongation of transcription (reviewed in [ 2 ]). The central component of PIC is the general transcription factor (GTF) TFIID, which contains the TATAbox- binding protein (TBP) and 12 to 15 TBP-associated factors (TAFs). TFIID acts as a “landing pad” for other GTFs and RNAPII to nucleate PIC assembly. Moreover, TAFs serve as coactivators to a diverse set of activators. Both an ordered stepwise assembly and the recruitment of the 100-plus-subunit “holoenzyme” have been proposed to be critical for the positioning of RNAPII at start sites of transcription. Next, the GTF TFIIH unwinds the DNA, opens the transcription bubble, and phosphorylates serines at position 5 in the C-terminal domain (CTD) of the RPB1 subunit of RNAPII (reviewed in [ 2 ]). This phosphorylation is critical for the recruitment of complexes that put a 7-methylguanylate cap on the 5′ end of nascent transcripts. After the transcription complex clears the promoter, the negative transcription elongation factor (N-TEF) is recruited to the RNAPIIa (reviewed in [ 3 ]). It consists minimally of 5,6- dichloro-1-β-D-ribofuranosylbenzimidazole riboside (DRB)- sensitivity-inducing factor (DSIF) [ 4 ] and negative elongation factor (NELF) [ 5 ]. They bind and arrest RNAPII distal to the promoter cooperatively. Such arrested transcription complexes have now been found on many inducible genes in Drosophila melanogaster (reviewed in [ 6 ]) and humans [ 7 ]. The transition to robust elongation depends on the positive transcription elongation factor b (P-TEFb) (reviewed in [ 3 ]). P-TEFb contains the cyclin-dependent kinase 9 (CDK9) and one of four possible C-type cyclins. When recruited to stalled transcription complexes, P-TEFb phosphorylates serines at position 2 in the CTD [ 8 ], the Spt5 subunit of DSIF [ 9 ], and the RD subunit of NELF [ 10 ]. These modifications result in heavily phosphorylated RNAPII (RNAPIIo), the recruitment of the Elongator, which contains splicing and polyadenylation machineries, and the conversions of DSIF and NELF into elongation factors. RNAPIIo now copies the gene and directs the cotranscriptional processing, i.e., splicing and polyadenylation, of primary transcripts. Upon successful polyA addition, the CTD phosphatase FCP1 dephosphorylates RNAPIIo. RNAPIIa dissociates from DNA, and the transcription cycle starts all over again (reviewed in [ 2 ]). Tat is unique among transcriptional activators in eukaryotic cells in that it functions via RNA rather than DNA promoter elements ( Figure 1 ). It binds the transactivation response element (TAR) that forms a stable RNA stem loop at the 5′ end of all viral transcripts. Thus, Tat requires minimally the transcription of TAR before it can stimulate HIV transcription from the long terminal repeat (LTR). Indeed, in the absence of Tat, RNAPIIa clears the HIV LTR successfully but soon arrests, yielding predominantly short viral transcripts [ 11 ]. Tat binds the 5′ bulge in TAR via its arginine-rich motif from positions 49 to 57, where a central arginine (R52) is key for this interaction. However, this binding is not sufficient for Tat's function in vivo. Adjacent to the arginine-rich motif lie N-terminal core and cysteine-rich regions, which form the activation domain of the protein. This activation domain binds cyclin T1 (CycT1) from P-TEFb, whose partner is CDK9 [ 12 ]. As a consequence, P-TEFb and Tat bind TAR cooperatively. The final proof that P-TEFb is the cellular cofactor for Tat came from studies of HIV transcription in murine cells, where the introduction of the human CycT1 protein restores Tat function [ 12 ]. The same effect can be achieved by substituting just the tyrosine with the cysteine at position 261, such as are found in murine and human CycT1 proteins, respectively [ 13 ]. A paper in this issue of PLoS Biology suggests that Tat and P-TEFb can also recruit TAF-independent transcription complexes to the HIV LTR [ 14 ] ( Figure 1 ). Possibly, this assembly reflects interactions between CycT1 and the unphosphorylated CTD of RNAPIIa [ 15 ]. Figure 1 Activation of HIV Transcription by Tat Activators (red circles) that bind the HIV LTR promoter (light-blue rectangle) assemble the PIC and recruit RNAPIIa to the start site of transcription. For simplicity, only RNAPIIa in the PIC is presented. The yellow sphere with two open circles, depicting serines at position 5 and 2 within the CTD (S5 and S2, respectively), represents the unphosphorylated CTD of RNAPIIa (white sphere). TFIIH, which performs DNA-helicase and CTD-kinase activities, melts the DNA and phosphorylates S5 (red circle in the CTD; P-S5), resulting in promoter clearance. RNAPIIa transcribes TAR (red hairpin) and is paused by the binding of N-TEF, DSIF, and NELF, which are presented as blue spheres. The RD subunit of NELF binds the bottom stem in TAR. P-TEFb (comprising the red [CDK9] and pink [CycT1] spheres), which binds TAR together with Tat (small red sphere), phosphorylates S2 (red circle in the CTD; P-S2) to form elongating RNAPIIo (large red sphere). It also phosphorylates Spt5 in DSIF and RD in NELF, which become elongation factors, with the latter dissociating from TAR. In addition, P-TEFb, possibly independent of its kinase activity, assembles PIC via recruitment of TBP and RNAPIIa (dotted arrow). The phosphorylated CTD in RNAPIIo now binds the Elongator, which contains splicing machinery and polyadenylation factors. The red sphere at the 5′ end of the HIV transcript (red line) represents its cap. Finally, p300 acetylates Tat (magenta circle) and dissociates it from TAR. Acetylated Tat binds P-CAF and transfers it to RNAPIIo, possibly facilitating chromatin remodeling. Collectively, efficient RNAPII elongation of viral transcription ensues. The assembly and disassembly of the complex between PTEFb, Tat, and TAR is a regulated process in vivo. Whereas the phosphorylation of CDK9 strengthens this complex [ 16 ], the acetylation of the lysine at position 50 in Tat weakens it [ 17 ]. Upon this disruption, acetylated Tat is liberated from P-TEFb and recruits the p300/CREB-binding protein– associated factor (P-CAF) to the elongating RNAPIIo, most likely facilitating chromatin remodeling. In this issue of PLoS Biology, Pagans et al. now demonstrate that acetylated Tat is deacetylated by SIRT1 [ 18 ] ( Figure 1 ). In this way, Tat can reassemble with P-TEFb on TAR. Clearly, P-TEFb plays a key role in the control of transcriptional elongation. Although Tat was the first activator known that could recruit P-TEFb to initiating RNAPII, additional members of this group were soon identified. They include the androgen receptor, c-Myc, the class II transactivator (CIITA), myoblast determination protein (MyoD), and nuclear factor κ-B (NF-κB). The last one is of great interest as it explains how the HIV genome can be transcribed before the synthesis of Tat [ 19 ]. Cellular activation triggers the nuclear translocation of NF-κB, where it binds the HIV enhancer, leading to the stimulation of viral transcription. It is not surprising that proviral latency, in which low levels of transcription or only short HIV transcripts containing TAR are observed, would in large part reflect the absence of these activators. Indeed, in many of these latently infected cells, the induction of NF-κB or the addition of Tat leads to the reactivation of viral replication and spreading of the infection [ 20 , 21 ]. Recently, important aspects of the regulation of P-TEFb have been revealed ( Figure 2 ). Of interest, P-TEFb exists in two complexes in cells [ 22 , 23 ]. The larger measures approximately 500 kDa and contains the hexamethylene bisacetamide (HMBA)–induced protein 1 (HEXIM1) and 7SK small nuclear RNA (snRNA) in addition to P-TEFb [ 24 , 25 ]. In this large complex, Cdk9 is enzymatically inactive. HEXIM1 was identified as the inducible gene following the exposure of vascular smooth muscle cells to a potent differentiating agent, HMBA [ 26 ]. 7SK snRNA is one of the most abundant snRNA species, whose function remained a mystery for over a decade. Of interest, targeting of P-TEFb by HEXIM1 and 7SK snRNA contributes significantly to the control of cell growth and differentiation. For example, growth signals liberate P-TEFb from the large complex in the course of cardiac hypertrophy in mice, a disease characterized by the enlargement of myocytes due to a global increase in mRNA synthesis [ 27 ]. Also, following stress, ultraviolet light, or the administration of actinomycin D and DRB to cells, the large complex is converted to the small complex to stimulate transcription [ 22 , 23 ]. Figure 2 Inhibition of P-TEFb by the Coordinate Actions of HEXIM1 and 7SK snRNA HEXIM1 (blue sphere) binds the 5′ half of 7SK snRNA (red structure with multiple hairpins). Upon this binding, P-TEFb joins this RNA–protein complex and becomes enzymatically inactive, depicted by CDK9 as a black sphere. For simplicity, only the CDK9/CycT1 heterodimer is presented. Multiple stimuli, including stress, ultraviolet light, actinomycin D, DRB, and hypertrophic signals, dissociate HEXIM1 and 7SK snRNA from P-TEFb, possibly by preventing the RNA–protein interaction. In this way, P-TEFb is rendered active, depicted by CDK9 as a red sphere. How central is P-TEFb to eukaryotic transcription? In Saccharomyces cerevisiae , there are two candidates for PTEFb, CTDK-1 and Bur1/2. CTDK1-negative but not Bur1/Bur2-negative yeasts still grow, albeit poorly and only on rich media (reviewed in [ 2 ]). In Caenorhabditis elegans , genetic inactivation of CDK9 or CycT1 and CycT2 resulted in the inhibition of all RNAPII transcription [ 8 ]. Moreover, in D. melanogaster , following heat shock, PTEFb is recruited upstream of activated promoters [ 28 ]. Although no murine knockouts of subunits of P-TEFb have been reported, DRB and flavopiridol, two ATP analogs that inhibit the kinase activity of CDK9, can inhibit nearly all transcription by RNAPII in human cells [ 29 ]. Indeed, as P-TEFb is a coactivator of potent activators that mediate effects of enhancers and can itself activate transcription when placed on sites distal to promoter elements [ 15 ], it might mediate many more signaling events than those of heat shock, ultraviolet light, stress, and hypertrophy. Conversely, the inhibition of P-TEFb could explain the mode of action of some transcriptional repressors. Indeed, the global transcriptional repressor PIE-1, the regulator of embryogenesis in C. elegans , binds the histidine-rich stretch in CycT1, thus decoying P-TEFb away from RNAPII and blocking the elongation of transcription [ 30 ]. These are exciting findings and suggest a plethora of future experiments, including the genetic inactivation of subunits of P-TEFb and isoforms of HEXIM1 in the mouse. Of special interest are questions as to where to place this mechanism of transcriptional regulation in the hierarchy of competing or complementary processes. What roles do different P-TEFb complexes play in the transcription of specific genes? How central will the regulation of P-TEFb be to cellular growth, proliferation, and differentiation, and what roles will it play in normal development and disease states? As to HIV, how can we use our knowledge of P-TEFb to slow down viral replication and/or to eliminate the state of proviral latency in the host? Obviously, we are only at the beginning of this journey, which promises to change radically our view of eukaryotic transcription.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC548956.xml
520817	Mitoxantrone pleurodesis to palliate malignant pleural effusion secondary to ovarian cancer	Background Advanced ovarian cancer is the leading non-breast gynaecologic cause of malignant pleural effusion. Aim of this study was to assess the efficacy of mitoxantrone sclerotherapy as a palliative treatment of malignant pleural effusions due to ovarian cancer. Methods Sixty women with known ovarian cancer and malignant recurrent symptomatic pleural effusion were treated with chest tube drainage followed by intrapleural mitoxantrone sclerotherapy. Survival, complications and response to pleurodesis were recorded. The data are expressed as the mean ± SEM and the median. Results The mean age of the entire group was 64 ± 11,24 years. The mean interval between diagnosis of ovarian cancer and presentation of the effusion was 10 ± 2,1 months. Eighteen patients (30%) had pleural effusion as the first evidence of recurrence. The mean volume of effusion drained was 1050 ± 105 ml and chest tube was removed within 4 days in 75% of patients. There were no deaths related to the procedure. Side effects of chemical pleurodesis included fever (37–38,5°C) chest pain, nausea and vomiting. At 30 days among 60 treated effusions, there was an 88% overall response rate, including 41 complete responses and 12 partial responses. At 60 days the overall response was 80% (38 complete responses and 10 partial responses). The mean survival of the entire population was 7,5 ± 1,2 months. Conclusions Mitoxantrone is effective in the treatment of malignant pleural effusion secondary to ovarian cancer without causing significant local or systemic toxicity.	Background Cancer accounts for 40% of all pleural effusions, especially in patients over 50 years old [ 1 ]. Bronchogenic and breast cancer account for 75% of malignant pleural effusions, with the remaining 25% represented by a cross-section of other neoplastic diseases [ 2 ]. Approximately two thirds of malignant pleural effusions occur in women because of the strong association with breast and ovarian cancer [ 3 ]. Advanced ovarian cancer is the leading non-breast gynaecologic cause of malignant pleural effusion. Pleural metastases were found in 48% of women who died from ovarian cancer [ 4 ]. The general approach to managing malignant effusions is determined by symptoms, performance status of the patient, expected survival and response of the known primary tumor to systemic treatment. Intervention options range from observation in the case of asymptomatic effusions through simple thoracentesis to more invasive methods such as thoracoscopy, pleuroperitoneal shunting and pleurectomy. In patients with reasonable survival expectancy and good performance status every attempt should be made to prevent recurrence of the effusion. Intercostal tube drainage with instillation of a sclerosing agent, resulting in the obliteration of the pleural space, is the most widely used method to control recurrent symptomatic malignant pleural effusions. Aim of this study was to study 60 patients with ovarian cancer who had a pleural effusion as a direct consequence of metastatic disease and to assess the efficacy of mitoxantrone as a sclerosing agent. Methods Over an 8-year period (1996–2003), all patients with known ovarian malignancy and recurrent symptomatic malignant pleural effusion referred to Thoracic Surgery Department of Theagenio Cancer Hospital for drainage and sclerotherapy, were eligible to participate in this study. This study was approved by the Theagenio Cancer Research Ethics Committee and patients were included after giving their informed consent. All patients satisfied the following eligibility criteria: 1. Known ovarian malignancy. 2. Recurrent symptomatic malignant pleural effusion. The diagnosis established by positive pleural fluid cytology on thoracentesis or positive pleural biopsy. 3. Evidence of expansion of the lung after fluid drainage and abscence of bronchial obstruction and/or fibrosis preventing lung expansion. 4. No previous intrapleural therapy. 5. Predicted survival of >1 month. Patients were ineligible if they had a history of cardiac disease, obstructive jaundice or surgery within the previous month. No patient had systemic chemotherapy immediately prior to or during the first 30-day interval following sclerotherapy. Sixty women fulfilled the above eligibility criteria. Pretreatment assessment was performed during admission and included history and physical examination, full blood count, liver biochemistry, electrocardiogram, a pre-drainage base line posteroanterior and lateral chest radiograph and other imaging as clinically indicated. A chest tube (28–32 F) was inserted into the midaxillary line through the 5 th or 6 th intercostal space under local anesthesia and in some case additional intravenous benzodiazepines and/or narcotics. The pleural effusion was drained to dryness initially by gravity and followed if necessary by suction from a wall-mounted suction pump using a pressure of 20 cm H 2 O usually for 12–24 hours to achieve complete drainage of the effusion and lung re-expansion. Daily tube outputs were recorded and when drainage fell below 100 ml in a 24 h period, posteroanterior and lateral chest radiographs were obtained to assure that the fluid had been sufficiently evacuated, there were no loculated collections and the lung had fully re-expanded. Then the patients were eligible for pleurodesis. Fifty ml of normal saline solution containing 2 mg/kg lidocaine were infused through the chest tube. After 15 minutes, a pleurodesis solution containing a mixture of 40 mg mitoxantrone and 20 ml normal saline was infused into the pleural cavity, after which the tube was clamped for 2 hours, while the patients changed position (rotated 90°) every 15 minutes. The tube then was re-opened. If the post-sclerotherapy drainage was <100 ml per day thetube was removed. Complications related to the procedure were recorded. Post-sclerotherapy posteroanterior and lateral chest radiographs were obtained immediately after tube removal in order to be compared with others obtained 30 and 60 days later. The radiographic response was determined on posteroanterior and lateral chest radiographs by observing the level of fluid meniscus overlying the costophrenic or vertebrophrenic angles and was determined as follows: complete response(CR) – no re-accumulation of pleural fluid, partial response (PR) – fluid recurrence less than 50% of the original level without symptoms or not requiring repeat drainage, progressive disease (PD) – re-accumulation to or above the original level with symptoms and requiring repeat drainage. Survival was calculated from the day of diagnosis of pleural effusion to the day of death or to the last day of follow up if alive. The data are expressed as the mean ± SEM and the median. Results Sixty women were included in this study. The mean age of the entire group was 64 ± 11,24 years. The interval between diagnosis of ovarian cancer and the development of a subsequent malignant pleural effusion ranged from 1 to 36 months (mean: 10 ± 2,1 months). Fifty one patients (85%) had unilateral effusion and 9 (15%) bilateral. Histology, degree of differentiation and TNM stage [ 5 ] at the time of diagnosis of the primary tumor are shown in Table 1 . Table 1 Histology, degree of differentiation and TNM stage at the time of diagnosis of the primary tumor. Histology Number of patients (n:60) Percentage Serous cystadenocarcinoma 38 63,3% Mucinous adenocarcinoma 8 13,4% Mixed 4 6,7% Clear cell 3 5% Endometrioid 2 3,4% Unknown 5 8,4% Degree of differentiation G1 16 26,7% G2 18 30% G3 26 43,3% TNM stage I 8 13,3% II 18 30% III 22 36,7% IV 12 20% Eigtheen patients (30%) had pleural effusion as the first manifestation of recurrent disease, whereas 42 patients (70%) were already diagnosed as having local or distant spread before the onset of pleural effusion. These 42 patients with preexisting metastases showed a variable pattern of secondary spread. Eighteen patients had parenchymal liver metastases and intraabdominal lymph nodes, 11 patients had liver metastases only, 8 had synchronous lung and liver metastases, 2 had lung metastases only, 1 patient had umbilical nodule, 1 had anterior wall abdominal wall infiltration and 1 had brain metastases. The mean volume of effusion drained was 1050 ± 105 ml (range: 450–1500 ml). Chest tube was removed within 4 days in 75% of patients (range: 3 – 10 days). There were no deaths related to the thoracostomy procedure. One patient experienced vasovagal reflex during the procedure with systemic hypotension and intense pleuritic pain. Hypotension was treated with intravenous fluids and the pain was controlled with narcotics. This episode lasted 20 minutes. The patient recovered without incident. The most frequent complications related to pleurodesis were fever (temperature > 37°C), chest pain, nausea and vomiting (Table 2 ). Table 2 Complications related to chemical pleurodesis with mitoxantrone Complications Number of patients (n:60) None 31 (51,6%) Fever 16 (26,6%) Chest pain 12 (20%) Nausea 11 (18,3%) Vomiting 9 (15%) Diarrhea 4 (6,6%) Alopecia 1 (1,6%) Skin Rash 1 (1,6%) Dyspnea 1 (1,6%) Myelosuppression 1 (1,6%) Three patients died within 1 month of pleurodesis due to rapid progression metastatic disease. At 30 days, 57 patients were alive and 41 out of them had a complete response and 12 had a partial response. The overall response to chemical pleurodesis with mitoxantrone was 88% (53/57 patients). Four patients had progressive disease and revealed reaccumulation of fluid to or above the original level. At 60 days 52 patients were alive and 38 out of them had a complete response and 10 had a partial response. The overall response was 80% (complete response 38/60 patients – 63,4%, partial response 10/60 patients – 16,6%). Follow up ranged from 10 days to 38 months with a mean of 10 ± 1,36 months. Eight patients out of the 41, who initially had complete response developed later recurrent pleural effusion and needed again tube thoracostomy and a second attempt of chemical pleurodesis. The mean survival of the entire study population was 7,5 ± 1,2 months (median: 5,4 months). Discussion Management of malignant pleural effusions depends on the underlying malignancy, extent of disease, potential effectiveness of treatment and performance status. In patients with lymphoma, small cell lung cancer or germ cell neoplasms, pleural effusions may be controlled initially by systemic therapy alone. In patients with metastatic breast or non small cell lung carcinoma, local palliative treatment is often required. Since malignant pleural effusions are frequently a preterminal event with a 30-day mortality rate of 29 to 50%, treatment is directed toward symptomatic relief with minimal discomfort, inconvenience and cost [ 6 - 8 ]. Local treatment options include repeated thoracenteses, chest tube drainage with sclerotherapy, pleuroperitoneal shunt or pleurectomy. Repeated thoracentesis is usually a temporizing measure and carries the risk for pneumothorax and pleural infection [ 9 ]. Inpatient drainage with large-bore tubes (28–36 F) is effective, with variable 30-day success rates reported between 55% and 95% [ 10 ]. For this reason, large-bore tube thoracostomy with sclerotherapy has become the most common palliative treatment for malignant effusions. It has to be mentioned that recent studies have shown that small drainage catheters (10 to 14 F) are as effective as large bore chest tubes in the treatment of malignant effusions [ 11 ]. Using imaging guidance, small tubes can be placed into loculated collections, are well tolerated and have complication rates less than the larger tubes [ 12 ]. Pleural effusion due to metastatic ovarian cancer is a frequent phenomenon and as shown in our study it can occur as early as one month or as late as 36 months with a median of 8,5 months. This is in complete accord with the study of Cheng et al who found a median interval of 9 months [ 13 ]. When effusion occurs within 1 month of diagnosis of ovarian cancer, one is probably dealing with IV stage and clinical experience has proved that these patients have an especially poor prognosis. Numerous sclerosing agents have been used to treat malignant pleural effusions. Until recently, tetracycline was the most commonly used sclerosing agent with response rates ranging from 25 to 100% [ 14 , 15 ]. Because the intravenous form of tetracycline is no longer available, doxycycline has been proposed as an alternative. Bleomycin has been studied extensively as a sclerosing agent [ 16 , 17 ]. Goff et al succesfully used bleomycin intrapleurally to treat malignant pleural effusions from gynecological cancer with a 71% overall response at 30 days and minimal adverse reactions. Intrapleural instillation is usually well tolerated but a few patients may report mild fever or transient nausea. Pleuritic pain and rigors are rarely reported side effects. This relative lack of systemic toxicity is likely due to limited absorption of bleomycin (approximately 40%) of the pleural cavity [ 18 ]. At 30 days bleomycin has been reported to be superior to tetracycline [ 19 ]. Talc has proved to be one of the most effective sclerosing agents for treating malignant pleural effusions. Talc causes severe pleuritis resulting in effective pleurodesis but can worsen dyspnea and can result in respiratory failure [ 20 ]. Other complications associated with talc pleurodesis include fever, acute pneumonitis, granulomatous pneumonitis and empyema [ 21 ]. Talc is instilled either as a slurry via chest tube or insufflated via thoracoscope. Many other chemotherapeutic agents such as doxorubicin, cisplatin and cytarabine combination, etoposide, fluorouracil and mitomycin have been used for sclerotherapy. In addition radioactive isotopes, corynebacterium parvum, interferon and recombinant interleukin-2 have been instilled in the pleural space for treatment of malignant pleural disease. Response rate have been variable and less than optimal. Side effects are not inconsequential and thus none of these agents have gained widespread use [ 22 ]. Mitoxantrone is a synthetic anthracenedione which has been demonstrated to be effective in the treatment of peritoneal and pleural effusion. From a pharmacological point of view, mitoxantrone may be an especially appropriate choice due to its higher molecular weight and polarity since this may be factor important in prolonging contact with the pleura. The mechanism of intrapleural action of mitoxantrone has not yet been established. Both the inflammatory and antineoplastic activity of mitoxantrone intrapleurally have been described [ 23 , 24 ]. Our findings are consistent with the findings of others. In a prospective study in 18 patients, Musch et al [ 25 ] reported a 30-day success rate of 75%. A comparative study including bleomycin and mitoxantrone showed almost an equal 30-day response of 64% and 67% respectively [ 26 ]. Van Belle et al [ 27 ] had an overall 30-day response of successful pleurodesis of 67% in patients with ovarian cancer (2/3 patients). Morales et al [ 28 ] treated a group of 21 patients with malignant pleural effusions, with instillation of mitoxantrone with a 100% response and no toxic effects. There is only one study which proved mitoxantrone ineffective. Groth et al. [ 29 ] presented a prospective randomized trial on the treatment of malignant pleural effusions with intrapleural mitoxantrone versus placebo (pleural tube alone with instillation of isotonic NaCl). Their data suggest no statistically significant difference between the two arms with respect to response and response duration. Our study confirmed the majority of previous reports that mitoxantrone is an effective agent in controlling recurrent malignant pleural effusions. The overall 30-day response rate was 88%. Side effects were mild and rare. To develop new treatment plans for the management of pleural effusions, one must consider several requirements. First, no treatment regimen should exacerbate patients' symptoms, since palliation is the main aim. Second, seriously ill patients should not be subjected to procedures associated with high mortality and morbidity. Third, since about half the patients with pleural effusion will have no other clinically apparent metastases, treatment should be local rather than systemic. To be successful, the local treatment has to be effective and given at the first sign of the effusion, because inadequate or delayed treatment may eliminate the possibility of any subsequent therapy being effective, by producing loculation of the effusion. Conclusions Pleural effusion often occurs during the course of ovarian cancer. Chemical pleurodesis via bedside thoracostomy has been shown to be effective and has become a common therapeutic approach. Using this approach we found mitoxantrone to be highly effective at controlling malignant pleural effusions and decreasing the associated symptoms of dyspnea and pain. Our data justify further studies in a controlled setting to elucidate the biological action and prognostic relevance of mitoxantrone in the treatment of malignant pleural effusions and to compare this agent with other treatment procedures. Competing interests None declared. Authors's contributions NB conceived the study.and performed thestatistical analysis. MV and RV participated in the design of the study. KK and CT conceived of the study and participated in its coordination. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC520817.xml
368163	Endosymbiosis: Lessons in Conflict Resolution	Endosymbiotic bacteria live within a host species. There are many and diverse examples of such relationships, the study of which provides important lessons for ecology and evolution	Symbiosis, an interdependent relationship between two species, is an important driver of evolutionary novelty and ecological diversity. Microbial symbionts in particular have been major evolutionary catalysts throughout the 4 billion years of life on earth and have largely shaped the evolution of complex organisms. Endosymbiosis is a specific type of symbiosis in which one—typically microbial—partner lives within its host and represents the most intimate contact between interacting organisms. Mitochondria and chloroplasts, for example, result from endosymbiotic events of lasting significance that extended the range of acceptable habitats for life. The wide distribution of intracellular bacteria across diverse hosts and marine and terrestrial habitats testifies to the continued importance of endosymbiosis in evolution. Among multicellular organisms, insects as a group form exceptionally diverse associations with microbial associates, including bacteria that live exclusively within host cells and undergo maternal transmission to offspring. These microbes have piqued the interest of evolutionary biologists because they represent a wide spectrum of evolutionary strategies, ranging from obligate mutualism to reproductive parasitism ( Buchner 1965 ; Ishikawa 2003 ) ( Box 1 ; Table 1 ). In this issue of PLoS Biology , the publication of the full genome sequence of the reproductive parasite Wolbachia allows the first genomic comparisons across this range ( Wu et al. 2004 ). Table 1 Examples of Bacterial Endosymbionts of Insects Lifestyle Extremes in Insect Endosymbionts At one end of the spectrum, beneficial endosymbionts provide essential nutrients to about 10%–15% of insects and provide models for highly specialized, long-term mutualistic associations ( Figure 1 ). These ‘primary’ endosymbionts are required for the survival and reproduction of the host, most of which feed on unbalanced diets such as plant sap, blood, or grain, and occur within specialized host cells called bacteriocytes (or mycetocytes) ( Baumann et al. 2000 ; Moran and Baumann 2000 ). Molecular phylogenetic analyses demonstrate stability of these obligate mutualists over long evolutionary periods, ranging from tens to hundreds of millions of years. By allowing their hosts to exploit otherwise inadequate food sources and habitats, the acquisition of these mutualists can be viewed as a key innovation in the evolution of the host ( Moran and Telang 1998 ). Owing to their long-term, stable transmission from generation to generation (vertical transmission), these cytoplasmic genomes have been viewed as analogs to organelles. Figure 1 A carpenter ant, Camponotus pennsylvanicus , Hosts the Mutualistic Bacterial Endosymbiont Blochmannia Like all species of the ant genus Camponotus , the wood-nesting C. pennsylvanicus (shown here) possesses an obligate bacterial endosymbiont called Blochmannia . The small genome of Blochmannia retains genes to biosynthesize essential amino acids and other nutrients ( Gil et al. 2003 ), suggesting the bacterium plays a role in ant nutrition. Many Camponotus species are also infected with Wolbachia , an endosymbiont that is widespread across insect groups. (Photo courtesy of Adam B. Lazarus.) By contrast, reproductive parasites of insects, including Wolbachia ( O'Neill et al. 1998 ) and the more recently discovered endosymbiont in the Bacteroidetes group (also called CFB or CLO) ( Hunter et al. 2003 ), propagate in insect lineages by manipulating host reproduction. These maternally inherited bacteria inflict an impressive arsenal of reproductive alterations to increase the frequency of infected female offspring, often at the expense of their hosts. Such mechanisms include cytoplasmic incompatibility, parthenogenesis, and male killing or feminization. As parasites, these bacteria rely on occasional horizontal transmission to infect new populations ( Noda et al. 2001 ) and, by directly altering reproductive patterns, may be a causative agent of host speciation ( Bordenstein et al. 2001 ). Comparative molecular analysis of insect endosymbionts over the past decade has provided new insights into their distribution across hosts, their varying degrees of stability within host lineages (ranging from cospeciation to frequent host-switching), and their impressive genetic diversity that spans several major bacterial groups. More recently, studies in genomics of obligate mutualists—and now Wolbachia —illuminate mechanisms of host–symbiont integration and the distinct consequences of this integration in various symbiotic systems. In addition, since hosts and symbionts often have different evolutionary interests, the diverse features of insect–bacterial associations can be understood as different outcomes in the negotiation of genetic conflicts. Some recent highlights and tantalizing research areas are described below. Endosymbiont Genomes: Spanning the Gamut from Static to Plastic The distinct lifestyle of endosymbionts has clear effects on rates and patterns of molecular evolution. Compared to free-living relatives, endosymbionts are thought to have reduced effective population sizes due to population bottlenecks upon transmission to host offspring and, in the case of obligate mutualists that coevolve with their hosts, limited opportunities for gene exchange. The nearly neutral theory of evolution ( Ohta 1973 ) predicts accelerated fixation of deleterious mutations through random genetic drift in small populations, a phenomenon that has been observed in endosymbionts ( Moran 1996 ; Lambert and Moran 1998 ). Over time, this lifestyle-associated accumulation of deleterious mutations may negatively affect the fitness of both the host and symbiont. It is increasingly clear the distinct lifestyle of endosymbionts also shapes the architecture and content of their genomes, which include the smallest, most AT-rich bacterial genomes yet characterized ( Andersson and Kurland 1998 ; Moran 2002 ). A common theme is substantial gene loss, or genome streamlining, which is thought to reflect an underlying deletion bias in bacterial genomes combined with reduced strength or efficacy of selection to maintain genes in the host cellular environment. As a result of gene loss, these bacteria completely rely on the host cell for survival. Because they cannot be easily cultured apart outside of the host for traditional genetic or physiological techniques, analysis of genome sequence offers a valuable tool to infer metabolic functions that endosymbionts have retained and lost and to elucidate the steps in the evolutionary processes of genome reduction. Since 2000, full genome sequences have been published for Buchnera of three aphid host species, Wigglesworthia of tsetse flies, and Blochmannia of ants ( Shigenobu et al. 2000 ; Akman et al. 2002 ; Tamas et al. 2002 ; Gil et al. 2003 ; van Ham et al. 2003 ). Parallels among these mutualist genomes include their small size (each smaller than 810 kb), yet retention of specific biosynthetic pathways for nutrients required by the host (for example, amino acids or vitamins). However, genomes also show signs of deleterious deletions. Early gene loss in Buchnera involved a few deletions of large contiguous regions of the ancestral genome and often included genes of unrelated functions ( Moran and Mira 2001 ). These ‘large steps’ imply that genome reduction involved some random chance (due to the location of genes in the ancestral chromosome) and selection acting on the combined fitness of large sets of genes, rather than the fitness of individual loci. Such deletions are apparently irreversible in obligate mutualists, which lack recombination functions and genetic elements, such as prophages, transposons, and repetitive DNA that typically mediate gene acquisition. The scarcity of these functions, combined with limited opportunities to recombine with genetically distinct bacteria, may explain the unprecedented genome stability found in Buchnera compared to all other fully sequenced bacteria ( Tamas et al. 2002 ) and a lack of evidence for gene transfer in other mutualist genomes. Stability also extends to the level of gene expression, as obligate mutualists have lost most regulatory functions and have reduced abilities to respond to environmental stimuli ( Wilcox et al. 2003 ). The Wolbachia genome presented in this issue allows the first genome comparisons among bacteria that have adopted divergent evolutionary strategies in their associations with insects ( Wu et al. 2004 ). Like other parasites, but unlike long-term mutualists, Wolbachia may experience strong selection for phenotypic variation, for example, to counter improved host defenses, to compete with distinct Wolbachia strains that coinfect the same host, or to increase its transmission to new host backgrounds. High levels of recombination in Wolbachia (for example, Jiggins et al. 2001 ) may allow rapid genetic changes in this parasite and may be catalyzed by the exceptionally high levels of repetitive DNA and mobile elements in its genome ( Wu et al. 2004 ). Other bacteria that colonize specialized niches for long periods and lack co-colonizing strains also possess high levels of repetitive chromosomal sequences. For example, among ulcer-causing Helicobacter pylori in primate guts, repetitive DNA mediates intragenomic recombination and may provide an important source of genetic variation for adaptation to dynamic environmental stresses ( Aras et al. 2003 ). The potential contributions of repetitive DNA and phage to intragenomic and intergenomic recombination in Wolbachia are exciting areas of research ( Masui et al. 2000 ). The Wolbachia genome also provides a valuable tool for future research to test whether plasticity extends to gene content variation among Wolbachia strains and labile gene expression patterns. Between these two extremes of obligate mutualism and reproductive parasitism lies a spectrum of secondary symbionts of insects, most of which have not yet been studied in detail. Such ‘guest’ microbes transfer among diverse host species ( Sandström et al. 2001 ), may provide more subtle or occasional benefits (for example, relating to host defense against parasitoids [ Oliver et al. 2003 ]), and could represent an intermediate stage between a free-living lifestyle and obligate endosymbiosis. Genome-level data from these secondary symbionts promise to shed light on the range of lifestyles between obligate mutualism and reproductive parasitism and on the early stages in the transition to each. Microarray-based comparisons of gene content among Escherichia coli , a facultative mutualist of tsetse flies ( Sodalis glossinidius ), and a relatively young mutualist of weevils ( Sitophilus oryzae primary endosymbiont [SOPE]) show that genome streamlining in the endosymbionts may preclude extracellular existence, and highlight modifications in metabolic pathways to complement specific host physiology and ecology ( Rio et al. 2003 ). In addition, these endosymbionts may employ similar mechanisms as intracellular parasites in overcoming the shared challenges of entering host cells, avoiding or counteracting host defense mechanisms, and multiplying within a host cellular environment ( Hentschel et al. 2000 ). The rapidly growing molecular datasets for secondary (or young primary) insect endosymbionts have identified pathways that are considered to be required for pathogenicity, such as Type III secretion ( Dale et al. 2001 , 2002 ). Such pathways may therefore have general utility for bacteria associated with host cells and may have evolved in the context of beneficial interactions. Genetic Conflicts and Host–Symbiont Dynamics Given their diverse evolutionary strategies, insect endosymbionts also provide a rich playing field to explore genetic conflicts ( Frank 1996a , 1996b ), which might involve the mode of symbiont transmission, the number of symbionts transmitted, and the sex of host offspring. Genetic conflicts described between organelle and nuclear genomes of the same organism ( Hurst 1995 ) can provide a context to understand the evolutionary dynamics of insect–bacterial associations and the diverse outcomes of these relationships. For example, the uniparental (maternal) mode of inheritance of both mitochondria and insect endosymbionts may reflect host defense against invasion by foreign microbes with strong deleterious effects, which spread more easily under biparental inheritance ( Law and Hutson 1992 ). Host–symbiont conflicts over offspring sex ratio are quite apparent in reproductive parasites ( Vala et al. 2003 ). While the bacteria favor more female offspring and employ a variety of mechanisms to achieve this, the host typically favors a more balanced sex ratio. This conflict may lead to changes in the host that counter the symbiont's effect on sex ratio. For example, the spread of Wolbachia in a spider mite population caused selection on host nuclear genes that decrease the symbiont-induced sex ratio bias ( Noda et al. 2001 ). Obligate mutualists also experience genetic conflicts with the host regarding transmission mode and number. In general, symbionts generally favor dispersal out of the host to avoid competition with their close relatives, while hosts are expected to restrict symbiont migration and thus reduce the virulent tendencies ( Frank 1996b ). In obligate mutualisms, there may be little room for negotiation. For example, the highly conserved, host-controlled determination of aphid bacteriocytes ( Braendle et al. 2003 ) and the phylogenetic congruence observed in numerous studies suggest that aphids have won this conflict over symbiont transfer. However, the number of bacteria transmitted may be more flexible and is known to vary among aphid taxa ( Mira and Moran 2002 ). Models indicate that the fixation rate for symbiont-beneficial (selfish) mutations increase with the number of symbionts transmitted, reflecting greater efficacy of selection among bacteria within a given host ( Rispe and Moran 2000 ). Prospects In sum, the past few years have witnessed a surge of new empirical and theoretical approaches to understand the dynamics of bacterial–insect relationships. These tools have shed light on the roles of recombination, selection, and mutation on endosymbiont genome evolution and have highlighted parameters that shape the outcome of genetic conflicts between hosts and symbionts. These data provide a foundation for studying the evolution of mutualism and parasitism and modes of transitions between them. In the near future, we can look forward to full genome sequences that span a broader ecological and phylogenetic diversity of endosymbionts and provide a richer comparative framework to test existing models and develop new ones. Developments in endosymbiosis are important not only to questions in basic research, but may have important practical applications. Blood-feeding insects such as mosquitoes and tsetse flies are vectors for parasites that cause significant global infectious diseases such as malaria, dengue virus, and trypanosomiasis, many of which have frustrated attempts at vaccine development. The same insects that transmit these devastating human parasites often possess a diversity of mutualistic and parasitic bacterial endosymbionts. A very promising and urgent area of endosymbiont research is the manipulation of these bacteria to control vector populations in the field. Current studies already provide evidence that endosymbiont manipulation is a promising strategy to reduce the lifespan of the insect vector or limit its transmission of disease-causing parasites ( Aksoy et al. 2001 ; Brownstein et al. 2003 ). Each advance in our understanding of endosymbiont genomics and evolutionary dynamics represents one step closer to that goal. Box 1. Glossary Endosymbiont: A symbiont that lives inside of its host, often within host cells (intracellular symbiont). Facultative mutualist: A beneficial symbiont that associates with the host, but can also live apart from it. Examples include Rhizobium spp. that associate with legumes, but also have a free-living stage to their life cycle. Obligate mutualist: A beneficial symbiont that lives exclusively with its host and depends on the host for survival. Examples include many nutritional endosymbionts of insects, which cannot survive outside of the insect host cell. These associations are reciprocally obligate when the host cannot survive without the endosymbiont. Parasite: A symbiont that has a negative effect on host fitness, in contrast to a mutualist, which increases host fitness. Reproductive parasite: A symbiont that manipulates host reproduction to its own benefit, but at the expense of host fitness. Reproductive parasites typically bias offspring toward infected females. Symbiosis: An association between two more species.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC368163.xml
406393	A Neutral Model of Transcriptome Evolution	Microarray technologies allow the identification of large numbers of expression differences within and between species. Although environmental and physiological stimuli are clearly responsible for changes in the expression levels of many genes, it is not known whether the majority of changes of gene expression fixed during evolution between species and between various tissues within a species are caused by Darwinian selection or by stochastic processes. We find the following: (1) expression differences between species accumulate approximately linearly with time; (2) gene expression variation among individuals within a species correlates positively with expression divergence between species; (3) rates of expression divergence between species do not differ significantly between intact genes and expressed pseudogenes; (4) expression differences between brain regions within a species have accumulated approximately linearly with time since these regions emerged during evolution. These results suggest that the majority of expression differences observed between species are selectively neutral or nearly neutral and likely to be of little or no functional significance. Therefore, the identification of gene expression differences between species fixed by selection should be based on null hypotheses assuming functional neutrality. Furthermore, it may be possible to apply a molecular clock based on expression differences to infer the evolutionary history of tissues.	Introduction Advances in microarray technology have made the systematic study of expression levels of thousands of transcripts possible. This has been heralded as a major step forward in understanding the function of genomes, since transcript expression levels are expected to correlate with biological functions. Although this is clearly the case for many genes that change their expression in response to environmental stimuli (e.g., Spellman et al. 1998 ; Hughes et al. 2000 ; Miki et al. 2001 ), it is not known whether evolutionary changes in gene expression are determined primarily by Darwinian selection or by stochastic processes. Indeed, the extent to which natural selection has shaped the properties of organisms has been hotly debated ever since Charles Darwin proposed that organisms are adapted to their environment as a result of natural selection. At the molecular level, the view that most changes are due to Darwinian selection was challenged by Kimura's neutral theory of molecular evolution ( Kimura 1983 ). This theory states that the vast majority of differences seen in nucleotide and amino acid sequences within and between species have no or only minor selective effects. Consequently, their occurrence within a species and the fixation of differences between species are primarily the result of stochastic processes. Thus, it is believed today that the evolution of the overwhelming majority of synonymous nucleotide changes within protein-coding exons, as well as changes in noncoding parts of genomes, are determined by mutational processes and random genetic drift ( Li 1997 ). In fact, even at the level of morphology, it has been argued that many features are not adaptive, but instead result from physical constraints or historical accidents ( Gould and Lewontin 1979 ). However, since selection acts at the level of the phenotype while variation is generated at the level of the genotype, the proportion of changes caused by selection can be expected to be largest at the phenotypic level and smallest at the DNA sequence level. As a corollary, we may expect the proportion of selected changes to gradually decrease at the proteome and the transcriptome levels, since these are located progressively further from the phenotype. Consequently, a large proportion of transcriptome changes might be explained by historical accidents rather than by selective events. To test whether this may be the case, we have investigated whether a neutral model can describe transcriptome differences observed among primate and mouse species as well as among various brain regions within a species. Results/Discussion Transcriptome Evolution among Species If the majority of evolutionary changes are caused by historical accidents rather than by natural selection, they will accumulate mainly as a function of time rather than as a function of morphological or behavioral change of organisms. Applied to transcriptome evolution, a neutral model therefore implies that the rate of transcriptome change is proportional to time. In particular, if we assume that mutations cause changes in the relative amounts of transcripts independently of the absolute expression level of the gene, then the squared difference of the logarithm of the expression level is expected to increase linearly with divergence time ( Lande 1976 ; Felsenstein 2004 ). To investigate whether this is the case, we have studied differences in the gene expression levels of around 12,000 genes in the prefrontal cortex of six humans, three chimpanzees (Pan trogodytes), one orangutan (Pongo pygmaea), and one rhesus macaque (Macaca mulatta) using oligonucleotide microarrays. To exclude the influence of DNA sequence differences on the hybridization results, at least between humans and chimpanzees, only oligonucleotide probes that matched perfectly to the chimpanzee DNA sequences were used in the analysis ( see Materials and Methods ). In Figure 1 A, we plot species divergence times against the average squared difference between the logarithm of the expression levels of 1,998 genes that had expression levels large enough to be detected in all primate samples. Although comparisons involving orangutan and rhesus were complicated by nucleotide sequence differences to array probes, the result shows that the squared differences represent an approximately linear function of time over at least 20 million years. When we apply the same analysis to published gene expression data for the livers of three humans, three chimpanzees, and one orangutan ( Enard et al. 2002 ), we again observe a linear relationship between gene expression differences and species divergence times ( Figure 1 B). Figure 1 Brain and Liver Transcriptome Change among Primates as a Function of Time Average expression differences within and between primates in brains (A), in liver (B), and for genes in brain for genes with high (red) and low (blue) variation among six humans (C). Colors: red, comparisons between and with humans; blue, comparisons between and with chimpanzees; purple, comparisons between humans and chimpanzees; orange, comparisons between orangutan and rhesus macaque; black, comparisons between experimental duplicates. Vertical error bars for expression indicate 95% confidence intervals calculated by 10,000 bootstraps over genes. Divergence times are according to Glazko and Nei (2003 ). Since oligonucleotide-based microarrays are sensitive to DNA sequence differences and the orangutan and rhesus macaque genome sequences are not yet known—so that we cannot delete oligonucleotides carrying mismatches between the species—we used arrays containing around 28,000 cDNAs ranging in length from 500 to 1,500 nucleotides to assay gene expression patterns in the prefrontal cortex of six humans, five chimpanzees, five rhesus macaques, and five crab-eating macaques (Macaca fascicularis). Due to the greater probe length, these arrays are much less sensitive to DNA sequence differences and therefore can be used to compare gene expression in humans and macaques ( Ranz et al. 2003 ). When we plot the extent of gene expression divergence for 5,829 genes whose expression was detected in all samples against species divergence time, we again observe that expression differences accumulate approximately linearly with time ( Figure 2 ). Figure 2 Brain Transcriptome Change as Measured by cDNA Arrays Colors and symbols as in Figure 1 except orange, which indicates comparisons between chimpanzee and both macaque species, and blue, which indicates comparisons between rhesus macaque and crab-eating macaque. Divergence times are according to Hayasaka et al. (1996 ) and Glazko and Nei (2003 ). In a recent study of gene expression in the brains of humans, chimpanzees, and orangutans, we found that the rate of expression change on the human lineage has been larger than on the chimpanzee lineage ( Enard et al. 2002 ). This is in apparent contradiction to the linearity observed here. However, the analysis of Enard et al. (2002 ) was based on less than 5% of all genes expressed in the brain because it was confined to genes that differed significantly in expression between humans and chimpanzees. In contrast, here we perform a transcriptome-wide analysis of all genes with detectable expression in several primate species. However, the slightly higher divergence of humans than chimpanzees from the two macaque species may reflect the previously reported higher rate of gene expression divergence on the human evolutionary lineage ( Enard et al. 2002 ; Caceres et al. 2003 ; Gu and Gu 2003 ). However, additional experiments are necessary to exclude the possibility that this is caused by experimental artifacts. The clocklike accumulation of expression differences between species observed for primates is in agreement with the recent observation that differences in gene expression are consistent with phylogenetic relationships among Drosophila species ( Rifkin et al. 2003 ), and both these observations are compatible with the predictions of the neutral model. However, under certain selection scenarios, positively selected changes would also accumulate linearly with time ( Felsenstein 2004 ). Therefore, linear accumulation of expression differences alone does not rule out selection. In addition to the clocklike accumulation of evolutionary changes, the neutral theory states that the same forces determine the rate of evolution both within and between species ( Kimura 1983 ). Thus, a neutral prediction with respect to transcriptome evolution is that genes that vary more within species should be more likely to change between species as well. In order to test this, we ranked 2,926 genes with detectable expression levels in six humans and three chimpanzees according to their variation within humans and calculated the species divergences for the 25% of genes that had the largest and the smallest human variation, respectively. Figure 1 C shows that the genes with high variation among humans changed significantly faster between species than the genes with low variation. The magnitude of observed expression differences may be influenced by DNA sequence mismatches affecting hybridization between orangutan and rhesus samples and array probes. However, the difference in divergence rates between genes with high and low expression variation within species is unlikely to be explained by hybridization artifacts, since this would require a difference in sequence divergence between the two groups of genes. We further considered the correlation between the average diversity within humans and chimpanzees and the divergence between the species for the 2,926 genes. This correlation is highly significant ( p < 0.001) as gauged by a permutation test ( see Materials and Methods ). Since all array probes that carried sequence differences between humans and chimpanzees were removed prior to analysis, this correlation is not affected by hybridization artifacts. The strength of the correlation (τ = 0.24) is of a similar magnitude as the one obtained for the correlation of diversity and divergence of random genomic DNA sequences in humans and chimpanzees (τ = 0.179, p = 0.028, n = 76), the vast majority of which are noncoding ( Hellmann et al. 2003 ). Thus, although the two measures are not directly comparable, the degree of correlation between intraspecific diversity and interspecific divergence is similar for brain transcriptomes and random genomic DNA sequences in humans and chimpanzees. To investigate whether gene expression differences accumulate as a function of time also in another group of mammals, we analyzed three mouse species. An advantage in this case is that post mortem artifacts are less likely to influence the results than in the case of autopsy material of humans and great apes. We determined differences in gene expression levels for around 9,000 genes in the frontal cortex of six outbred Mus musculus, three outbred M. spretus, and one M. caroli. As shown in Figure 3 A, the squared transcriptome differences accumulated linearly with time among the mouse species. To test if divergence rates differ for the genes with high and low variation within species, we investigated the 25% of the 2,742 genes detected in all samples with the highest and the lowest variation within M. musculus, respectively, as was done in the primates. Figure 3 B shows that genes that vary more within M. musculus diverged faster among mouse species than genes that vary less. As in the case of primate species, imperfect matches of M. spretus and M. caroli mRNAs to the array oligonucleotides may partly influence the observed expression differences between species. Nonetheless, as for primates, the difference in divergence rates between genes with high and low expression variation within species is unlikely to be explained by hybridization differences since there is no indication that genes that vary more in expression within species diverge faster between species with respect to their DNA sequence. The correlation between diversity and divergence for M. musculus and M. spretus for genes detected in both species is highly significant (τ = 0.29, p < 0.001, n = 3,139), although in this case we cannot correct for DNA sequence differences. A correlation between gene expression differences within and between species was recently demonstrated also in teleost fish ( Oleksiak et al. 2002 ). Thus, in agreement with the neutral model, genes that vary more within species tend to vary more between species in three vertebrate groups. Figure 3 Brain Transcriptome Change among Mice as a Function of Time Average expression differences within and between the mouse species (A) and for genes with high (red) and low (blue) variation among M. musculus individuals (B). Colors: red, comparisons between and with M. musculus; blue, between and with M. spretus; purple, between M. musculus and M. spretus . Vertical error bars for expression indicate 95% confidence intervals calculated by 10,000 bootstraps over genes. Divergence times are according to She et al. (1990 ). A Test for Neutrality One way to test whether gene expression differences between species accumulate at a rate consistent with neutral expectation is to compare them to the expression differences observed for a class of genes that can reasonably be expected to not be the direct targets of positive or negative selection. Since expressed pseudogenes do not produce any functional gene products, they can be viewed as such a class of genes. Thus, if a substantial proportion of intact genes accumulate expression differences faster than pseudogenes, this would indicate that they are positively selected. Such an observation would falsify a neutral model. To test this, we considered the expression patterns in four regions of the brain in three humans and three chimpanzees using the Affymetrix U95 array set interrogating approximately 40,000 genes (Philipp Khaitovich, unpublished data). In order to identify all probe sets on these arrays that interrogate expressed pseudogenes, we aligned the probe sequences, as well as published lists of human pseudogenes, to the human genome ( see Materials and Methods ). In total, 889 probe sets that overlap with pseudogenes were identified. Thirty-three of these were detected (detection p -value < 0.05) in at least one of four brain regions in either the chimpanzees or the humans after masking all probes carrying DNA sequence differences between the species. Of these, 28 contained at least one mutation that leads to a loss of function in both humans and chimpanzees. We therefore assumed that these pseudogenes were nonfunctional in the common ancestor of humans and chimpanzees. Finally, we checked whether these probe sets may crosshybridize with any intact genes by aligning them to the human genome. This left us with 23 expressed pseudogenes. We compared the distributions of the squared differences between the mean expression levels of each gene in humans and in chimpanzees for the 23 pseudogenes and 12,647 intact genes for each of the four brain regions. In each case, only the genes detected in a given brain region were used for the calculation. In all four brain regions the distribution of expression distances among intact genes did not differ significantly from that among pseudogenes in either a Kolmogorov-Smirnov test or a Wilcoxon rank sum test. These tests would have been significant if more than 5% (1/23) of the genes had a distribution radically different from that of the pseudogenes. When the data for four brain regions were combined, no visual difference between the two distributions was apparent ( p = 0.16 and p = 0.69, respectively) ( Figure 4 A). Figure 4 Comparison between Intact Genes and Pseudogenes (A) shows the distributions of expression divergence between humans and chimpanzees for intact genes and pseudogenes. (B) shows the distributions of the ratio of expression divergence between humans and chimpanzees and expression diversity within humans for intact genes and pseudogenes. Thus, we failed to detect any significant excess of intact genes that diverged faster in expression than pseudogenes. This indicates that the fraction of gene expression differences between the species that are fixed by positive selection is small. Interestingly, there was also no detectable excess of intact genes that diverged slower than pseudogenes. This may seem unexpected, since the expression of many intact genes might be thought to be stabilized by negative selection and therefore to change more slowly than pseudogenes. This may indicate that purifying selection as well is a weak force affecting gene expression. However, it should be noted that the small number of expressed pseudogenes analyzed limits the power to detect positive and negative selection. A targeted effort to study expressed pseudogenes in closely related species would be a worthwhile undertaking. A Test for Positive Selection The fact that the overall accumulation of expression differences conforms to a selectively neutral model does not mean, of course, that all expression differences between species are selectively neutral. As for nucleotide changes, some changes in gene expression will have had phenotypic consequences and some of these will have become fixed due to positive selection. To identify such gene expression differences, we propose to use the ratio of divergence between species to diversity within species, akin to the tests suggested for quantitative genetic traits ( Charlesworth 1984 ; Lynch and Hill 1986 ; Turelli et al. 1988 ) and in agreement with recent suggestions by Rifkin et al. (2003) or Hsieh et al. (2003 ). However, to do this it is necessary for each gene considered to distinguish the gene expression diversity caused by genetic differences between individuals from the diversity caused by environmental factors. This is crucial since the environmental component is likely to be much larger than the genetic component. For example, under strict neutrality and no environmental influence, we expect a divergence to diversity ratio that is equal to the ratio of time of divergence of the species to the average time to the common ancestors of the individuals sampled within a species. This would be about 1:10 for humans and chimpanzees ( Chen and Li 2001 ; Lander et al. 2001 ). However, the observed ratio is approximately 1:3, suggesting that the environmental component is on the order of three times bigger than the genetic component. Studies of gene expression differences among individuals with different genetic relatedness will eventually allow an estimation of the genetic component of expression variation. Since we are unable to tease apart genetic and environmental contributions to expression diversity, we instead used pseudogenes to estimate the distribution of divergence to diversity ratios observed in the absence of selection and compared these ratios to intact genes. No significant difference was found (Kolmogorov-Smirnov test, p = 0.388; Wilcoxon rank sum test, p = 0.134), and both distributions appeared to center around roughly the same values ( Figure 4 B). Note that this observation has to be taken cautiously since it is based on a small number of pseudogenes and the gene expression diversity is calculated from only three human individuals. Nevertheless, this result indicates that there is no drastic difference between the expression patterns of intact genes and expressed pseudogenes, since our tests would have been significant if 5% or more of the genes had had a radically different divergence to diversity ratio than that observed among the pseudogenes. Transcriptome Evolution among Brain Regions Different anatomical brain structures appeared at different times during vertebrate evolution. These time points can be viewed as divergence times between brain regions extending millions of years back in the past ( Figure 5 A). If gene expression changes between different brain regions have a large random component, gene expression differences between brain regions within species could potentially be used as a molecular clock to time the divergences of tissues. To investigate whether this may be the case, we compared expression patterns for Brodmann's area 44, the prefrontal cortex, the anterior cingulate cortex, the primary visual cortex, the caudate nucleus, and the cerebellum in three adult human and three adult chimpanzee males (Philipp Khaitovich, unpublished data). All comparisons were performed between brain regions within the same individual. This has two advantages. First, such comparisons are unaffected by nucleotide sequence variation between and within species. Second, environmental differences and post mortem changes have little effect when expression differences within one individual are studied. In Figure 5 B, we plot the average squared distances between the six brain regions in humans and chimpanzees against the time when these brain regions emerged during vertebrate evolution ( Butler and Hodos 1996 ; Nieuwenhuys et al. 1998 ) for 2,297 and 2,525 genes detected in all human and all chimpanzee samples, respectively. It can be seen that the expression differences increase approximately linearly with time over more than half a billion years. To investigate if this finding holds also in another mammalian species, we used published expression data for 1,346 genes with detectable expression in eight brain regions in the mouse ( Su et al. 2002 ). In this case as well there is an approximately linear relationship between transcriptome differences and evolutionary divergence times ( Figure 5 C). Figure 5 Transcriptome Change among Brain Regions as a Function of Evolutionary Time (A) Schematic evolutionary tree for six human brain regions: B.44, Brodmann's area 44; PFC, prefrontal cortex; ACC, anterior cingulate cortex; PVC, primary visual cortex; CN, caudate nucleus; and CB, cerebellum. Numbers indicate approximate divergence time in millions of years ( Butler and Hodos 1996 ; Nieuwenhuys et al. 1998 ). (B) Average expression differences among brain regions in humans (red) and in chimpanzees (blue). (C) Average expression differences among brain regions in M. musculus. Error bars for expression indicate 95% confidence intervals calculated from 10,000 bootstrap replications over genes. If gene expression differences between the brain regions were largely adaptive, one would expect them to correlate with tissue function and not with evolutionary divergence time. Our data show that tissues that diverged recently have very similar gene expression profiles irrespective of the differences in function. For instance, the transcriptome of Brodmann's area 44 in the left hemisphere (Broca's area) is very similar to that of the prefrontal cortex in both humans and chimpanzees, although it is known to be involved in speech processing in humans while it must have another function in chimpanzees ( Kandel et al. 2000 ). This is what we would expect if the time since divergence rather than the extent of functional differences determined the magnitude of transcriptome change. Thus, although a number of expression differences between brain regions surely correspond to functional differences, our findings suggest that a sizeable proportion of the differences are functionally neutral. A noteworthy finding is that the accumulation of expression differences between brain regions within a species is much slower than the accumulation of expression differences within a brain region between species. In fact, the expression differences that have accumulated among the primate species over 20 million years (see Figure 1 A) are approximately as extensive as those that have accumulated among brain regions over 500 million years (see Figure 5 B). This is likely to result from the fact that all expression differences seen between brain regions within an individual are caused by changes in regulatory networks established during development by cells that carry the same genome. In addition, expression differences between brain regions reflect the different cell-type compositions of these regions. In contrast, transcriptome differences between species are the result of changes in regulatory networks and cellular composition of tissues, as well as nucleotide sequence differences between species that affect promoters and other genomic elements that determine transcript levels. Our results show that the latter type of changes are much more common than the former. A possible alternative explanation for the correlation between differences in gene expression and evolutionary divergence time among brain regions could be that differences in gene expression do not correlate with evolutionary divergence time, but instead with divergence time during fetal development. Our observations would then result from the fact that both developmental divergence times and expression differences correlate with evolutionary divergence. A correlation between developmental and evolutionary divergence times has been hypothesized before (for a review, see Gould 1977 ). In fact, gene expression analyses now provide a quantitative approach to address this question and may also provide a tool to date the evolutionary emergence of brain regions that cannot be discerned in the fossil record. Conclusions We show that a neutral model of evolution can predict the main features of transcriptome evolution in the brains of primates and mice. A neutral model is also in agreement with published observations in Drosophila ( Rifkin et al. 2003 ) and fish ( Oleksiak et al. 2002 ). Although selective scenarios that explain some or even most of these observations can be found, the combined evidence presented leads us to conclude that a neutral model is the most adequate null model for transcriptome evolution. This suggests that the majority of gene expression differences within and between species are not functional adaptations, but selectively neutral or nearly neutral. The main challenge now is to develop a mathematical model of transcriptome evolution that allows quantitative predictions of transcriptome changes. Such a model, combined with experimental data estimating the normal variation of gene expression within a species and the relative contributions of genetic and environmental factors to this variation, should allow adaptive gene expression changes to be identified. Further work is also needed to reveal whether proteome evolution is also dominated by changes that are largely selectively neutral. Finally, the finding that gene expression differences can be used as a molecular clock to date tissue divergences opens the prospect of reconstructing the evolutionary history of organs and tissues based on gene expression measurements in a single species. Materials and Methods Tissue samples and microarray data collection For the primate samples, approximately 200 mg of gray matter was collected from post mortem brain samples from prefrontal cortex region corresponding to Brodmann's area 9 in the left hemisphere from six male humans who were 45, 45, 63, 65, 70, and 70 years old; five male chimpanzees that were 7, 12, 12, 12, and approximately 40 years old; one 16-year-old male orangutan; five approximately 10-year-old male rhesus macaques; and five approximately 15-year-old male crab-eating macaques. All individuals had no history of brain-related diseases and suffered sudden deaths without associated brain damage. For the mouse samples, approximately 50 mg of gray matter was collected from the frontal cortex regions of six M. musculus (three of which are previously described in Enard et al. 2002 ), three M. spretus, and one M. caroli individuals. All mice were outbred, older than 14 weeks, and healthy. Total RNA was isolated using the TRIzol reagent (GIBCO, San Diego, California, United States) according to manufacturer's instructions and purified with Quiagen RNeasy kit (Quiagen, Valencia, California, United States) following the “RNA cleanup” protocol. RNAs were of high and comparable quality as gauged by the ratio of 28S to 18S ribosomal RNAs visualized on agarose gels and by the signal ratios between the probes for the 3′ and 5′ ends of the mRNAs of GAPDH and β-actin genes used as quality controls on Affymetrix microarrays (Affymetrix, Santa Clara, California, United States). For Affymetrix microarrays, labeling of 5 μg of the RNA, hybridization, staining, washing steps, and array scanning were carried out following Affymetrix protocols. Expression data were collected using Affymetrix HG U95Av2 arrays for the primate samples and Affymetrix MG U74Av2 arrays for the mice samples. The Affymetrix CEL files containing expression data for the different regions of the mouse brain, including amygdala, cerebral cortex, hippocampus, hypothalamus, cerebellum, olfactory bulb, and two regions of spinal cord were provided by John Hogenesch. Arrays containing 51,000 cDNAs corresponding to approximately 40,000 UniGene clusters were manufactured in the laboratory of W.A. as described elsewhere (Anonymous 2003). Labeling, hybridization, staining, washing, and array scanning were carried out as described by Cortes-Canteli et al. (2004 ) with slight modifications. All samples were hybridized twice with dye reversal, using a mixture of all samples as a common reference. All primary expression data were submitted to the Array Express database ( http://www.ebi.ac.uk/arrayexpress/ ). Masking of sequence differences between humans and chimpanzees In order to exclude all oligonucleotide probes that did not match perfectly between humans and chimpanzees, we aligned all Affymetrix target sequences ( http://www.affymetrix.com/analysis/index.affx ) first to the human genome (build 33) and then to a draft version of the chimpanzee genome (the assembly was given courtesy of David Jaffe in June 2003). Using BLAT ( Kent 2002 ), we matched chimpanzee sequences with Affymetrix target sequences containing the 16 oligonucleotide probes and determined the best hit using a scoring function. The chimpanzee sequence was then aligned to the human genome to determine whether the best match coincided with the match obtained from alignment of Affymetrix target sequences with the human genome. To identify insertion and deletions (indels), we compared the alignment of the Affymetrix target sequence to the human genome and to the chimpanzee genome, and differences in the indel structure relative to the target sequence were identified as indels. We then identified all oligonucleotide probes within target sequences that matched the chimpanzee sequence perfectly. These probes were used for the analysis while the rest of the probes were masked. Microarray data analysis Affymetrix microarray image data were analyzed with Affymetrix Microarray Suite v5.0 using default parameters. Arrays were scaled to the same average intensity using all probes on the array. Detected genes were defined as those with a detection p -value less than or equal to 0.05. For calculation of the expression values, data were processed with the Bioconductor “affy” software package ( Ihaka and Gentleman 1996 ) using the quantile normalization procedure ( Bolstad et al. 2003 ). cDNA arrays were analyzed using the TM4 software package ( Saeed et al. 2003 ). Detected genes were defined as those with a spot intensity exceeding the background intensity by more than 2-fold. All slides were normalized to the common reference using the LOWESS normalization algorithm. For calculation of diversity and divergence, signal to reference ratio measurements were transformed into standardized signal intensities by multiplying them by the average reference intensity for each gene. Divergence was defined as the squared difference between the mean expression of two groups of samples averaged over (all detected) genes. Diversity was defined as the expression variance within a group of samples. Correlation significance test We measured the divergence between human and chimpanzee by looking at the squared difference between the mean expression values in humans and chimpanzees. This estimate of divergence includes the errors in our estimates of the two means, which is proportional to the variance in each of the species, and thus to the diversity in each species. Therefore, even if no correlation between divergence and diversity existed, our measured divergence and diversity estimates would correlate, and the smaller the divergence is relative to diversity, the stronger the correlation would be. To estimate if the observed correlation is larger than that expected from this effect alone, we performed a randomization test, in which we computed how much correlation between diversity and divergence would be generated from the above effect even if no correlation between diversity and divergence exists. To be conservative, we first generated a distribution that deliberately underestimated the real divergence between humans and chimpanzees. This was done by first generating a distribution of the expected observed differences (X) in gene expression between humans and chimpanzees if the real divergence is zero. Then using this distribution and the observed distribution of differences (Z), we generated a distribution (Y) that-added to values from X-would give Z. In order to underestimate the divergence, we generated Y assuming that the correlation of X and Y is one. We then generated random samples in the following way: For each gene (g), we chose a random difference of expression (d) from our generated distribution. We then drew six samples from a normal distribution whose mean is zero and whose variance is the diversity in humans for gene g, and three samples from a normal distribution whose mean is d and whose variance is the diversity of chimps for gene g. For these expression values we then calculated the correlation between diversity and divergence. We repeated the whole procedure 1,000 times. None of these randomizations generated a correlation that is as strong as the observed one. To make sure that the whole test is conservative, we generated 100 datasets of three types, all of which had a similar diversity, but had a “real” divergence distribution of (1) zero, (2) the underestimated divergence, or (3) the measured divergence, and had uncorrelated diversity and divergence. We then performed the whole test described above, doing just one randomization test. If the test was not conservative, one would expect the correlation in the dataset to be higher than the correlation after randomization in 50% of the cases. Instead, the correlation after randomization was higher in 98, 98, and 99 cases respectively-showing that our test is indeed conservative. Expressed pseudogenes We retrieved sequences of all pseudogenes as determined by Torrents et al. (2003 ), Zhang et al. (2003 ), and the VEGA project ( http://vega.sanger.ac.uk ). These sequences, as well as the Affymetrix target sequences, were mapped to the human genome (build 34) using BLAT ( Kent 2002 ), and the best hit was determined using the following parameters: match, +1; mismatch, −3; gap-opening penalty only for gaps ≤ 20, −5; and gap extension, −1. Next, using BLAT, we determined the Affymetrix target sequences where the best-matching sequence did not overlap with the genomic region of a known gene ( http://genome.ucsc.edu ). Thus, we identified 889 probe sets that overlapped with a pseudogene, but not with a known gene. Combined with gene expression data collected in four brain regions (anterior cingulate cortex, Broca'a area, caudate nucleus, cerebellum; Philipp Khaitovich, unpublished data) in three humans and three chimpanzees, 33 of these probe sets had detectable expression levels in at least one brain region in either three chimpanzees or three humans. For these probe sets, we checked whether at least one of the identified interruptions of the human pseudogene was also present in the chimpanzee, indicating that the pseudogene was already nonfunctional at the time of the chimpanzee–human divergence. This left us with 28 probe sets that were checked for crosshybridization with other genes by aligning oligonucleotide probes from these probe sets to the human genome. Finally, we were left with 23 expressed pseudogenes that did not match perfectly to any other gene by more than seven out of 16 probes in the probe set.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC406393.xml
545057	Catheter-related bacteremia due to Kocuria rosea in a patient undergoing peripheral blood stem cell transplantation	Background Micrococcus species may cause intracranial abscesses, meningitis, pneumonia, and septic arthritis in immunosuppressed or immunocompetent hosts. In addition, strains identified as Micrococcus spp . have been reported recently in infections associated with indwelling intravenous lines, continuous ambulatory peritoneal dialysis fluids, ventricular shunts and prosthetic valves. Case presentation We report on the first case of a catheter-related bacteremia caused by Kocuria rosea , a gram-positive microorganism belonging to the family Micrococcaceae , in a 39-year-old man undergoing peripheral blood stem cell transplantation due to relapsed Hodgkin disease. This uncommon pathogen may cause opportunistic infections in immunocompromised patients. Conclusions This report presents a case of Kocuria rosea catheter related bacteremia after stem cell transplantation successfully treated with vancomycin and by catheter removal.	Background Long-term indwelling central venous access devices are indispensable to the supportive care of cancer patients. Infection remains a major complication of Hickman type catheters [ 1 ]. Intravascular catheter-related infections can be the cause of morbidity and mortality in patients undergoing autologous peripheral blood stem cell transplantation (PBSCT). Staphylococcus aureus , coagulase-negative staphylococci, aerobic gram-negative bacilli and Candida albicans are the most frequent causes of catheter related bloodstream infection [ 2 ]. Kocuria rosea , is gram-positive, strictly aerobic microorganisms belonging to the family Micrococcaceae that usually grow on simple media and generally considered as non-pathogenic commensals that colonize the oropharynx, skin and mucosa. However, it can be opportunistic pathogen in the immunocompromised patient [ 3 , 4 ]. We describe the first case of central venous catheter (CVC) related infection associated with Kocuria rosea in a patient undergoing autologous PBSCT. He had a chills, high fever, tachycardia and increased respiratory rate. Persistent bacteremia, which was unresponsive to vancomycin, was successfully treated by catheter removal. Case presentation A 39-year-old man with relapsed Hodgkin disease was admitted to Erciyes University Hospital for autologous PBSCT. A Hickman-type central venous catheter was implanted and the patient received a conditioning regimen consisting ifosfamide, carboplatin and etoposide. On day 3 after PBSCT, he was acutely ill with a chills, high fever (temperature 38.5°C), tachycardia (heart rate 108 beats/min) and increased respiratory rate (26 breaths/min). Clinical, radiological and microbiological diagnostic procedures were performed, and at three blood cultures were obtained from peripheral veins and catheter half an hour apart. The patient was empirically treated with intravenous imipenem/cilastatin, 2 g/day; and amikacin, 1 g/day. On day six after PBSCT, he had a persistent fever and the temperature spiked to 39.5°C. There was no sign of a catheter exit-site infection. Chest radiography was normal. Serial blood cultures taken from the catheter (two cultures) and a peripheral vein (four cultures) were positive for Micrococcus spp ., which were identified as Kocuria rosea . An antibiogram showed that the pathogen was sensitive to ampicillin/sulbactam, erythromycin and vancomycin. Imipenem and amikacin were discontinued, and intravenous vancomycin, 2 g/day was begun. The patient did not respond satisfactorily to vancomycin therapy for five consecutive days. The catheter was removed and a catheter tip culture was obtained. The cultures demonstrated a massive growth of Kocuria rosea . Blood cultures were thereafter negative. The patient was discharged from the hospital after 14 days of intravenous vancomycin. He was followed for six months; the patient is alive and well and is in complete remission of the primary disease. Microbiological diagnosis Cultures of blood from the peripheral veins and the CVC were performed with a BACTEC system (BACTEC 9210; Becton Dickinson). The CVC tip was cultured by using the Maki roll technique. A growth of more than 15 cfu on the blood agar plate was considered positive. It was identified as Kocuria rosea on the basis of Bergey's Manual of Systematic Bacteriology [ 5 ]. This identification was confirmed with the commercially available tests (ID32 Staph ATB system, and mini-API, Biomerieux). Discussion Micrococcus spp . are usually regarded as non-pathogenic skin commensals and their clinical relevance is questionable. However, Micrococcus spp . can be opportunistic pathogens in immunocompromised patients such as PBSCT. It has been reported that Micrococcus luteus has been implicated as the causative agent in cases of intracranial abscesses, meningitis, pneumonia and septic arthritis in immunosuppressed or immunocompetent hosts [ 6 - 12 ]. In addition, strains identified as Micrococcus spp . have been reported recently in infections associated with indwelling intravenous lines, continuous ambulatory peritoneal dialysis fluids, ventricular shunts and prosthetic valves [ 13 - 18 ]. We performed a Medline search using the terms "Micrococcus" and/or " Kocuria rosea " and/or "Kocuria". However, we were unable to find published studies on catheter infections caused by Kocuria rosea in patients undergoing HSCT in febrile neutropenic period. In this patient, the repeated isolation of Kocuria rosea from different blood cultures in the absence of other microorganisms, together with isolation of the same organism from the catheter were primarily responsible for the episodes of febrile neutropenia. The catheter was then removed and cultures demonstrated a massive growth of Kocuria rosea . The clinician should not underestimate the importance of the repeated isolation of a Kocuria rosea from blood cultures. Although uncommon, the possibility that a central venous line may be the portal of entry should be evaluated. Adherence of bacteria to the silastic tube would possibly explain the failure of treatment by antibiotics alone. Conclusions This report emphasizes that Kocuria rosea should be considered as a nosocomial pathogen, which may cause of catheter infections in febrile neutropenic patients. In these patients, persistent Kocuria rosea bacteremia unresponsive to medical management should be treated by catheter removal. Competing interests The author(s) declare that they have no competing interests. Authors' contributions FA, OY and KG carried out the clinical study of the patient. BS carried out the culture and specific identification of the bacterium. BE and MC drafted the manuscript. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545057.xml
538269	Characterisation of weak magnetic field effects in an aqueous glutamic acid solution by nonlinear dielectric spectroscopy and voltammetry	Background Previous reports indicate altered metabolism and enzyme kinetics for various organisms, as well as changes of neuronal functions and behaviour of higher animals, when they were exposed to specific combinations of weak static and alternating low frequency electromagnetic fields. Field strengths and frequencies, as well as properties of involved ions were related by a linear equation, known as the formula of ion cyclotron resonance (ICR, abbreviation mentioned first by Liboff). Under certain conditions already a aqueous solution of the amino acid and neurotransmitter glutamate shows this effect. Methods An aqueous solution of glutamate was exposed to a combination of a static magnetic field of 40 μT and a sinusoidal electromagnetic magnetic field (EMF) with variable frequency (2–7 Hz) and an amplitude of 50 nT. The electric conductivity and dielectric properties of the solution were investigated by voltammetric techniques in combination with non linear dielectric spectroscopy (NLDS), which allow the examination of the dielectric properties of macromolecules and molecular aggregates in water. The experiments target to elucidate the biological relevance of the observed EMF effect on molecular level. Results An ion cyclotron resonance (ICR) effect of glutamate previously reported by the Fesenko laboratory 1998 could be confirmed. Frequency resolution of the sample currents was possible by NLDS techniques. The spectrum peaks when the conditions for ion cyclotron resonance (ICR) of glutamate are matched. Furthermore, the NLDS spectra are different under ICR- and non-ICR conditions: NLDS measurements with rising control voltages from 100–1100 mV show different courses of the intensities of the low order harmonics, which could possibly indicate "intensity windows". Furthermore, the observed magnetic field effects are pH dependent with a narrow optimum around pH 2.85. Conclusions Data will be discussed in the context with recent published models for the interaction of weak EMF with biological matter including ICR. A medical and health relevant aspect of such sensitive effects might be given insofar, because electromagnetic conditions for it occur at many occasions in our electromagnetic all day environment, concerning ion involvement of different biochemical pathways.	Background Weak magnetic fields and extremely low frequency electromagnetic fields (EMF) are omnipresent in natural environmental and increasingly man-made factors. A possible influence on life processes was already mentioned in the late 19 th century [ 1 ]. It is now recognized, that many organisms are capable of perceiving such fields, while less is known on the elementary perception. Three types of mechanisms are considered therefore, the orientation of ferromagnetic particles in tissues [ 2 ], singlet-triplet mixing states of macromolecules building radical pairs [ 3 ], and the ICR, whose persistent investigation began with the works of Liboff [ 4 ]. Ferromagnetism has been implicated in animal navigation (e.g. compass mechanism of migratory birds [ 5 ], and the magnetotaxis of certain bacteria [ 6 ]. The radical pair mechanism is independent of ferromagnetism and has putatively a higher magnetic sensitivity. It has been primarily studied in photosynthetic reaction centers and the respiratory chain [ 7 ], where triplet yields are modulated by electromagnetic interaction with fields as low as about 50 μT [ 8 ]. Already two decades ago effects were described by Blackman et al. [ 9 ], and later by [ 10 - 12 ], which require a combination of static and alternating magnetic fields. It turned out, that the magnetic field strength B of the static component and the frequency f of the alternating EMF relate to the "ion cyclotron resonance (ICR) formula": whereas m is the mass and q the charge of ions involved. The explanation of the mechanism of this effect in an aqueous, more or less viscous environment seems to be difficult, nevertheless there are some efforts. Liboff [ 13 ] suggested that magnetic fields can interact in a resonant manner with endogenous AC electric fields in biological systems, instead of a direct interaction with external AC magnetic fields. Binhi [ 14 ] reviewed the mechanisms of magnetobiological effects, and tried to estimate the sensitivities and involved molecular topologies. Adair [ 15 ] questioned a model involving altered transition rates of excited ions by weak EMF, while others [ 16 ] consider the ionic environment, eg. properties of the water, with Ca 2+ as the most investigated ion. An altered Ca 2+ -transport was found in human lymphocytes [ 4 ]. The motility of benthic diatoms is effected, if ICR conditions are matched for Ca 2+ and K + in the range of 8–64 Hz, and static field strengths comparable to geomagnetic fields [ 17 ]. The germination rate of Raphanus sativus was altered, when the ICR conditions for Ca 2+ , K + and Mg 2+ were applied to the seedlings [ 18 ]. ELF effects on macromolecules indicate an ICR effect possibly caused by additionally involved alternating electric fields [ 19 ]. It is noteworthy remarkable that ICR conditions can be matched by combinations of the local geomagnetic field and man-made electromagnetic fields, especially the frequency range of power lines (50 or 60 Hz). Liboff et al. [ 20 ] suggest to consider ICR effects for the evaluation of epidemiological childhood leukaemia studies. The assessment of elevated brain cancer risk has been evaluated by Aldrich et al. [ 21 ] on the assumption of interactions of the geomagnetic field and a 60 Hz field component from power lines. NLDS was developed during the past decade in order to investigate dielectric properties of small particles in aqueous solutions, using relatively simple electrochemical equipment. In the simplest case, a sinusoidal alternating electric field is applied to the solution by 2 electrodes, using peak to peak voltages up to 1.5 V and frequencies of 1 to 1000 Hz. Particles with a dielectric constant different from that of their environment (generally water) distort the field. This induces alternating voltages over and currents through the solution, which are detected by 2 auxiliary electrodes in order to avoid polarisation effects. Phase shifts and distortions of the obtained signals, as compared to the input signal, contain information on damping and relaxation kinetics. Therefore, the signals are Fourier -transformed and evaluated as power spectra in the frequency domain [ 22 - 24 ]. Usually, the sample is compared to a reference, which lacks the solute, but otherwise is identical. Sample and reference can either be measured one by one in a single chamber device, or simultaneous with a "dual-chamber" setup, which also needs a two channel data acquisition, and allows a real-time differential-NLDS (DNLDS). The data are usually calculated using the decibel (dB) scale for the intensity (power) P n : Where U(n) sample is the signal output intensity of the n th harmonic from the sample measuring channel, and U(n) ref the corresponding value from the reference channel. Zhadin et al. [ 25 ] reported the alteration of electric properties of an electrolyte under ICR conditions. They found an increasing ion current through an aqueous glutamic acid (Glu) solution in narrow frequency bands (resonance), which could be described by equation (1). These results are the starting point for the present work, which is aimed to further elucidate this conduction mechanism. The influence of the concentration of Glu has been investigated, and the time resolved electric current through the solution is analyzed using "non linear dielectric spectroscopy" (NLDS), which indicate microcolloidial properties of the solvent-solute system. The NLDS was amplified by two features: The option of simultaneous data acquisition in two cuvettes (DNLDS), and the frequency resolved voltammetry (FRV), whereby simultaneous a AC voltammetry is performed [ 26 ]. By recording NLDS spectra at varying electrode voltages from e.g. 100–1100 mV, additional information was obtained on redox potentials. The electrode current never increases proportionally with the applied voltage but remains constant in the range of the counter voltage to an existing redox potential given by the investigated electrode-electrolyte system. This was used to improve the method by recording differential spectra (DNLDS). The integral over the spectrum represents one data point of a simple (not frequency resolved) AC voltammetry, while the intensity course of corresponding spectral data points provide information about the dielectric state of the redox reaction, e.g. its capacitive, time-dependent properties. Methods Preparations All preparations were performed with doubly de-ionized water. The solutions were degassed and stored under Argon, in order to avoid oxidation of the solute and increased electrode fouling during the subsequent measurements. An acidic solution of 2.24 mM Glu was adjusted to pH = 2.85 ± 0.03 with a stock solution of 5 mM HCl. Equilibration was assumed, when the pH varied less than ± 0.03 for at least one minute. All procedures were performed at 20°C. For yielding a reference signal, an aqueous solution of HCl was provided by diluting the HCl stock solution with water to pH = 2.85. All solutions were stored at 4°C under Argon. Apparatus The experimental arrangement for differential non-linear dielectric spectroscopy (DNLDS) is shown in Figure 1 . It allows the simultaneous evaluation of a sample and a reference under same conditions. A double cuvette (K) is built up by two standard photometric plastic cuvettes (1 × 1 × 4.3 cm). Both contain electrode arrays (E 1 , E 2 ) consisting each of 4 gold wires (Au 99.9%, Johnson Matthey, Karlsruhe) with a diameter of 0.25 mm, mounted parallel at a distance of 2 mm on a teflon frame. The required sample volume was 1 ml. These electrode carriers are mounted on a stable socket for electric connection and mechanical adjustment (not shown). The cuvettes are enclosed by a hermetically sealable plastic tank (T) with a copper bottom, which is filled at a height of 2 cm around the cuvettes with water for thermal coupling to an outer temperature controlled water bath. The setup is kept under Ar atmosphere throughout the experiment. Thermic control (20 ± 0.1 °C) of the cuvettes is provided by a water thermostat (Haake "G", Karlsruhe-Berlin, Germany) with a sequential home built temperature fine controller, ensuring highly stable working conditions for the electrodes. Once assembled, these components form a mechanically stable unit, with in- and outlets for gas and samples by small teflon hoses (not shown). The assembly is placed in the center of a solenoid (S), consisting of two cylindrical coils with a inner diameter of 16 cm and a height of 7 cm for applying the vertically orientated EMF (B). The coil for the static field component consisted of 300 turns of coated copper wire (diameter 0.5 mm), the other coil was winded above and had 50 turns. Figure 1 Experimental facility. Schematic sketch of the arrangement for the differential NLDS (DNLDS) experiments (left, components not drawn to scale) and photograph of the opened permalloy shielding box with the assembled sample carrier (right): Two arrays of 4 gold electrodes ( E 1 , E 2 , length 10 mm, distance 2 mm) each are located in two adjacent perspex cuvettes ( C ) of 1 × 1 × 4.3 cm, enabling simultaneous acquisition of two liquid samples (used volume 1 ml each) under the same environmental conditions. The cuvettes are enclosed by a tank ( T ) for providing an Argon protection gas atmosphere. This all is mounted on a socket housing water temperature control and magnetic field monitoring, and is centered inside a cylindrical solenoid ( S ) consisting of 2 coils with a inner diameter of 16 cm and a height of 7 cm for independent generating the static and the alternating magnetic fields of vertical direction ( B ). The input signal to the sample is applied by the electrodes labeled "in", the probe signals are taken by the electrodes labeled "out" and connected to preamplifiers with symmetric inputs. The complete arrangement is enclosed by a shielding box of 1 mm Permalloy, which is bonded inside with perspex. For electric and magnetic shielding the complete setup resides in a grounded double-walled Permalloy box with a total wall thickness of 1 mm. A overall inhomogeneity ≤ 0.3 % of the generated fields was determined inside the box with a triaxial CXM539 magnetometer (CMT GmbH, Herrsching, Germany) over the cuvette locations. For coil calibration the relation of field strength to coil current could be ascertained directly in measurement series with the magnetometer for 0.1–100 μT, showing a overall deviation from linearity ≤ 0.2 % (DC and AC), so currents corresponding to even lower field strengths were obtained by extrapolation. Signal processing was mostly done as previously described [ 27 ]. Figure 2 shows the schematic circuit diagram of the special NLDS measurement setup used here: The sinusoidal controlling voltage (100–1100 mV) for NLDS with a frequency of 2 Hz was applied to the two outer electrodes by a symmetric amplifier (output impedance 50 Ω). The inner two electrodes were connected to the input ports of a differential preamplifier. Because a simultaneous examination of two samples under same conditions is required, a second identical electrode array with preamplification must be available. The resulting signals were digitized by a computer controlled multi channel DA/AD-converter (Lab-PC+, National Instruments, Austin TX U.S.A.). This board also supplied the voltages for the NLDS and the control of the EMF. A function generator (Krohn-Hite Model 5200) generated the sine curve for the AC magnetic field with a frequency accuracy of 0.1 %. The two operational power amplifiers of a OPA 2541 chip drove the solenoids generating the constant as well as the variable magnetic field components, which were monitored by the coil currents and the magnetometer. Figure 2 NLDS measurement setup (schematic). The voltage control signal is applied by a symmetric amplifier to the outer two of a plane 4 gold electrode array. The NLDS signal generated by the sample is clamped by the inner two electrodes. It will also be preamplified symmetrically, digitized by a fast computer controlled analog-digital converter and fourier analyzed by the data acquisition software. The static and dynamic magnetic field component is directed parallel to the electrode plane. The measuring station provides two such NLDS setups, enabling a simultaneous examination of two samples under same conditions. For cleaning, the electrodes were first treated with chromosulfuric acid for 1 h at room temperature and intensively rinsed with de-ionized water. This procedure was repeated approximately once per week. An improved long-term electric stability was obtained by slight modifications of the treatments described by Woodward et al. [ 23 ] and Yardley et al. [ 28 ]: The electrodes were additionally washed with chloroform, sonicated for 20 min in a detergent solution (0.5 % Triton X-100 in water), treated with CaCl 2 (0.5 M in water) in a ultrasonic bath (Bachhofer, Reutlingen), and finally rinsed with de-ionized water (<2 μS). This treatment resulted in amplitude deviations ≤ 5% over an experimental session of up to 2 h. If electrodes were not used for DC measurements, but for NLDS, they were additionally coated with a thin polymer film in order to improve noise reduction and stability [ 24 ]. Measurement techniques The cuvettes could be charged with the test solutions, discharged and rinsed through the teflon hoses by a syringe. A sample volume of 1 ml was used. Device specific, systematic errors were routinely checked by exchanging the electrode arrays used for sample and reference measurements and testing several cuvettes of the same type. After loading they were flooded with Argon for about 10 min. in order to remove O 2 from the solutions, avoid oxidation reactions and subsequent arising of reactive oxygen species (ROS) in the solute, then the hoses were sealed with rubber caps. After reaching a stable temperature of 20 ± 0.2°C, measurements were started. First 10 "dummy" scans were performed, in order to obtain a dynamic equilibration of the electrodes. B dc = 40 μT was selected as static magnetic field component for the ICR condition, because it is of comparable intensity as the natural geomagnetic field of the earth. A new sample was used for every experiment, an "aging effect" of the test solutions was observed, similar to an earlier seen effect, which resulted in a decreasing reproducibility for experiments with magnetic field exposed lipid vesicles [ 27 ]. Three types of techniques for measuring the electric currents in the solutions were applied, always using the gold electrode array described above: 1) For the validation of the ICR parameters of the Glu-HCl solution, the experiment of Zhadin et al. [ 25 ] was repeated. The DC voltage of 80 mV was applied to the outer electrodes (+40 mV and -40 mV), and the current through the solution was calculated from the resulting voltage between the inner electrodes. The current calibration was earlier performed with 10 mM HCl and the Glu-HCl solution. By that way, used by many established voltammetric techniques [ 29 ], superimposing electrode transition potentials can be widely avoided, in contrary to a direct current measurement with a two electrode system. A constant magnetic field B dc = 40 μT or 50 μT and a frequency sweep of the alternating magnetic field B ac = 50 nT (parallel to B dc ) from 2 to 7 Hz with 0.025 Hz/s and a resolution of 0.05 Hz were used. 2) For the investigation of the ICR transition with NLDS the same magnetic field setup is used like described under 1), the NLDS sine wave was applied on the electrodes (instead of the DC-voltage) and a constant magnetic field B dc = 40 μT was used. 3) Finally the FRV setup allowed the frequency analysis of the electric signals with variable amplitudes using the DNLDS technique described above. Glu-HCl samples were exposed to constant ICR conditions ( B dc = 40 μT and B ac = 50 nT, 4.14 Hz fixed), for reference experiments only the static component ( B dc = 40 μT) was applied with B dc switched off. The amplitude of the sinusoidal scanning voltage was increased in each experiment from 100–1100 mV in steps of 10 mV, record by record, the duration of each cycle was 4 s. The two data sets (from Glu-HCl and HCl sample) yielded by every single record were seperately Fourier transformed in order to get the spectra, these two spectra were divided by themselves (Glu-HCL spectrum by HCl spectrum) and the ratio spectrum was subsequently attached to a data file on a harddisc for later evaluation. Results Exploring Zhadin et al.'s experiment Applying a constant magnetic field of B dc = 40 μT at pH 2.85 and scanning the alternating magnetic field B ac from 2–7 Hz in steps of 0.05 Hz, a sharp peak was observed at 4.15 Hz. The peak current is about 20% larger than the mean ionic current of 7.4 nA, the peak width at half-height is 0.3 Hz (Figure 3 ). Equation (1) was validated by repeating the experiment ten times at an altered static magnetic field strenght of B dc = 50 μT. The current peak shifted to 5.2 ± 0.05 Hz with a height of 9.08 ± 0.3 nA, which lies approximately 22% over the mean ionic background current. These data verify the results of Zhadin et al. [ 25 ], and the field-dependence is in agreement with Eqn. 1. The signal was observed over a concentration of 2–10 mM. The signal became too small at c Glu < 2 mM, and there was insufficient solubility c Glu >10 mM (at 20°C). Subsequently, the pH-dependence was determined under identical magnetic field and scanning conditions mentioned above. Resonance effects are only seen in a narrow pH range of Glu-HCl (pH 2.75 – 2.90), with an maximum at 2.85, and vanishes outside this range. Figure 3 Current increase at ICR (DC). Current increase through the glutamic acid /HCl solution (2.24 mM, pH = 2.85) at and near ICR conditions. The static magnetic field strength is B dc = 40 μT, the amplitude of the alternating field B ac is 50 nT, the frequency resolution Δf = 0.05 Hz. Course using a constant electrode voltage of 80 mV ("Zhadin's experiment"). After this verification of the experiment of Zhadin et al. [ 25 ], these electric measurements were accompanied by some UV-VIS light scattering investigations, which should give information about possible colloidal properties of the sample. Glu-HCl solutions were investigated at a wavelength of λ = 260 nm with the pH adjusted from pH 2.55 to 3.25, showing a significant scattering maximum around pH 2.8 (data not shown). Further some DC voltage scans were performed with the gold electrode array for Glu at pH 2.85, and for dilute HCl adjusted to pH 2.85, applying only a static magnetic field B dc = 40 μT (no B ac ). A voltage range of 100–1000 mV was selected to allow a comparison with the voltammetric information out of the frequency resolved voltammetry (FRV). Again, maxima of conductivity were obtained, they lie at 250 ± 10 mV for Glu-HCl and 280 ± 10 mV for water/HCl pH 2.85 (data not shown). NLDS spectroscopy Next, the solutions were investigated by NLDS spectroscopy, in order to investigate in which way the frequency composition of the current spectra will change, when the predicted ICR condition for Glu-HCl is matched ( B dc = 40 μT and a B ac with f = 4.15 Hz). 15 experiments were performed and averaged. Figure 4 shows the power of the 2 nd harmonic (referenced against dilute HCl, pH 2.85). The full dataset is shown in Figure 5 on an absolute current scale, for magnetic frequencies of 4.00–4.30 Hz in a 3d-representation. The 1 st harmonic is split up into 2 closely spaced peaks around the ICR frequency. This is also well seen in Figure 4 , an effect not seen in the "Zhadin's" DC experiments [ 25 ] without frequency resolution. Furthermore an increase of the 2–6 harmonics is seen in Figure 5 for 4.10 and 4.20 Hz magnetic frequency, closely flanking the ICR value. The average standard deviation of these experiments was 8.2 % of the average Power of all DNLDS spectra. Figure 4 Current increase at ICR (AC). Course of the 2 nd harmonics of NLDS spectra taken for every scanned frequency of B ac . Data were related to reference scans with B dc = 40 μT, but without B ac . The grey bars indicate standard deviations. Other conditions like Figure 2. Figure 5 NLDS spectra on ion cyclotron resonance (ICR) transition. 3D-representation of the NLDS resolved current through a glutamic acid / HCl solution (2.24 mM, pH 2.85) during transition of the ICR condition (static magnetic field B dc = 40 μT, alternating field B ac = 50 nT, f BAC = 4.14 Hz) in steps of 0.05 Hz. Kinetics The following kinetic experiment should clarify, in which way the conductivity of the Glu solution is affected by repeated transitions through the ICR conditions. 12 experiments were performed, each with a new Glu-HCl sample. 100 DNLDS spectra were recorded with single 2 Hz sinus signals with 100 mV amplitude. B dc = 40 μT was permanently applied in all experiments, while B ac with f = 4.15 Hz was applied only during measurement no. 20–39 and 60–79. Subsequently, the courses of the lowest 5 harmonics (for 2, 4, 6, 8 and 10 Hz) were normalized to ± 1, and all 12 experiments were averaged, Figure 6 therefore represents the kinetics averaged over a total of 60 datasets. Because of the standardization, data are scaled in arbitrary units (a.u.). The power difference between "on" (exposure) and "off" periods is 1.38 ± 0.34 dB, standard deviations are drawn as bars. Figure 6 Current kinetics of switched ion cyclotron resonance (ICR) condition. Kinetics in arbitrary units (a.u.) of the ICR condition to a glutamic acid / HCl solution (2.24 mM, pH 2.85). The static magnetic field B dc (40 μT) was applied permanently, the alternating field B ac (50 nT, 4.14 Hz) was applied as indicated by "on" and "off". One experiment consists of a set of 101 DNLDS spectra performed by a 2 Hz sinus signal with 100 mV Amplitude. The data first 5 harmonics (2, 4, 6, 8, 10 Hz) where normalized and then averaged. Data are calculated out of 12 independent experiments. The grey lines mark the standard deviations, the dotted straight line shows the linear regression of the negative drift, represented by the equation y = -0.0039t + 0.5015 . Changes of the signal intensity become obvious, when switching the alternating magnetic field on or off. Over the entire experiment there seems to be a constant drift which we take as an indication for irreversible processes. This drift is indicated by the dotted line, which results from a linear regression of the entire dataset (-0.0039 t + 0.5015). The course seems to reach a new steady value after on/off switching of the alternating magnetic field with a time delay, which seems larger, when ICR is switched off. The average current change after the switching processes is -0.2 nA/s, the negative values result from comparison with a reference. Differential NLDS experiments with variable control voltages (FRV) Finally, the FVR method should show the intensity distributions of the harmonics of the DNLDS spectra and their dependence from the used amplitude of the electrode input voltage. Again Glu-at pH = 2.85 was investigated, using diluted HCl (pH 2.85) as the reference. The ratio of the resulting two NLDS power spectra was calculated according to equation (2), resulting in a logarithmic DNLDS spectrum. 101 such scans (4 s each) were performed for every single experiment, during which the amplitude of the applied course of 4 periods of a 2 Hz sine voltage increased from 100 to 1100 mV in 10 mV steps. Corresponding datapoints of the successive single DNLDS spectra generated one AC voltammogram each, for the respective frequency. Altogether, a set of 201 frequency resolved voltamogramms was obtained, because every spectrum contains 201 data values. Subsequently 20 such experiments were performed in which the solution was exposed to ICR conditions, alternating with 20 experiments, were only the static field was applied ( B dc = 40 μT), but not B ac . Each of the two groups of experiments were averaged separately. Then the two resulting datasets were subtracted (ICR experimental data minus data of the experiments with ICR condition switched off). This differential dataset had a total amplitude of 2.03 ± 0.38 dB, presenting just the contribution of Glu, because the voltammetric background from HCl was subtracted. Subsequent data normalization should allow a better comparison of spectra recorded with different amplitudes and likewise of voltammograms at different frequencies. Therefore in Figures 7 , 8 the full dataset is shown, again after standardization in a range from – 1 to 1. Figure 7 presents the data with standardization on the voltage axis for the voltammograms belonging to the individual frequencies. Figure 8 contains the same dataset, but with standardized spectra. The intensity maximum shifts with rising frequency from approx. 250 mV to 500 mV for frequencies <40 Hz, it then remains constant around 500–700 mV for higher frequencies. So most information will be contained in the low harmonic orders. Figure 9 shows the voltage dependent behaviour at the NLDS fundamental frequency (2 Hz) and three harmonics in the lower range (4, 8, and 12 Hz). Broad maxima are obvious, which seem to shift to higher voltages with increasing harmonic order by about 60 mV/Hz. The intensities increase to a local maximum at approx.25 Hz. At higher frequencies, the amplitude effects caused by the exposure to ICR conditions have a local maximum at 480 mV and merge into a continuum beyond 750 mV for all higher frequencies, according to the predominating capacitive damping of aqueous solutions with rising frequency. Figure 7 DNLDS resolved voltammogram dataset (normalized to spectral axis): Normalizations of the DNLDS resolved voltammogram dataset (sinewave 2 Hz with amplitude rising from 100–1100 mV, details of gaining data see text) of a Glutamic acid / HCl solution (2.24 mM, pH 2.85) under ICR Conditions (B dc = 40 μT, B ac = 50 nT, 4.14 Hz). Datapoints are colored resp. shaded according to the scale on the right border. Normalization of the spectra for each Amplitude shows a rising proportion of higher frequencies with a local (at about 500 mV) and a total maximum (at about 700–800 mV). By contrast, the proportions of the base frequency (2 Hz) and the lower harmonics decline. Figure 8 DNLDS resolved voltammogram dataset (normalized to voltage axis): The same dataset and representation style like Figure 9, but with normalization of the single voltammograms for each frequency. For low frequencies (<5 Hz) Voltammograms have a maximum at about 250 mV, comparable to the pure DC volt scans. But with rising spectral harmonics voltammetric maxima occur at about 700 mV with overlaying intensity patterns of 4 and 16 Hz in distance. Worthy of remark are 62, 78 and 94 Hz, these all are four folds of the used base ICR resonance frequency 4.14 Hz. Figure 9 Extracted voltage courses of the DNLDS resolved voltammogram dataset. Voltammograms for some harmonics of the DNLDS resolved voltammogram dataset (sine wave 2 Hz with variable amplitude 100–1100 mV, not normalized here, see text for details) of a glutamic acid / HCl solution (2.24 mM, pH 2.85) under ICR Conditions (B dc = 40 μT, B ac = 50 nT, 4.14 Hz). Discussion All results suggest the existence of a sensitive magnetic field effect on the conductance of a aqueous Glu solution. The effect shows no linear dependency of magnetic field parameters, it is rather peaking in a narrow range of combinations of static magnetic field strengths and frequencies of additional alternating magnetic fields, described by Eqn. 1. Several precautions were applied, in order to avoid artefacts as best as possible. So it has been shown, that the signal to noise ratio will be improved significantly by clamping the voltage drop inside the electrolyte and, if needed, by a subsequent calculation of the current by calibration functions, instead of a direct current measurement. These techniques are wide spread in voltammetry [ 26 ] and obligatory in NLDS [ 23 ]. Because the voltage clamping ideally should work without any electric current flow, the electrode surface transition potentials could more likely be excluded for causing the observed EMF effect ("electrode effects"). It should be least then apply, if a cell voltage is used bellow the electrochemical potentials of the electrode-electrolyte system, and independent from the other experimental setup. Different explanations are recently discussed for the kind of EMF effect observed here, all of them suppose a non linear oscillator principle described by quantum mechanical terms, allowing energetic interactions with the environment far below the thermal equilibrium of life processes. This search for "wave functions fitting in a properly sized box" should consequently provide an explanation for the repeatedly observed effects of effectiveness windows, regarding specific field strengths and frequencies of the EMF, e.g. seen on green algae grown in a magnetic gradient [ 30 ]. Ion channels of biological membranes were proposed in a early work of Liboff [ 31 ] for a suitable environment supporting ICR. A model of Binhi et al. [ 14 , 32 ] is based on an interference mechanism of quantum states of ions within protein cavities. The quantum dynamic description of an ion is given for the case of ion-protein complexes that rotate in magnetic fields. The individual molecular rotation is taken into account. The spatial distances considered here are in the size of the molecules involved, cavities built by proteins, and their bond lengths. A quantum electrodynamic description needing no additional supporting structures like protein molecules or lipid membranes was worked out by Giudice et al. [ 33 ], as an attempt to explain the experimental results of Zhadin et al. [ 25 ]. It is based on an underlying two-phase domain model of the solvent water, in which at room temperature ~40% of its volume is organized in spheres with a diameter of approximately 100 nm providing coherence for the included water molecules. These spheres should establish a stable frontier region with a thickness of ~4 nm, which allows a undisturbed ion movement, separated by an energy gap 0.26 eV against the surrounding, non coherent water phase. The circulation frequency of the ions in the frontier region should be given by equation (1) and be dependent on the external magnetic field strength. An additional superimposed alternating field B ac with the same frequency will modulate the radii of the orbits. As a consequence, the ion orbits fit no longer the frontier region and the ions escape into the surrounding water phase, where they increase the conductance. This model also tries to describe the results of [ 25 ] quantitatively, but takes therefore in account the electrode geometry of the original experiment. Further attention should turned to the comparably long persistence time of the ICR state (see Figure 6 ) implying a comparable long lifetime. Considering the existence of supramolecular orders of liquid water, such long lifetimes (>10 s up to hours) have been predicted for these states sensitive to weak EMF at biologically relevant temperatures. Ponomarev et al. [ 34 ] propose linearly ordered chains and clusters like a liquid crystal phase in water which interact with EMF. The soliton theory was applied for description. Studies on the electromagnetic "memory effect" of water implicate even high sensitivity and long lifetimes [ 35 ], and are probably caused by the same mechanism as the effects observed here. An more hypothetical two phase model also providing boundary layers has been emphasized by Colic et al. [ 36 ]. The authors discuss the presence of micro-dispersed gas bubbles. But this possibly can be discarded more than likely in our experiments, because degassed solutions were used throughout. Special attention deserve the obvious frequency dependent amplitude windows of the dielectric currents, which are observed in the NLDS experiments (FRV) with variable amplitudes. Two explanations for this effect would be possible. The additional electric field caused by the AC signal of the NLDS could modulate the charged particles inside a "quantum box", whatever will be the reason for its existence. An indication for such a mechanism could be the more or less ordered local maxima of conductivity in spectral as well as in the voltammetric domain of the data. But the frequency dependent conductivity band shifts of the FRV experiments (Figure 9 ) could either result from a "simple" interference with the frequency of the B ac field, which can be tuned on its part in discrete multiples, corresponding to possible "overtones" of the ICR (orbital) frequency of the ions. Interactions of the internal electric and external magnetic field could probably cause side band modulations. They are probably responsible for the seen splitting up of the ICR resonance peak (figure 4 ) when using a AC instead of a DC probe voltage. For progressed investigation of the observed effects some additional properties of the electric charge environment of the Glu ion should be known. The isoelectric point of Glu is at pH 3.22, the pK of the α-COOH-group is 2.19, that of the β-COOH at 4.25, the small optimum for the EMF effect around pH 2.85 does not coincide with any of these points. The Debye-Hueckel radii for Glu are about 5 nm, they determine the free ion movement, and influence consequently the current. Moreover, they could be responsible for a proper fit of the spatial ion distribution to the environing structure whatever, which enables a resonant EMF effect. Conclusion The results strengthen the idea, that weak electromagnetic fields can cause an resonance effect on molecular or even supramolecular scale in electrolyte solutions [ 33 , 35 ], and thereby possibly, influence biological processes, which involve these electrolytes. In this work, the electric currents in a glutamic acid solution were investigated with frequency resolution after applying weak EMF. The resonance peaks and the overtone-analysis in response to weak static plus alternating EMF support the existence of the ICR phenomenon in aqueous solutions containing electrolytes. A analysis of the data is possible under the basic assumption of a far reaching principle of arrangement (realized e.g. by the solvent matrix), which allows quantum electrodynamic processes on the nano-physical scale or larger. In general any kind of a suitable coherence mechanism should be essential for the observed effects in a dense medium like water, which had to support an energy gap against the thermal fluctuations of the environment, and enable a movement of charged particles which are only magnetically coupled to their outer environment. Not at least, the high sensitivity of the ICR to weak electromagnetic fields should be regarded. It makes the modulation of biological processes by the weak EMF of our everyday environment conceivable [ 37 ], possibly inducing likewise health risks and chances for new therapies, hardly minded till this day. Especially concerning the earlier [ 25 ] and the present study, glutamate is a neurotransmitter and is involved in a couple of other biological processes. The geomagnetic field, with all its anomalies and regional differences [ 38 ], in combination with all the natural and civilizing EMF, overlap with a wide range of possible ICR of biologically relevant ions. But also technical applications basing on the ICR are imaginable, as a potential direction of future research. Its further investigation will be worthwhile, by new experiments, comparing field studies of health phenomena, and not at least a further clear up of its physical principle. List of abbreviations EMF: (low frequency) electromagnetic field ICR: ion cyclotron resonance NLDS: non linear dielectric spectroscopy DNLDS: differential non linear dielectric spectroscopy. FRV: frequency resolved voltammetry Glu-HCl: A glutamate solution adjusted to pH 2.85 with hydrochloric acid (HCl). Authors' contributions The author itself carried out all experiments and drafted the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538269.xml
544349	Streptococcal necrotising fasciitis from diverse strains of Streptococcus pyogenes in tropical northern Australia: case series and comparison with the literature	Background Since the mid-1980's there has been a worldwide resurgence of severe disease from group A streptococcus (GAS), with clonal clusters implicated in Europe and the United States. However GAS associated sepsis and rheumatic fever have always remained at high levels in many less developed countries. In this context we aimed to study GAS necrotising fasciitis (NF) in a region where there are high background rates of GAS carriage and disease. Methods We describe the epidemiology, clinical and laboratory features of 14 consecutive cases of GAS NF treated over a seven year period from tropical northern Australia. Results Incidence rates of GAS NF in the Aboriginal population were up to five times those previously published from other countries. Clinical features were similar to those described elsewhere, with 7/14 (50%) bacteremic and 9/14 (64%) having associated streptococcal toxic shock syndrome. 11/14 (79%) had underlying chronic illnesses, including all four fatalities (29% mortality overall). Important laboratory differences from other series were that leukocytosis was absent in 9/14 (64%) but all had substantial lymphopenia. Sequence typing of the 14 NF-associated GAS isolates showed no clonality, with only one emm type 1 and two emm type 3 strains. Conclusions While NF clusters can occur from a single emergent GAS clone, this was not evident in our tropical region, where high rates of NF parallel high overall rates of GAS infection from a wide diversity of strains. The specific virulence factors of GAS strains which do cause NF and the basis of the inadequate host response in those patients who develop NF on infection with these GAS require further elucidation.	Background Necrotising fasciitis (NF) is an aggressive and rapidly destructive soft tissue infection resulting in high mortality and significant long term morbidity. Two groups of infectious agents have been described – Type I which are polymicrobial, involving anaerobic bacteria and streptococci other than serogroup A, and Type II from Streptococcus pyogenes (Group A streptococci, GAS) alone, or in association with Staphylococcus aureus or Staphylococcus epidermidis [ 1 ]. The latter was originally described by Meleney in 1924 as 'haemolytic streptococcal gangrene' [ 2 ] and has since been considered a rare entity. Since the mid-1980's a worldwide increase in infections due to GAS has been noted. In affluent populations where GAS disease is uncommon aside from pharyngitis in childhood, increasing numbers of necrotising fasciitis and streptococcal toxic shock syndrome (STSS) have been seen, as well as an upsurge of acute rheumatic fever apparently restricted to parts of the United States [ 3 - 6 ]. This has been attributed in some locations to dissemination of a virulent M1 serotype GAS clone[ 7 ]. In developing nations the pattern of GAS disease is different, with continuing high rates of streptococcal pyoderma and post-streptococcal disease[ 8 ]. Recent case series of GAS NF have been published from Norway[ 4 ] and Ontario, Canada[ 5 , 6 ]. The Northern Territory (NT) of Australia consists of a combination of affluent, urban residents, as well as residents in remote communities, who are predominantly indigenous Aboriginals with high levels of poverty and overcrowding. As in developing nations, these remote communities have very high rates of GAS disease in the form of streptococcal pyoderma, rheumatic fever and post-streptococcal glomerulonephritis[ 9 ]. We examine the epidemiology, clinical features and streptococcal sequence typing of a series of cases of GAS NF from the tropical top end of the NT and compare the results with the literature. Methods The Royal Darwin Hospital services 135,000 inhabitants of the NT, with around 100,000 located in the city of Darwin. Indigenous Aboriginal Australians account for 24% of the population[ 10 ]. Cases of possible GAS NF treated at Royal Darwin Hospital were prospectively identified. Case definition for inclusion in the analysis required GAS to be cultured from sterile sites, notably operative tissue specimens +/- blood cultures. NF was defined by necrosis of subcutaneous tissue, specifically fascial oedema and necrosis detected either macroscopically at surgery, or microscopically on histopathology. Fourteen cases fulfilling the case definition were identified between May 1994 and April 2001. A chart review was conducted to identify epidemiological characteristics, clinical features, laboratory results, treatment and outcomes. Age, gender, ethnicity, urban or rural residence and any concurrent medical conditions were documented. The dates of presentation, presumptive diagnosis of NF and of surgery were documented. Symptoms, signs, hemodynamic status and temperature were documented for these dates, as was the presence of STSS as defined by the Working Group on Severe Streptococcal Infections[ 11 ]. Antibiotic treatment, surgery, and any adjuvant therapies were documented. Baseline hematological and biochemical parameters were recorded for the date of presentation, and blood and tissue culture results were documented. Duration of Intensive Care Unit (ICU) admission, and type of ICU support required were recorded. Incidence rates of GAS NF for this region of the NT were calculated using regional population numbers from census data. There was possibly an increase in case ascertainment after 1998 and incidence rates were also calculated by year. Procedures for emm sequence typing and analysis of emm -specific PCR products were carried out as previously described[ 12 ]. DNA sequences were subject to homology searches against all emm sequences deposited in GenBank and in the Centers for Disease Control (CDC) Streptococcus pyogenes emm sequence database[ 13 ]. Approval for the study was obtained from the Health Research Ethics Committee of Royal Darwin Hospital and the Menzies School of Health Research. Results Table 1 summarizes the details for each patient. From 1994 to 2001, 14 cases of confirmed GAS NF were treated at Royal Darwin Hospital. The number of cases per year ranged from zero in 1995 and 1998 to five in 2001. This gives a maximum yearly incidence of 3.8 cases per 100,000. Half of the cases occurred in Aboriginal patients, with an incidence of 5.8 per 100,000 for Aboriginals in the 2001 cluster of cases. Eight cases occurred in women (57%), five of whom were Aboriginal. Ages ranged from 27–61 years in males (mean 41.5 yrs) and 27–59 years in females (mean 41.5 yrs). Eleven cases were community acquired and three nosocomial. All seven Aboriginal patients had multiple concurrent medical conditions, whereas three of the seven non-Aboriginals had no additional pathology. Patients presented most consistently with pain in the affected area. The other common complaints were swelling, and systemic symptoms of fever and rigors. Signs on presentation were typically induration and tense swelling of the area. Only four of the fourteen cases (29%) were noted to have extensive skin erythema, but nine (64%) had local pre-existing wounds or ulceration. Ten patients had only limb involvement, two had chest wall involvement, one had extension from right leg to the abdominal wall and one had extension from the left buttock to perineum/labia. STSS occurred in nine patients. Ten patients required inotropic support, six had a documented coagulopathy and eleven had renal impairment. Thirteen patients (93%) required ICU stay of 1–24 days. Eleven patients were ventilated, and six received renal replacement therapy. Total white cell count (WCC) was found to be raised (>11 × 10 9 /L) in only 5/14 (36%) patients, although lymphocyte count was globally reduced, being <1.0 × 10 9 /L in all patients, and <0.5#215;10 9 /l in nine cases (64%). Renal function was frequently impaired, with 11/14 (79%) patients having raised urea (mean 17.7 mmol/L, NR 3.0–8.0 mmol/L) and 12/14 (86%) having raised creatinine (mean 301 μmol/L, NR 50–100 μmol/L). All patients were hypoalbuminemic with 11 having albumin <30 g/L (mean 24.5, range 11–34, NR 35–45 g/L) Plasma creatine kinase (CK) ranged from 25 U/L to 2526 U/L (NR <220 U/L), being elevated in 5/9 (56%) patients measured. C-Reactive Protein (CRP) was substantially raised in all 11 where measured, being between 108–522 mg/L (NR<10). Blood cultures were positive for GAS in seven cases (50%). GAS was isolated from all operative tissue specimens, with four also positive for S. aureus . One GAS isolate could not be sequenced. GAS from the other 13 patients showed a large diversity of emm sequence types, with no clonality. Where blood and tissue GAS were both recovered from an individual, the emm sequence types were the same. One patient's GAS isolate was a previously uncharacterized emm sequence type and there was only one case with emm sequence type 1 GAS isolated. The two emm 3 isolates were from the 2001 cluster of five cases. However they were very divergent on random amplified polymorphic DNA analysis (data not shown), one being emm sequence type 3.2, and the other three isolates in this cluster were all different emm types. All patients underwent surgery 0–7 days after presentation. Debridement alone was conducted in nine patients, and combined with fasciotomy in one case. Fasciotomy alone was done in one case. Three patients underwent limb amputation. At surgery it was established that three patients (21%) had concurrent myonecrosis. Antibiotic therapy was instituted with a beta lactam in all patients (benzylpenicillin in nine, meropenem in three, cephazolin in one and flucloxacillin in one). Clindamycin was also given in eleven patients. Adjuvant hyperbaric oxygen therapy was given in four cases, and intravenous immunoglobulin (IVIG) in one. Four patients (29%) died, all being Aboriginal females with co-morbidities. Two of the four patients with truncal infection died in comparison to only two of the ten with only peripheral involvement. The mean time to surgery from presentation was 4.25 days in patients who died and 3.7 days in survivors (not significant). Blood cultures were positive in 3/4 fatalities with STSS in 3/4. All four were in ICU for 1–5 days until death, with all requiring inotropes and mechanical ventilation, two given renal replacement therapy and one IVIG. Two received clindamycin in addition to beta-lactam antibiotic therapy. Surgery in the fatal cases consisted of debridement and/or fasciotomy, with no amputation. Discussion Demographics We describe a series of 14 cases of GAS necrotising fasciitis in a region with a high background incidence of GAS disease[ 9 ]. Our incidence of NF in 2001 of 3.8/100,000 and 5.8/100,000 in the Aboriginal population is higher than previously published rates which range from 0.4/100,000 in Canada[ 5 ] to 2/100,000/yr in Norway [ 4 ]. However the population denominators in the Canadian and Norwegian studies are far greater than in our study. Nevertheless, previous data from our region showed that rates of GAS bacteremia in the Aboriginal population were five times those in the Caucasian population[ 9 ]. Therefore NF in our population parallels an overall greater rate of GAS infection, most notably in the Indigenous population where conditions reflect those of developing countries[ 8 ]. The mean age of our patients (41.5 years) is lower than other reports showing an average age of 56–58 years, with notable increases in incidence with increasing age[ 4 - 6 ]. Unlike all other series which had a male predominance, 57% of our cases were female, with 71% females amongst the Aboriginal cases. The age and sex pattern seen in our series reflects the high rates of chronic diseases in the Indigenous population, with mortality from all causes being higher in all age cohorts, but especially so in females[ 14 , 15 ]. 79% of our patients had concurrent chronic disease, compared with 46–71% in previous series[ 4 - 6 , 16 ]. In this series 64% had pre-existing wounds, consistent with reported rates of 47–66%[ 4 - 6 , 16 ]. None of our patients had preceding varicella infection. Laboratory results Two studies have looked at the value of early blood test results in predicting NF as opposed to non-NF soft tissue infection. One found that CK >600 U/L achieved PPV of 58% for NF[ 17 ]. However only 1/9 cases in this series had CK >600 U/L. CRP of >16 mg/L was also thought to be indicative of NF[ 16 ]. CRP was over 100 in all those measured in our series. Raised WCC has also been suggested as useful in a predictive model of early NF[ 18 ], with rates of 66–73% in other series[ 4 - 6 ]. However, only 36% of our patients had a raised WCC. We found lymphopenia to be a more consistent factor, with 100% of patients having a lymphocyte count of <1.0 × 10 9 /L, 64% of those being <0.5 × 10 9 /L. Low albumin was also universal. Clinical features STSS occurred in 64% of our cases in comparison to 40–46% in other series[ 4 - 6 ]. 71% of our patients were hypotensive at presentation, with 86% having renal impairment in comparison to 35–61% in other series. Similarly to the other series, 93% of our patients were admitted to the ICU but more of our patients required mechanical ventilation (79%) and inotropic support (71%). Positive blood cultures are not a prerequisite for GAS NF, as stated by Meleney in the initial descriptions[ 2 ], and reinforced in subsequent studies showing bacteremic rates of 38–60%[ 4 - 6 , 16 ], consistent with 50% in this study. Case-fatality rates for NF previously reached 50%, with mortality of 80–100% in GAS myositis[ 3 ]. Recent series report lower case-fatality rates of 20–34%[ 4 - 6 ], with our mortality being 28%. As in previous studies, deaths were more common in those with underlying illness, truncal infection, bacteremia, STSS, myonecrosis and delay in diagnosis and appropriate therapy. Therapy Immediate, extensive surgical debridement of all necrotic tissue is considered essential for optimal treatment of NF, with one early observational study showing a mortality of 0% with aggressive early surgery in comparison to 50% in historical controls[ 19 , 20 ]. Surgery was conducted on the day of presumptive diagnosis of NF in all our patients, but none of the four fatal cases underwent amputation and one with upper limb involvement had only a fasciotomy performed. Combination therapy with benzylpenicillin and clindamycin is now recommended treatment for proven GAS NF and STSS[ 21 - 23 ]. Experimental data suggest that penicillin is not as bactericidal when there are large numbers of GAS organisms present[ 24 ], with decreased expression of penicillin binding proteins when large inocula reach a stationary growth phase[ 25 ]. The potential benefits of clindamycin have been supported by a murine model of streptococcal myositis[ 26 ]. Clindamycin is a protein synthesis inhibitor, potentially also suppressing bacterial toxin and M protein production[ 25 ]. Furthermore, clindamycin may modulate the host immune response, by reducing lipopolysaccharide-induced monocyte production of TNFα[ 27 ]. A combination of β-lactam antibiotic and clindamycin has been shown to be superior to β-lactams alone in a retrospective review of invasive GAS infection in 56 children[ 28 ]. A recent retrospective review of notified invasive GAS infections in Florida, USA showed the use of clindamycin to be associated with lower mortality in NF cases, but not in other invasive GAS infections[ 29 ]. However there have been no randomised clinical trials. Hyperbaric oxygen (HBO) therapy has been advocated as adjuvant therapy for both microbiological types of NF. Although there are no randomised trials, NF is an approved Undersea and Hyperbaric Medical Society indication for HBO therapy[ 30 ]. It is suggested that increased tissue oxygen partial pressures increases bacterial killing by increased respiratory burst and increased formation of oxygen free radicals. HBO is postulated to facilitate wound healing by supporting fibroblast proliferation and angiogenesis[ 30 , 31 ]. In a murine model of GAS myositis, the combination of penicillin and HBO therapy exerted at least additive effects in decreasing bacterial counts in vivo and increasing survival[ 32 ]. An observational study of 29 patients with NF showed significantly lower mortality, and need for fewer debridements in those receiving adjunctive HBO, despite the HBO group being more seriously ill[ 33 ]. Patients with non-clostridial fasciitis appeared to have greater benefit. However more recently a retrospective evaluation of 37 patients treated for NF showed higher mortality with greater need for debridement in those receiving HBO[ 34 ]. The clinical picture of STSS has been attributed to superantigens produced by GAS, including pyrogenic exotoxins, streptococcal superantigen SSA and a mitogenic exotoxin[ 35 ]. This is supported by selective depletion of Vβ-bearing T cells in patients with STSS[ 36 ]. However GAS isolates from both severe and uncomplicated disease can produce large amounts of toxins. Furthermore, in comparison to patients with uncomplicated infection the sera of patients with severe GAS disease had low antibody levels against erythrogenic toxins[ 37 ], suggesting an important role for host humoral immunity. IVIG has therefore been proposed as therapy to increase neutralizing antibody levels in patients with severe GAS disease. A study of 12 patients, 11 with STSS and one with GAS NF, showed increased capacity of plasma to neutralize superantigenic activity following IVIG[ 38 ]. Post-IVIG plasma from each patient completely blocked cytokine production elicited by their respective GAS culture supernatants or by purified streptococcal pyrogenic exotoxins. GAS superantigens and cytokines suggestive of superantigen response have been isolated from tissue samples in patients with GAS NF[ 39 ]. More severe clinical disease correlated with significantly higher bacterial load in biopsy samples, and bacterial load in turn correlated with expression of superantigenic toxins. This supports a role for IVIG in GAS NF, with or without STSS. Adjunctive IVIG therapy in STSS was evaluated in a multicenter, randomized, double-blind, placebo-controlled trial[ 40 ]. Although prematurely terminated due to slow patient recruitment, there was a non-significant 3.6 fold higher mortality rate in the placebo group and a significant decrease in sepsis-related organ failure assessments on days two and three in the IVIG group. There are case reports and two case series of four and seven patients describing clinical improvements of patients with STSS following IVIG[ 41 - 43 ]. In a series of 20 cases of GAS NF and myonecrosis, 16 were treated with IVIG[ 6 ]. Although the case fatality rate was not significantly lower in those receiving IVIG, the overall survival rates of 80% in patients with NF and 63% in patients with myonecrosis were much higher than previously reported[ 6 ]. In summary, there is in vitro and in vivo experimental data to suggest a benefit for clindamycin, HBO therapy and IVIG in GAS NF and myositis. This is supported by limited case series but not by randomized controlled trials. The primary role of early, aggressive surgery in GAS NF remains probably the most critical factor in minimizing mortality. Clindamycin is recommended for GAS NF but whether clindamycin alone is as effective as penicillin plus clindamycin remains unclear. Benefit from HBO remains unproven, while there is emerging evidence for the use of IVIG, especially if STSS is present. Molecular epidemiology Following a reported resurgence in invasive GAS infections in the 1980's, epidemiological studies proposed that dissemination of a new, more virulent strain of M1 serotype GAS was the cause for a large proportion of these infections[ 7 ]. In reports from the United States, Europe and New Zealand the proportion of this strain appeared to increase, both in those with uncomplicated pharyngitis and in invasive GAS infections[ 37 , 44 - 47 ]. The M1 serotype has also been reported to have a stronger association with STSS[ 48 ]. In contrast to these findings, other groups have noted a significant diversity amongst the organisms causing severe streptococcal infections, as well as significant clinical variation amongst patients infected with identical organism clones, suggesting that the severe infections are not related to one specific virulent clone[ 49 - 51 ]. Our findings are consistent with the latter, with GAS sequence typing showing all NF isolates to be different and only 1/13 to be emm type 1. This reflects previous findings of enormous diversity amongst organisms causing GAS bacteremia in our region, with no evidence of a dominant clone[ 9 ]. As well as finding no correlation between NF and specific M types, a recent Dutch study also found that no specific toxin genes or genes encoding matrix binding proteins were associated with NF[ 48 ]. Conclusions In conclusion we have described a case series of necrotising fasciitis due to GAS for a unique mixed population of patients living in both affluent and disadvantaged society. Our overall incidence rates of GAS NF are higher than previously reported, with rates in the Aboriginal population up to five times those previously published. The majority of those with NF had underlying chronic illnesses and mortality was also higher in this group. Important laboratory findings were that leukocytosis was absent in more than half of cases but all had substantial lymphopenia. Emm sequencing typing of GAS isolates showed none to be the same and only 1/13 to be emm 1. GAS NF in our region is not clonal but occurs from a wide diversity of GAS strains that reflect the very high background rate of both uncomplicated GAS infection and invasive disease. While NF case clusters can occur from a single emergent GAS clone, this is not evident in our tropical region. The specific virulence factors of the GAS which do cause NF and the basis of the inadequate host response in those patients who develop NF on infection with these GAS require further elucidation. Competing interests The authors declare they have no competing interests. Authors' contributions MH collated and analysed the case data and was principal writer of the manuscript, PF carried out the molecular studies, PC participated in the study design and patient management, BC conceived the study and participated in the study design and coordination and patient management. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544349.xml
555536	Can Zipf's law be adapted to normalize microarrays?	Background Normalization is the process of removing non-biological sources of variation between array experiments. Recent investigations of data in gene expression databases for varying organisms and tissues have shown that the majority of expressed genes exhibit a power-law distribution with an exponent close to -1 (i.e. obey Zipf's law). Based on the observation that our single channel and two channel microarray data sets also followed a power-law distribution, we were motivated to develop a normalization method based on this law, and examine how it compares with existing published techniques. A computationally simple and intuitively appealing technique based on this observation is presented. Results Using pairwise comparisons using MA plots (log ratio vs. log intensity), we compared this novel method to previously published normalization techniques, namely global normalization to the mean, the quantile method, and a variation on the loess normalization method designed specifically for boutique microarrays. Results indicated that, for single channel microarrays, the quantile method was superior with regard to eliminating intensity-dependent effects (banana curves), but Zipf's law normalization does minimize this effect by rotating the data distribution such that the maximal number of data points lie on the zero of the log ratio axis. For two channel boutique microarrays, the Zipf's law normalizations performed as well as, or better than existing techniques. Conclusion Zipf's law normalization is a useful tool where the Quantile method cannot be applied, as is the case with microarrays containing functionally specific gene sets (boutique arrays).	Background DNA microarrays have become a widely used biotechnology for assessing expression levels of tens of thousands of genes simultaneously in a single experiment [ 1 , 2 ]. Whether microarrays are being used for global tissue profiling or for differential expression studies, data normalization is an essential preliminary step before statistical analysis methods can be applied. The purpose of all normalization techniques is to transform the data to eliminate sources of variability stemming from experimental conditions, leaving only biologically relevant differences in gene expression for subsequent analysis. Normalization can be divided into two stages, intra-array normalization and inter-array normalization. Intra-array normalization deals with variability within a single array caused by factors such as differences in print-tip characteristics, channel differences in two-dye systems, and spatial heterogeneity across the array surface [ 3 - 5 ] and should be carried out using accepted methods before inter-array normalization is applied. This paper assumes intra-array normalization has been performed and presents an inter-array normalization method for comparison of gene intensity levels between multiple microarrays to deal with variation caused by such factors as differences in RNA isolation efficiency, labeling efficiency, hybridization conditions, exposure times, and detection efficiencies. It is now clear that simple inter-array normalization techniques, such as simple scaling to housekeeping genes or normalizing to a global mean, are not adequate for microarray data [ 6 ]. Housekeeping genes have been found to be more susceptible to modulation than previously thought [ 7 ]. Along with others [ 5 ], this paper underscores the potentially serious drawbacks of the global mean and other such methods. Recent literature has thus provided a plethora of more sophisticated normalization and analysis techniques as researchers struggle to cope with the task of microarray data analysis, some of which include maximum likelihood analysis [ 5 ], centralization [ 6 ], principal component analysis [ 8 ], analysis of variance [ 9 ] and Bayesian network analysis [ 10 ]. Analysis of publicly available large-scale SAGE gene expression data sets [ 11 , 12 ] and an intra-phyletic survey of genome wide Affymetrix microarray experiments [ 13 ] have indicated that the large majority of expressed genes exhibited power-law distributions, while some microarray expression data exhibit a more log-normal distribution [ 14 ]. Our normalization procedure was inspired by the observation that the intensities measured on our microarray system also followed a power law distribution and can therefore be described by a simple mathematical model. Zipf's law [ 15 ] is a power law function that states that the magnitude of an intensity measurement ( y ) is inversely proportional to the rank ( r ) of that data point in the data set, y ∝ r c (1) where c is a coefficient close to -1. Our microarray data can be classified as a generalized form of Zipf's law because the coefficient ( c ) is not always close to -1 and, in fact, varies between individual microarrays, making simple linear normalization procedures, such as global normalization to the same mean, inappropriate. However, the normalization procedure proposed here demonstrates that by taking Zipf's law into account, it is possible to apply a simple intra-array normalization procedure such that all filters have the same coefficient c and proportionality. We demonstrate the Zipf's law based normalization technique on microarray data sets representing both single channel and two channel technologies. In the single channel category, we produced two radio-labeled, nylon membrane based cDNA data sets, one commercial and one generated "in-house". Both systems contain a selection of genes chosen without regard to functional or pathway considerations, which make them especially appropriate for normalization using Zipf's law. These data sets were also normalized to a global mean (the mean of all microarrays) [ 16 ], and the quantile normalization method [ 17 ]. In addition we produced a two channel, fluorescently labeled, glass slide, oligo-based microarray data set generated 'in-house'. This microarray can be classified as a 'boutique' microarray because it consists of a selection of genes involved in apoptosis. This data set was normalized with a variant of the Zipf's law normalization method that uses a subset of the distribution as a proxy for normalizing the entire microarray. A comparison was then conducted against a variant of the loess normalization method that uses an a priori selection of 'housekeeping' genes as a proxy for normalization. The finding that our microarray data distributions conform to a power law distribution agrees with predictions based on genome wide gene expression studies [ 11 - 13 ], however Hoyle, et. al. [ 14 ] observed that microarray distributions were log normally distributed with possible power law tails. To investigate this discrepancy, and to verify that our normalization technique could be useful in the normalization of data sets from other microarray systems, we also surveyed publicly available data sets from the NCBI Gene Expression Omnibus [ 18 ]. The two assumptions upon which the normalization method are based are the same as those used in other normalization methods [ 5 , 6 ], namely that in comparisons between similar tissues or cell lines under different experimental conditions i) most genes are not, or only moderately, regulated, and ii) approximately equal numbers of genes are up regulated as down regulated. Systems which conform to these two assumptions will be referred to as 'well-behaved' in this paper. While these assumptions probably hold for microarrays derived from a diverse sampling of genes, for example an EST library survey, they may not hold for microarrays containing genes specifically selected based on function or pathway (so called 'boutique' microarrays) as it is likely that most genes will be affected by the experimental treatments. One way to circumvent the restrictions resulting from these assumptions is to use a subset of data, or proxy, from the boutique array data set which fulfils the 'well-behaved' criteria. In developing a boutique microarray normalization technique, Wilson et. al. [ 4 ] have devised a method for selecting a subset of genes within a microarray data set that have low variation between arrays and are well representative of the spectrum of intensities measured on the microarray. They term this a priori selected subset 'housekeeping' genes, however it should not be confused with the a posteriori set of genes typically envisioned when the term is used. Another possible proxy that could meet the 'well-behaved' criteria are control spots which are included on the microarray during it's manufacture. We tested our normalization method on data from a two channel boutique microarray experiment using two types of control spots as proxies for normalization (Positive and negative internal controls, and housekeeping genes). The Zipf's law normalization methods were then compared with the variant of the loess method developed by Wilson et. al. [ 4 ] using housekeeping genes. Results Verifying Zipf's Law Before applying the described normalization method, the adherence of the reference curve (the median gene intensity data versus rank) to Zipf's law was verified. The most common method of verifying conformity to Zipf's law is a linear regression on the log e -log e transformed data set. Our regression showed a good fit, with a correlation coefficient of -0.98 and a slope of -0.56 for microarrays representing human colon (Figure 1a , Figure 6A , Table 1 set A), a correlation coefficient of -0.99 and a slope of -0.78 for rat brain microarrays (Figure 6B , Table 1 set B), and a correlation coefficient of -0.99 and a slope of -0.60 for the mouse apoptosis microarrays (Figure 6H , Table 1 set H). It should be noted that while the low ranking intensities may show a marked deviation from the regression line, this data typically accounts for a very small proportion of the total data and does not have a large affect on the regression curves. Normalization results – single channel microarrays A comparison of the Zipf's law normalization method to the simple method of setting all arrays to a global mean (the mean of all microarrays) and to the quantile method was conducted on the single channel microarray data sets. Five human Unigene microarrays from the panel of thirty-two microarrays used in the sigmoidal colon experiments were selected to represent the greatest variability in pre-normalized data observed in the experiment (Figure 1b ). Normalization to a global mean (Figure 1c ) yielded data sets that displayed a higher variability in the coefficient c of the Zipf's power function (formula 1) than that observed after normalization by the Zipf's law method (Figure 1e ) or the quantile method (Figure 1d ). The Zipf's method showed the lowest variation in the Zipf's exponent and had the lowest spread of the data around the ln(rank) vs. ln(intensity) line. Results of an identical log e intensity versus log e rank plot comparison in Clontech rat microarrays showed little difference between the quantile and Zipf's methods [see Additional file 1 ]. However it should be mentioned that this method of data plotting provides one view of the data which is especially favorable to the Zipf's law normalization method. Next we examine the results of the MA-plots, a technique that is especially favorable to the quantile normalization method. In order to access the effectiveness of the normalization method, pairwise comparisons using MA-plots (sometimes called RI plots, or log ratio vs. log mean intensity plots) [ 19 ] were carried out on the raw data, and data normalized with the global mean method, quantile normalization and Zipf's law on both data set A & B (Figure 2 & 3 respectively). With the raw data, the distribution of log-intensity ratios is not centered around zero which is as expected in an un-normalized data set. There is a noticeable intensity dependent effect, sometimes described as a 'banana' curve, which is characteristic of many microarray data sets. Normalization with the global mean method results in a shift of the center of the log-intensity ratio distribution closer to zero, one important criterion for well normalized data, however, especially in the low log mean range, the bulk of the data points still deviate appreciably from zero. The intensity dependent effect is evident, with the low intensity end of the loess fit curving away from the zero axis. The intensity dependent effect is removed using the quantile method. The log intensity ratios of the data distributions normalized using Zipf's law are well centered around zero, but the intensity dependent effect is still apparent. In this case however, the bulk of the data lies very close to zero on the log-ratio scale. [see Additional file 2 ] This is due to the fact that Zipf's law normalization not only shifts the data distribution on the log ratio scale, but also rotates the whole distribution in log-ratio log-intensity space. The Kolmogorov-Smirnov test is often used to determine whether data distributions differ significantly and provides a test statistic that measures the proportion of overlap between distributions which ranges from 0 (in the case of identical distributions) to 1 (for non-overlapping distributions) [ 20 ]. Mean Kolmogorov-Smirnov values (Table 2a, b ) showed the expected trend, with the high values for raw, unnormalized data decreasing when global median normalization was applied, decreasing again after Zipf's law normalization, and reaching zero for both data sets under quantile normalization. It should be noted that the Kolmogorov-Smirnov test statistic will always be zero after quantile normalization because this method forces the data distributions of all microarrays to be identical. Normalization Results – Two Channel Boutique Microarray Plots of log e intensity versus log e rank fitted with linear regressions show that the Zipf's law normalization based on internal controls (Figure 4a ) and on selected housekeeping genes (Figure 4c ) have relatively similar coefficients c according to Zipf's power function (formula 1) as evidenced by the similarity in slopes of the regression lines. Loess normalization using selected housekeeping genes (Figure 4b ) showed slightly more variation in c coefficients. The unnormalized raw data is also depicted (Figure 4d ) along with two other normalization results, the loess method (Figure 4e ) and the quantile method (Figure 4f ). These are provided for reference only. Neither method can be validly applied to boutique arrays because both rely on the 'well-behaved' genes assumption. It should be noted that much of the variation in c coefficients under the various normalization regimes is due to one channel (Cy3) on one microarray which had low median intensity and high variance due to low labelling efficiency (depicted in black in Figure 4 ). When normalized with the loess techniques (Figure 4c and 4f ) the second channel (Cy5) on this array is adjusted to have a similar median intensity and variance, possibly skewing the results in favour of the Zipf's normalization techniques. To make the normalization method comparison unbiased, we eliminated this array from the analysis [see Additional file 3 ]. The Zipf's normalization based on internal controls (a) showed the lowest variation in c coefficients, the methods based on selected housekeeping genes (b, c) performed approximately equally well. Here again, raw (d), quantile normalized (e), and loess normalized (f) plots are provided for reference only. We generated MA plots for each of the normalization methods we compared (Figure 5 ). Typically, MA plots are produced from data from each channel of a single microarray. In addition to these 'within-array' plots (the first three rows of graphs in Figure 5 ), we also examined 'between-array' plots to evaluate the potential of the normalization methods to allow us to perform across array comparisons. The Zipf's using internal controls was slightly more well centered around the zero log ratio axis than the methods using selected housekeeping genes, especially in between-array plots. The raw and loess normalized plots are provided for reference only. Finally, to quantify the differences between distributions after normalization, pairwise Kolmogorov-Smirnov values were computed for both the complete boutique array data set (Table 2c ) and after eliminating the array which contained a low median intensity and high variance due to low labelling efficiency (Table 2d ). In addition to computing the Kolmogorov-Smirnov values for all possible between-array pairwise combinations, we also summarized just the within-array pairwise comparisons (in parenthesis in Table 2 ). Of the normalization methods which can be applied to boutique microarrays, the Zipf's method using internal controls produced the most similar data distributions when all possible between-array comparisons are taken into consideration. When only within-array comparisons are considered, the Zipf's method using internal controls was better after the low labelling efficiency array was eliminated. The Zipf's method using selected housekeeping genes did not perform as well as the other methods in within-array comparisons, and was the middle performer when all possible between-array comparisons were computed. Kolmogorov-Smirnov values were computed from the global mean, Zipf's general, quantile, and loess normalization methods and are provided for reference only. Microarray platform comparison In a survey of seventeen microarray data sets, Hoyle et. al . [ 14 ] reported that microarray data follow a log-normal distribution with power-law tails. The three data sets presented in this paper exhibited distinct power-law distributions (Table 1 , data sets A, B and H). To ascertain whether the data sets we used were unusual, we surveyed nine additional data sets (Table 1 , data sets C-G, I-K) to determine their conformity to Zipf's law and the log-normal distribution respectively. The microarray data sets fell into two broad categories, power law distributions (Figure 6 , data sets A-E) and log normal distributions (Figure 6 , data sets I-K). Of the six power law data sets, two (B and C) clearly followed Zipf's law distributions. The remaining four (data sets A, D, E, and H), while still power-law distributed, showing noticeable deviations from the distribution at the lower rank (higher intensity) portion of the distribution. Of the platforms that where recognizably log normal in distribution, two fluorescent dye labeled, oligo-based Affymetrics platforms (data sets K and L) followed the distribution most closely and two dye labeled, cDNA systems (data sets I and J) were perceptibly log normal. The two SAGE experiments (data sets F and G) which were included for comparison purposes, exhibited Zipf's law distributions. Coefficients of determination (r 2 ) of the log mean intensity vs. log rank are a measure of conformation to a power-law distribution and ranged from 0.9968 to 0.7773 for microarray data sets, 0.9982 and 0.9978 for the SAGE experiments (Table 1 ). Discussion Zipf's law is based on observations made by linguist George Kingsley Zipf that the frequency of word occurrences in natural languages is proportional to the negative power of the rank order of the word. Beside the original findings in natural languages [ 15 ], Zipf's law has been found to apply to a plethora of natural phenomena, from the populations of cities to the impact factors of scientific journals as well as a variety of biological data, of which a review made available by Wentian Li [ 21 ] is an excellent online resource. It is important to point out, that being a phenomenological principle, Zipf's law does not imply that there is a universal underlying physical process at work. However, in much the same way that the Gaussian-Normal distribution occurs naturally in data and can be used to statistically test or otherwise manipulate the data, the fact that microarray data conforms to Zipf's law can be adapted for the purpose of microarray normalization. Zipf's law is a power law function that states that the magnitude of an intensity measurement is inversely proportional to the rank of that data point in the data set, where c is a coefficient close to -1. Ranking is a method common in statistics, which has previously been used to analyze microarray data. Hoyle el. al. [ 14 ] used ranking as a method for evaluating microarray data and proposed the use of several statistics including χ 2 to quantify the agreement of the distribution to Benford's Law [ 22 ], and σ 2 as a quality control measure to detect such factors as low signal to background ratio, or mRNA probes extracted from mixed cell types. Ranking also figured prominently in the evaluation of a survey of inter-array normalization methods [ 23 ] where the statistics 'absolute rank deviation' and 'relative rank deviation' were used to select the method that produces the most 'well-normalized' data. The normalization procedure described in this paper is the first to combine these two ideas, namely that ranking can be used to judge the effectiveness of a normalization method, and that microarray data conforms to Zipf's law. We evolved these ideas into a novel and easily applicable normalization method and compared this method with existing methods to eliminate non-biological variation from microarray data sets. In order to implement an appropriate data normalization technique, it is important to know the distribution of a given data set. Several publications have examined the data distributions that typically result from microarray experiments. In a survey of seventeen microarray data sets, sixteen of which were fluorescent dye labeled, Hoyle et. al. [ 14 ] reported that microarray data were found to have a log normal distributions with power law tails. More recent publications have reported that the abundance of expressed genes exhibit power-law distributions [ 11 , 13 , 24 ]. Results from our own data sets and a subsequent survey of publicly available data sets from both radioactively and fluorescently labeled platforms suggest that both types of distributions can be manifested in microarray data. Comparisons between the Zipf's law and quantile normalization methods using MA plots showed that the quantile method effectively removes intensity dependant effects, sometimes referred to as 'banana' curves, from microarray data sets, while the Zipf's law method has no effect on the curved nature of the intensity dependent effect. This is not altogether unexpected as the quantile method was specifically designed to remove such effects. While the Zipf's method does not remove the curve from the intensity dependent effect, it does minimize negative consequences by rotating the data distribution such that the maximal number of data points lie on the zero of the log ratio axis. In this respect, the Zipf's law normalization technique can be considered inferior to the quantile method, however, it may still be a useful tool where the quantile method cannot be applied. One such case, in which quantile normalization is inappropriate, is with so called 'boutique' microarrays where the genes spotted on the array represent a selected set of genes, for example from a specific pathway or those involved with a particular biological process or disease state. In such systems, most genes are expected to be differentially regulated when control and experimental samples are compared and the expected data distribution of control samples may be significantly different than that of experimental samples (in mean intensity for example). The quantile normalization method would effectively remove this difference by replacing the data distribution of each microarray with the mean distribution of all arrays. In contrast, the principle of normalization according to Zipf's law can also apply to arrays of this type if a group of control spots are included on the microarray. These control spots could be an external reference probe which hybridises to a concentration gradient of matching spots on the array, or internal positive (highly expressed genes) and negative (spotting buffer) control spots on the microarray, or an a priori selected set of housekeeping genes using a method such as that described by Wilson et. al. [ 4 ] or Schadt et. al. [ 25 ]. A linear model can be fitted to the control spots alone, and the normalization procedure can then be applied using the control spots as a proxy for the entire data distribution. The critical assumption in using control spots in normalization is establishing their relationship to the experimental spots. The results of our comparison between methods which are designed to normalize boutique microarray data show that Zipf's law normalization using internal control spots results in a relatively well normalized data set when compared to Zipf's law normalization using selected housekeeping genes and the modified loess method using selected housekeeping genes. In addition, the Zipf's law method produced data distributions which are more similar between arrays allowing for between-array comparisons which are advantageous in terms of both cost, because of the reduced number of microarrays that need to be run, and, statistical power, by allowing for greater numbers (n), experimental design permitting. Conclusion In summary, we examined the applicability of using Zipf's law as the basis for a novel normalization technique, which is applicable to both one channel microarray data and two channel microarrays. This method is shown to out-perform such methods as global normalization to the mean but would appear to be inferior to quantile normalization. The quantile method was superior to Zipf's law in removing intensity dependent effects commonly seen in microarray data. While the latter method cannot be applied to boutique arrays, we show that the Zipf's normalization method used with internal positive and negative controls or with selected housekeeping genes normalizes boutique arrays as well as currently existing methods. Additionally, data normalized with the Zipf's method using internal control spots seems more amenable to between-array gene intensity comparisons when compared to other methods. Methods Data acquisition Data set A (Table 1 ) was generated using a global genome-wide cDNA clone set (Human UniGene clone set RZPD 1 Build 138, NCBI [ 26 ]), which consisted of ~33,792 cDNA clone inserts spotted in duplicate onto membranes [ 16 ]. These microarrays (n = 31) were hybridized with 33 P-labeled cDNA derived from total RNA extracted from biopsy material from the sigmoidal colon of normal (control, n = 11), and patients with Crohn's disease (condition A, n = 10) and ulcerative colitis (condition B, n = 10). To emphasize that our normalization technique can be used to normalize other array systems, the second array set used was a smaller, but widely used, commercially available microarray system. Data set B (Table 1 ) was generated by using Atlas Rat cDNA microarrays (Clontech, 588 genes) probed with rat brain tissue, from control (cerebellum n = 10, olive n = 10) and harmaline treated (cerebellum n = 10, olive n = 9) animals. A third microarray data set, data set H (Table 1 ) was included to demonstrate the normalization method on two channel fluorescent based (Cy3/Cy5) oligonucleotide systems. These custom produced boutique microarrays (n = 5) contained 1024 spots, and were used in a study to identify differences in apoptotic mechanisms in two different mouse cell lines. Microarrays were probed according to established protocols and exposed to imaging plates overnight (BAS-MS 2325) and scanned at a 50 μm resolution on a FLA-3000G phosphoimager (Raytest, Germany). Image gridding was carried out using VisualGrid ® software [ 27 ], and intensity data was stored in a relational database and normalized and analyzed using database stored procedures and Perl scripts. All data was normalized from raw data, no background subtraction or other inter-array normalization was performed. Plots were generated using the Grace software package [ 28 ]. Normalization Normalization was accomplished by transforming the data such that the coefficient c and proportionality of the Zipf's power function (formula 1) are identical for all microarrays. This is easily achieved using a regression model on the log e intensity versus log e rank transformed data, which has the general form, ln ( y ) = a + b ln ( r ) + e (2) where y is the intensity, r is the rank, a is the regression constant (corresponding to proportionality in Zipf's power function), b is the regression coefficient (corresponding to the coefficient c in Zipf's power function), and e is an error coefficient, which is assumed to be normally distributed. The first step in this three step procedure was to compute the median intensity of each gene over all microarrays to establish ranks, which were used as the 'reference' to which all microarrays were normalized. This was done by taking the median intensity ( y med ) of each gene, over all microarrays on which it was measured, and sorting the resulting list of medians to obtain their median ranks ( r med ). The regression model (2) is applied to the log e median intensities and their ranks to estimate a med and b med using the least squares method, The ranking of genes by their median intensities effectively groups genes of similar overall expression level along the log rank axis. Under the assumptions that most genes are not differentially expressed, the reference curve generated from the median intensities should have an identical regression coefficient and constant to that of each individual microarray plotted using the ranks determined by the medians. For the genes which are differentially expressed, the median value represents a 'center' around which expression levels on each individual array may vary, and the neighbouring (by rank) genes, which do not (or only slightly) vary, act to stabilize the regression line and allow normalization to be performed. In the second step of the normalization procedure, the regression model was applied individually to each microarray using the same ranking as the reference curve, This results in a set of coefficients a k and b k which are estimated individually for each array using the least squares method, where k is equal to the number of microarrays in one channel systems, and equal to 2 time the number of microarrays (one for each channel) in two channel systems. Data from two channel arrays were treated in the same way as one channel systems, i.e. each channel was treated independently. In the third step, the difference between the expected gene intensity value on the k th array and that of the reference curve was applied as the normalization factor, A scaling factor was applied to the raw data before normalization such that the values y k , and were always greater than one to avoid negative values after log transformation. After normalization, the same scaling factor was applied to the data to back transform to their original magnitude. For example, if the smallest raw value in the data set was 0.1, the unlogged raw data was multiplied by a scaling factor of 10 before normalization, and the unlogged normalized data was divided by the same scaling after normalization. In the special case of our third microarray data set (see Methods: Data Acquisition) which was a boutique array, the same procedure as described above was applied with the following modifications. Each microarray contained 32 spots each of internal positive controls (GAPDH, glyceraldehyde-3-phosphate_dehydrogenase) and internal negative controls (spotting buffer). The medians of all gene intensities were computed (including internal positive and negative controls), and median ranks were assigned as described. However, only the medians of the 64 internal control spots were used to estimate a med and b med , and only the 64 internal control spots from each array were used to estimate a k and b k . In both cases, the ranks generated from the entire data set, were used. The normalization factor was then applied over the entire data set as described above. An alternative to the used of internal control spots for the normalization of boutique microarrays was also explored. Wilson, et. al. [ 4 ] described a method wherein a set of 'housekeeping' genes is selected a priori from the data set by virtue of their low variance in intensity and such that the entire range of intensities observed on the microarrays is uniformly represented. We also applied the Zipf's law normalization technique to our boutique microarrays using the set of housekeeping genes selected using the method of Wilson, et. al. In addition to the normalization method based on Zipf's law, all data sets were normalized to a global mean (the mean of logged intensities from all microarrays) and the quantile method. The quantile method is applied by ranking the genes in each array by intensity, taking the median intensity at each rank, and replacing each gene intensity with the median intensity corresponding to the same rank. All normalization methods were compared to each other and to the raw data distribution using box plots and MA plots (pairwise array comparisons of the log-intensity ratio (M) to the mean log-intensity (A)). The two channel boutique microarray data set allowed further normalization methods not possible on one channel array systems to be applied. We normalized this data set using the popular loess method [ 19 ], and a modified Loess method specifically designed for boutique arrays using selected housekeeping genes described by Wilson, et. al. [ 4 ]. Software The Zipf's normalization procedure was initially implemented as an SQL stored procedure in a relational database. However, because this is not easily transferable to other systems, we provide two further implementations, a Perl script and an Excel macro [see Additional files 4 , 5 ]. Implementations are available for download from our website [ 29 ] and as additional files accompanying this paper. Both the Perl script and Excel macro implement matrix algebra style computation, using either built-in functions or the Perl PDL module [ 30 ]. Normalization of two channel arrays with the loess method was performed using the marray package from R's Bioconductor [ 4 ]. Loess normalization using selected housekeeping genes and the selection of the housekeeping genes themselves was done with the tRMA package [ 19 ] which is publicly available for download on the internet. Sample data sets are also provided with this paper [see Additional files 6 , 7 , 8 ]. Normalization method comparison To compare and evaluate the effectiveness of the various normalization methods applied in this paper, several well established methods were used along with some less common techniques. MA plots [ 19 ] are a convenient way to examine differences in fluorescent marker efficiency and other dye effects in two channel microarray systems. In addition to the standard practice of generating within-array MA plots, we apply them additionally to one channel systems and between arrays in two channel systems to evaluate the extent to which a normalization procedure allows for multiple pairwise comparisons between microarrays. Plots of log e intensity versus log e rank fitted with linear regressions are a way to visually evaluate the normalization procedure according to the criteria of the Zipf's Law normalization. Specifically, all arrays have identical coefficients c and proportionality for the Zipf's power function when the slops and y-intercepts of the regression lines are identical. Finally, to quantify the similarity between microarray distributions after normalization, the mean Kolmogorov-Smirnov value was calculated over all possible pairwise combinations of microarrays within an experiment. In the case of two channel arrays, the mean of within-array Kolmogorov-Smirnov values was also computed (n = the number of arrays). It should be emphasized that even though the Kolmogorov-Smirnov values are technically a test statistic, no statistical test is performed. The values are here used only as a measure of similarity between microarray distributions. Microarray platform comparison The underlying premise of the Zipf's normalization method is that microarray data distributions follow a power law distribution such that the relationship between the log intensities and the log ranks is clearly linear. While this assumption holds true for the three data sets we present in this paper, to evaluate the general applicability of the method we also examined eight publicly available data sets (Table 1 , data sets C-G, I, K-L) from the NCBI Gene Expression Omnibus [ 18 ], and one unpublished data set from an independently maintained website [ 31 ] (Table 1 , data set J). The survey contains a variety of microarray system types (cDNA vs. Oligo based, radioactivity vs. dye labeled systems, academic vs. commercially produced) and two SAGE experiments for comparison. Two plots were generated for each data set to ascertain the conformity to the Zipf's power law distribution and the log normal distribution respectively. For each data set, a representative array was constructed by ranking the intensities within each array, and then mean over ranks were taken. To determine how well data sets follow the Zipf's power law distribution, log intensity vs. log rank plots were constructed and linear regressions were performed. Data distributions, which were very linear in form, closely follow the power law distribution. A second plot of the distribution of (log y – μ) / σ, where y is the mean intensity over ranks, and μ and σ 2 are the mean and variance, was made for each data set to visualize the conformity to log normal distribution. List of abbreviations EST – Expressed Sequence Tag MA – log ratio (M) vs. mean log intensity (A) NCBI – National Center for Biotechnology Information RZPD – Deutsches Ressourcenzentrum für Genomforschung GmbH SAGE – Serial analysis of gene expression SQL – Structured Query Language Authors' contributions TL conducted the data analysis and implementation of algorithms, participated in the development of the normalization method and is principle author of this manuscript. CMC generated the Unigene and Clontech microarray data set, participated in the development of the normalization method and participated in manuscript preparation. PJPC conceived of and participated in the development of the normalization method. RH participated in the generation of microarray data sets and participated in the development of the normalization method. GD conceived of and coordinated neurology related aspects of this study. SS conceived of and coordinated gastrointestinal related aspects of this study. Supplementary Material Additional File 1 Clontech microarray log plots Five rat Clontech microarrays from the panel of thirty-nine microarrays probed with rat-brain tissue. Upper left to lower right: a . Log e median gene intensity vs. log e rank – conformity to Zipf's law is demonstrated by the linear regression line (in red) b . Five microarrays chosen to maximize pre-normalization variability, each plotted according to the gene ranks determined by their median gene intensity levels. c . The same five microarrays, normalized to a global median, with regression lines. d . The same five microarrays, normalized with the quantile method, with regression lines. e . The same five microarrays normalized taking Zipf's law into account, with regression lines. For plots b-d, a sub-sample of 50% of the data points are plotted for readability. Click here for file Additional File 2 Mean of squared log ratios from MA plots in Figure 2 In Figure 2 , it is difficult to see that the distribution of the Zipf's normalized data is more closely centered around zero on the log ratio axis than the Globally normalized data. To quantify this, the mean of squared log ratios was computed for each MA plot. The positions of the values in this table correspond exactly to the positions of the plots in Figure 2 . In 6 out of 8 cases, the mean of squared log ratio is smaller in the Zipf's normalized data than in the corresponding Globally normalized data. Click here for file Additional File 3 Boutique microarray log plots Four mouse apoptosis boutique microarrays used in the mouse cell line experiments. This is the same data set as shown in Figure 4 , with the array containing one channel with low expression intensities and high variability removed. Upper left to lower right: Log e median gene intensity vs. log e rank – a . Normalized according to Zipf's law, using internal positive and negative controls as proxies for the whole data set. b . Normalized with a loess curve fit using a selected set of housekeeping genes as proxies (see Methods). c . Normalized according to Zipf's law, using the same selected set of housekeeping genes as in b. as proxies d . The raw data. e . For comparison purposes only, normalized using the quantile method. f . For comparison purposes only, normalized using the standard loess method. Click here for file Additional File 4 Requires: Microsoft Excel (Does not handle missing data values.) Click here for file Additional File 5 Requires: Perl (which runs on many platforms), the PDL perl module (Handles missing data values if PDL is compiled correctly.) Click here for file Additional File 6 Microarray type: Filter based cDNA from the RZPD Number of genes: 33,792 Number of microarrays: 31 Probed with: Total RNA from human sigmoidal colon. Within microarray normalization: None Click here for file Additional File 7 Microarray type: Clonetech Atlas Rat cDNA 7738-1 Number of genes: 558 Number of microarrays: 33 Probed with: Total RND from rat cerebellum and olive. Within microarray normalization: None Click here for file Additional File 8 Microarray type: custom made glass slide Number of genes: 1024 Number of microarrays: 5 Probed with: Total RND from mouse cell lines. Within microarray normalization: None Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC555536.xml
523885	An Immune Basis for Lung Parenchymal Destruction in Chronic Obstructive Pulmonary Disease and Emphysema	ABSTRACT Background Chronic obstructive pulmonary disease and emphysema are a frequent result of long-term smoking, but the exact mechanisms, specifically which types of cells are associated with the lung destruction, are unclear. Methods and Findings We studied different subsets of lymphocytes taken from portions of human lungs removed surgically to find out which lymphocytes were the most frequent, which cell-surface markers these lymphocytes expressed, and whether the lymphocytes secreted any specific factors that could be associated with disease. We found that loss of lung function in patients with chronic obstructive pulmonary disease and emphysema was associated with a high percentage of CD4 + and CD8 + T lymphocytes that expressed chemokine receptors CCR5 and CXCR3 (both markers of T helper 1 cells), but not CCR3 or CCR4 (markers of T helper 2 cells). Lung lymphocytes in patients with chronic obstructive pulmonary disease and emphysema secrete more interferon gamma—often associated with T helper 1 cells—and interferon-inducible protein 10 and monokine induced by interferon, both of which bind to CXCR3 and are involved in attracting T helper 1 cells. In response to interferon-inducible protein 10 and monokine induced by interferon, but not interferon gamma, lung macrophages secreted macrophage metalloelastase (matrix metalloproteinase-12), a potent elastin-degrading enzyme that causes tissue destruction and which has been linked to emphysema. Conclusions These data suggest that Th1 lymphoctytes in the lungs of people with smoking-related damage drive progression of emphysema through CXCR3 ligands, interferon-inducible protein 10, and monokine induced by interferon.	Introduction Chronic inhalation of tobacco smoke causes progressive lung destruction in susceptible individuals, resulting in chronic obstructive pulmonary disease (COPD) and emphysema, two well-described clinical syndromes with poorly understood pathogenesis [ 1 , 2 , 3 ]. A role for T helper cells in the pathogenesis of obstructive lung disease has been established with asthma, where T helper 2 (Th2) cells are strongly linked to both human and experimental disease [ 4 , 5 , 6 , 7 ]. A potential role for T cells in COPD has also been suggested in several recent studies that show CD8 + T cells are increased in the lungs of people who smoke [ 8 , 9 , 10 , 11 ]. T cells cause tissue injury through their secreted products such as cytokines; in mice, overexpression of interleukin (IL)-13, a T cell cytokine that is strongly implicated in the pathogenesis of experimental asthma, resulted in increased production of proteases and enlargement of airspaces reminiscent of emphysema [ 12 ]. Further, airway limitation, another characteristic of human asthma, is clinically linked to an accelerated rate of loss of lung function in smoker individuals [ 13 ]. It has been suggested, therefore, that asthma and COPD may involve the same type of recruited inflammatory cells, differing only in their location within the lung [ 14 ]. Chemokines, their receptors, and cell adhesion molecules regulate migration of immune cells into inflamed tissue [ 15 , 16 , 17 , 18 ]. T helper 1 (Th1) cells have been shown to secrete interleukin 2 and interferon gamma (IFN-γ), and express a distinct repertoire of chemokine receptors such as CCR5 and CXCR3 [ 19 , 20 , 21 ]. In contrast, Th2 cells that are biased to produce IL-4 and IL-5 express mainly CCR4 and CCR3 [ 22 , 23 , 24 , 25 ]. Immunofluorescent analysis of airway mucosal biopsies in patients with asthma showed that most T cells co-express IL-4 and CCR4, but, in contrast, T cells in airways of patients with COPD and pulmonary sarcoidosis produce IFN-γ and express high levels of CXCR3, while lacking CCR4 expression [ 26 ]. In addition to T cells, a wide variety of other inflammatory cells have been shown to express distinct chemokine receptors that are critical for their homing, suggesting a universal mechanism for regulating immune responses. Interferon-inducible protein 10 (IP-10), monokine induced by interferon gamma (MIG), and interferon-inducible T cell alpha chemoattractant (I-TAC) are three known ligands for CXCR3 produced by normal and injured epithelial cells and T cells that are required for homing of Th1 cells [ 27 , 28 , 29 ]. In addition to regulation of chemotaxis and homing, other functions have been ascribed to chemokines, including modulation of T cell fate by direct effects on differentiating T cells, and regulation of proteolysis in blood monocytes [ 19 , 30 ]. In this study we determined the dominant T helper phenotype in lung samples from ex-smoker individuals with moderate to severe COPD and emphysema and control individuals with no evidence of smoking-related lung disease. Analysis of chemokine receptor expression on isolated peripheral lung lymphocytes from ex-smokers with COPD/emphysema indicated that both CD4 and CD8 T helper cells are strongly polarized to the Th1 phenotype compared to T cells isolated from lung tissue of normal individuals or individuals with non-smoking-related obstructive lung disease. The same cells spontaneously secreted more IFN-γ and CXCR3 receptor ligands MIG and IP-10 in the COPD and emphysema group than in the group without emphysema. Further, IP-10 and MIG, but not IFN-γ, upregulated macrophage metalloelastase (matrix metalloproteinase [MMP]-12) from isolated lung macrophages. Together, our findings reveal the strong association between COPD/emphysema- and Th1-driven adaptive immunity, suggesting a link to lung destruction mediated by IFN-γ, MIG, and IP-10. Methods Participants Twenty-eight non-atopic ex-smoker individuals (see Table 1 ) undergoing medically necessary lung resection were serially entered into the study: ten individuals with no COPD and no evidence of emphysema (control group) and eighteen individuals (diseased group) with moderate to severe COPD and evidence of emphysema as determined by pulmonary function tests, high-resolution computed tomography (CT), or conventional CT scan. All participants were ex-smokers who had quit smoking for a mean (SD) of 7 (2) y and 4 (2) y in COPD/emphysema and control groups, respectively. COPD was diagnosed according to the criteria recommended by the National Institutes of Health/World Health Organization workshop summary [ 31 ]. Participants in the control and COPD/emphysema groups had similar (mean [SD] of 54 [ 6 ] and 45 [ 5 ], respectively) “pack-year” smoking histories, where smoking one pack of cigarettes per day each year is defined as one pack-year. Table 1 Clinical and Demographic Characteristics of Participants a p -Values are for the comparison of emphysema with control participants (Student's T test) b Histological diagnoses included lipoma, papiloma, and benign scar LVRS, lung volume reduction surgery; SE, Standard Error; percent FEV1 l/s, percent predicted FEV1 in liters/second; QT, the number of years since the individuals had stopped smoking All participants were recruited from the surgical clinic at the Michael E. DeBakey Veterans Affairs Medical Center and the Methodist Hospital, and were undergoing lung resection for diagnostic or therapeutic purposes ( Table 1 ). Study protocols were approved by the institutional review board for human studies, and informed consent was obtained from all participants. Participants had no history of allergy or asthma and had not received oral/systemic corticosteroids during the last 6 mo. At the time of study, all participants had been free of acute symptoms suggestive of upper or lower respiratory tract infection for the 6 wk preceding the study. CT-Based Evaluation for Emphysema High-resolution CT (two in emphysema group and two in control group) or conventional CT analysis was used to detect emphysema, characterized by the presence of areas of low attenuation contrasted with surrounding normal lung parenchyma [ 32 , 33 ]. CT scans were used by a radiologist to separate participants on the basis of the presence or the absence of any objective evidence for centrilobular, panacinar, or paraseptal emphysema with a detection limit of greater than 3-mm low attenuation density [ 34 ]. Isolation of Lung Lymphocytes Lung lymphocytes were isolated by modifying established protocols, using a combination of mechanical fragmentation, enzyme digestion, and centrifugation procedures described previously [ 35 , 36 , 37 ]. Viable lymphocytes were separated from whole lung inflammatory cells (macrophages, eosinophils, and neutrophils) using an immunomagnetic positive separation technique (autoMACS, Miltenyi Biotec, Auburn, California, United States). Briefly, lung leukocytes were labeled with paramagnetic bead-conjugated anti-CD3, -CD19, and -CD56 to positively select T, B, and NK cells, according to the manufacturer's instructions. Each of the harvested cell populations was used directly for in vitro assays or was cryopreserved in aliquots of 1 × 10 7 cells for future analysis. Antibodies The following monoclonal antibodies were purchased from BD Biosciences Pharmingen (San Diego, California, United States): FITC-, Cy5-, and PE-conjugated anti-CD4, -CD8, -CD3, -CD14, -CD69, -CXCR3, -CCR3, -CCR4, and -CCR5. For enzyme-linked immunosorbent assay studies, anti-human antibodies to IFN-γ, IL-4, IP-10, MIG, I-TAC, and the appropriate secondary reagents were purchased from R&D Systems (Minneapolis, Minnesota, United States). Quantification of Polarized Peripheral Blood and Lung Lymphocyte Subsets Phenotypic characterization of T cells was done by two-color flow cytometry (Epic XL FL, Beckman Coulter, Allendale, New Jersey, United States) using combinations of the following monclonal antibodies: FITC-conjugated anti-CD4, -CD8, and -CD14; PE- and Cy5-conjugated anti-CCR4, -CCR3, -CCR5, and -CXCR3. Freshly isolated lung lymphocytes were resuspended to 1 × 10 7 cells/ml, and 50 μl of cells was incubated with antibodies to CD3 and CD4 or CD8. Intracytoplasmic Cytokine Staining Lung lymphocytes were cultured in the presence or absence of phorbol myristate acetate (PMA)/ionomycin and brefeldin A for 12 h. Cells were harvested, fixed with formaldehyde, permeabilized with saponin, and intracellularly labeled for IFN-γ and IL-4, in addition to staining for surface CD69, CD4, and CD8 according to the manufacturer's recommendations (Fastimmune, BD Biosciences Pharmingen). In Vitro T Cell Culture and Cytokine Assay Lung lymphocytes were isolated from surgical tissue and cultured in vitro in triplicate for 4 d. Supernatants were collected and stored at –80 °C for future analysis. Standard antibody-based enzyme-linked immunosorbent assay was used to measure supernatant concentrations of IP-10, MIG, IL-4, and IFN-γ according to the manufacturer's instructions (R&D Systems and BD Biosciences Pharmingen). Detection of MMP12 by Western Blotting, and Real-Time PCR Peripheral blood mononuclear cells and lung macrophages were isolated by positive selection using immunomagnetic beads conjugated with anti-CD14, and cultured in serum-free medium (RPMI, L-glutamine, and Pen/Strep) prior to overnight stimulation with 0, 50, 250, or 500 ng/ml of IFN-γ, IL-4, MIG, I-TAC, and IP-10. Supernatants were collected, and MMP12 was detected using anti-human MMP12 (R&D Systems) by Western blotting according to the manufacturer's instructions. Total cellular RNA was extracted from CD14 + lung macrophages stimulated overnight with rIP-10 (500 ng/ml) in the presence or absence of blocking anti-CXCR3 antibodies (5 μg/ml, R&D Systems). Two-step real-time reverse transcription PCR was used to determine the relative expression of mRNA using the ABI Perkin Elmer Prism 5700 Sequence Detection System (Applied Biosystems, Foster City, California, United States) as described previously [ 38 ]. Immunostaining and Histopathology Paraffin-embedded, and fresh-frozen lung sections (5 μm) were immunostained using monoclonal antibodies against human MMP12 (R&D Systems) or non-immune antisera by an immunoperoxidase protocol (Vectastain Elite, Vector Labs, Burlingame, California, United States) and counterstained with hematoxylin as recommended by the manufacturer. Statistical Analysis The Mann-Whitney test (non-parametric, two-tailed) and Student's T test (two-tailed) were used to compare differences between the two groups of subjects. p < 0.05 was considered statistically significant. Results Th1 Immune Bias of Peripheral Lung Lymphocytes in Emphysema Inflammatory chemokines, cytokines, and their receptors are upregulated at sites of inflammation and play a key role in the recruitment of leukocytes to peripheral tissues in response to injury [ 17 , 39 ]. To detect Th1 polarization, we assessed lung lymphocytes for expression of CCR5 (a receptor for several Th1 chemokines) and CXCR3 (the receptor for IP-10, I-TAC, and MIG). We screened for the presence of Th2 cells by assessing T cell expression of CCR4—a receptor for eotaxin/CCL11, macrophage chemoattractant protein 3 (CCL7), and thymus- and activation-regulated chemokine (CCL17) [ 40 , 41 ]—and CCR3, a receptor for eotaxin and related chemokines. Flow cytometry revealed very low expression of CCR3 and CCR4 (1%–3%) in control ( n = 10) and emphysema ( n = 18) groups, and did not discriminate between these populations ( Figure 1 A and 1 B; data not shown). These findings were in sharp contrast to the enhanced expression of both CCR5 and CXCR3, as shown in representative histograms ( Figure 1 A). These Th1-specific chemokine receptors were expressed prominently on lung lymphocytes from all participants, but their expression was significantly enhanced in the setting of emphysema ( Figure 1 A– 1 C). Further, both CD4 and CD8 T cells expressed CCR5 at the same level ( Figure 1 C). In contrast, we found highly variable expression (0.5%–30%) of CCR4, CXCR3, and CCR5 on peripheral blood lymphocytes isolated from the same participants, and this variation did not correlate with the presence of disease in either group (data not shown). Furthermore, we compared the lung lymphocyte CCR5 and CXCR3 profiles among the eight participants with emphysema alone (lung volume reduction surgery for emphysema; non-cancer) and ten participants with emphysema and accompanying cancer (lung resection for treatment of small peripheral cancer), and found that these two groups cannot be distinguished based on these indices ( Figure 1 D; data not shown). Figure 1 Chemokine Receptor Expression on Peripheral Lung Lymphocytes (A) Single color histograms showing expression of chemokine receptors CCR4, CCR5, and CXCR3 from representative control and emphysema participants. (B) Pooled data from all participants (control, n = 10; emphysema, n = 18) showing percent (median ± SD) of total lung lymphocytes expressing CCR4 and CCR5. (C) Pooled data from same participants showing percent (median ± SD) CCR5 expression on CD4 (top) and CD8 (middle) T cells, and CXCR3 expression on unfractionated T cells (bottom) from the same participant groups. (D) Analysis of total lung lymphocyte chemokine receptor (median ± SD) profiles among participants with emphysema. Participants had either (1) lung volume reduction surgery for emphysema (non-cancer, n = 8) or (2) lung resection for treatment of small peripheral cancer ( n = 10). Participants showed similar inflammatory indices as determined by CCR5 expression. In (B) and (C), , p < 0.001; †, p = 0.01; ‡, p = 0.02; ∫, p = 0.007 (Mann-Whitney test) for the comparison of emphysema and control groups. Although human lung macrophages are not known to express CXCR3, we suspected based on the immunohistochemical localization of this chemokine receptor that CD14 + cells in the lungs of ex-smoker individuals with emphysema accounted for much of the total lung CXCR3 + immunoreactivity ( Figure 2 A; data not shown). To confirm this, we determined the percent of total lung cells expressing CD14 and CD11b—which are both markers of monocytes/macrophages—and CXCR3. We found that over 40% of CD14 + cells from participants with emphysema but not control participants were also positive for CXCR3 ( Figure 2 ). In addition, there was a significant negative association between CXCR3 expression on lung T cells and the percent of predicted forced expiratory volume in 1 s (FEV1), based on an R 2 goodness-of-fit statistic of 0.27 ( Figure 2 C; p = 0.0089, r = −0.52). Together, these data indicate that a strong type 1 bias is characteristic of the T cells isolated from the peripheral lung of participants with COPD and emphysema and that this immune phenotype correlates with the lung destruction that is characteristic of this disease. Further, we have shown for the first time, to our knowledge, that CXCR3 expression, a marker of Th1 inflammation, extends to lung monocytes and macrophages. Figure 2 Expression of CXCR3 in Lungs of Control and Emphysematous Smoker Individuals (A) Representative forward and side-scatter characteristics of whole lung cells from a participant with COPD and emphysema. Anti-CD11b PE-conjugated and anti-CD14 FITC-conjugated antibodies detect lung macrophages (middle), and histogram of mean fluorescence intensity showing anti-CXCR3-Cy5 and control antibodies (cIg) detects lung macrophages in the patient with emphysema. (B) Pooled data from control individuals without ( n = 5) and with ( n = 8) emphysema. Columns are median, bars represent SD. , p = 0.009 (Mann-Whitney test) for the comparison of emphysema and control participants. (C) Negative association between CXCR3 expression on CD3 + T cells and FEV1 percentage predicted based on an R 2 goodness-of-fit statistic of 0.27 ( p = 0.0089, r = −0.52, n = 24). IFN-γ, IP-10, and MIG But Not IL-4 Are Expressed by Lung Lymphocytes We sought additional functional data to confirm the apparent Th1 bias of peripheral lung inflammatory cells isolated from ex-smoker individuals. Freshly isolated lung lymphocytes that were not otherwise manipulated secreted high levels of IFN-γ, MIG, and IP-10, with significantly greater secretion of both cytokines from lymphocytes of participants with emphysema ( Figure 3 A– 3 C). Interestingly, we could not detect appreciable amounts of I-TAC, another known ligand for CXCR3, in lung lymphocytes of control participants or those with emphysema (data not shown). Similar results were obtained using intracytoplasmic cytokine staining of the same cells ( Figure 3 D), in which PMA/ionomycin stimulation strongly induced IFN-γ production from CD69 + /CD8 + lung lymphocytes. Surface staining for CD4 was not feasible with this protocol; however, the percentage of CD8 − /IFN-γ + cells was approximately equal to that of CD8 + /IFN-γ + cells (median [SD], 19[ 6 ] versus 16[ 4 ], respectively). Because total numbers of CD4 + and CD8 + T cells were approximately equivalent, this suggests that non-CD8 + /IFN-γ + cells are largely CD4 + , and therefore Th1 cells. Finally, the typical Th2 cytokine, IL-4, was not detected in either group, as determined by enzyme-linked immunosorbent assay or intracytoplasmic cytokine staining ( Figure 3 E; data not shown), confirming the marked Th1 bias of the immune response that underlies smoking-related lung inflammation and emphysema. Figure 3 IFN-γ, MIG, and IP-10 Production by Isolated Lung Lymphocytes (A–C) Lung lymphocytes from control individuals and participants with emphysema were cultured without additional stimulation for 3 or 4 d and assessed for secretion of (A) IFN-γ, (B) MIG, and (C) IP-10 (control, n = 8; emphysema, n = 12). Columns are median, bars represent SD. , p= 0.007; †, p = 0.01; ‡, p = 0.02 for the comparison of emphysema and control participants. (D) The same cells from a representative ex-smoker individual with emphysema were either left unstimulated (No ST) or treated with PMA/ionomycin (PMA/I) for 24 h and assessed for surface CD8 and CD69 expression and the intracytoplasmic accumulation of IFN-γ by flow cytometry. (E) Production of IL-4 by lung lymphocytes. Lung lymphocytes from a representative ex-smoker individual with emphysema were cultured for 24 h with or without PMA/ionomycin stimulation (PMA/I) and assessed for intracytoplasmic IL-4 and IFN-γ accumulation by flow cytometry. IP-10 and MIG But Not IFN-γ Directly Upregulate MMP12 through CXCR3 Emphysema and irreversible airway limitation that is characteristic of chronic tobacco smoking are related to the destruction of elastin and the resulting loss of lung elastic recoil. Therefore, to be relevant to the pathogenesis of airway obstruction, type 1 inflammation must be shown to promote lung elastolysis. Because loss of elastin is regulated by proteinases [ 42 ], we next determined if expression of MMPs, in particular the elastases MMP9 and MMP12, was regulated by IP-10, MIG, and IFN-γ, the principal cytokines detected in emphysematous lung. Indeed, isolated peripheral lung macrophages, but not isolated blood monocytes, secreted MMP12 in response to IP-10 and MIG, but not IFN-γ ( Figure 4 A; data not shown). These findings reflect a specific receptor–ligand interaction because in the presence of a CXCR3 function-blocking antibody, IP-10 failed to induce MMP12 ( Figure 4 B). Furthermore, immunohistochemical studies revealed that lung macrophages of participants with emphysema, but not control participants, specifically express MMP12 ( Figure 4 C and 4 D). Together, these findings indicate that Th1, but not Th2, cytokines and related chemokines are required for establishing the pro-elastolytic lung environment that underlies human emphysema. Figure 4 Regulation of MMP12 by Type 1 Cytokines (A) CD14 + , lymphocyte-depleted lung leukocytes were cultured with and without the indicated amounts of recombinant human IP-10 and IFN-γ, and supernatants were assessed for the presence of MMP12 by Western blotting. (B) Fold increase relative to unstimulated of MMP12 mRNA from lung macrophages stimulated without ( – ) and with ( + ) 500 ng/ml of IP-10 in the presence or absence of a function-blocking antibody to CXCR3 as determined by real-time PCR. (C and D) Lung tissue from a participant with emphysema (C) shows strong immune staining for MMP12 localized to macrophages (arrows), and (D) shows lung tissue from a control participant without emphysema and with undetectable MMP12. The insets show a high-power view of lung macrophages staining positive (C) and negative (D) for MMP12 (×60) , p = 0.04. Discussion In this investigation, we characterized T cells and lung macrophages isolated from emphysematous and non-emphysematous human lungs. Three principal findings emerge from our study. First, rather than being functionally diverse, as suggested by the heterogeneous nature of humans, lung T cells of ex-smoker individuals with emphysema are relatively homogeneous and characterized by a marked Th1 bias. Second, the principal Th1 chemokines, MIG and IP-10, are linked to a pro-elastolytic lung environment because these cytokines upregulate the elastase MMP12, which is associated with emphysema. Finally, we found no significant expression of Th2 chemokine receptors, such as CCR3 and CCR4, or IL-4 production in lung lymphocytes. Together, our findings demonstrate the role of the adaptive immune response in COPD and suggest a primary role for Th1 cells in controlling the main smoking-related physiologic and structural changes of the lung. Upregulation of CCR5 and CXCR3 on T cells and accumulation of these cells in the lung periphery suggest that aberrant, unremitting pulmonary recruitment of these activated T cells is unique to people with smoking-related lung disease, despite cessation of exposure to the inciting agent, tobacco smoke. We showed that ex-smoker individuals without obstructive lung disease or emphysema have comparatively little Th1-biased inflammation in their lungs; thus, our findings reflect the inflammatory changes that are unique to the COPD microenvironment. Additionally, lung lymphocytes isolated from four lifelong non-smoker individuals with severe obstructive lung disease due to cystic fibrosis or bronchiolitis obliterans did not show a Th1 inflammatory bias of the lung (S. Grumelli, F. Kheradmand, D. B. Corry, unpublished data). This information confirmed our finding that the predominant Th1 bias in COPD/emphysema reflects the microenvironment unique to the lungs of ex-smoker individuals. The prevalence of asthma among people who smoke is currently not known, but in order to study COPD/emphysema in a population without other confounding variables, people who might have had asthma were excluded, and thus our findings are restricted to non-asthmatic individuals with emphysema. Our use of T cell chemokine receptor expression analysis to determine recruitment of lung T cells is not without precedent. Analysis by immunohistochemistry of airway mucosa of people with atopic asthma after antigen challenge revealed that large numbers of CCR4 + and CCR8 + T cells express IL-4, and CCR4 expression was prominent in people with severe atopic dermatitis, which decreased upon abatement of disease activity [ 26 , 43 ]. Immunostaining of T cells in synovial fluid from individuals with rheumatoid arthritis showed that virtually all of the T cells associated with inflamed joints expressed CXCR3 and CCR5, representing significant enrichment compared to blood T cells from the same participants. Furthermore, previous studies of smoker individuals with COPD and normal lung function showed the presence of CD8 + /CXCR3 + T cells in the airway epithelium and submucosa [ 44 ]. We extend these findings by showing CXCR3 expression on lung macrophages and CD4 + T cells in emphysema patients and the functional interplay between Th1-related chemokines and elastolytic MMPs. In addition to detailing surface chemokine receptor expression, we have functionally confirmed the marked Th1 bias of peripheral lung T cells, demonstrating that either at rest or following stimulation, these cells secrete IFN-γ and not IL-4. Our findings therefore confirm the utility of chemokine receptor expression patterns in the initial assessment of T cell effector phenotype. Destruction of lung parenchyma in emphysema is thought to occur through excessive proteolysis mediated by the elastin-degrading enzymes MMP2, MMP9, and MMP12 from the MMP family, and by neutrophil elastase from the serine proteinase family [ 45 , 46 ]. Cytokines and chemokines are substrates for MMPs, but they also regulate expression of MMPs under pathological conditions [ 47 , 48 ]. We have shown here that IP-10 and MIG, two chemokines that are secreted from lung lymphocytes of participants with emphysema, upregulate specifically MMP12 and thus favor a proteolytic microenvironment that facilitates lung destruction. Strengthening the association between lung macrophages and IP-10/MIG-dependent MMP12 secretion is the fact that we have demonstrated that in humans macrophages, like T cells, express CXCR3 and that this receptor is required for MMP12 secretion in response to IP-10/MIG stimulation. In addition to defining the predominant immune phenotype of emphysematous lung, these additional findings implicate the principal cell (macrophage), MMP (MMP12), and effector cytokines (IFN-γ, IP-10, and MIG) as likely underlying smoking-induced lung destruction. We have further shown that these enzymes may be regulated by proximal immune events driven by Th1 cells or Th1-associated cytokines. A question of major importance for future study is, therefore, the nature of the antigens and adjuvant factors that ultimately drive this inflammatory response. Although this was an entirely human study, our findings show remarkable parallels with studies performed in mice. MMP12 deficiency has been shown to protect mice against emphysema after chronic exposure to cigarette smoke, implying that MMP12 may be the key proteinase in the development of emphysema in this species [ 49 , 50 ]. Studies from both humans and mice therefore firmly suggest the importance of MMP12 in the pathogenesis of emphysema. Interestingly, in addition to solubilizing elastin, MMP12 is the MMP most efficient at degrading α1-antitrypsin, the primary physiological inhibitor of human leukocyte elastase [ 51 , 52 ]. Thus, chemokine-induced upregulation of MMP12 may orchestrate lung matrix degradation both directly and indirectly through inactivation of α1-antitrypsin. The therapy of COPD and emphysema is currently limited to pharmacologic bronchodilation to relieve dyspnea, antibiotics for intercurrent respiratory tract infection, and vaccination against prominent respiratory pathogens. Aside from efforts to prevent smoking or encourage cessation, there exist no measures that prevent development of emphysema or treat the specific causes of airway obstruction. By providing insight into the immunopathogenesis of COPD, our findings provide genuine hope that future therapies capable of preventing or halting smoking-related lung disease may be possible. Patient Summary Background Many people develop long-term lung problems after smoking, including a condition called emphysema. At the very end of the airways are tiny air sacs. In healthy people, the air sacs stretch and relax easily on breathing in and out. But in people with emphysema, the air sacs fill up with air but can't empty out properly, so air gets trapped, making breathing difficult. While the symptoms of emphysema can be treated, there are no treatments that can reverse the damage to the lung. What Did the Researchers Find? The researchers studied two groups of patients, all ex-smokers who had been admitted to a hospital to have part of their lung removed—some because of cancer, some for other reasons. The researchers studied the lung samples and looked to see exactly what type of immune cells the patients with emphysema had in their lungs and found that most of the immune cells were of one particular type. The researchers also showed that the immune cells could tell other lung cells to produce chemicals that can damage the lung. What Does This Mean for Patients? Lung damage in emphysema may not be caused directly by toxins in cigarette smoke. Instead, if you have emphysema, your body may react to the toxins and produce a special kind of immune cell that is key in causing the lung damage. So perhaps if doctors can find a way to change how this cell behaves, it might be possible to reduce or limit the lung damage. Obviously, not smoking, or stopping smoking, is the best way to prevent COPD or emphysema. What Are the Problems with the Study? The study is quite small, which means that the results may not be completely accurate; in particular, the study did not include detailed information from patients who had never smoked. So it is too soon to say for sure whether these special immune cells really are the link between smoking and lung damage in emphysema. Researchers will need to study many more patients with emphysema as well as people who have never smoked. Where Can I Find More Information? Two places to start are the patient Web pages of the following professional organizations. American Association for Respiratory Care: http://www.yourlunghealth.org/diseases_conditions/copd/ The British Thoracic Society: http://www.brit-thoracic.org.uk/public_content.asp?pageid=9&catid=21&subcatid=177	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC523885.xml
545043	Effects of PM10 in human peripheral blood monocytes and J774 macrophages	The effects of PM 10 , one of the components of particulate air pollution, was investigated using human monocytes and a mouse macrophage cell line (J774). The study aimed to investigate the role of these nanoparticles on the release of the pro-inflammatory cytokine TNF-α and IL-1α gene expression. We also investigated the role of intracellular calcium signalling events and oxidative stress in control of these cytokines and the effect of the particles on the functioning of the cell cytoskeleton. We showed that there was an increase in intracellular calcium concentration in J774 cells on treatment with PM 10 particles which could be significantly reduced with concomitant treatment with the calcium antagonists verapamil, the intracellular calcium chelator BAPTA-AM but not with the antioxidant nacystelyn or the calmodulin inhibitor W-7. In human monocytes, PM 10 stimulated an increase in intracellular calcium which was reduced by verapamil, BAPTA-AM and nacystelyn. TNF-α release was increased with particle treatment in human monocytes and reduced by only verapamil and BAPTA-AM. IL-1α gene expression was increased with particle treatment and reduced by all of the inhibitors. There was increased F-actin staining in J774 cells after treatment with PM 10 particles, which was significantly reduced to control levels with all the antagonists tested. The present study has shown that PM 10 particles may exert their pro-inflammatory effects by modulating intracellular calcium signalling in macrophages leading to expression of pro-inflammatory cytokines. Impaired motility and phagocytic ability as shown by changes in the F-actin cytoskeleton is likely to play a key role in particle clearance from the lung.	Introduction Increased exposure to PM 10 particles is associated with adverse health effects [ 1 , 2 ]. Much of the mass of PM 10 is low in toxicity and it has been suggested that, combustion-derived nanoparticles (ultrafine particles) [ 3 - 5 ] are a key component that drives these effects, especially inflammation. In individuals with pre-existing lung disease, inhalation of nanoparticles may induce inflammation and exacerbate respiratory and cardiovascular effects through the induction of oxidative stress and inflammation [ 4 , 6 , 7 ]. Rat inhalation studies using nanoparticles of various types, at high exposure, have demonstrated pulmonary fibrosis, lung tumours, epithelial cell hyperplasia, inflammation and increased cytokine expression [ 8 - 11 ]. The alveolar macrophage plays an important role in particle-mediated inflammation by phagocytosing particles and release of pro-inflammatory mediators such as the cytokine tumor necrosis factor-alpha (TNF-α) [ 12 ]. The signalling mechanisms for transcription of the TNFα gene includes calcium-related pathways in diseases such as sepsis [ 13 - 15 ]. Calcium is released from the endoplasmic reticulum stores on stimulation of the cell, leading to a calcium influx across the plasma membrane via calcium channels [ 16 ]. Various pathogenic particles have been shown to produce such changes in calcium flux within the cell [ 17 , 18 ] and a large number of pathological responses could be stimulated via such calcium signalling. In order for macrophages to migrate and phagocytose foreign material, an intact functional cytoskeleton is necessary. The cytoskeleton is sensitive to ROS and oxidative stress, due to the presence of thiol groups located on the actin microfilaments. On oxidation, these filaments cross-link, leading to reduced cell motility, impaired phagocytosis and hence clearance of foreign material from the lung. The cytoskeleton mediates several basic cell functions: chemotaxis, migration, phagocytosis, phagosome-lysosome fusion, and intracellular signalling [ 19 - 21 ]. Several lines of evidence suggest that changes in actin filament organisation play an important role in macrophage motility, adherence to surfaces and phagocytosis. Cellular dysfunctions associated with the cytoskeleton can cause retarded phagocytosis [ 22 ] and impaired phagosome-lysosome fusion [ 23 ], which may result in a diminished cellular killing and clearance of particles and pathogens from the lung. The pro-inflammatory cytokine interleukin 1 (IL-1) is not normally produced by the cells of healthy individuals, exceptions being skin keratinocytes, some epithelial cells and some cells of the central nervous system. In response to inflammatory stimuli, however, there is a dramatic increase in the production of IL-1 by macrophages and other cell types [ 24 ]. There are two distinct proteins, IL-1α and IL-1β which are the products of two distinct genes but which recognise the same cell surface receptors [ 25 ]. IL-1 possesses a wide variety of biological activities. As well as inducing its own synthesis, IL-1 stimulates the secretion of TNF-α and IL-6 from macrophages/monocytes [ 26 , 27 ]. Normal production of IL-1 is vital for host responses to injury and infection, while prolonged secretion has been linked with a number of pathological conditions [ 28 , 29 ]. Our hypothesis in this study was that PM 10 particles produce cytokine release and cytokine gene expression in macrophages by a process which involves calcium signalling and reactive oxygen species (ROS). Furthermore, we hypothesise that other effects of PM 10 , such as alterations in the cytoskeleton, are also mediated via signalling processes involving both ROS and calcium. Materials and Methods Particle Characteristics Collection of PM 10 samples was co-ordinated by Casella Stanger, London, England. Particles were collected onto TEOM filters in Marylebone Road, London, a site which had particularly high levels of traffic and therefore high levels of primary, combustion-derived nanoparticles. Ultrafine carbon black (UfCB) was obtained from Degussa (Printex 90), the average particle size was 14 nm. The characteristics and details of UfCB particles have been published previously [ 30 ]. Particle Quantification A single PM 10 filter was placed into a bijou bottle and 0.5 ml phosphate buffered saline (PBS) added. The bottle was vortexed for 4 minutes to remove the particles from the filter and the resulting suspension transferred to a clean bijou bottle. The mass of particles was assessed by densitometry. As standards, a series of dilutions of UfCB particles were made, ranging from 15.625 μg/ml to 1 mg/ml in saline, sonicated for 5 minutes, and 75 μl of each concentration was added into triplicate groups of wells in a 96-well plate. Seventy-five microlitres of PM 10 sample were added into a separate triplicate group of wells. The samples and standards were then read on a plate reader at 340 nm and the mass of particles calculated from a linear regression of the UfCB standards. J774.A1 Cell Culture The mouse macrophage cell line J774.A1 (a kind gift from Dr W Muller GSF, Gauting, Germany) was routinely cultured in RPMI medium (Sigma) containing 5% foetal calf serum (FCS) and Penicillin/Streptomycin. Cells were cultured until confluency was reached and then scraped from the surface of the flasks using a cell scraper. The cells were counted and adjusted to 5 × 10 5 /ml in RPMI plus 5% FCS. Sterile 10 mm glass cover slips were placed in each well of a 24-well plate and 1 ml of cell suspension added to each well. Cells were incubated at 37°C for 24 hours prior to particle treatment. Isolation of Human Peripheral Blood Mononuclear Cells Human peripheral blood mononuclear cells were prepared according to the protocol of Dransfield et al, [ 31 ]. In brief, two separate volumes of 40 ml of blood were withdrawn from healthy consenting volunteers and transferred to 50 ml sterile Falcon tubes containing 4 ml of 3.8% sodium citrate solution. Tubes were gently inverted and centrifuged at 250 g for 20 minutes, the plasma removed from each tube and pooled without disturbing the cell pellet. Dextran (Pharmacia), prepared as a 6% solution in saline was warmed to 37°C, before adding to the cell pellet (2.5 ml/10 ml cell pellet) and the volume made up to 50 ml with sterile saline. Tubes were gently mixed and the cells allowed to sediment at room temperature for 30 minutes. In order to prepare autologous serum, calcium chloride solution (220 μl 1 M/10 ml), was gently mixed with the plasma and incubated in a glass tube at 37°C until the clot retracted. Percoll (Pharmacia) gradients were made from a stock solution of 90% (18 ml Percoll + 2 ml 10x PBS, (Life Technologies, Paisley) without calcium or magnesium) to give final concentrations of 81%, 70% and 55% using 1x PBS. The separating gradient was prepared by layering 2.5 ml of 70% percoll over 2.5 ml 81% percoll. The leukocyte-rich fraction from the dextran sedimentation was transferred to sterile falcon tubes, 0.9% saline added to give a final volume of 50 ml and the tubes centrifuged at 250 g for 6 minutes. The pellet was resuspended in 55% percoll and 2.5 ml layered over the previously prepared separating gradients. Tubes were centrifuged at 290 g for 20 minutes and the mononuclear cells collected from the 55/70 layer. Cells were washed twice with PBS, counted, and resuspended in RPMI medium at a concentration of 5 × 10 6 cells/ml and 1 ml added to each well of a 24 well plate. For calcium imaging, cells were also set up in 6-well plates containing a 26 mm diameter sterile glass coverslip. The cells were incubated for 1 hour at 37°C, the medium removed and replaced with RPMI plus 10% autologous serum and incubated for 48 hours at 37°C. After the second incubation, the medium was replaced and the cells incubated for a further 72 hours prior to treatment. Cell Treatments PM 10 particles were diluted to give a final concentrations ranging from 5 μg/ml to 40 μg/ml in RPMI medium without serum and the suspension was sonicated for 5 minutes to disperse the particles. Cells which had been set up as described above, were washed twice with sterile PBS and 250 μl of particle suspension added to appropriate wells. UfCB particles were quantified as described for the PM 10 and set up in parallel with PM 10 particles at similar mass concentrations with J774 cells to investigate TNF-α release. One well received medium only (-ve control) and one received 250 μl of 1 μg/ml LPS (+ve control). The calcium antagonists were added concomitantly with the particles to give final concentrations of verapamil (100 μM), BAPTA-AM (50 μM), W-7 (250 μM), trolox (25 μM), and nacystelyn (5 mM). The cells were then incubated at 37°C for 4 hours and the supernatants removed and stored at -80°C until required. The cells cultured on 10 mm cover slips were fixed by the addition of 3% formaldehyde. J774.A1 Intracellular Calcium Measurements J774.A1 cells were cultured and removed from flasks as described above. Cells were pooled into a single tube, adjusted to 4.5 × 10 6 cells/ml in RPMI plus 10% FCS and incubated at 37°C until required for the assay. One millilitre of cell suspension was transferred to an Eppendorf tube, centrifuged at 145 g for 2 minutes, the medium removed, the cell pellet resuspended in 1 ml PBS and again centrifuged at 145 g for 2 minutes. The PBS was removed and cells resuspended in serum-free RPMI medium containing 23 mM Hepes buffer. Cells were loaded with 1 μg/μl Fura 2-AM (Sigma) in DMSO, 2 μl/ml cell suspension, the tube wrapped in foil and incubated in a shaking water bath for 20 minutes at 34°C. After incubation, the tube was centrifuged at 145 g for 2 minutes at 4°C, the medium removed and replaced with 1.5 ml fresh RPMI without serum. The Fura 2-AM-loaded cells were transferred to a quartz cuvette with stirrer and placed immediately into a fluorimeter with heated block and basal fluorescence measurements obtained over a 100 second period. The fluorimeter was set up with to give excitation wavelengths of 340 nm and 380 nm, emission 510 nm and excitation and emission slit widths set at 5 nm. During the experiments, the cuvette temperature was kept constant at 37°C. After 100 seconds, 10 μl appropriate treatment in RPMI medium was added to the cuvette and the experiment allowed to run for a further 1700 seconds. Treatments consisted of PM 10 to give a final concentration of 10 μg/ml with and without the calcium antagonists at the concentrations described above. Twenty microlitres of 5% Triton solution were added to the cuvette to lyse the cells to give the maximum fluorescence (Rmax) and the experiment continued for 500 seconds. To give the minimum fluorescence value (Rmin), 15 μl of 0.5 M EGTA in 3 M Tris buffer were added to the cuvette. The experiment was terminated after a further 500 seconds. The ratio of the fluorescence measurements at excitation wavelengths of 340 and 380 nm were converted to calcium concentration values using the method of Grynkiewicz et al, [ 32 ]. Human and mouse TNF-α ELISA The supernatants previously prepared were assayed for TNF-α protein content using a commercially available human TNF-α kit (Biosource) or mouse TNF-α kit (R&D Systems) according to the manufacturer's instructions. Briefly, each well of a 96-well plate was coated overnight with capture antibody, before washing with PBS containing 0.05% tween, and then adding test supernatant to the appropriate wells in triplicate groups. After incubation for 2 hours at room temperature, the wells were washed, a detection antibody added and incubated for a further hour at room temperature. The wells were then washed with PBS/tween before addition of Horseradish peroxidase (HRP)-conjugated streptavidin and incubated for 45 minutes at room temperature. Finally, the colour was developed by adding peroxidase substrate to each well, before reading the absorbance at 450 nm using a Dynatec plate reader. mRNA Extraction The experiments described above were also used to generate cells for total RNA extraction. After removal of the supernatant, 400 μl Tri reagent (Sigma) was added to each well. The lysed cells were then scraped from the surface of the plate using a cell scraper and transferred to Eppendorf tubes. Two hundred microlitres of chloroform were added to each Eppendorf, vortexed for 15 seconds and allowed to stand at room temperature for 15 minutes. The resulting mixture was centrifuged at 12000 g for 15 minutes at 4°C. The colourless upper phase was transferred to a fresh Eppendorf, before adding 450 μl isopropanol. The mixed samples were allowed to stand for a further 10 minutes at room temperature. Again the tubes were centrifuged at 12000 g for 10 minutes at 4°C, the supernatant removed and the RNA pellet washed in 1 ml of 75% ethanol. The resulting samples were then vortexed briefly, centrifuged at 7500 g for 5 minutes at 4°C and the RNA pellet air-dried for 10 minutes. The RNA was then suspended in 50 μl diethylpyrocarbonate (DEPC)-treated water and stored at -70°C until required for quantification and Reverse Transcriptase-Polymerase Chain Reaction (RT-PCR). RT-PCR The RT-PCR procedure was carried out using the Promega Access Kit. Briefly, a master mix of the kit reagents was prepared according to the manufacturers instructions. Ten microlitres of RNA at 0.03 μg/ml was added to 40 μl of the master mix, containing 10 μl of the appropriate human primers Glyceraldehyde Phosphate Dehydrogenase (GAPDH) or IL-1-α (MWG AG Biotech, Ebersberg, Germany). Tubes were placed in a thermal cycler which was programmed for the following temperatures and times. Following an initial 45 minute incubation at 48°, samples were cycled as follows: 94°C for 2 minutes, 95°C for 30 seconds, 60°C for 1 minute, 68°C for 2 minutes. This cycle was repeated 25 times for GAPDH and 30 times for IL-1 alpha. To conclude, the sample was incubated at 68°C for 7 minutes and then cooled to 4°C. The resulting RT-PCR products were separated by electrophoresis using a 2% agarose gel containing 1 μg/ml ethidium bromide and viewed under UV light. The RT-PCR bands were quantified by densitometry using Syngene software and the IL-1α band intensity was expressed as a percentage of the corresponding GAPDH band. These results were then expressed as a percentage of the untreated control. Calcium Imaging Human mononuclear cells were isolated from blood as previously described followed by adhesion onto 26 mm glass coverslips contained in 6-well plates. Cells were seeded in RPMI medium containing 0.1% BSA and penicillin/streptomycin at a density of 5 × 10 5 cells/ ml and incubated at 37°C, 5% CO 2 for 1 hour before washing with 1 ml PBS. Prior to particle treatment and digital enhanced video microscopy (Roper scientific), cells were loaded with the calcium-sensitive dye, Fura 2-AM (2 μg/ml in RPMI) (Sigma) for 30 minutes at 37°C. The coverslips were then washed with PBS, assembled into the microscope holder and 400 μl RPMI medium without phenol red (Sigma) added. The fluorescence ratio was observed (excitation 340 and 380 nm, emission 510 nm) at a magnification of 63× (Zeiss Axiovert microscope). Images were captured every 2 seconds by a Coolsnap fx Photometrics (Roper Scientific) camera controlled by Metafluor software. After 100 seconds particle treatment was added to the cells (100 μl of a 250 μg/ml stock solution of particles to give a final concentration of 50 μg/ml) contained in phenol red-free RPMI medium. F-Actin Staining The cells cultured on cover slips and fixed with formaldehyde were washed three times with PBS and permeablised with 0.1%Triton for 4 minutes. Cells were then washed three times with PBS and the F-actin stained using 33 ng/ml Phalloidin-FITC (Sigma) in PBS for 30 minutes at room temperature. Cells were washed three times with PBS before staining with propidium iodide (10 μg/ml in PBS) for 5 seconds. Cells were further washed three times with PBS before being mounted on glass slides using Citifluor mounting medium. The cells were then examined using an Axiofluor fluorescence microscope. Images were captured and quantified using Metamorph software (Universal Imaging Corporation). Seven fields of view were captured from each treatment and the images deconvolved using the image software. The staining intensity of each cell was then measured using the analysis software. Statistical Analysis Data from all of the experiments were analysed using analysis of variance with Tukey or Fishers multiple comparison test. Significance was set at p < 0.05. Results Intracellular Calcium Concentration in PM 10 -Treated J774.A1 Cells The effects of PM 10 on J774 murine macrophages was investigated at final concentrations of 5, 10 or 25 μg/ml PM 10 . Treatment of the cells with these particles produced a dose-dependent increase in cytosolic calcium concentration [Ca 2+ ] c up to a concentration of 10 μg/ml. At 25 μg/ml the [Ca 2+ ] c decreased to a value similar to the 5 μg/ml particle concentration. (Figure 1a ). Subsequent treatment with thapsigargin to release the endoplasmic reticulum calcium store produced a further increase in cytosolic Ca 2+ indicating that the cells remained viable and confirming previous studies [ 33 ]. There was a statistically significant difference between control and PM 10 treatments at 10 μg/ml (p < 0.05). The [Ca 2+ ] c following concomitant treatment of cells with particles and calcium antagonists was reduced (figure 1b ). Both the calcium channel blocker verapamil, and the calcium chelator BAPTA-AM significantly reduced (p,0.05) the intracellular calcium compared with PM 10 alone. In contrast, the antioxidant Nacystelyn did not significantly reduce the PM 10 -stimulated [Ca 2+ ] increase, with Ca 2+ concentration remaining significantly greater than the control value (p < 0.05). In our previous studies [ 34 , 35 ] we demonstrated that the antagonists used at the same concentrations used here caused no toxic effects to cells and that the drug treatment produced similar results to the untreated controls. Figure 1 The cytosolic calcium concentration (nM) in J774 cells on treatment with 5–25 μg/ml PM 10 particles for 1700 seconds (a) and with 10 μg/ml PM 10 particles plus calcium antagonists (b). There was a significant difference only at the 10 μg/ml particle dose from the control (p < 0.05). With calcium antagonist treatment, there was a significant difference between all of the treatments and the control (p < 0.05). Data represents the mean ± SEM of the intracellular calcium concentration (nM). (n = 3). Intracellular Calcium Concentration in PM 10 -Treated Human Monocytes PM 10 also induced a significant increase in cytosolic calcium in the primary human monocytes (p < 0.05). The dose of 10 μg/ml final concentration was chosen as the dose at which a significant increase in [Ca 2+ ] c had previously been observed (figure 1a ). At time points from 600 to 800 seconds after the addition of particle/antagonist treatment, there was a rapid increase in [Ca 2+ ] c with particles alone compared with the antagonists. In contrast to the antagonist treatment effects reported in figure 1 , the antioxidant nacystelyn significantly inhibited the [Ca 2+ ] c changes induced in the human monocytes treated with 10 μg/ml PM 10 (figure 2 ). Both the intracellular calcium chelator BAPTA-AM and the calcium channel blocker verapamil also significantly inhibited the [Ca 2+ ] c rise compared with particles alone (p < 0.05). Figure 2 The intracellular calcium concentration (nM) in human monocytes on treatment with PM 10 at a concentration of 10 μg/ml. Particles and calcium antagonists were added at zero time and the experiment run for 800 seconds in total (first 800 seconds shown). There was a significant difference between PM 10 -treated cells and PM 10 and calcium antagonist treatment at each time point tested (p < 0.05). Data represents the mean ± SEM of the ratio of 340/380 nm. (n = 3). Effect of UfCB and PM 10 particles on TNF-α release by J774 cells A comparison of the gram for gram dose effect of PM 10 and UfCB particles on TNF-α release by J774 macrophages is shown in figure 3 . The data show that PM 10 particles caused significantly more TNF-α release as the same mass of UfCB particles by the mouse macrophage cell line. Figure 3 TNF-α protein release in J774 cells treated with UfCB or PM 10 for 4 hours. There was significantly more TNF-α release in PM 10 treated cells compared with an equal mass of UfCB. Data represents the mean ± SEM pg/ml TNF-α release (n = 3). TNF-α release in PM 10 -treated Human Monocytes The release of TNF-α protein by human monocytes after treatment with varying concentrations of PM 10 is shown in figure 4 . The dose of particles ranged from 5 μg/ml to 40 μg/ml with concomitant treatment with calcium antagonists. There was a clear dose response of particle treatment from 10 μg/ml to 40 μg/ml. Within this range, the TNF-α concentration was approximately 170 pg/ml to 1000 pg/ml. At higher particle doses, the calcium antagonists reduced TNF-α release only marginally with the most dramatic and significant effect being seen with a particle concentration of 10 μg/ml with verapamil (V) and BAPTA-AM (B) treatments which reduced TNF-α release to 29 pg/ml and 7 pg/ml respectively (p < 0.05). There was no reduction in PM 10 induced TNF-α release with W-7 (W), Trolox (T) or Nacystelyn (N) at any particle dose tested. Figure 4 TNF-α release by human monocytes after treatment with PM 10 (5–40 μg/ml) and with concomitant treatment with calcium antagonists for 4 hours verapamil (V), BAPTA-AM (B), W-7 (W), trolox (T), and nacystelyn (N). There was a significant difference between the untreated control and PM 10 treatment only for the 10 μg/ml dose (p < 0.05). Data represents the mean ± SEM pg/ml TNF-α release (n = 5). IL-1 mRNA Expression Treatment of human peripheral blood monocytes with 10 μg/ml PM 10 for 4 hours produced a significant increase in IL-1α mRNA content compared with unstimulated cells (p < 0.05) (figure 5 ). The IL-1α band intensities were expressed as a percentage of the GAPDH band intensities and then normalised to the unstimulated control. PM 10 induced a five fold increase in IL-1α mRNA expression compared with the control and on treatment with the calcium antagonists, this was reduced to values similar to the control. There was a significant difference between the PM 10 exposed cells and concomitant treatment with all of the calcium antagonists and antioxidants tested (p < 0.05). Figure 5 IL-1α mRNA expression in human monocytes treated with 10 μg/ml PM 10 particles with and without calcium antagonists for 4 hours. The top panel shows a typical gel. The graph shows the IL-1α expression as a percentage of the GAPDH and normalised to the control. There was significantly greater expression of IL-1α mRNA in the PM 10 treatment which was reduced to control levels with calcium antagonist treatment. Data represents the mean ± SEM of the mRNA intensity. (n = 3). F-Actin Staining The fluorescence intensity of cells stained for F-actin after treatment with PM 10 particles and calcium antagonists is shown in figure 6 . Particles alone significantly increased the phalloidin-FITC fluorescence and hence the F-actin content of the cells compared with untreated cells. All of the calcium antagonists tested inhibited the PM 10 induced increase in F-actin intensity to control levels and this was significantly different from particle only treatment (p < 0.05), although the increase in the fluorescence intensity of the PM 10 -treated cells was modest (a 5% difference). Figure 6 The fluorescence intensity of F-actin stained J774 cells after 10 μg/ml PM 10 treatment and with calcium antagonist treatment for 4 hours. There was a significant difference in the intensity of PM 10 -treated cells compared with the untreated control (p < 0.05). There was no significant difference between the control and any other treatment. Data represents the mean ± SEM of the fluorescence intensity of the cells. (n = 3). Discussion There is evidence that increases in particulate air pollution correlate with increased morbidity and mortality from respiratory and cardiovascular causes [ 1 , 36 - 38 ] and the pro-inflammatory effects of PM 10 are considered to drive these effects [ 39 , 40 ]. The present study aimed to investigate the effect of PM 10 particles on oxidative stress- and calcium-related cytokine regulation in human monocytes and on the cytoskeleton in mouse J774 cells. We have previously shown that ultrafine or nanoparticles enhanced the calcium influx into cells of a monocytic cell line (MM6) [ 19 , 34 ] and that these [Ca 2+ ] c changes lead to production of the proinflammatory cytokine TNF-α [ 35 ]. We demonstrate here using calcium imaging, that PM 10 particles can also stimulate entry of extracellular calcium into both J774 macrophages and human macrophage derived monocytes, and that this process is inhibited by a calcium channel blocker suggesting that the PM 10 , in a similar fashion to UfCB induces opening of plasma membrane calcium channels leading to a calcium influx. The results obtained using the antioxidant nacystelyn were confusing. In the J774 macrophages nacystelyn was unable to inhibit PM 10 induced increases in cytosolic calcium concentration, whereas the same antioxidant was very effective in the human monocyte derived macrophages. This difference could be due to a species difference or a comparison between a cell line and primary cells. A number of cell lines have been demonstrated to exhibit aberrant calcium signalling pathways. Our previous studies using human macrophages suggest that ultrafine particle-induced increases in cytosolic calcium can be mediated by ROS [ 35 ] and since a large proportion of the particles within PM 10 are ultrafine, it is conceivable that much of the calcium increase is ROS mediated, at least in part. However, PM 10 also contains other substances, such as metals, that could influence this pathway. Metals would in fact be expected to increase the ROS production by the PM 10 particles [ 41 ]. The present study clearly shows that the same dose of PM 10 (10 μg/ml) that induces calcium elevation also stimulates significant increases in both TNF-α protein release and IL-1α mRNA production by macrophages. The calcium channel blocker verapamil and the intracellular calcium chelator BAPTA-AM reduced the calcium increase, TNF-α protein release and IL-1 mRNA expression by human monocytes when stimulated with PM 10 particles. This is strong evidence to suggest that influx of extracellular calcium plays a key role in upregulating the proinflammatory response induced by PM 10 that could lead to disease. However, the calmodulin inhibitor W-7 had little effect on TNF-α release, while it did inhibit IL-1 mRNA expression. The antioxidants also had variable abilities to block cytokine expression, inhibiting IL-1 mRNA production but not TNF-α protein release. These differences could be explained either by divergent pathways controlling expression of the two cytokines, or that TNF-α protein was measured in comparison to IL-1 mRNA. However, clearly both calcium and ROS are important in the regulation of IL-1α mRNA expression while only calcium is important in controlling TNF-α expression in macrophages exposed to PM 10 . These studies indicate that on an equal mass basis PM 10 is far more potent that UfCB in terms of its ability to induce TNF-α protein release. This is likely to be due to other components, such as metals and organic compounds other than the carbon core, within PM 10 that can promote inflammation. It is also possible that components such as the UF particles and metals could interact to enhance toxicity as has been shown for ROS production in vitro and inflammation in vivro [ 41 ]. Our previous studies have failed to detect LPS in the PM 10 particles, therefore it is unlikely that cytokine release, changes in intracellular calcium, and IL-1α gene expression can be explained solely by endotoxin. As explained previously, the cytoskeleton is the scaffold of cells, and in the case of motile cells such as macrophages it is responsible for controlling movement. Disruption of the cytoskeleton, particularly via oxidative stress, is thought to disrupt cellular structure and hence function [ 42 ]. We have previously demonstrated that PM 10 generates ROS [ 43 ]. The ability of the antioxidants trolox and nacystelyn to prevent the PM 10 induced increase in F-actin staining in this study demonstrates that particle-derived ROS impact on the macrophage cytoskeleton. Our previous studies also demonstrate that Uf particle-induced ROS play a role in elevating the cytosolic calcium concentration of macrophages leading to increased TNF-α production [ 35 ]. The results of this study also suggest that both calcium signalling and ROS are important in modulating the F-actin cytoskeleton in response to PM 10 exposure. As has been shown by other workers [ 44 , 45 ], in both the macrophage cell line and primary cells that F-actin is distributed as microfilaments around the cell, with special prominence at the leading edge of the cells. The microtubules in contrast are situated throughout the cell. Microtubules and actin filaments have previously been studied as targets of antitumour drugs [ 46 ] which mainly work by acting on microtubules and alter the dynamics of actin filaments. Changes in the distribution of actin filaments and their expression compared with normal cells may indicate alterations in the phagocytic ability of macrophages which may eventually lead to impaired particle clearance from the lungs. We show here that treatment of macrophages with PM 10 particles increased the F-actin fluorescence signal in cells stained with FITC-labelled phalloidin, although changes in the distribution of actin filaments was not apparent from microscope analysis there appeared to be more cortical staining. In accord with the role of calcium and ROS in the induction of IL-1 expression, both of these factors appeared to play an important role in modulating the F-actin cytoskeleton. The present study has shown that PM 10 particles may exert their increased pro-inflammatory effects by modulating intracellular calcium signalling in macrophages leading to expression of proinflammatory cytokines. An additional consideration is the effects of particles on the cytoskeleton of the cell. Impaired cellular motility and phagocytic ability is likely to play a key role in particle clearance from the lung, thus perpetuating the effects of PM 10 . The role of calcium and ROS in other cellular responses are under investigation.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545043.xml
509249	Adenoid cystic carcinoma of the parotid metastasizing to liver: case report	Background Adenoid cystic carcinoma is a rare malignant parotid tumor. Metastasis can occur even a decade or more after initial treatment of the primary. Case presentation We report a 60 year old female patient who presented with adenoid cystic carcinoma of the parotid gland. She underwent a total conservative parotidectomy followed by adjuvant radiotherapy. While on follow up, patient developed multiple liver metastases which manifested three years later. Patient lived for another two years before she died of her disease. Conclusions Although distant metastases of adenoid cystic carcinoma develop frequently, isolated metastasis to liver is unusual. Even after manifestation of distant metastasis, patients can be expected to live for a number of years. Palliative chemotherapy can be considered in symptomatic cases while the usefulness of metastatectomy is controversial.	Background Adenoid cystic carcinoma (ACC) is a rare malignant neoplasm of the salivary gland. Salivary gland neoplasms constitute 3% of cancers of all sites, of which, 10–15% are malignant [ 1 , 2 ]. Though ACC is the most common malignant tumor of the submandibular, sublingual and minor salivary glands, it accounts for only 15% of parotid cancers [ 3 ]. They are generally slow growing and spread relentlessly to adjacent structures. Hematogenous spread is more common than lymphatic spread, the common sites of metastasis being the lung, bone and viscera [ 4 , 5 ]. We present a case of multiple liver metastases occurring 3 years after surgery for ACC of the parotid gland. The primary therapy, metastasis and outcome of ACC are discussed. Case presentation A 60 year old woman presented with a small swelling beneath the right ear lobe of 4 months duration. The swelling measured 2 × 1 cm placed in the superficial part of the parotid and was not fixed. There was no facial nerve palsy or palpable cervical nodes. A fine needle aspiration cytology (FNAC) was carried out which showed the lesion to be ACC [ 6 ]. The clinical staging was T1, N0, M0. The patient underwent a total conservative parotidectomy after metastatic work up. Histopathology revealed ACC with cribriform pattern and perineural invasion (Figure 1 ). 60 Gy adjuvant external beam radiotherapy was administered post-operatively to the parotid area and the neck. Patient was placed on regular follow-up. Three years after primary surgery, patient presented with heaviness and pain in the right hypochondrium of 15 days duration. Patient was anicteric and abdominal examination revealed firm nodular and non tender enlargement of the liver. There was no ascites. Surgical site and neck were clinically normal. Chest roentgenogram was normal. Ultrasonography (US) of the abdomen revealed multiple metastatic lesions scattered in both lobes (Figure 2 ). Liver function tests were normal and an US guided FNAC revealed metastatic ACC (Figure 3 ). Since the lesions were multiple and scattered over both lobes of liver, surgical option was not considered and the patient was offered palliative chemotherapy which she declined. She developed pedal edema and abdominal distention 20 months after detection of liver metastasis. On clinical examination, patient was anicteric but liver had increased in size and abdomen showed evidence of a little free fluid. Chest CT scan was normal. Bone scan did not suggest any metastatic focus. Patient died a month later still without evidence of local recurrence or pulmonary metastasis. Conclusions Although ACC is the second most common malignant salivary gland neoplasm and constitutes approximately one third of all salivary gland malignancies it constitutes only 2% of parotid neoplasms [ 3 ]. As ACC is neurotropic, frozen section analysis of nerve margins is suggested specially when nerve is grossly involved by the tumor [ 7 ]. A total conservative or a radical parotidectomy is advocated for ACC though the main intent is to obtain a tumor free area of at least 1 cm [ 8 ]. ACC, with its often unusually slow biologic growth, tends to have a protracted course and ultimately a poor outcome, with a 10-year survival reported to be less than 50% for all grades [ 9 , 10 ]. These carcinomas typically show frequent recurrences and late distant metastases [ 11 ]. In a retrospective review of 92 cases, a tumor size greater than 4 cm was associated with an unfavorable clinical course [ 12 ]. Cribriform and solid patterns seen histologically were thought to predict more biological aggressiveness while tubular pattern represented more differentiated pattern of ACC. Over long periods of patient follow-up such grade based prognostication is less valid. Currently, stage and tumor location are the only factors considered prognostically significant [ 13 ]. Radiotherapy has been used as a primary modality for patients with surgical contraindications and in those with unresectable neoplasms. Though no improvement in survival is reported, the use of adjuvant radiation improves locoregional control and disease free survival. This patient received adjuvant radiation and did not have any locoregional recurrence. Regional metastasis is less common occurring in about 17% while systemic failure occurs in 33 to 50% of the patients [ 4 , 8 ]. Though involvement with distant metastases are unpredictable, organs involved in the order of decreasing frequency are lung, bone, brain and the liver [ 3 ]. Other rare metastatic sites of parotid and non parotid ACC include stomach, toe, choroids, brain and skin [ 14 - 18 ]. The initial site of metastasis is usually the organ containing the first capillary bed (first filter) and hence lungs would be the common site of metastasis [ 19 ]. Clinical observations from various malignancies have indicated that metastasis from certain types of tumor tend to occur in specific target organs leading to the famous 'soil and seed' hypothesis where metastatic cells 'home' to the organ [ 20 ]. Though liver metastasis has been reported, most of the liver metastases reported are of non parotid ACC [ 3 , 5 ]. The occurrence is usually metachronous or synchronous with metastasis to other organs like the lung as it is the first filter. In the series of Spiro, of the 74 patients developing metastasis from salivary gland ACC, 23 did so without loco-regional recurrence while 5 had isolated bone metastasis [ 5 ]. Sung and colleagues found metastases in 46 out of 94 head and neck ACC [ 21 ]. In that study, only one patient developed liver metastasis and that patient had metastasis to both lung and bone. In this case, the patient manifested with multiple metastatic foci in the liver as the first and only metastatic organ which is very unusual. Surgical options of metastatectomy were not explored as patient had multiple metastases involving both lobes of liver. Although the patient lived with disease for a further two years, she did not show any evidence of lung metastasis or loco-regional recurrence. In this patient, the liver metastasis could have occurred prior to treatment as an organ of preference; evidenced by the fact that there were no other organs showing metastasis nor was there a loco-regional recurrence later. Studies in ACC have shown long tumor doubling times of pulmonary metastasis and late recurrences up to 10 years after primary treatment [ 22 ]. The estimated doubling time of lung metastasis in ACC ranges from 200 to 600 days [ 22 ]. There is even a suggestion that metastasis at the cellular level could occur many years prior to clinical presentation of primary tumor. FNAC was done to prove the metastatic foci in this patient and can be a useful tool for diagnosis [ 23 ]. Although this patient declined chemotherapy, chemotherapeutic responses have been reported in ACC [ 24 ]. ACC carries a mortality of 75–80% over a 30 year period and most patients who die of their disease do so between 5 and 10 years after initial treatment [ 10 ]. Our patient died 5 years from diagnosis with metastatic disease that developed 3 years after initial treatment. ACC is a rare malignant tumor of parotid gland. Metastasis can manifest very late and hence a long term follow-up and a high index of suspicion is necessary to diagnose them early. An annual ultrasound study of abdomen would be desirable on follow-up. Unlike metastasis from other malignancies, these grow indolently and long term survival can be expected even with multiple metastases as also evidenced in the present case. Chemotherapy could be considered in selected patients as a therapeutic option in metastatic disease. List of abbreviations used ACC: Adenoid cystic carcinoma FNAC: Fine needle aspiration cytology Gy: Gray Competing interests None declared Authors' contributions KH was the principal clinician who planned the evaluation and procedure, in addition to conceptualizing and drafting the article. SRMG was the pathologist. Both the authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC509249.xml
544361	The effect of ethnicity on outcomes in a practice-based trial to improve cardiovascular disease prevention	Background Health disparities are a growing concern. Recently, we conducted a practice-based trial to help primary care physicians improve adherence with 21 quality indicators relevant to the primary and secondary prevention of cardiovascular disease and stroke. Although the primary concern in that study was whether patients in intervention practices outperformed those in control practices, we were also interested in determining whether minority patients were more, less, or just as likely to benefit from the intervention as non-minorities. Methods Baseline (fourth quarter 2000) and follow-up (fourth quarter 2002) data were obtained from 3 intervention practices believed to have at least 10% minority representation. Two practices had a black (non-Hispanic) population sufficient for analysis, while the other had a sufficient Hispanic population. Within each practice, changes in the 21 indicators were compared between the minority patient population and the entire patient population. The proportion of measures in which minority patients exhibited greater improvement was calculated for each practice and for all 3 practices combined, and comparisons were made using non-parametric methods. Results For all black patients, the observed improvement in 50% of 22 eligible study indicators was better than that observed among all white patients in the same practices. The average changes in the study indicators observed among the black and white patients were not significantly different (p = 0.300) from one another. Likewise for all minority patients in all 3 practices combined, the observed improvement in 14 of 29 (43.3%) eligible study indicators was better than that observed among all white patients. The average changes in the study indicators among all minority patients were not significantly different from the changes observed among the white patients (p = 0.272). Conclusions Among 3 intervention practices involved in a quality improvement project, there did not appear to be any significant disparity between minority and non-minority patients in the improvement in study indicators.	Introduction In 2002, the Institute of Medicine (IOM) issued a report suggesting that minorities are more likely than non-minorities to receive a lower quality of healthcare [ 1 ]. Because of the issues such as those raised in the IOM report, health disparities are a growing concern. This concern is reflected in many ways, including the development by National Institutes of Health of a program of action to confront these disparities and the Healthy People 2010 goal of eliminating these disparities. Disparities are particularly evident in the area of chronic diseases. Although blacks are more likely than whites to have blood pressure monitoring, cholesterol screening, and smoking counseling, coronary heart disease is more prevalent among blacks than among whites [ 2 ]. Additionally, among all ethnic groups, blacks experience the highest mortality rates associated with heart disease, cancer, cerebrovascular disease, and HIV/AIDS. Although the overall mortality rate among blacks has been declining over the past 50 years, rates for cancer and diabetes were actually higher in 1995 than in 1950. On a similar note, Hispanics are significantly more likely as non-Hispanic whites to die from diabetes and HIV/AIDS [ 3 ]. In hopes of improving health outcomes and prevention practices for all patients, much focus has recently been given towards the improvement in quality of healthcare. For example, the 7 th report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure outlines specific guidelines for preventing and managing hypertension, hyperlipidemia, and coronary heart disease [ 4 ]. The 2nd report of the National Cholesterol Education Program (NCEP) expert panel on detection, evaluation, and treatment of high blood cholesterol in adults makes specific recommendations on the prevention and care of hyperlipidemia and coronary heart disease [ 5 ]. Other quality indicators also exist for the prevention and management of other chronic diseases, including heart failure, atrial fibrillation, and diabetes, diseases which were all targeted in the current study. Recently, there have also been a number of practice-based interventions aimed at improving the quality of healthcare for patients. For example, researchers have shown that a practice-based intervention (the Healthy Steps for Young Children Program) can enhance the quality of care for families of young children [ 6 ]. Additionally, a practice-based telephone intervention was proven to improve pneumococcal vaccine immunization for seniors [ 7 ]. We have also reported on a practice-based intervention to help primary care physicians improve adherence with 21 quality indicators relevant to the primary and secondary prevention of cardiovascular disease and stroke [ 8 , 9 ]. What these earlier interventions have lacked, however, are analyses examining whether the interventions have improved the quality of care for all patients, regardless of ethnicity. Because these types of interventions are heavily dependent on physician and/or clinical staff interaction with patients, because ethnic minorities may have less trust in their healthcare providers [ 10 ], and because barriers in the patient physician relationship may contribute to the ethnic disparities in the quality of the healthcare experience [ 11 ], there exists the possibility that poor cultural competency could result in a lack of effectiveness of the intervention among ethnic minorities. If such quality improvement efforts do not improve care for all ethnic groups equally, then there may be significant healthcare policy implications related to the refinement of existing interventions and to the development of future interventions. The aim of this study was to examine whether or not a multi-method quality improvement (QI) intervention was equally successful among patients of different ethnicities. Some of the findings from this QI intervention have been previously published [ 8 , 12 ], and they suggest that primary care practices that use electronic medical records and receive regular performance reports can improve their adherence with clinical practice guidelines for cardiovascular disease and stroke prevention. Methods The multi-method QI intervention added practice site visits (for academic detailing and QI facilitation) and network meetings (for sharing of best practices) to the approach of guideline dissemination and audit and feedback, employed in a less intensive intervention. Ten sites received the intensive multi-method QI intervention, and ten sites received the less intensive intervention. The study was conducted in a practice-based research network (PPRNet) among users of a common electronic medical record (Practice Partner Patient Records, Seattle WA), which historically provided audit and feedback to its practice members. As a supplement to the original study, we were also interested in whether minority patients were more, less, or just as likely to benefit from the intervention as non-minorities. The study presented here focused on outcome and process measures for minorities within 3 primary care practices, all of which received the intensive intervention. These 3 practices (labeled A, B, and C) were selected because they each had a significant (i.e. > 10%) proportion of minority patients and had recorded patient ethnicity in their electronic medical record. Practice A is an urban internal medicine practice in the Midwestern U.S. with 5 healthcare providers. Practice B is a rural family medicine practice in the Northeastern U.S. with 8 healthcare providers. Practice C is an urban family medicine practice in the Southeastern U.S. A total of 21 study indicators (see Table 1 ) were obtained from each practice at baseline (fourth quarter 2000) and at the end of the study (fourth quarter 2002). These indicators were derived from published sources [ 4 , 5 , 13 - 16 ] and were deemed to be the most appropriate indicators for measuring quality of prevention and management of cardiovascular disease and stroke. Fourteen of the study indicators are process measures, reflecting whether recommended tests were done, appropriate diagnoses made or medication prescribed. Seven indicators are outcome measures, reflecting whether patients achieved recommended treatment goals. Some of the measures represent primary prevention, e.g., screening for hypertension or hyperlipidemia. Others represent secondary prevention, e.g., reaching treatment goals for glycosylated hemoglobin, low-density lipoprotein (LDL) cholesterol, and blood pressure in patients with diabetes. The institutional review board at the Medical University of South Carolina approved the study. Table 1 Study indicators CONDITION MEASURES Hypertension 1 Process measures: • BP measurement in prior 12 months • Diagnosis of hypertension for 3 measurements >= 140/90 in prior 12 months • BP measurement in prior 3 months for patients with diagnosis of hypertension Outcome measures: • Most recent BP measurement < 140/90 for all patients • Most recent BP measurements < 140/90 for patients with diagnosis of hypertension Hyperlipidemia (General Population screening) 2 Process measures: • Measure of total cholesterol in prior 36 months • Measure of HDL-C in prior 36 months Coronary Heart Disease 1,2,3 Process measures: • Measurement of LDL-cholesterol in prior 12 months • Recorded diagnosis of hyperlipidemia for LDL-cholesterol > 130 mg/dl • Medication for hyperlipidemia for LDL-cholesterol > 130 mg/dl • Prescription of beta-blocker in patients with history of myocardial infarction Outcome measures: • Most recent LDL-cholesterol < 100 mg/dl • Most recent BP measurement < 140/90 Heart Failure 4 Process measure: • Prescription of angiotensin converting enzyme inhibitor or angiotensin receptor blocker Atrial Fibrillation 5 Process measure: • Prescription of oral anticoagulant Diabetes Mellitus 6 Process measures: • Measurement of glycosylated hemoglobin in prior 12 months • Measurement of LDL-cholesterol in prior 24 months • BP measurement in prior 3 months Outcome measures: • Most recent glycosylated hemoglobin < 7 % • Most recent LDL-cholesterol < 100 mg/dl • Most recent BP measurement < 130/85 1 Adapted from The Sixth Report of the Joint National Committee on Detection, Evaluation and Treatment of High Blood Pressure (JNC VI). National Institute of Health, National Blood Pressure Education Program, NIH publication 98-4080, November 1997. 2 Adapted from Summary of the Second Report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel II). JAMA 1993; 269(23): 301523. 3 Adapted from American Heart Association Scientific Statement: Smith SC, Blair SN, Criqui MH, etal: Preventing Heart Attack and Death in Patients with Coronary Disease. Circulation . 1995;92:2–4. Although the AHA recommendation is to measure total cholesterol at least every 60 months, we were forced to restrict our measure to 36 months due to restrictions of the historical data. 4 Adapted from American Heart Association Scientific Statement: Williams JF, Bristow MR, Fowler MB, etal: Guidelines for the Evaluation and Management of Heart Failure. Circulation . 1995;92:2764–2784 5 Adapted from American Heart Association Scientific Statement: Prystowsky EN, Benson DW, Fuster, V, etal: Management of Patients with Atrial Fibrillation. Circulation 1996;93:1262–1277. 6 Adapted from American Diabetes Association, Diabetes Quality Improvement Project, Initial Measure Set (Final Version) August 14, 1998 To determine practice performance on the study indicators, participating practices ran a computer program to extract patient activity during the previous quarter from their electronic medical record. To protect patient confidentiality, the extract program assigned an anonymous numerical identifier unique to each patient. The extract program obtained demographic information such as age, ethnicity, and gender, and diagnoses, medications, laboratory data, and vital signs. Text of consultation reports, progress notes, and discharge summaries were not extracted. The data were copied to diskettes and mailed to PPRNet or sent electronically via a secure server. In the PPRNet offices, data were bridged to standard data dictionaries and converted to SAS ® (Statistical Analysis System, Cary NC) data sets on standard microcomputers for analyses. In each patient's electronic medical record, ethnicity was recorded as white, black/African American, American Indian/Alaskan native, Asian, native Hawaiian/other Pacific islander, and "some other ethnicity", while ethnicity was recorded as Hispanic/Latino and non-Hispanic/Latino, all in concordance with the 2000 U.S. Census ethnicity categories. Currently, these physician practices allow the patient to designate their ethnicity categorization. However, because this process for collecting ethnicity data began in the middle of our study, some ethnicity categorizations were made by the office staff within each of the practices. Ethnicity data was only available on approximately 42% of patients, due to the fact that the electronic medical record software program did not require physicians to enter patients' ethnicity data until its most recent version was released, which occurred during the study time frame. Improvements in process and outcome measures were compared between minority and non-minority patients. Minority was defined as any ethnic designation other than white non-Hispanic. Changes in the process and outcome measures were of primary interest in this study. Within each practice, these changes were compared between the minority patient population and the white patient population. Measures were deemed eligible for comparison if at least 10 minority patients were included in the rate calculations. For example, if practice A only had 8 minority patients with a diagnosis of having had myocardial infarction (MI), then the measure of the percentage of MI patients who had been prescribed a beta blocker could not be compared between the minority and white patient population. The proportion of eligible measures in which minority patients exhibited greater improvement was calculated for each practice and for all 3 practices combined. A Wilcoxon signed rank test (the non-parametric equivalent of the paired t-test) was used to test the hypotheses that minority patients exhibited changes similar to those of the non-minority patients. This study had approximately 80% power (2-sided hypothesis testing, α = 0.05) to detect a 6.6 percentage point difference between average improvement in the study indicators among all minority and non-minority patients. Results Baseline characteristics of the patients from the 3 practices are listed in table 2 . In practice A, black (non-Hispanic) patients were the only sizable minority. Although practice B did contain 10 black non-Hispanic patients, this sample was not large enough for substantive comparisons. There were enough Hispanic patients in Practice B to compare with the entire groups of patients within that practice. In practice C, there were 117 black patients used for comparison. There were several significant differences of note between the minority patients to the overall population of patients within that same practice. Compared to the white patient population in practice A, the minority patients were significantly younger and significantly less likely to be diagnosed with hyperlipidemia. Compared to the white patient population in practice B, the minority patients were significantly more likely to be male and to have diabetes. In practice C, the minority patients were significantly younger, more likely to be female, and less likely to have a diagnosis of hyperlipidemia. Table 2 Baseline characteristics of the patients within the 3 practices of interest Characteristic Practice A Practice B Practice C Black patients (n = 179) White patients (n = 1,079) Hispanic patients (n = 254) White patients (n = 2,526) Black patients (n = 117) White patients (n = 491) Demographics Age (mean ± s.d.) 54.2 ± 15.2** 60.7 ± 17.8 34.5 ± 17.9 33.7 ± 18.9 42.0 ± 16.7* 51.2 ± 16.9 Gender (% female) 68.7 69.2 46.1*** 57.9 67.5* 56.4 Medical conditions Hypertension (%) 50.8 51.1 7.1 7.8 21.4 21.0 Hyperlipidemia (%) 33.5*** 47.5 5.5 4.1 0.9* 5.3 Diabetes (%) 12.9 9.7 4.3* 2.0 12.0 7.5 Coronary disease (%) 7.8 11.7 1.2 1.0 0.0 0.0 Heart failure (%) 2.2 4.9 0.0 0.2 1.7 0.6 Atrial fibrillation (%) 1.1 3.9 0.0 0.1 0.0 0.2 * p < 0.05 when compared to white patients within the particular practice p < 0.01 when compared to white patients within the particular practice * p < 0.001 when compared to white patients within the particular practice **** p < 0.0001 when compared to white patients within the particular practice Additional file 1 lists the baseline and end-of-study measurements for each of the 21 study indicators, for minority patients and white patients within each of the 3 practices. In practice A, the improvement in 7 of 16 eligible study indicators was better among black patients than among white patients in that practice. (For 9 of these 16 indicators, the improvement was worse among the black patients.) In practice B, the improvement in 3 of 7 eligible study indicators was better among Hispanic patients than among white patients in that practice, and worse for 4 of the 7 indicators. In practice C, the improvement in 4 of 6 eligible study indicators was better among black patients than among white patients in that practice. Thus for all black patients in practices A and C, the observed improvement in 11 of 22 (50.0%) eligible study indicators was better than that observed among white patients. On average, indicators improved 4.4 and 9.3 percentage points among black and white patients, respectively. These changes were not significantly different (p = 0.300) from one another. Likewise for all minority patients in all 3 practices combined, the observed improvement in 14 of 29 (48.3%) eligible study indicators was better than that observed among non-minority (white) patients. On average, indicators improved 4.6 and 8.3 percentage points among minority and non-minority patients, respectively, and these changes were not significantly different (p = 0.272) from one another. Discussion In these 3 physician practices, all of which were in the intervention arm of a randomized trial aimed at improving primary and secondary prevention of cardiovascular disease and stroke, we found that results for minorities were relatively similar to the results experienced by the overall practice populations. Change from baseline was greater among minority patients than among white patients for 48.3% of the 29 eligible study indicators, and the average changes in the study indicators among all minority patients were not significantly different from the changes observed among the white patients. There are some limitations of this study which should be noted. As noted earlier, the ethnicity status was only available on 42% of patients within the practices of interest; thus the results may not truly represent what occurred in these practices overall during the study. Given the relatively small number of eligible indicators for comparisons across ethnicities, this statistical power to detect subtle differences was not optimal. Nevertheless, the overall findings suggest that any true differences in this intervention's effectiveness across ethnicities are small. These findings are encouraging, and they suggest that the quality improvement strategies that have been developed to date for physician practices that use electronic medical records have a similar impact on minorities and non-minorities. Future studies should continue to address whether the effectiveness of interventions such as ours is cross-cultural, and whether interventions tailored to be more culturally appropriate can improve the overall effectiveness of interventions. List of abbreviations used IOM: Institute of Medicine HIV: Human Immunodeficiency Virus AIDS: Acquired Immunodeficiency Syndrome QI: Quality Improvement LDL: Low-Density Lipoprotein MI: Myocardial Infarction Competing Interests The author(s) declare that they have no competing interests. Authors' Contributions PJN helped design the study, perform the analyses, and write the manuscript. SMO helped design the study, perform the site visits, and edit the manuscript. RGJ helped design the study, perform the analyses, and write the manuscript. LFR helped design the study, assisted with data acquisition, and edited the manuscript. LMD helped design the study, perform the site visits, and edit the manuscript. CF helped perform site visits and edit the manuscript. All authors read and approved the final manuscript. Supplementary Material Additional File 1 Study indicators as measured at baseline (B) and the end (E) of the study for all patients and minority patients within each of the 3 practices. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544361.xml
517711	Fingersomatotopy in area 3b: an fMRI-study	Background The primary sensory cortex (S1) in the postcentral gyrus is comprised of four areas that each contain a body map, where the representation of the hand is located with the thumb most laterally, anteriorly and inferiorly and the little finger most medially, posteriorly and superiorly. Previous studies on somatotopy using functional MRI have either used low field strength, have included a small number of subjects or failed to attribute activations to any area within S1. In the present study we included twenty subjects, who were investigated at 3 Tesla (T). We focused specifically on Brodmann area 3b, which neurons have discrete receptive fields with a potentially more clearcut somatotopic organisation. The spatial distribution for all fingers' peak activation was determined and group as well as individual analysis was performed. Results Activation maps from 18 subjects were of adequate quality; in 17 subjects activations were present for all fingers and these data were further analysed. In the group analysis the thumb was located most laterally, anteriorly and inferiorly with the other fingers sequentially positioned more medially, posteriorly and superiorly. At the individual level this somatotopic relationship was present for the thumb and little finger, with a higher variability for the fingers in between. The Euclidian distance between the first and fifth finger was 17.2 mm, between the first and second finger 10.6 mm and between the remaining fingers on average 6.3 mm. Conclusion Results from the group analysis, that is both the location of the fingers and the Euclidian distances, are well comparable to results from previous studies using a wide range of modalities. On the subject level the spatial localisation of the fingers showed a less stringent somatotopic order so that the location of a finger in a single subject cannot be predicted from the group result.	Background The first somatotopic maps of the homuncular organisation of the primary somatosensory cortex (S1) were established in 1937 by using intra-operative electrical stimulation of the brain surface [ 1 ]. Subsequent, non-invasive investigations in humans on the hand representation in S1 have described a somatotopic organisation along the central sulcus with the thumb located laterally, anteriorly and inferiorly to the little finger [ 2 - 6 ]. Studies in non-human primates have revealed the cytoarchitectonic subdivisions of S1, namely areas 3a, 3b, 1 and 2, that outline the cortex in the postcentral gyrus [ 7 ]. Area 3a occupies the fundus of the central sulcus, area 3b the anterior wall of the postcentral gyrus, area1 its crown and area 2 its posterior wall. Each area contains a fairly complete map of the body surface and is the cortical representation of different somatosensory receptors. In area 3b the neurons are predominantly responsive to stimulation of cutaneous receptors. As opposed to neurons in area 1, that also receive input from cutaneous receptors, those in area 3b possess discrete receptive fields with a homuncular organisation that may be more distinct [ 8 ]. Previous studies on somatotopy in the hand area using functional Magnetic Resonance Imaging (fMRI) have yielded varying results. Gelnar et al. failed to show a somatotopy in S1 when applying vibratory stimuli to three of the fingers of the right hand [ 5 ]. Maldjian et al. demonstrated somatotopy in 3 out of 5 subjects [ 3 ]. Similarly, Kurth et al., using electrical stimulation of two fingers, found somatotopically arranged activation patterns in 5 out of 20 subjects [ 9 ]. In a follow-up study where activation of all fingers in area 3b was found in 7 out of 10 subjects, the same authors reported a general somatotopy, without further specification [ 4 ]. Methods differ considerably between these three studies with regard to anatomical considerations, number of subjects studied, and the field strength used. Maldjian et al. used the highest field strength, 4 Tesla (T), while both Kurth et al. and Gelnar et al. used 1.5 T [ 3 - 5 ]. However, Maldjian et al. did not contribute activations to any area in S1. Also Maldjian et al. included the smallest number of subjects (5); data from one were discarded due to motion artefacts and group analysis was based on data from the remaining 4. In the present study we readdressed the issue of somatotopy in the hand area as assessed with fMRI. Our aim was to optimise results by including a larger number of subjects, by focussing on area 3 b, where homuncular organisation expectedly is most distinct and by performing fMRI at 3 Tesla (T). Results Tactile stimulation of the fingers of the dominant hand yielded activation in contra-and ipsilateral S1, contra- or bilateral secondary sensory cortex (S2), ipsilateral cerebellum, and in some subjects the contralateral thalamus. Significant activation for all five fingers in area 3b was present in seventeen subjects and these data were further analyzed. In the one volunteer excluded from the analysis, activation was present for four fingers. The spatial distribution of the activations in the contralateral S1 for tactile stimulation versus rest in one subject is shown in Figure 2 . A somatotopic organisation with the representation of the thumb located laterally to the little finger was present in 16 out of 17 subjects, with the thumb located anteriorly to the little finger in 14 out of 17 subjects and with the thumb located inferiorly to that of the little finger in 16 out of 17 subjects. Group averages of the distances from D2 to D1 (D2-D1), D3 to D1(D3-D1), D4 to D1 (D4-D1) and D5 to D1 (D5-D1) are presented in the Table and shown as graphs in Figure 3 . Combined these indicate a strict somatotopy with the distance to D1 increasing for every finger in each of the three directions. Distances to D1 were compared for neighbouring fingers. In the medial-lateral direction, the distance D4-D1 was different from D3-D1. In the anterior-posterior direction a significant difference was observed between D4-D1 and D3-D1. Finally, in the superior-inferior direction the distances D2-D1 and D3-D1 as well as D3-D1 and D4-D1 differed; the location of D2, as determined by its distance to D1, was different from the location of D1 [0, 0, 0]; no difference was found between the distances D5-D1 and D4-D1. Considering that the three coordinates [x, y and z] together define one point in the 3D Cartesian space, the coordinates of D3, D4 and D5, differed from those of D1, p now <0.05/3, corrected for multiple comparisons (not in Table). The Euclidian distance from D1 to D2 was 10.6 mm (SEM ± 1.5). The distance from D2 to D3 was 5.5 mm (± 0.9), from D3 to D4 7.4 mm (± 1.1) and from D4 to D5 6.8 mm (± 1.2), resulting in an average for D2-D3, D3-D4 and D4-D5 of 6.6 mm. The spatial extension of the representation of the hand in area 3b, defined as the Euclidian distance between D1 and D5 was 17.2 mm (± 2.0 mm). Discussion In the group average from the present study the strict somatotopic organisation in the primary sensory cortex known from studies using a variety of modalities, was reproduced [ 2 - 6 ]. The fingers' average activations were laid out on the body map, with the thumb located most laterally, anteriorly and inferiorly and the little finger most medially, posteriorly and superiorly and the remaining fingers in between, the distance to the thumb increasing for every finger in each of the three directions. In individual subjects the arrangement in the hand representation with the thumb located laterally, anteriorly and inferiorly to the little finger is frequently found, while the remaining fingers may or may not display the orderly lateral-to-medial, anterior-to-posterior and inferior-to-superior organisation 'D1-D2-D3-D4-D5' [ 2 , 3 , 10 , 11 ]. We chose to present group averages as it is our belief our results would have greater significance if a regular somatotopy was present at the group level. Body maps in non-human primates demonstrating the regular sequence mentioned above, were established with cortical single unit recordings and are supposedly the golden standard. The present study is based on data from 20 subjects, generating results available for analysis from 18 of these; the spatial representation of all fingers in area 3b of the primary somatosensory cortex was localised in 17 subjects. The average extension of the hand representation in area 3b of 17 mm with a somatotopic arrangement of fingers 1-5 as described above is consistent with results from previous studies using a range of modalities [ 2 - 4 , 7 ]. Also the mean distance between D2-D3, D3-D4 and D4-D5 of 6.6 mm and 6.3 mm for main and differential effects, respectively, is in good agreement with human electrophysiological and fMRI data [ 2 , 4 ]. The larger distance between the thumb and index finger as compared to distances between subsequent fingers suggests a larger representation for the thumb. This finding is in agreement with results from a study using electrocorticography with subdural electrodes in three patients [ 1 , 12 ]. With the spatial resolution used in this study, 3 × 3 × 3 mm 3 , resampled to 1.5 × 1.5 × 1.5 mm 3 , activation for all five fingers was found in 17 of 18 subjects. For comparison, Kurth et al. used a resolution of 1.7 × 1.7 × 1.7 mm3 and electrical stimulation with ring electrodes and found activation in area 3b for all five fingers in 7 out of 10 subjects (70%) [4]. The present study showed activations for all fingers in 94 % of subjects. This difference might be due to the higher field strength used in this study as the higher magnetisation vector and sensitivity to changes in susceptibility increase the signal-to-noise ratio. In one subject excluded from the analysis, activation was present for four fingers and another was excluded due to general lack of activation. Lack of activation may be due to subjects being 'low-activators' in fMRI experiments, due to inadequate stimulation or to some other, unknown factor. The type of stimulation used is decisive of what area in S1 can be expected to be activated. For example, neurons in area 3a are responsive to deep receptor and proprioceptive stimulation and in one study punctate tactile stimulation did not activate area 3a [ 13 ]. Also, receptive fields are maximally focused in area 3b, while in area 1 receptive fields become larger and more complex. In area 2, receptive fields are even more complex with reduplications. This combined knowledge made area 3b the area of our choice to study somatotopy and explains why we chose not to report on activations in areas 3a, 1 and 2. We found activation in the anterior wall of the postcentral gyrus, defined as area 3b according to our operational definition during tactile stimulation (Fig 1 ). More pronounced activation was noticed frequently in the crown or posterior wall of the postcentral gyrus, defined as areas 1 and 2 (Fig 2 ). Similar observations have been made in other studies using both fMRI and PET [ 5 , 14 ]. According to studies in non-human primates the representation of the distal fingertips in area 1 points posteriorly, a finding confirmed in a recent fMRI-study on humans [ 15 ]. Area 2 then is the mirror-image of area 1 with the fingertips pointing anteriorly. The activation in the posterior wall might represent activation in both areas 1 and 2 localised at their meeting point, i.e. the fingertips. The larger cluster size of this activation is explained by the clusters arising from area 1 and 2 being contiguous and therefore additive. In the present study the activation of area 1 and 2 is probably due to hierarchical processing in the rostrocaudal direction within S1. A previous observation that electrical stimulation of the cutaneous afferents of the median nerve resulted in evoked potentials in area 3b after 30 msec while a potential in area 1 was seen after another 5 msec lends support to this assumption [ 16 ]. Conclusion In the group analysis, a somatotopic organisation for all the fingers in the hand representation of area 3b could be demonstrated using fMRI; the Euclidian distance between the thumb and the little finger was well comparable to that determined in previous studies. On the subject level the cortical somatosensory representation of the thumb was located laterally, anteriorly, and inferiorly to that of the little finger in 14 out of 17 subjects. The spatial localisation of the remaining fingers showed a less stringent somatotopic order when compared individually. Methods Subjects Twenty healthy, self-reportedlly right-handed volunteers (6 male and 14 female, age 21–43 years, mean 29.4 years) were included in the study. The protocol was approved by the local ethics committee and written informed consent was obtained. All volunteers had normal images on a fluid-attenuated inversion recovery (FLAIR) sequence. Stimulation Tactile stimulation consisted of brushing the glabrous skin of the two distal phalanges of each finger continuously forwards and backwards with a commercially available tooth brush. During the experiment, volunteers were positioned on the MR table with their right arm from the elbow down in a padded cast, that also provided support for the dorsal part of the hand. They were instructed to rest their arm against the magnet bore so that both arm and hand were relaxed. Pieces of soft cloth were placed between the fingers in order to avoid that stimulation also involved a neighbouring finger. The frequency was 1 Hz; no forced pressure was exerted. Consistency was tested on a finger model firmly taped onto a computerized electronic scale (Biopac Systems, DA 100B, MP 100A; Macintosh Powerbook G3 with software AcqKnowledge 3.5): the intra-examiner error was 18% based on a mean pressure of 6.64 g, standard deviation 1.199 g. Imaging MRI was performed using a 3 T head scanner (Siemens Allegra) with a quadrature birdcage coil. Morphological T1-weighted images with a resolution of 1 × 1 × 1 mm 3 using a magnetisation prepared gradient echo sequence (MPRAGE) were acquired. Functional echo-planar image volumes of the whole brain (number of slices = 49, thickness = 3 mm, voxel size = 3 × 3 × 3 mm 3 ) sensitised to the Blood Oxygenation Level Dependent (BOLD)-effect (echo time = 30 ms) were acquired. Five scanning sessions were performed. Each session included 92 functional volumes with a temporal resolution of 3 seconds. The first two volumes in each session were discarded from further analysis to allow for initial T1-equilibrium effects. Experimental protocol Fingers 1 to 5 (D1 = thumb, D5 = little finger) of the right hand were stimulated sequentially in separate sessions according to a block design that included four periods of stimulation and five of rest for each finger. The epoch length for both stimulation and rest periods was 30 seconds. Postprocessing Image processing and analysis were carried out using the SPM99 soft ware package [ 17 ]. All functional images were resliced to a voxel size of 1.5 × 1.5 × 1.5 mm 3 and then realigned to the first image and coregistered to the T1-weighted image volume. All data were spatially filtered using an isotropic 4 mm, full-width, half-maximum Gaussian kernel. A high pass filter (cut off frequency 0.008 Hz) was applied to eliminate low frequency signal fluctuations. In order to preserve each subject's somatotopic arrangement in area 3b no normalization to a common brain atlas was performed. Data analysis Functional data from two subjects were excluded due to major motion artefacts (subject no. 7) and global lack of activation (subject no.18). Task specific effects were estimated using the general linear model (GLM) that included a box car function convolved with the canonical hemodynamic response function in SPM99. The effect of sensory stimulation of each finger versus rest was determined using a one-sample t-test of pertinent linear contrasts of parameter estimates in each subject with a significance level of p < 0.001 (uncorrected). Then the spatial coordinates of the peak activation voxel in area 3b were determined. Due to lack of neuroanatomical landmarks, exact delineation of the cytoarchitectonically defined areas within S1 cannot be achieved in MR images. Therefore we used an operational definition based on cytoarchitectonic studies of S1 on 10 post-mortem brains. In >50 % of these brains area 3a was located in the fundus of the central sulcus, area 3b in the rostral bank of the postcentral gyrus and area 1 on its crown reaching down into the postcentral sulcus [ 18 , 19 ]. Fig 1 illustrates these locations, that continue along the central sulcus. As interareal borders vary across brains, the same authors constructed probability maps for each area by superimposing histological volumes of the individual brains on a computerized reference brain. Volumes of interest (VOIs) were defined for each area in which >50 % of the brains had a representation of that area. Despite close relationship of areas 3a, 3b and 1 in the postcentral gyrus, the three VOIs overlapped by <1% of their volumes. These probability maps were at hand when the spatial coordinates of the peak activation voxel in area 3b were determined. Activated peak voxels were labelled as belonging to area 3b when they were located within the anterior wall of the postcentral gyrus (Figure 1 ). The spatial coordinates of the peak voxel for the thumb (D1) was defined as being at origo [0, 0, 0] in a 3D Cartesian coordinate system.When the spatial coordinates for all fingers were known in all subjects, somatotopy was assessed by determining the average distances for the whole group from each finger to the thumb. Euclidian distances between fingers as well as from each finger to the thumb (D1) were calculated as: with x DI , y DI and z DI representing the coordinates of the finger DI and x DII , y DII and z DII representing the coordinates of DII in the three directions x, y, z in a system where the coordinates of D1 are at origo [0, 0, 0]. Statistics Each finger's distance to the first finger was compared to that of the directly neighbouring fingers using the Wilcoxon matched pairs test with a significance level of p < 0.05. The Euclidian distances to the first finger were compared for each finger using the Wilcoxon matched pairs test with a significance level of p < 0.05. Authors' contributions DvW conceived of the study, performed the image analysis and drafted the manuscript. PF guided the image analysis. JO designed the MR-parameters. EML participated in the design of the study. BR and GL initiated and conducted the stimulation consistency study. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517711.xml
514556	Assessing computer skills in Tanzanian medical students: an elective experience	Background One estimate suggests that by 2010 more than 30% of a physician's time will be spent using information technology tools. The aim of this study is to assess the information and communication technologies (ICT) skills of medical students in Tanzania. We also report a pilot intervention of peer mentoring training in ICT by medical students from the UK tutoring students in Tanzania. Methods Design: Cross sectional study and pilot intervention study. Participants: Fourth year medical students (n = 92) attending Muhimbili University College of Health Sciences, Dar es Salaam, Tanzania. Main outcome measures: Self-reported assessment of competence on ICT-related topics and ability to perform specific ICT tasks. Further information related to frequency of computer use (hours per week), years of computer use, reasons for use and access to computers. Skills at specific tasks were reassessed for 12 students following 4 to 6 hours of peer mentoring training. Results The highest levels of competence in generic ICT areas were for email, Internet and file management. For other skills such as word processing most respondents reported low levels of competence. The abilities to perform specific ICT skills were low – less than 60% of the participants were able to perform the core specific skills assessed. A period of approximately 5 hours of peer mentoring training produced an approximate doubling of competence scores for these skills. Conclusion Our study has found a low level of ability to use ICT facilities among medical students in a leading university in sub-Saharan Africa. A pilot scheme utilising UK elective students to tutor basic skills showed potential. Attention is required to develop interventions that can improve ICT skills, as well as computer access, in order to bridge the digital divide.	Background Developments in information and communication technology occur at an astonishing rate. The World Wide Web (WWW) doubled in size during the first 6 months of 2000 and by 2005 the number of Internet users is likely to pass the one billion mark [ 1 ]. This has had huge implications for medical practice throughout the world. One estimate suggests that by 2010 more than 30% of a physician's time will be spent using information technology tools [ 2 ]. But these developments are occurring in a world that many of our colleagues cannot access. The International Labour Organization's World Employment Report for 2001 noted that barely 6% of people in the world had ever logged onto the Internet, and 85–90% of these are in the industrialized countries [ 3 ]. In September 2000, the digital divide was highlighted by the World Health Organization as 'more dramatic than any other inequity in health or income' [ 4 ]. In a world afflicted by poverty, debt and HIV, why has the digital divide continued to trouble so many academics and development policymakers? The basic concern is that the spread of information and communication technologies in developed countries is leaving the rest of the world behind. The development of online databases allows medical professionals throughout the developed world immediate access to hundreds of e-journals at the touch of a button, a striking contrast to the plight of many of their colleagues in developing countries who are forced to trawl empty libraries. Highlighting one of the greatest tragedies of the digital divide, it threatens the very communities that could benefit the most from the developments in ICT. Many programmes have concentrated on increasing the number and spread of telephones and computers [ 5 ]. Other schemes have minimised cost barriers to accessing Internet resources [ 6 ]. Beyond this classic access gap, several other factors have been identified as contributing to the divide. These include a gap in ability to use ICT, measured as the skills base; a gap in the actual use, measured as amount of time spent utilising ICT facilities [ 5 ]; and a gap in the impact of use, measured by financial, economic and clinical returns. In other words, equipment alone is useless unless people are able to use it effectively and informed of the potential benefits of its use. During January and February 2003 we studied ICT skills of medical students at Muhimbili University College of Health Sciences (MUCHS), University of Dar es Salaam, in Dar es Salaam, Tanzania. We also report a pilot intervention of peer mentoring training in ICT by medical students during the elective period. Methods Setting MUCHS is the largest medical school in east Africa and the only public medical school in Tanzania. A lack of resources has resulted in the university library being filled with out-dated textbooks. The majority of its graduates go on to serve rural communities throughout Tanzania and its neighbouring countries, often as the only qualified doctor serving populations of over 100,000. Access to computer facilities is a key problem. The University of Dar es Salaam is aiming for a ratio of one computer for every ten students. One-hundred-and-twenty computers would therefore be required for the 1200 MUCHS students. Help from international donors has allowed MUCHS to secure the presence of 40 computers, but only 25 are fully functional at any one time. Of these, less than half were connected to the Internet or loaded with basic software, bringing the real ratio of students to adequately equipped computers to around 100:1. Most MUCHS students found commercial Internet cafes too expensive to use on a regular basis. The cost of one hour at an Internet café can often be as high as $1, an important limiting factor considering that over 50% of an estimated 36 million people live in extreme poverty, surviving on less than US $1 per day [ 7 ]. Methods The abilities and attitudes of the fourth year MUCHS medical students (MD4s) towards ICT was assessed using Questionnaire 1 [see Additional File 1 ], an adapted version of a questionnaire developed by Jeannette Murphy j.murphy@chime.ucl.ac.uk at the Centre for Health Informatics and Multiprofessional Education (CHIME, ) in London, UK, to assess ICT skills amongst first year medical students (MD1s) attending University College London (UCL). The questionnaires were distributed to all MD4 students by Tanzanian student representatives, to be filled in independently, and were then collected by the representatives. The questionnaire addressed different ICT-related variables associated with generic skills (Figure 1 ) and specific skills (Figure 2 ). A self-reporting assessment of competence (none, very basic, average or advanced, the equivalent of 0, 1, 2 or 3 points respectively) on several topics was evaluated as part of a generic ICT score. A specific ICT score aimed to address similar abilities (presence or not, 0 or 1 point) but asking about their abilities to perform such tasks. The generic ICT score has a range between 0 and 33 (11 items studied), and the specific one between 0 and 16 (16 items in total). Figure 1 Generic ICT Skills of 92 Medical Students at Muhimbili University College of Health Sciences. The data are shown as percentages of students reporting average and advanced competences. Figure 2 Specific ICT Skills of 92 Medical Students at Muhimbili University College of Health Sciences. The data are shown as percentages of students reporting to have the abilities to perform these tasks. Information related to frequency of computer use (hours per week) and years of computer use was also gathered. Access to computers at home or at educational facilities, last time of computer use, reasons for use and resources for reference in students' medical studies was also evaluated. The data were analysed using SPSS v.11 to calculate frequencies and percentages. Pearson Rho coefficient was used to look at the correlation between variables, with logarithmic transformation where necessary. Comparisons of frequencies before and after tutoring were performed using Wilcoxon's test. Data are shown as mean (± SD) for normally distributed data and median (interquartile [IQ] range) for skewed data. Results Ninety two (72.7%) of 120 Tanzanian MD4s completed the questionnaires. 76% of them did not have a computer at home and 74% never use a computer as part of any course either at school or university. Only 48 students (52%) felt that they understood the basic terminology and concepts of computing. The mean (± SD) of the generic and specific ICT scores were 11.1 (± 7.6) (out of a maximum score of 33) and 7.7(± 4.1) (out of a maximum score of 16) respectively. The two scores were significantly correlated, r = 0.81, p < 0.001. The highest levels of competence, assessed using the generic ICT score, were for email, Internet and file management (see Figure 1 ). For the remaining items most respondents reported low levels of competence. Over 60% of the Tanzanian students in each of the generic areas indicated that they had taught themselves these skills. Of the remainder most students had learnt the skills at school, with a small number learning them at work. The results for the tasks evaluated in the specific ICT score are shown in Figure 2 . Of 16 skills evaluated, only 2 (12.5%) were present in 90% of the participants. The majority of skills were found in between 40 and 60% of participants. The majority of the students claimed to use the available computers very regularly, 25% of students using them daily and nearly two-thirds at least once a week. The median hours per week of computer use was 3.8 (2–10). The median years of computer use was 3 (2–5) years. The main reasons for using a computer during the last year was to communicate by email (75%), Internet navigation (33%), learning purposes (27%), and to prepare reports (22%). Only 21 students (23%) had ever consulted an electronic journal, and nearly 70% did not use any electronic resource. The participants unanimously agreed that medical students should receive training in the use of computers. Piloting a peer mentoring training scheme During our time in Tanzania we piloted an ICT peer mentoring training scheme aimed at the fourth year Tanzanian medical students whose skills had been assessed. We utilized tutorials that had been developed by CHIME for underskilled UCL medical students [ 8 ]. Topics covered included file management, Internet, email, word processing, Excel and PowerPoint. The willingness of the Tanzanian students to participate was identified as vital to the success of the scheme, since one of our main concerns was that we would we be seen as a neo-colonial force trying to impose our values. This was not the case: the response from the students was overwhelmingly positive. The majority of those who returned the questionnaires wished to receive peer mentoring training. Four tutorials (1.0–1.5 hours) were conducted utilising MUCHS computers. The course covered computing skills such as file management, word processing, spreadsheet use, Internet access and email. Competence levels were compared for 12 students before and after 4 to 6 hours of peer mentoring training using Questionnaire 2 [see Additional File 2 ]. The generic ICT score in these students increased from 8.8(+5.2) to 18.4 (+4.8) out of a maximum score of 33 (p = 0.007). The specific ICT score increased from 6.8 (+3.6) to 12.1(+2.2) out of a maximum score of 16 (p = 0.003). There were significant improvements (p < 0.05) in 7 of the 11 components of the generic ICT score and 5 of the 16 items of the specific score. Discussion We have assessed ICT competence of a representative sample of 4th year medical students in a Tanzanian medical school and have found substantial limitations in computing skills. A mean generic score of 11.1 would be equivalent to approximately 1 point (very basic) in each of the 11 skills studied. The corresponding score among first year UK medical students in 2002 was 18.5. Initially at UCL, students with an overall score of less than 10 are considered to have low skills and are offered peer mentoring training. Using this criterion, around 50% of the Tanzanian students would fall into the low skills category compared with 9% of first year UCL medical students in 2002. Scores amongst UCL medical students have been rising in recent years (J Murphy, personal communication). The mean of the specific ICT score was 7.7, in comparison with 11.9 for first year UCL medical students. (J Murphy, personal communication). The strong correlation between generic and specific scores suggests that the students were good at judging their own ability, thus decreasing a potential bias from self-assessment in our results. Two studies from Nigeria show that there is poor knowledge of computer use. Ajuwon [ 9 ] reports that only 42.6% of the sample studied could use the computer. In Lagos [ 10 ], 79% of students had little or no computer skills. Although we did not evaluate self-reporting ability to use a computer, we can assume that this figure is similar considering the low averages of the scores reported in our study. The Internet was the most common application utilised in our sample in Tanzania, especially for email communications (75%). This is concordant with the reports from Nigeria where 76.4% of first year clinical and nursing students in Ibadan [ 9 ] and 58% of final year medical and dental students in Lagos [ 10 ] have used email. This high rate of Internet and email use amongst medical students is also similar in other countries, such as Denmark [ 11 ], Finland [ 12 ], India [ 13 ], Malaysia [ 14 ] and the United Kingdom [ 15 ]. Around half of the Tanzanian students were able to print a document, cut and paste information from one application to another and word process an essay, letter or their CV. These results, clearly addressed through items in the specific ICT score, show us that even though the students are able to communicate by email and use the Internet, they cannot perform some basic skills necessary for working with files. The lack of confidence in use of packages such as PowerPoint and Excel was also noticeable in the reported score and in the reported reasons for using the computer. The median duration of computer use was 3 years, meaning that most doctors in training had their first contact with a computer at university. Access is one of the main issues when one considers the fact that the current student/computer ratio at MUCHS is 100:1, compared to 35:1 in Portugal, 9:1 in the UK and 5:1 in Norway [ 16 ]. Seventy-six percent of students in our sample did not have a computer at home. This figure is in stark contrast with the 71.7% of first year medical students that do have access to a computer at home in Aarhus, Denmark [ 11 ] and 86% in California, USA [ 17 ]. These findings have substantial implications for addressing the digital divide in this population. These range from a marked limitation of knowledge of basic packages to the possibility of making incorrect assumptions that students who are able to communicate electronically are automatically able to perform basic tasks such as word processing. Graduates recruited by foreign institutions to pursue postgraduate training would face additional difficulties in their learning experience due to their problems with ICT. Also important is the fact that several academic institutions are now looking to expand their teaching programmes using online courses. Again, course organisers could erroneously enrol participants with a lack of basic ICT skills, based solely on the ability to communicate by email. The peer mentoring training piloted in the present study shows promising results, achieving a doubling of competence scores. This programme [ 18 , 19 ], which uses foreign medical students visiting MUCHS to teach local ones, is a cheap intervention and feasible to replicate in the same context, and perhaps in other countries, in a similar way to the buddy system that has been suggested by Ajuwon in Nigeria [ 9 ]. Conclusions Our study has found a low level of ability to use ICT facilities among medical students in one university in sub-Saharan Africa. These findings reinforce the idea that more is needed to bridge the digital divide than simply increasing the number of computers. A pilot scheme utilizing visitors from the developed world to tutor basic skills showed potential, but it would be naïve to think that volunteers alone can bridge the gap. Some would argue that increasing the number and distribution of computers will eventually result in an improved skills base. Our experiences would support this. Tanzanian medical students are keen to learn new skills and are aware of the implications of being left behind in the technological revolution. The main concern of such a scenario is that without direct intervention the time required to attain these skills increases, to the further detriment of some of the world's most vulnerable societies. Competing interests None declared. Authors' contributions JCC, RM, MS, EJWY and JJM took part on the initial planning of the project. JCC, RM, MS and EJWY did the field work and the peer mentoring training to Tanzanian medical students. PA and JJM did the data input and analysis. MS, JJM and JCC wrote the initial version of the paper. All the authors contributed to subsequent versions of the paper. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 Questionnaire 1 Instrument used to assess baseline ICT skills amongst MUCHS medical students. ©Jeannette Murphy Click here for file Additional File 2 Questionnaire 2 Instrument used to assess skills after the ICT peer mentoring training amongst MUCHS medical students. © Jeannette Murphy Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514556.xml
526282	Is nebulized saline a placebo in COPD?	Background Many trials of nebulized therapy have used nebulized saline as a "placebo". However, nebulized isotonic saline is sometimes used to assist sputum expectoration and relieve breathlessness in COPD patients. We designed this study to establish if nebulized saline had a placebo effect or a clinical effect. Methods 40 patients were studied following hospital admission for exacerbated COPD (mean FEV1 30% predicted). Patients were randomised to single-blind administration of either 4 mls of nebulized isotonic saline using an efficient nebulizer (active group n = 20) or an inefficient nebulizer (placebo group n = 20). Spirometry and subjective breathlessness scores (Modified Likert Scale) were measured before nebulized treatment and 10 minutes after treatment. Results There was no significant change in FEV1 after active or placebo nebulized saline treatment. Patients reported a 4% improvement in mean breathlessness score following placebo (Wilcoxon test; p = 0.37) compared with 23% improvement following active nebulized saline (p = 0.0001). 65% of patients given active nebulized saline but only 5% of the placebo group reported that mucus expectoration was easier after the treatment. Conclusions This study lends support to the current use of nebulized saline to relieve breathlessness (possibly by facilitating sputum clearance) in COPD patients. Lung function was not affected. Nebulized saline can therefore be used as a placebo in bronchodilator studies involving COPD patients but it cannot be used as a placebo in trials assessing symptom relief.	Background Nebulized saline is used by some doctors and physiotherapists to assist mucus clearance and to relieve breathlessness in patients with COPD, bronchiectasis and Cystic Fibrosis. This practice is justified by a small number of studies which have demonstrated enhanced sputum expectoration or improved breathlessness after nebulized saline or humidified oxygen [ 1 - 4 ]. Nebulized hypertonic or isotonic saline has been used to obtain induced sputum specimens from patients with asthma and COPD for diagnostic and experimental purposes [ 5 - 9 ]. For example, Vlachos-Mayer and colleagues [ 5 ] used increasing strengths of nebulized saline (from isotonic up to 5%) to induce sputum in 304 patients with asthma and 25 patients with COPD. Satisfactory specimens were obtained in 93% of cases, 17% of asthmatic patients and 56% of COPD patients required only isotonic saline to achieve sputum induction. However, nebulized saline has also been used as a placebo in several trials involving nebulized bronchodilator therapy [ 10 , 11 ]. For example, Jenkins et al [ 10 ] found in a double blind study that patients reported clinical benefit from nebulized saline (with MDI bronchodilator therapy) which was similar to the subjective response to nebulized bronchodilator therapy given with placebo MDI therapy. It was assumed that these patients had a placebo response to nebulized saline but it is also possible that they may have experienced a non-bronchodilator benefit from nebulized saline. We have designed a trial to determine whether the symptomatic benefit associated with nebulized saline use in clinical trials is a placebo effect or a non-bronchodilator therapeutic effect. Methods 40 patients were studied during a hospital admission for an exacerbation of COPD. Patients were recruited at a time when their condition had stabilized prior to their planned discharge from hospital. Clinical details of the patients are summarised in table 1 . Six patients undertook both limbs of the study (partial crossover design). Table 1 Patient Characteristics Active group Placebo group Gender 13 Male, 7 Female 12 Male, 8 Female Age (Mean, SD) 68.1 +/- 7.2 Range 54–79 67.3 +/- 7.4 Range 58–79 Mean FEV1 95% CI 0.77 Litres 0.65–0.89 0.78 Litres 0.61–0.95 FEV1 as % predicted (Mean and SD) 29.9 (10.0) 30.6 (14.8) Patients were randomised to receive 4 mls of 0.9% saline using an efficient nebulizer system (active group) or an inefficient nebulizer system (placebo group). The active nebulizer was a System 22 Acorn nebulizer (Medic-Aid, Bognor Regis UK Ltd) driven by the hospital's piped oxygen supply at a flow rate of 9 l/min for 10 minutes. This nebulizer system was found to deliver 95% of particles in the size range 2.5 to 2.8 microns using a Malvern laser system. (Measurement courtesy of Dr Steve Newman, Principal Physicist, Royal Free Hospital, London, UK). This small particle size was selected to achieve effective delivery to the airways. The placebo nebulizer was an old model (1980s) Bard Inspiron nebulizer (no longer manufactured) driven by oxygen at a flow rate of 3 l/min. This nebulizer system delivered 95% of particles in the size range 9.5 to 9.9 microns. This particle size was selected to achieve a placebo effect with deposition in the tubing of the system and in the pharynx but little penetration to the airways [ 12 ]. Both nebulized treatments were administered by mouthpiece to avoid nasal deposition of saline droplets and to make it less likely that patients would notice that the output from the placebo system was different to previous nebulized treatment which they had received. The trial was conducted in a single-blind manner. 40 slips of paper were labelled "Treatment A" or "Treatment B" and placed in opaque brown envelopes. These were shuffled in random order and each patient was asked to select one envelope. This was then opened by the investigator and the appropriate treatment was administered (A active, B Placebo). For the six patients who took part in the study twice, the second treatment consisted of whichever treatment they had not received previously. Patients were told that we wished "to observe the effects of a nebulized treatment which is not a new or experimental drug". They were not informed of the exact nature of the nebulized treatments as this might have led patients to try to guess if the treatment which they received was a "placebo". The Ethics Committee agreed that it would not have been possible to measure a true placebo effect if patients were made aware that both treatments were saline (not a bronchodilator) and one of the nebulizers was deliberately made to run inefficiently. Patients were recruited on the Respiratory Wards of a University hospital. We recruited patients who had diagnosis of COPD confirmed by a respiratory consultant (patients with asthma or bronchiectasis were excluded from the study). Patients were approached by one of the investigators whilst in a relatively stable phase prior to discharge from hospital following an admission for exacerbated COPD. All testing was undertaken between 12.00 and 16.00, at least four hours after bronchodilator treatment. Prior to participating in the study, patients gave informed consent and undertook baseline measurement of FEV1 and FVC using the best of 3 blows on a Microlab 3300 Spirometer (Micro-Medical LTD, Rochester, UK. Peak Expiratory Flow (PEF) was measured using a Wright's Peak Flow meter. Each patient also recorded an assessment of their perceived level of breathlessness using a seven point modified Likert scale (1 = Not breathless, 2 = Very mild breathlessness, 3 = Mild breathlessness, 4 = Moderate breathlessness, 5 = Severe breathlessness, 6 = Very severe breathlessness, 7 = Worst possible breathlessness). Ten minutes after completion of nebulized therapy, FEV1, FVC and PEF measurements and subjective breathlessness scores were repeated. Patients also recorded a subjective assessment of benefit using the following modified Likert scale. (1 = No benefit from this treatment, 2 = Very slight benefit, 3 = Slight benefit, 4 = Moderate benefit, 5 = Good benefit,, 6 = Very good benefit, 7 = Best possible benefit). Patients then received 4 puffs of salbutamol (400 mcg) using a Metered Dose Inhaler and 750 ml Volumatic spacer (Glaxo Smith Kleine UK). Fifteen minutes later, FEV1, FVC and PFR measurements and subjective breathlessness scores and symptom relief scores were repeated. All data was entered on a SPSS version 9 statistical package. Mann Whitney U-test was used to compare lung function tests and symptom relief scores. Wilcoxon Signed Rank test was used to compare the change in breathlessness scores for matched pairs before and after nebulized saline. The study was approved by Salford and Trafford Research Ethics Committee. All patients gave written informed consent to partake in the study and to receive a single dose of nebulized treatment (in addition to all usual treatment). Results 34 patients completed the study; patient details are summarised in table 1 . 6 patients took part in the study twice (once in each limb). This allowed 20 treatments with each nebulizer system to be compared. The baseline FEV1 of the two treatment groups was well matched. Both groups had a non-significant fall in FEV1 after nebulized saline therapy and a small rise in FEV1 after 400 mcg salbutamol from MDI-spacer (Table 2 ). FVC, and PEF changes (not shown in table) were similar to FEV1 changes. Table 2 Results All results expressed as Medians in top line and Mean (and 95% CI) in second line. Active group Placebo group P value ( Mann Whitney ) FEV1 0.77 0.80 0.84 Pre-treatment 0.77 (0.65–0.89) 0.78 (0.61–0.95) FEV1 0.75 0.73 0.74 Post nebulized saline 0.73 (0.62–0.84) 0.75 (0.59–0.91) FEV1 0.80 0.77 0.63 Post salbutamol MDI 0.81 (0.67–0.94) 0.79 (0.62–0.96) Breathlessness 4 4 0.34 Score 1 (Pre-treatment). 3.9 (3.6–4.3) ( 1 = Not breathless, 7 = Worst possible breathlessness ) 3.5 (3.0–4.0) Breathlessness 3 4 Score 2 (Post nebulized saline) 3.0 (2.6–3.5) 3.3 (2.8–3.9) 0.85 Wilcoxon test Score 1 V Score 2 <0.0001 0.37 Breathlessness 3 3 0.35 Score 3 (Post salbutamol MDI) 2.9 (2.5–3.3) 3.0 (2.6–3.5) Wilcoxon test Score 2 V Score 3 0.43 0.014 Symptom Relief 3 1 0.0002 Score Post nebulized saline 3.1 (2.7–3.6) ( 1 = No benefit, 7 = Best possible benefit ) 1.7 (1.2–2.3) Symptom Relief 3 3 0.37 Score Post salbutamol MDI 3.2 (2.7–3.7) 2.9 (2.4–3.4) The placebo group had a 4% improvement in breathlessness after treatment (Wilcoxon p = 0.37) compared with a 23% improvement after active nebulized saline (Wilcoxon p = 0.0001). This corresponded to a reduction from 4/7 (moderate breathlessness) before treatment to 3/7 (mild breathlessness) after treatment in the active treatment group. The mean symptom relief score (patient's assessment of benefit) for the active treatment was 3/7, (slight benefit) almost identical to the response to 400 mcg salbutamol from MDI. The placebo group had a score of 2/7(very slight benefit) after nebulized placebo and 3/7(Slight benefit) after 400 mcg salbutamol from MDI. 15 patients in the active group felt better after nebulized saline, 5 felt the same and no patient felt worse. Six patients in the placebo group felt better after nebulized treatment, 12 felt the same and 2 felt worse (Chi Squared test p = 0.013). Patients were asked if the nebulized treatments had any effect other than relief of breathlessness. 13 patients in the active group (65%) said that the nebulized treatment assisted sputum expectoration. Only 1 patient in the placebo group reported this effect (Difference between groups: -Fisher exact test, p = 0.0001). No patient reported any adverse effects from either nebulized treatment. Discussion This is the first study which has compared active nebulized saline with placebo nebulized saline. The results suggest that nebulized saline has non-bronchodilator therapeutic effects that are possibly explained by airway-moistening and sputum-inducing effects of nebulized saline. Sputum volume was not measured in the present study but two thirds of patients who were given nebulized saline through an efficient nebulizer system reported that it helped them to expectorate sputum. This finding is consistent with the results of previous studies which have shown improved sputum clearance and decreased breathlessness following the open administration of nebulized saline [ 1 , 3 ] The results of the present study may be explained by airway-moistening and sputum-inducing effects of nebulized saline, both isotonic and hypertonic [ 1 - 9 ] The study of Vlachos-Mayer and colleagues [ 5 ] showed that most asthmatic patients required hypertonic saline to achieve sputum induction but more than half of COPD patients achieved sputum induction with nebulized isotonic saline (similar to the finding that 65% of COPD patients in the present study reported enhanced sputum clearance following nebulized saline). Previous studies have shown that nebulized saline can have a bronchoconstrictor effect in some patients which is greater with hypertonic saline than with isotonic saline and greater in asthma patients than COPD patients [ 5 , 6 , 8 , 9 ] Nebulized isotonic saline had no significant effect on FEV1 in the study of Poole et al [ 3 ] or in the present study. The main strength of the present study is inclusion of a placebo limb using an inefficient nebulizer system. Patients in the placebo group believed that they were receiving a nebulized treatment because a placebo effect could have been abolished if patients were told that both treatments involved no active medication and one of the patients involved an inefficient nebulizer system. This issue was discussed fully with the ethics committee and found to be acceptable because the patients did not miss any of their regular medication and they did not receive any pharmacological treatment. The use of a mouthpiece ensured that patients could not see or feel that the output from the experimental system was different to the nebulized bronchodilator therapy which they had received during their hospital admission (usually via a facemask). Furthermore, only 6 patients took part in both limbs of the study so most patients could not have tried to guess which treatment was more effective. The 23% improvement in breathlessness in the active group was equivalent to the subjective benefit following 400 mcg of salbutamol from MDI-spacer. This improvement in breathlessness occurred without bronchodilation, mirroring the findings of Poole et al [ 3 ]. Based on the patients' observations and the results of previous studies, we believe that the therapeutic effect of nebulized saline may be produced by enhanced sputum clearance. A previous study at this hospital showed a similar subjective response to nebulized saline (given at 7 am) but the previous study also reported an improvement in FEV1 and PEF [ 13 ]. Patients in the previous study received nebulized saline on awakening, prior to their first bronchodilator treatment of he day. In these circumstances, it is likely that the nebulized saline assisted the expectoration of copious overnight secretions in the airways with some subsequent improvement in airflow. Patients in the present study were treated at about mid-day, having had bronchodilator therapy on awakening. It is therefore not surprising that the beneficial effects of nebulized saline were more modest in the present study. However, this study lends support to the common clinical practice of allowing patients with COPD to have nebulized saline "as required" as a supplement to regular nebulized bronchodilator therapy. This may assists sputum expectoration and relieve breathlessness without the side-effects that would occur if additional beta agonist treatment were given. This study is in agreement with previous studies which have shown no bronchodilator effect (or a small bronchoconstrictor effect) when nebulized saline is given to patients with COPD [ 3 ]. This justifies the continuing use of nebulized saline as a placebo treatment in clinical trials of bronchodilator therapy which measure rise in FEV1 or PEF as the primary outcome measure. However, as nebulized saline has non-bronchodilator therapeutic effects, it cannot be used as an inert placebo treatment in clinical studies where breathlessness or quality of life are to be measured. For example, Jenkins et al concluded that nebulized treatment had a strong placebo effect because patients expressed a preference for nebulized treatment even though the same bronchodilator effect could be achieved for most of their patients when nebulized saline was given with active bronchodilator therapy from a MDI device [ 10 ]. It is likely that many of these patients experienced a non-bronchodilator therapeutic benefit such as enhanced mucus clearance during nebulized saline therapy. It would be possible to co-administer nebulized saline with MDI bronchodilator therapy as an alternative to nebulized bronchodilator therapy for some patients with COPD who report difficulties with mucus clearance. However, this would be more inconvenient than nebulized bronchodilator therapy (and at least as expensive). For future clinical trials it would be possible to have two control groups, one receiving nebulized saline using an efficient system and one group using an inefficient system such as that used in the present study. This would allow investigators to assess whether nebulized saline had any therapeutic effect on their patients and it would also assess the true placebo response rate. British and European nebulizer guidelines state that most patients with airflow obstruction should be treated with hand-held devices unless they have demonstrated clear additional benefit from the use of nebulized treatment in carefully monitored domiciliary studies [ 12 , 14 ]. The present study supports these recommendations, especially the provision that some patients may be commenced on nebulized treatment on the basis of substantial subjective benefit even if an additional bronchodilator response cannot be demonstrated. This study also supports the present practice of many physiotherapists and doctors who use nebulized isotonic saline to assist sputum clearance for patients with COPD who have difficulty in expectorating sputum. Abbreviations FEV1 Forced Expiratory Volume in 1 second FVC Forced Vital Capacity PEF Peak Expiratory Flow COPD Chronic Obstructive Pulmonary Disease MDI Metered dose inhaler Competing interests The authors declare that they have no competing interests. Authors' contributions BROD developed the concept for the study and both authors designed the study protocol. SYK recruited patients and performed all study measurements. Both authors assisted in analysis of the data and preparation of the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526282.xml
547898	Large scale hierarchical clustering of protein sequences	Background Searching a biological sequence database with a query sequence looking for homologues has become a routine operation in computational biology. In spite of the high degree of sophistication of currently available search routines it is still virtually impossible to identify quickly and clearly a group of sequences that a given query sequence belongs to. Results We report on our developments in grouping all known protein sequences hierarchically into superfamily and family clusters. Our graph-based algorithms take into account the topology of the sequence space induced by the data itself to construct a biologically meaningful partitioning. We have applied our clustering procedures to a non-redundant set of about 1,000,000 sequences resulting in a hierarchical clustering which is being made available for querying and browsing at . Conclusions Comparisons with other widely used clustering methods on various data sets show the abilities and strengths of our clustering methods in producing a biologically meaningful grouping of protein sequences.	Background With the overwhelming growth of biological sequence databases, handling of these amounts of data has increasingly become a problem. Protein sequences constitute one such data type. The number of unique entries in all protein sequence databases together exceeds now about a million. However, biological evolution lets proteins fall into so-called families, thus imposing a natural grouping. A protein family contains sequences that are evolutionarily related. Generally, this is reflected by sequence similarity. Therefore, one aims at organizing the set of all protein sequences into clusters based on their sequence similarity. Clustering a large set of sequences as opposed to dealing only with the individual sequences offers several advantages. A frequent problem is the identification of sequences that are similar to a new query sequence. This task can be executed much quicker when only one comparison to an entire cluster has to be performed rather than one comparison per database sequence. Another application lies in the possibility of analyzing evolutionary relationships among the sequences in a cluster and of the species they come from. Moreover, the presence or absence of sequences of a group of species can give useful information about their evolutionary relationship, if their complete set of protein sequences is known. The aim of clustering protein sequences is to get a biologically meaningful partitioning. One of the simplest well-studied and computationally cheap methods to construct a clustering of data points is single linkage clustering . Starting with the pair of data points of least distance, one incrementally merges single data points or already existing clusters. Such a hierarchical clustering can be viewed as a tree, called the single linkage tree . The leaves represent the individual data points, while the root of this tree corresponds to just one large cluster representing the whole data set. All other layers in between can be seen as cluster sets at different levels of similarity. However, it is not clear which layers give a meaningful partitioning of the data. They should be chosen so that they neither produce small trivial clusters nor form huge uninformative clusters. Several approaches already deal with the problem of partitioning a protein sequence database into protein families. Automatically generated cluster sets like ProtoMap [ 1 ], ProtoNet [ 2 ], or CluSTr [ 3 ] typically provide a hierarchical classification at several different levels of similarity. Others, like iProClass [ 4 ] or PIRSF [ 5 ] include further knowledge, e.g., from domain based classifications, or require manual interaction. Kawaji et al . [ 6 ] recently developed a graph-based clustering method for the detection of distantly related sequences of a protein family. TribeMCL [ 7 ] is a method for clustering proteins into 'protein families' using a Markov Clustering method. It is primarily used for comparing protein sequence sets of completely sequenced genomes. Reviews of currently available cluster sets can be found in: Heger et al . [ 8 ] and Liu et al . [ 9 ]. In our approach we first exploit the branching structure of the single linkage tree, which elucidates an unexpected structuring of the sequence space. Traversing the tree from a leaf towards the root we inspect the sizes of the merging subtrees. First one notices relatively small increases that correspond to very similar proteins. Later on, sequences merging in correspond to weakly related proteins. At one point, however, we observe an enormous increase in the size of the subtree, where a large part of the database merges in. All sequences below this point in the tree are assumed to belong to the same superfamily . Each superfamily typically covers several closely related protein families. They can be determined by revealing the connectivity of the sequences belonging to a superfamily. Since the single linkage tree is built using only the smallest distances connecting subtrees, information about the connectivity within these subtrees is lost in the hierarchy. For each superfamily, we construct a superfamily distance graph by including only those nodes labeled with sequences belonging to the respective superfamily. These graphs are then split at reasonable cut sites into highly connected subclusters . For historical reasons [ 10 ], we call our procedure as well as the resulting cluster set SYSTERS, which is short for SYSTEmatic Re-Searching . Up to that point, the hierarchy consists of superfamily and family clusters. However, protein sequences are built up of smaller entities, called domains . They again can be grouped independently of a certain order in a protein sequence. For this level we rely on one of the currently established domain databases, i.e., the Pfam database [ 11 ]. To allow the user to explore protein sequence space through the complete hierarchy, we present an interface to our cluster set on the Internet. It is possible to enter the hierarchy at each of the layers through various entry points and change to another layer whenever desired. Additional information like a multiple alignment or a phylogenetic tree is given for each of the family clusters. Here, we explain in more detail the SYSTERS algorithms developed to determine superfamily and family clusters. Each step is illustrated by an example. We report our results on clustering the non-redundant protein sequence space consisting of about 1,000,000 sequences. An overview of the availability and accessibility of the cluster set is given. Finally, we present a comparison of our clustering method with two other currently available and widely used clustering methods, namely single linkage clustering and TribeMCL. Results and discussion Clustering We have applied our algorithms as described in the Methods Section to a sequence set consisting of all known protein sequences from the Swiss-Prot Rel. 41 and TrEMBL Rel. 23 databases [ 12 ], and from several completely sequenced organisms [ 13 - 16 ]. The original set contains 1,168,542 sequences. Sequences which are too short to yield a result in the database search are removed from this set. Sequences which are identical (sub-)sequences of other sequences are sorted together and only the longest sequence is retained as the representative. In a pairwise comparison of all remaining 969,579 non-redundant sequences, this results in a triangular matrix sparsely filled with 775,133,144 E-values better than or equal to 0.05. Comparisons of a sequence to itself are not considered. By temporarily removing all those sequences which are at least 80% identical over at least 80% of their entire length to another sequence, this number decreases. These sequences are considered redundant, and are added to the cluster set again later in order to retain their annotation. By reducing the number of sequences to 546,538 non-redundant sequences, the remaining number of pairwise comparisons decreases significantly. Fortunately, the resulting triangular distance matrix turns out to be sparsely filled with only 52,618,818 values (0.035% of all possible pairs). Constructing the distance graph with these values, the data splits into 93,918 connected components with 76,347 components consisting of only one sequence. The resulting single linkage tree divides into 147,796 superfamilies with 110,308 of them consisting of only one sequence. The subclustering splits the data further into 158,153 family clusters with an overall number of 110,322 single sequence clusters. Access to the cluster set The SYSTERS cluster set [ 17 ] is available over the Internet at . There it is possible to explore the protein sequence space by navigating through the complete hierarchy consisting of superfamilies, family clusters, and domains. For the last layer in the hierarchy, the domain level, we rely on one of the currently established domain databases, namely the Pfam collection of protein domains. It is possible to enter the hierarchy at any layer, e.g., by searching for a keyword, choosing a species, or selecting a domain. For each family cluster a consensus sequence is generated. All consensus sequences together build a database searchable by BLAST. Thus, a clear assignment of a new protein or nucleotide sequence to a family and a superfamily is possible. Additional information like a multiple alignment or a phylogenetic tree is given for each of the family clusters. Whenever possible, links to external resources are provided to allow for further information, e.g., about structural properties or underlying genes. Validation For the validation of our clustering procedures we needed on one hand a "true" biologically verified cluster set and on the other hand results of other clustering procedures on this data set. Unfortunately, for large scale analyses such validated data is not available. Thus, we decided on performing our evaluations on two biologically meanigful data sets, namely well characterized sequences from Swiss-Prot and TrEMBL with (a) Pfam domain annotations and (b) ENZYME annotations. Clustering of such large data sets is not an every day routine. Normally the software to handle such data sets is not publicly available and only the results of their application are published. Although these results are mostly publicly available for browsing on the web the underlying primary data differs in all of these data sets. Additionally a systematic, unbiased and independent comparison would be intractable on a large scale by querying the web. One of the simplest well-studied and computationally cheap methods to construct a clustering is single linkage clustering. We implemented procedures to perform a single linkage clustering on the two data sets at various different cutoffs. This corresponds to performing single sequence searches with a certain E-value cut-off for all sequences in the data set with subsequent determination of the connected components of the results. Additionally we decided to compare our clustering procedure to one of the most widely used and publicly available methods for large scale protein sequence clustering, namely TribeMCL. We applied the single linkage clustering as well as the SYSTERS clustering to the Pfam data set and computed the Jaccard coeffcient, the sensitivity and the selectivity of the clustering results in comparison to the "true" cluster set as described in the Methods Section. All clusterings were performed on the non-redundant data set as described under Preprocessing in the Methods Section. After the clustering, redundant sequences were added again to the cluster sets to allow for a comparison with the "true" cluster sets. For the Pfam cluster set the best single linkage clustering with respect to the "true" cluster set can be achieved at an E-value cutoff of 1e-53 (cf. Table 1 ). The SYSTERS clustering results in a slightly higher Jaccard value. Note that the "best" single linkage clustering result can not be determined from the clustering itself, but was selected after comparison with the "true" cluster set, which is not available when clustering new sequence data. Thus, the SYSTERS clustering turns out to be superior to the single linkage clustering in the sense that it is able to determine the correct cluster granularity without manual interaction. In total we get only weak results for the Pfam data set. One of the reasons is the choice of the "true" cluster set. Figure 1 shows an example where sequences composed of the same domains and belonging to the same family of adenylate cyclases end up in different "true" clusters. The repetition of one domain and the presence/absence of another domain lets them fall into different "true" clusters. These sequences are in a biological sense correctly clustered by SYSTERS but cause a problem when comparing them to the "true" cluster set. In this case the "true" clusters build subsets of the SYSTERS subclusters. Another reason for the weak results in comparison with the Pfam data set are fusion proteins. They bring together sequences belonging to otherwise unrelated families. We applied the single linkage clustering, the SYSTERS clustering and the TribeMCL clustering to the ENZYME data sets and computed the Jaccard coeffcient, the sensitivity and the selectivity of the clustering results in comparison to the "true" cluster sets as described in the Methods Section. For this data set the SYSTERS clustering turns out to be superior to both the single linkage clustering and TribeMCL (cf. Table 1 ). In both ENZYME data sets the TribeMCL clustering shows the best ability to reject unrelated sequences but at the expense of finding distantly related sequences. As expected, the SYSTERS subclustering shows the best result on the lowest level of the ENZYME data set where individual enzymes are identified. In total all methods perform significantly better on the ENZYME data set. This data set is much smaller than the Pfam data set and contains well annotated enzymes. In contrast to the Pfam data set, the "true" cluster set was chosen on the basis of enzyme annotation, namely EC numbers, as described in the Methods section. Sequences belonging to the same "true" cluster thus may show the same domain composition but may also differ in this sense. Although this is a somehow weaker definition of a "true" cluster set it is more focussing on the functional properties of the proteins. Conclusions We have presented a hierarchical clustering of protein sequences into biologically meaningful superfamily and family clusters. A combination of an upward sweep with dynamic determination of superfamily cutoffs and a downward pass that divides superfamilies into families has been introduced. We determine a superfamily by detecting the largest increase in the size of the merging subtree traversing from a leaf in a single linkage tree to the root. We assume that at this point the twilight zone begins because suddenly a large number of supposedly unrelated sequences enters the cluster. Each of the superfamilies is further cut into family clusters by detecting weak connections in the underlying distance graph. It is interesting that both the superfamilies as well as the family clusters are generated solely from the structure of the single linkage tree (respectively the underlying distance graph), without any knowledge of the biological information represented. Such self-structuring properties have also been observed in other large data sets such as phone-call or web-link graphs [ 18 ]. An alternative approach for cluster determination is presented by Sharan et al . [ 19 ]. Their CLICK algorithm (Cluster Identification via Connectivity Kernels) uses graph-theoretic and statistical techniques to first identify tight groups of highly similar elements (kernels), which are likely to belong to the same cluster. Several heuristic procedures are then used to expand the kernels into the full clustering. In our much simpler approach, we produce a hierarchical clustering based on the partitioning into superfamilies, which already results in a biologically meaningful set of family clusters. Although the vast majority of cases we looked at are in agreement with biological knowledge, there exist some inconsistencies due to peculiarities in the data. Distinct protein families may end up erroneously in the same superfamily because of a fusion protein covering sequence information from both families. The same effect can be seen at multidomain protein families linked together by a single highly conserved common domain. Although the subclustering in most cases splits these superfamilies again into distinct families, we would prefer to take care of these cases already in the process of superfamily determination. Nevertheless, comparisons with other methods showed that our clustering methods are able to produce a biologically meaningful clustering. Thus far, our hierarchy consists of two layers representing protein superfamilies and families. For the third layer located at the domain level, we currently rely on well-established domain databases, but intend to follow our methodology also in the direction of deriving so far unknown domains. Future plans also include a regular update of the SYSTERS cluster set. Since the most time consuming part are the all-against-all sequence searches, new sequence similarities will be incrementally added instead of recalculating all similarity values. The clustering procedures themselves rely on the topology of the whole sequence space and can be run on the whole data set whenever the underlying sequence set changes. Other future developments will be in the direction of the so called tree of life . We plan to combine the evolutionary information given by each of the protein clusters to extend the knowledge about the relationship between different groups of species. Methods Clustering procedures Here we present the methods that we use to compute our clustering of protein sequences, i.e., selecting superfamilies and dividing them into reasonable family clusters. Figure 2 shows a schematic overview. Preprocessing The total number of entries in all protein sequence databases together now exceeds about a million. This number includes fragmental as well as identical (sub-)sequences from different resources. To reduce the amount of data without losing information we exclude redundant information in the form of identical and nearly identical (sub-)sequences from the data set prior to the clustering. We model the remaining protein space as a weighted undirected graph with pairwise distances attached to the edges. We decided on using E-values computed from pairwise local sequence alignments [ 20 ] as distances (all-against-all database searches were carried out on a Paracel GeneMatcher™ machine [ 21 ]). The E-value (short for Expectation value) is the number of alignments with similarity scores equivalent to or better than the score S that one expects to find in a database search by chance. Thus, the lower the E-value, the more significant is the score. Typically, matches with an E-value lower than 1e-20 are assumed to be relevant, while those sequence pairs with an E-value higher than 0.01 need further experimental evidence to be accepted as being distantly related. Values in between belong to the so called twilight zone , and a clear statement about relatedness cannot be made for them. All sequence pairs whose E-value was worse than 0.05 were assumed to be unrelated and their distance was set to infinity. We are aware that we may miss distantly related sequences with this E-value threshold in a single sequence search. However, by using each sequence in the data set as query in a database search and combining all results we hope to overcome this problem. The resulting symmetric distance matrix D contains all pairwise distances d ( s i , s j ) for each pair of protein sequences s i and s j , 1 ≤ i , j ≤ n , for which d ( s i , s j ) < ∞. Single linkage tree The distance matrix D can be represented by an undirected weighted graph G , which we call the distance graph . G = ( V , E ) is defined as follows: V = { v i \| v i = { s i }, i ∈ {1, ..., n }} and E = {( v i , v j ) \| i , j , ∈ {1, ..., n }, i ≠ j , d ( s i , s j ) < ∞}. The weight w ( v i , v j ) of an edge ( v i , v j ) ∈ E is given by w ( v i , v j ) = d ( s i , s j ). The single linkage tree is built based on the distance graph G in an agglomerative manner. The algorithm starts with a forest (collection of trees) F where each sequence corresponds to a distinct tree. As long as there are edges in the graph G , the edge with the smallest weight is selected and the adjacent nodes in G are merged. Edges linking this newly created node to adjacent ones in the graph receive the weight of the smaller of the two original edges. The two corresponding trees in F are collected together in a new tree rooted by a parental node labeled with the connecting edge weight. Finally, to allow for a better handling of the data, the resulting unconnected trees are rooted by connecting their roots to an artificial overall root node with weight infinity. Superfamily determination Different protein superfamilies display a different degree of conservation. Therefore, for each superfamily, the twilight zone starts at a different cutoff. A crucial problem thus lies in the determination of an appropriate E-value threshold for each superfamily. To this end we have devised the following procedure. For an edge of the tree linking, say, a parent p and a child q , we compute the quantity J represents the ratio between the size of all the subtrees below p without the child q and the size of the subtree below q . Watching the development of J as one walks up the tree from a leaf towards the root, one can observe that J tends to increase dramatically as one leaves the superfamily to which the leaf belongs, and then decreases again. This intuition is captured by our algorithm. For each leaf, we determine the maximum J as one proceeds from the leaf to the root of the single linkage tree. This strategy is applied to all leaves in the tree, assigning a superfamily to each leaf. In the end, inclusions are resolved by keeping the largest superfamilies. We call the internal node induced by a superfamily the superfamily root . The E-value linked to this node is called the superfamily cutoff . Refer to Algorithm 1 in Figure 3 for more details. Figure 4 shows an example of the superfamily determination. Only a part of the complete single linkage tree consisting of 290,811 leaves and 186,176 internal nodes is shown. The superfamily procedure correctly determines the ephrin family of sequences. Ephrins are membrane-attached proteins involved in the development of the nervous system and can be further distinguished into type A and type B ephrins depending on their membrane binding mechanism. Subclustering Stepping down the hierarchy of the single linkage tree starting at a superfamily root usually splits off one sequence after another, but does not lead to a meaningful partitioning into families. Since the single linkage tree is built using only the best (lowest) E-values connecting subtrees, information about the connectivity within these subtrees is lost in the hierarchy. For each superfamily we construct a distance graph that includes only those nodes labeled with sequences belonging to the respective superfamily and those of the induced edges which are labeled with a distance better than or equal to the superfamily cutoff. Let SF be the set of sequences belonging to the superfamily sf and c the corresponding superfamily cutoff. We call the connected weighted graph G = ( V , E ) with V = { v i \| v i = { s i }, s i ∈ SF } and E = {( v i , v j ) \| w ( v i , v j ) = d ( s i , s j ), s i , s j ∈ SF , i ≠ j , d ( s i , s j ) ≤ c } the superfamily distance graph of sf . To split a superfamily distance graph into family clusters, we use an algorithm that can be seen as a weighted version of a method presented by Hartuv et al . [ 22 ]. First, we review some standard graph-theoretic definitions. The edge-connectivity k ( G ) of a graph G is the minimum number k of edges whose removal results in a disconnected graph. A cut in a graph is a set of edges C whose removal disconnects the graph into two disjoint components H 1 and H 2 . A minimal cut is a cut with a minimum number of edges. The length p ( u , v ) of the shortest path between nodes u and v in G is the minimum length of a path from u to v , if such a path exists; otherwise p ( u , v ) = ∞. The diameter of a connected graph G is the maximum shortest path length p ( u , v ) over all pairs of nodes u and v in G . The key definition of the algorithm in [ 22 ] is the following: An undirected unweighted graph G with n > 1 nodes is called highly connected , if k ( G ) > . A highly connected subgraph (HCS) is an induced subgraph H ⊆ G , such that H is highly connected. In an unweighted graph this definition results in the following property: The diameter of every highly connected subgraph is at most two. Thus, these subgraphs are compact clusters which need not meet the constraint of being fully connected. The original HCS algorithm in [ 22 ] recursively splits a connected graph at a minimal cut site until a disjoint set of highly connected subclusters is reached. For our purposes we had to modify the algorithm to be able to handle a weighted graph. Precisely, in our weighted HCS algorithm, if the edge weights covered by the minimal cut are approximately the same as in the remaining graph, the graph is assumed to be already highly connected and is not further split into subclusters (see Algorithm 2 in Figure 3 ). The E-values in our data set range from 0 (corresponding to any E-value better than 1e-180) to 0.05. To be able to find a minimal cut in our graph, edge labels should be positive values with a low value representing a weak connection and a high value representing a strong connection. Instead of using the raw E-values we label the edges in our graph with the negative logarithm of the corresponding E-value each. Since the logarithm of 0 is not defined, we use an arbitrary value (e.g., 181) for these edges instead of the logarithm. The running time of both HCS algorithms is bounded by 2 N * f ( n , m ), where N is the number of clusters found and f ( n , m ) is the time complexity of computing a minimum cut in a graph with n nodes and m edges. We use the implementation of the "mincut" algorithm given in the LEDA [ 23 ] distribution, which has a time complexity of ( nm + n 2 log n ). To apply this algorithm to our data set we added a preprocessing as well as a postprocessing step as shown in Algorithm 3 in Figure 3 . First, we describe the preprocessing. Cuts consisting of only one edge in the graph will be found first by the mincut algorithm, but are as time consuming to find as other cuts. Sequences being connected with the remaining graph by only one edge are either fragmental or are the so far sole representative of a protein family in the sequence database. The underlying data of our clustering is known to contain lots of fragmental sequences. Before applying the HCS algorithm to our graph, we repeatedly merge all nodes connected to the remaining graph with only one edge with their respective adjacent node. Nevertheless, the HCS algorithm may split off single sequences as subclusters. Thus, in a postprocessing step, sequences which ended up after the subclustering as a single sequence cluster are assigned to their closest neighboring cluster ( singleton adoption ), if there is no contradiction. When there are several minimum cuts in a graph, the original as well as our weighted HCS algorithm might choose a minimum cut which, from the clustering point of view, is not optimal. In many cases this process will break clusters into singletons. In the original algorithm in [ 22 ] iterations were introduced to handle these cases. Since we are working on a weighted graph these cases occur very rarely and mostly are compensated by the subsequent singleton adoption step. Figure 5 shows an example of splitting the superfamily distance graph of the ephrin superfamily (see Fig. 4 ) into two subclusters representing ephrin types A and B. Validation Pfam sequence set For our analyses we used all sequences from the Swiss-Prot and TrEMBL databases annotated with Pfam domains (Rel. 9). This data set consists of 5,724 single domain families assigned to 733,830 sequences. Since our aim is not to cluster single domains but full-length sequences, we define a "true" cluster consisting of all sequences having the same domain composition. Fragmental sequences will cause a problem in our analyses by showing a different domain composition than complete sequences. We restrict our analyses to sequences not annotated as being fragmental in the Swiss-Prot or TrEMBL databases. The resulting "true" cluster set thus consists of 442,872 sequences sorted into 16,990 distinct families. ENZYME sequence set The ENZYME database [ 24 ] stores data of a functional classification system based on function rather than sequence or structure. Each enzyme of known function is given an EC (Enzyme Commission) Number of the form A.B.C.D with A : type of reaction catalyzed (at present 6 classes) B : subclass, information about type of compound or group involved C : sub-subclass, further specifies the nature of the reaction D : serial number to identify individual enzyme within sub-subclass Although several distinct proteins may catalyze the same reaction, they are all ascribed the same EC number, since the naming system is based upon the reaction catalyzed. Thus, sequences given the same EC number do not necessarily show sequence similarity. For our analyses we used all sequences from the Swiss-Prot and TrEMBL databases annotated with a unique EC number. We define two different "true" cluster sets representing different levels of granularity as follow: (1) sequences having A, B, C, and D in common build a cluster, (2) sequences having A, B, and C in common build a cluster. The data set consists of 84,405 sequences. Clustering coeffcient Assuming we have a well defined cluster set, we can compare our cluster set with this "true" cluster set based on the following numbers: Number of sequence pairs clustered together in a : both cluster sets ("true positives"). b : the "true" cluster set, but not in our cluster set ("false negatives"). c : our cluster set, but not in the "true" cluster set ("false positives"). As similarity measure we decided on the Jaccard similarity [ 25 ] defined as follows: . A perfect clustering which is identical to the "true" cluster set would result in S = 1. Additionally we calculated the sensitivity (the ability to detect distantly related sequences: ) and the specificity (the ability to reject non-related sequences: ) for all cluster sets. Single linkage clustering We performed a single linkage clustering at various static E-values from 1e-02 to 1e-180. All resulting cluster sets have in common that when plotting the number of clusters against the cutoff E-value, one observes a continuous, smooth curve, indicating that there is no obvious (biologically given) choice of a cutoff (data not shown). TribeMCL TribeMCL [ 7 ] is a method for clustering proteins into 'protein families' using a Markov Clustering method. It is primarily used for comparing protein sequence sets of completely sequenced genomes. We performed TribeMCL clustering (Version 03–276) with different inflation value settings ranging from 1.1 to 5 for all data sets. The inflation parameter is part of the core MCL algorithm and influences the granularity (or size) of the output clusters. For very small or 'tight' protein families an inflation value setting of 4.0 or 5.0 is recommended. For larger (broader) protein families settings of 1.1, 2.0 and 3.0 can be used. For the Pfam data set we were not able to perform TribeMCL clustering due to memory allocation problems while executing the program. Authors' contributions MV had the initial ideas for SYSTERS. JS developed the superfamily determination. AK developed the subfamily determination, implemented the workflow from the raw sequence databases to a working web-server and made the validations. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC547898.xml
546238	Nutrition education: a questionnaire for assessment and teaching	It is generally recognized that there is a need for improved teaching of nutrition in medical schools and for increased education of the general population. A questionnaire, derived in part from a study of physician knowledge, was administered to first year medical students in order to assess their knowledge of various aspects of nutrition and metabolism, and as a teaching tool to transmit information about the subject. The performance of first year students was consistent with a generally educated population but there were surprising deficits in some fundamental areas of nutrition. Results of the questionnaire are informative about student knowledge, and immediate reinforcement from a questionnaire may provide a useful teaching tool. In addition, some of the subject matter can serve as a springboard for discussion of critical issues in nutrition such as obesity and markers for cardiovascular disease. A major barrier to improved teaching of nutrition is the lack of agreement on some of these critical issues and there are apparent inconsistencies in recommendations of government and health agencies. It seems reasonable that improved teaching should address the lack of knowledge of nutrition, rather than knowledge of official guidelines. Student awareness of factual information should be the primary goal.	Background Like many medical schools, SUNY Downstate Medical Center has been trying to improve the teaching of nutrition in the curriculum. The study presented here had two goals. First, we wanted to assess the knowledge of the first year class on the subject and, second, we hoped to use the questionnaire method, with immediate feedback, as a mechanism for imparting information on the areas covered. Whereas some form of nutrition education is part of the curriculum of most medical schools [ 1 ], it is generally believed that both medical professionals and the general public have serious limitations in their knowledge of the field [ 2 , 3 ]. Four of the 14 questions presented to the first year medical class were included in a recent survey of physicians published in this journal [ 4 ]. That paper was critical of the level of physician knowledge regarding nutrition. Therefore it seemed appropriate to compare performance of students with that of practitioners. In addition, the mechanism of providing information by presenting an anonymous quiz with immediate feedback could address one of the problems cited as a barrier to introducing nutrition into the medical curriculum – that is, that the first year basic science curriculum is already very concentrated, leaving little room for new material. The quiz format provides motivation and, because it is anonymous, does so in a relatively unstressful way. Perhaps the most important problem in introducing nutrition into the medical curriculum is the lack of agreement on what should be taught. There is strong, even contentious debate about the most fundamental issues such as obesity, cardiovascular disease, and diabetes. Thus, concerns about inadequate physician knowledge frequently refer to ability to counsel patients according to standard guidelines [ 4 - 6 ]. Here we suggest that a more appropriate goal would be the understanding of basic nutritional information and that this information is not always in agreement with official guidelines. The details of these problems are left to Results and Discussion since we first want to offer the reader a chance to take the questionnaire as given to students. The quiz is presented twice: first, as given to medical students, and then, a second time, with the answers and the comments given to students immediately upon completion of the quiz. A follow-up email that was sent to students after results were tabulated has also been reproduced. Results and Discussion provides data on student performance and additional discussion. Methods The following is a verbatim reproduction (with addition of references) of the questionnaire given to first year medical students at SUNY Downstate Medical Center. The first four questions were taken from Flynn, et al . [ 4 ]. Following Flynn, we did not attempt to define the term "low fat diet" which is usually used without elaboration. No specific time limit was given for the quiz but a class of 111 students finished in about 20 min with a few stragglers. Students were given the answer sheet (shown after the quiz itself) in exchange for their questionnaire. Students were in the middle of the metabolism course, part of a subdivision characterized as Gastrointestinal Block. They had been taught bioenergetics and carbohydrate and lipid metabolism, and would be expected to know the answers to some of the questions based on that material. Other questions tested general nutrition knowledge while questions 12–14 emphasized material taught later in the course. Questionnaire This questionnaire is anonymous and does not affect your grade. Make up a User Name (in case there is a follow-up) that only you know and that is not common, like Sherlock-37) User Name _________________________________________________ The questionnaire is designed to test your general knowledge, not anything you have learned in the course. 1. A good source of monounsaturated fat is: (check all that apply) _____ Butter _____ Canola Oil _____ Corn Oil _____ Flaxseed Oil _____ Olive Oil _____ Safflower Oil _____ Soybean Oil _____ don't know 2. The diet component that is most likely to raise triglycerides is (select one) _____ Fat _____ Carbohydrate _____ Protein _____ don't know 3. In general, what effect does a low-fat diet have on triglycerides? _____ Increase _____ Decrease _____ no change _____ don't know 4. In general, what effect does a low-fat diet have on HDL-c (high density lipoprotein-cholesterol) ? _____ Increase _____ Decrease _____ No change _____ Don't know 5. In the past thirty years, the per cent fat in the American Diet has: _____ Increased _____ Decreased _____ Stayed about the same. 6. The most energy dense food (most calories/gram) is: _____ Carbohydrate. _____ Protein. _____ Fat. 7. High total blood cholesterol can be lowered significantly by: _____ Diet _____ Drugs such as statins _____ Diet or drugs are equally effective _____ Neither 8. The dietary change that is most likely to increase the risk of cardiovascular disease: _____ unsaturated fat → saturated fat (that is, replace unsaturated fat with saturated fat) _____ unsaturated fat → carbohydrate _____ carbohydrate → unsaturated fat _____ carbohydrate → saturated fat _____ saturated fat → carbohydrate _____ saturated fat → unsaturated fat 9. Glycemic Index measures the increase in blood sugar over 2 hours per gram of carbohydrate ingested, compared to glucose (=100). For each food indicate the approximate glycemic index as: H, high (70–100), M, Medium (40–70) or L, Low (< 40). You may enter a number if you think you know or can figure it out: _____ white bread _____ whole wheat bread _____ ice cream _____ carrots _____ sucrose (table sugar) _____ fructose _____ bran muffin _____ banana 10. The substances in the following list that either are themselves or are considered to contain large amounts of complex carbohydrates. _____ white bread _____ whole wheat bread _____ ice cream _____ fructose _____ sucrose (table sugar) _____ corn starch _____ fiber 11. In the first list check the vitamins that are generally considered to have antioxidant activity. In the second list check the vitamins that are precursors for oxidative coenzymes. (check all that apply) ANTIOXIDANTS _____ ascorbic acid (vitamin C) _____ niacin _____ riboflavin _____ thiamine _____ pyridoxal phosphate (Vitamin B 6 ) _____ vitamin B 12 _____ vitamin D _____ vitamin E REDOX PRECURSORS _____ ascorbic acid _____ niacin _____ riboflavin _____ thiamine _____ pyridoxal phosphate _____ vitamin B 12 _____ vitamin D _____ vitamin E 12. Megaloblastic anemia is a prominent feature of deficiencies of: _____ Vitamin B 12 _____ Folic Acid _____ Neither _____ Deficiencies of either 13. Addition of folic acid to the diet can relieve all the symptoms due to deficiencies of: _____ Vitamin B 12 _____ Folic Acid _____ Both _____ Neither 14. Vitamin B 12 deficiency is most commonly seen in: _____ children due to poor nutrition. _____ children due to poor absorption. _____ the elderly due to poor nutrition. _____ the elderly due to poor absorption. Answers and feedback Answers and comment were given to students immediately in exchange for their questionnaires. Students were polled about the quiz verbally and by Email. There were few responders although all were positive and included comments that they might not have read the material in the answers had they not taken the quiz first. The Answer Sheet is reproduced verbatim below and it should be understood that it contains some comments that are more colloquial than would be included in a more formal document. Answer Sheet The first four questions appeared in a recent paper (Flynn M, et a l. Nutr J 2003, 2 :19) that reported the results of a questionnaire designed to assess the level of knowledge of physicians. The authors were critical of the responses; the paper was entitled "Inadequate physician knowledge of the effects of diet on blood lipids and lipoproteins." In particular, "Physicians showed a poor understanding of the effects of changing the relative intake of carbohydrates and fats on triglycerides and HDL." The rationale for the study were guidelines recommended by the Third Report of the National Cholesterol Education Program Adult Treatment Program (ATP III, Circulation 2002, 106 :3143–3421). The ATP III, compared to previous version, advocates lowering triglycerides as a secondary target to lowering LDL. The major conclusions: • Half of the physicians did not know that canola oil is a good source of monounsaturated fat; 26% did not know that olive oil is also. • Ninety-three percent (84% of cardiologists vs. 96% of internists) did not know that a low-fat diet, in general, would increase blood triglycerides. • Approximately three-quarters (70% of cardiologists vs. 77% of internists) did not know a low-fat diet would decrease HDL; almost half (45%) thought that a low-fat diet would not change HDL. • About one-half (47%; 22% of cardiologists vs. 53% of internists) did not know carbohydrate was the diet component most likely to raise triglycerides. 1. A good source of monounsaturated fat is: (check all that apply) _____ Butter _ X _ Canola Oil _____ Corn Oil _____ Flaxseed Oil _ X _ Olive Oil _____ Safflower Oil _____ Soybean Oil _____ don't know It is generally considered that monounsaturated fats are protective of cardiovascular disease (as in the Mediterranean Diet). The effect of monounsaturated fat explains the anomaly in the so called Seven Countries study [ 7 ] which launched low fat recommendations. The two countries among the highest consumers of fat had roughly the highest (Finland) and the lowest (Crete) incidence of cardiovascular disease. This finding immediately led to a change in recommendations to lower saturated fat, rather than fat across the board. The persistence of a recommendation to lower total fat is now controversial. Some people may be surprised when they actually look at the data (Figure 1 ). High concentrations of monounsaturated fats are found in olive oil (73 %) and canola oil (58 %), but the highest is in avocado oil, and there are fairly high levels in beef tallow and lard (44% and 47 %). Interestingly, beef tallow is what MacDonald's previously used to fry their French fries (which at the time got thumbs up from Julia Child) until pressured to switch to vegetable oil. Also of interest is that most of the saturated fat in beef fat itself is stearic acid which is not considered atherogenic [ 8 ]. Figure 1 Fatty acid composition of common dietary fats and oils. Data from Figure 1.9 of reference [44]. You should also know that canola oil is named for CANadian OiL (there is no canola tree) and is a highly processed version of rapeseed oil (which is a plant in the Brassica family) and is the subject of popular controversy because of the possible production of trans -fatty acids during processing. 2. The diet component that is most likely to raise triglycerides is (select one) _____ Fat _ X _ Carbohydrate _____ Protein _____ don't know 3. In general, what effect does a low-fat diet have on triglycerides? _ X _ Increase _____ Decrease _____ no change _____ don't know The phenomenon of carbohydrate-induced hypertriglyceridemia is well established in the literature [ 9 - 15 ]. It is the primary reason we ask patients to fast prior to having blood drawn for serum lipid analysis. It is also one of the arguments for low carbohydrate diets since the most predictable effect of such diets is a dramatic decrease in triglycerides compared with low fat diets [ 16 - 22 ]. Low carbohydrate diets also generally increase HDL (good) cholesterol while low fat diets lower HDL although this is somewhat less reliable than the triglyceride effect. An argument against low carbohydrate diets is that, whereas, on average, LDL frequently goes down, individual results are highly variable and LDL increases for some subjects [ 16 - 22 ]. 4. In general, what effect does a low-fat diet have on HDL-c (high density lipoprotein-cholesterol)? _____ Increase _ X _ Decrease _____ No change _____ Don't know 5. In the past thirty years the per cent fat in the American Diet has: _____ Increased _ X _ Decreased _____ Stayed about the same. This is the major point of attack on current guidelines since the obesity epidemic correlates with a decreased percentage of fat in the diet (for men in some years, there was a decrease in total amount of fat) and increased carbohydrate consumption [ 23 , 24 ]. Defenders of current guidelines maintain it is simply that the wrong kind of carbohydrate (i.e., sugars and refined starches) is being consumed, and that affluence has increased the availability of food and portion sizes, and people are overeating "because it is there." On the other hand, published results of low carbohydrate diets show that they can be effective and therefore there is no great over-consumption even though portion sizes are unlimited. It is also argued that the wrong kind of fat (i.e., saturated and trans fats) is being consumed although no effect of fat per se on obesity, independent of calories, has been found. 6. The most energy dense food (most calories/gram) is: _____ Carbohydrate. _____ Protein. _ X _ Fat. The operational numbers in kcals/g are 4, 4, and 9 for carbohydrate, protein, and fat. This is the basis of traditional recommendations for low fat diets for obesity. The reduction in percentage of dietary fat and the obesity epidemic noted above, however, suggests that this is not a good universal principle. The role of macronutrient composition on satiety, taste, total consumption and effect on weight loss is largely unknown due to the multiple factors and individual differences, and some experimental support can be found for just about any idea. 7. High total blood cholesterol can be lowered most significantly by: _____ Diet _ X _ Drugs such as statins _____ Don't know _____ Diet or drugs are equally effective _____ Neither Drugs such as statins (HMGCoA reductase inhibitors) are very effective at reducing cholesterol. Diet can also be effective but usually less so. Combination diet and drugs, however, may be most effective but there is, again, not universal agreement as to what that diet should be. 8. The dietary change that is most likely to increase the risk of cardiovascular disease: _____ unsaturated fat → saturated fat (that is, replace unsaturated fat with saturated) _ X _ unsaturated fat → carbohydrate _____ carbohydrate → unsaturated fat _____ carbohydrate → saturated fat _____ saturated fat → carbohydrate _____ saturated fat → unsaturated fat In addition to the effect on risk factors, epidemiologic evidence suggests that replacing fat with carbohydrate is deleterious. Replacing unsaturated fat with saturated fat will also increase the risk of cardiovascular disease [ 8 , 9 , 14 , 25 ]. Of course, in terms of obesity, reducing calories by removing fat and not replacing it with anything is good. Removing carbohydrate, however, may be better, at least in terms of cardiovascular risk as in references above. 9. Glycemic Index (GI) measures increase in blood sugar over 2 hrs per gram of carbohydrate, compared to glucose (=100). For each food indicate the approximate glycemic index as: H, high (60–100), M, Medium (40–60) or L, Low (< 40). You may enter a number if you think you know it: H (70) white bread M (52) whole wheat bread M (50) ice cream M (47) carrots H (70) sucrose (table sugar) L (20) fructose HM (60) bran muffin M (50) banana The GI is a very rough indicator of rise in blood sugar and is influenced by absorption and the concentration of glucose. Fructose has a low GI (20) indicating slow conversion to glucose in 2 hrs but far from being considered a "good" sugar, at high levels may be very deleterious. The concept of glycemic load (GL) which corrects for the amount of carbohydrate per serving may be a better parameter but runs into problems about serving size. So muffins and candy bars have GL = 15 and carrots only 3 but 80 g of carrots may not be a lot for some people [ 26 - 28 ]. 10. The substances in the following list that either are themselves or are considered to contain large amounts of complex carbohydrates. _ ? _ white bread _ ? _ whole wheat bread _ NO ice cream _ NO fructose _ NO sucrose (Table Sugar) _ ? _ corn starch _ ? _ fiber If you had trouble with this, it is because nobody knows how the term should be used and, in fact, we recommend it not be used at all. The original chemical definition – to some extent still used in organic chemistry – is that a complex carbohydrate is a polysaccharide (not mono- (glucose, fructose) or di- (sucrose, lactose)). Any starch, e.g. corn starch or white bread (poly-glucose: amylose or amylopectin) fits the definition, and it used to be nutritional dogma that, at least in terms of raising blood glucose, complex carbohydrates (that is, polysaccharides) were better than simple sugars. When the dogma was finally tested this turned out not to be true and the concept of glycemic index arose. It is clear from question 9 that white bread is nutritionally similar to pure glucose. Probably because of the evocative nature of the word, the term "complex" is still used. Sometimes it means foods that have a low glycemic index due to poor absorption usually due to the presence of fiber, but it is never precise. When people say complex carbohydrate they usually mean the carbohydrate recommended in their diet and missing from somebody else's diet. Suggested nomenclature is "polysaccharides, starch, high fiber," although fiber itself is a heterogeneous category. 11. In the first list check the vitamins that are generally considered to have antioxidant activity. In the second list check the vitamins that are precursors for oxidative coenzymes. (check all that apply) ANTIOXIDANTS _ X _ ascorbic acid (vitamin C) _____ niacin _____ riboflavin _____ thiamine _____ pyridoxal phosphate (Vitamin B 6 ) _____ vitamin B 12 _____ vitamin D _ X _ vitamin E REDOX PRECURSORS _____ ascorbic acid _ X _ niacin _ X _ riboflavin _ X _ thiamine _____ pyridoxal phosphate (Vitamin B 6 ) _____ vitamin B 12 _____ vitamin D _____ vitamin E 12. Megaloblastic anemia is a prominent feature of deficiencies of: _____ Vitamin B 12 _____ Folic Acid _____ Neither _ X _ Deficiencies of either 13. Addition of folic acid to the diet can relieve all the symptoms due to deficiencies of: _____ Vitamin B 12 _ X _ Folic Acid _____ Both _____ Neither 14. Vitamin B 12 deficiency is most commonly seen in: _____ children due to poor nutrition. _____ children due to poor absorption. _____ the elderly due to poor nutrition. _ X _ the elderly due to poor absorption. The most obvious symptom of a folic acid deficiency, anemia, is due to a requirement for folic acid in the synthesis of DNA. Deficiencies lead to poor maturation of red blood cells (megaloblasts). Megaloblastic anemia can also be caused by a B 12 deficiency, which, indirectly, has the same effect. There are only two reactions in humans requiring vitamin B 12 . First, vitamin B 12 is a cofactor in formation of the amino acid methionine from homocysteine and the folic acid derivative, methyl-tetrahydrofolic acid (methyl-THF): Homocysteine + methyl-THF → Methionine + THF This explains the relation between dietary folic acid and high homocysteine which is a marker for cardiovascular disease and potential birth defects. A deficiency in folic acid or the cofactor, B 12 , will prevent this reaction from occurring. The effect of B 12 deficiency on folic acid is indirect: if methionine synthesis cannot be carried out, methyl-THF will build up ("methyl trap"). This is effectively a folic acid deficiency and anemia is the outcome. The second requirement for B 12 involves organic acids and deficiencies can lead to neurologic damage. The anemia in a B 12 deficiency may be successfully treated with folic acid, swamping out the methyl trap. Neurologic damage, however, may still occur unless the B 12 deficiency is also treated. A dietary deficiency of B 12 is rarely seen since little is needed and it is stored well. Deficiency is usually detected in the elderly due to decreased production of intrinsic factor, a protein required for absorption. The considerations above bear on the recommendation to add folic acid to manufactured food. Critics point out that by preventing anemia, a B 12 deficiency could be masked. Since the major deficiency is not dietary but absorptive, the problem can't be solved by simply adding B 12 as well. Follow up The following analysis was sent as an Email to students after the scores on the quiz were tabulated. Results of Nutrition Questionnaire As indicated in the answer sheet, the first four questions are taken from a recent paper (Flynn, M., et al. (2003) Nutrition Journal 2 : 19) that reported the results of a questionnaire designed to assess the level of knowledge of physicians. Their conclusion was that "Physicians showed a poor understanding of the effects of changing the relative intake of carbohydrates and fats on triglycerides and HDL." Interestingly the rationale for the study was determining "Physicians' ability to effectively counsel patients with elevated cholesterol to initiate a Therapeutic Lifestyle Changes Diet (TLC)" (as proposed by the Third Report of the National Cholesterol Education Program Adult Treatment Program (ATP III) which recommends lowering triglycerides as a secondary target to lowering LDL. The TLC Diet recommends total fat as 25–35 % and carbohydrate at 50–60 %. Paradoxically, Flynn, et al.'s assessment of physician knowledge focused on the deleterious effect of carbohydrate on triglycerides. Given this association, it is puzzling that ATP III would council people trying to lower triglycerides to undertake such a diet. In any case, your performance compared to their sample is shown in Figure 2 . Figure 2 Performance of physicians and first year medical students on questionnaire. Data on physicians from reference [4]. Two questions of importance: Question 5: In the past thirty years, the per cent fat in the American diet has declined by about 10 %, close to the target level of 30% of total calories set back in the 70's. In order to explain why this has been associated with an obesity epidemic, many have blamed portion-size, the fast-food industry and consumers themselves. In any case, most people did not know that percent fat consumption had gone down. Answers were: Increased: 76 % Decreased 22 % Same 3 % Question 6. Again, it was surprising that so many people did not know the relative energy density. The answers: Carbohydrate 19 % Protein 7 % Fat 74 % Analysis Questionnaires were collected from 111 students, all but 6 of whom answered all the questions. Student answers were tabulated and discrimination coefficients were determined by point biserial correlations [ 29 ] using LXR-TEST™ software (Logic eXtension Resources ). The discrimination coefficient varies between -1 and +1 and measures the extent to which performance on a particular question reflects performance on the quiz as a whole, that is, whether a question discriminates high performers from low performers. A typically high value of +0.4 means that the question was answered correctly by most students who did well on the exam and answered incorrectly by most who did not. Because the questions had different goals (knowledge assessment vs. motivation for accepting new information), and because some (e.g. questions 2–4) are interrelated, no measure of performance or statistics were carried out for the quiz as a whole. Results and Discussion Student and physician knowledge of nutrition Questions 1–8 Results from the first eight questions and the data from Flynn, et al. [ 4 ] are shown in Tables 1 and 2 and in Figure 2 . Flynn, et al . used four questions that succinctly identified both practical and conceptual knowledge bearing on the ability to implement dietary recommendations from the Adult Training Program (ATP III) of the National Cholesterol Education Program (NCEP)[ 6 ]. Table 1 Student and physician responses (%) Cardiologists Internists Students discrim coeff 1. A good source of monounsaturated fat Butter 4 4 8 0.39 Canola Oil 43 51 26 0.21 Corn Oil 13 16 22 0.32 Flaxseed Oil 12 10 25 0.32 Olive Oil 82 73 58 0.35 Safflower Oil 24 32 25 0.26 Soybean Oil 18 16 38 0.35 don't know 6 6 16 2. Diet component most likely to raise triglycerides Fat 16 47 63 -0.2 Carbohydrate 78 47 32 0.17 Protein 0.8 0.6 0 don't know 5 5 2 3. Effect of low-fat diet on triglycerides Increase 16 4 14 0.22 Decrease 52 73 68 -0.17 no change 26 26 15 0.01 don't know 6 4 3 4. Effect of low-fat diet on HDL-c Increase 11 24 31 -0.08 Decrease 30 23 32 0.25 no change 52 44 23 -0.18 don't know 7 9 14 Discrimination coefficients indicate discrimination of high and low performers on quiz overall as described in the text. (correct answers in bold). Table 2 Student Responses (%) Students discrim coeff 5. Past thirty years, per cent fat in American diet Increase 76 -0.10 Decrease 22 0.14 Same 3 -0.10 6. Most energy dense food Carbohydrate 19 -0.15 Protein 7 -0.10 Fat 74 0.19 7. High blood cholesterol lowered significantly by Diet 14 0.06 Drugs such as statins 21 0.17 Diet or drugs equal 61 -0.15 Neither 0 Don't know 0 8. Most likely to increase risk of CVD UF t>gF 46 0.02 UF -> CHO 7 0.11 CHO -> UF 3 -0.27 CHO -> SF 23 0.09 SF -> CHO 2 0.00 SF -> UF 12 -0.08 First year students did not do as well as physicians at identifying sources of monounsaturated fats. On the other hand, the good discrimination coefficient indicates that knowledge of fat composition is a good indicator of overall knowledge (at least as assessed by general performance on this quiz). Although a substantial fraction of cardiologists polled by Flynn knew that carbohydrate raised triglycerides (Figure 2 ), most internists and most medical students did not. Likewise, a very small fraction of first year students or physicians were aware of the association between low fat diets and two markers of CVD, triglycerides and HDL-c. As discussed in the student answers (see Methods), there is some irony in that the questions chosen by Flynn bring out the unfavorable effect of carbohydrate on triglycerides while the ATP III recommendation is to maintain 50 % carbohydrate in the diet. Student responses attest to the success of continued popular and government recommendations favoring low fat diets but the content of the answers raises the question of whether sufficient information is being disseminated. It further raises the question as to whether these recommendations, rather than the basic nutritional knowledge, should be communicated. Along the same lines, our questionnaire went beyond the area covered by Flynn to consider the changes in diet that have accompanied the epidemic of obesity. It has to be considered very surprising that only 22 % of an educated population knew that the per cent fat in the American diet has decreased (Figure 3 ); for men, in fact, the total amount of fat has decreased, whereas for women there has been a slight increase consistent with the much larger increase in caloric intake among women [ 24 ]. The observation of a decrease in fat and an increase in carbohydrate in parallel with the obesity epidemic remains as a serious challenge to traditional dietary recommendations. The reduction in fat in the diet from the 1970s to 1995 has been noted by one author [ 24 ] to provide a benefit in reduction in serum cholesterol from 213 to 205 mg/dL! Although more difficult to quantify, a decline in exercise is also a likely contributor to the epidemic, but it seems inappropriate, without further evidence, to ignore the prima facie evidence of the effect of macronutrients. Despite the clear correlation between higher carbohydrate, lower fat and obesity, government and health agencies rarely question the appropriateness of the original guidelines and have continued to recommend still higher carbohydrate and still lower fat [ 6 , 30 , 31 ]. Such recommendations have to be considered controversial and likely to change. For this reason we feel that one of the points brought out by our quiz and Flynn's is that nutritional facts rather than official recommendations should be the goal of nutrition education. Figure 3 Changes in fat and carbohydrate between 1977 and 1995. Data from USDA as reported in reference [24]. Question 6 We were surprised by some lapses in student knowledge revealed by the questionnaire. The most basic question – which macronutrient is the most energy dense – had only 74 % correct answers. To understand how surprising a response this is, it should be understood that the mean score on exams in this section of the medical course is typically 80 % and, on most exams, several questions are answered correctly by 98–100 % of the class. The National Board of Medical Examiners assumes knowledge of caloric value of macronutrients, and we had expected that it was common knowledge. The questionnaire result indicates that no fact in nutrition is too basic to be excluded from course material. Questions 7–8 – Diet and Cholesterol An overwhelming number of medical students believe that diet is as effective as drugs in lowering total cholesterol. This again attests to the pervasive message that diets control blood cholesterol, an idea continually reinforced by media advertisements, for example, for the cholesterol lowering effect of breakfast cereal. Whereas it is likely that diet is an important influence on CVD, there is, again, the problem of which diet and the question would probably have been better framed, as in questions 2–4, on specific lipid components. It is likely, for example, that many medical students would not know that dietary cholesterol is largely without effect on serum cholesterol. In any case, it is generally acknowledged that, on average, drugs such as statins have a greater impact on cholesterol than currently reported diet interventions. The general effectiveness of statins and the promotion by pharmaceutical companies has, most recently, led to a movement to reinforce the idea that genetics (which can't be controlled by diet) also plays a role. The competing financial interests have produced, in our view, bizarre and unpatriotic (?) television commercials blaming mother and apple pie for high cholesterol. Figure 4 Effect of substitution of 5 % of calories on incidence of cardiovascular disease. Data from Hu, et al . [8] Few students in the first year medical class knew that replacing unsaturated fat with carbohydrate was the most damaging substitution in terms of an association with CVD risk. The data from Hu, et al., [ 8 , 25 ] shown in Figure 4 represent yet another reason to reevaluate low fat recommendations. Similar results have been found in the analysis of risk factors [ 14 ]. In other words, whereas everybody agrees that removing fat from the diet as a mechanism of calorie reduction is a good thing, replacing fat with carbohydrate correlates with an increase in CVD risk and is likely worse for weight loss. Questions 9–10 – Glycemic Index and Complex Carbohydrates These two questions (Table 3 ) point out the confusion that exists in characterizing dietary carbohydrates. The glycemic index (GI), and the glycemic load (GL) which corrects for total carbohydrate in individual foods, are indicators of rise in blood glucose. Glycemic control is a major variable in the analysis of metabolic syndrome and obesity, and dietary strategies based on the glycemic index [ 28 ] have the same rationale as low carbohydrate diets: reduce fluctuations in insulin and associated anabolic effects. A low carbohydrate diet might be described as a very low glycemic load diet. Nonetheless, the concepts of GI and GL have become part of the political controversies surrounding dietary strategies and proponents usually urge a low GI diet as an alternative rather than a variation of low carbohydrate diets ([ 15 ]) despite the fact that in at least one isocaloric comparison of high GI and low GI meals, the low GI meal was, in fact, lower in carbohydrate [ 27 ]. An important limitation on the concept of GI is that fructose and therefore fructose-containing products such as sucrose and high-fructose corn syrup may have low values, although these substances may not be desirable. The atherogenic qualities of fructose [ 32 ] is one of the ideas that we bring out in the lectures in the medical school course. Table 3 Student Responses (%) on Carbohydrates 9. Glycemic Index Low Medium High discrim coeff White Bread 5 22 70 0.2 Whole Wheat Bread 39 49 10 0.18 Ice Cream 5 16 77 -0.1 Carrots 68 19 9 0.07 Sucrose 1 10 86 0.16 Fructose 3 22 73 -0.04 Bran Muffin 25 49 23 Banana 20 53 23 0.23 10. Complex Carbohydrates Yes No White Bread 60 38 Whole Wheat Bread 68 32 Ice Cream 31 68 Fructose 11 88 Sucrose 18 81 Corn Starch 48 51 Fiber 43 56 We recommend that the term complex carbohydrates not be used since, in practice, it has lost its original meaning of polysaccharide. It is interesting that, to some extent, student answers followed the original definition. Most students picked both white bread and whole wheat bred as complex although, with a slight preference for picking whole wheat over white bread as many health professionals and the lay public might. Question 11 – Vitamins We credit the popular media with the generally good knowledge about the antioxidant vitamins shown in Table 4 . In our view, however, this may be a mixed blessing because it shifts the emphasis from macronutrient composition, a major factor in health, to micronutrients, which, at least for the American population, has to be considered secondary. The relatively low performance and good discrimination coefficient in the question on redox precursors is somewhat discouraging, especially in that students had been exposed to the involvement of the three oxidative coenzymes in glycolysis and the TCA cycle. Moreover, the origin of NAD coenzymes in dietary niacin was explicitly taught. We think this apparent deficiency likely results from a lack of emphasis on integration of nutritional information with biochemistry. Table 4 Student Responses (%) on vitamin question 11. Vitamins with indicated activity. yes discrim coeff ANTIOXIDANTS Ascorbic Acid 77 0.47 Niacin 23 0.39 Riboflavin 19 0.45 Thiamine 9 0.57 Pyridoxal Phosphate (Vitamin B-6) 21 0.31 Vitamin B-12 31 0.41 Vitamin D 16 0.38 Vitamin E 69 0.45 REDOX PRECURSORS Ascorbic Acid 27 0.47 Niacin 51 0.31 Riboflavin 43 0.23 Thiamine 34 0.31 Pyridoxal Phosphate 33 0.28 Vitamin B-12 27 0.45 Vitamin D 11 0.44 Vitamin E 5 0.43 Questions 12–14 – Questionnaires as a teaching method: folate metabolism These questions were presented as a preview for an upcoming lecture on folate metabolism; therefore it was expected that students would not score very highly (Table 5 ). We identify folate metabolism as one of the critical areas of biochemical nutrition. The importance of homocysteine and use of dihydrofolate reductase inhibitors such as methotrexate are two of the most obvious examples of how biochemistry is a practical part of medicine. At the same time, the biochemical pathways are among the most complex, and because folate spans different areas of metabolism, it is difficult to teach. The key nutritional issues are covered both in lecture and in a case-based learning session. Table 5 Student Responses (%) on folic acid questions Students discrim coeff 12. Megaloblastic anemia: deficiencies of Vitamin B-12 39 -0.11 Folic Acid 31 0.17 Neither 6 0.09 Either 19 -0.09 13. Folic acid: relieve deficiencies of Vitamin B-12 7 -0.1 Folic Acid 49 0.12 Both 38 -0.07 Neither 3 14. Vitamin B-12 deficiency commonly seen in children due to poor nutrition. 32 -0.11 children due to poor absorption. 23 -0.04 the elderly due to poor nutrition. 9 0.1 the elderly due to poor absorption. 25 0.1 Nutrition in the Medical School Many papers have been written on the need for, and the difficulty in implementing, improvements in teaching nutrition in medical schools [ 2 , 3 , 33 ]. Some of the major problems frequently cited are 1) inflexibility in the curriculum due primarily to time constraints and 2) inability to define what aspects of the subject needs to be taught. There is also considerable disagreement on the best method of teaching the subject. The current study bears on some of these questions. Adding nutritional material to the curriculum With respect to point 1) above, the first year medical school curriculum is undoubtedly very dense in content. Adding new material is difficult, especially if it is of the strictly factual type, e.g. macronutrient composition of particular foods. The "low pressure" quiz used here can, in theory, impart a certain amount of specific knowledge and generate student interest without interrupting the general flow of course work. The quiz provides a venue in which interested students can absorb the information, and become aware of the general area if they need to find the information later. Also, in our view, many subjects taught in basic science courses already have nutritional relevance, e.g., cofactors that come from vitamins, and these ideas should be better emphasized. We point out, when the NAD cofactors are introduced, that one would expect global effects of a deficiency disease because of the number of different enzymes that use these cofactors. Although vitamin deficiencies are rare in the absence of gross malnutrition, the emerging role of hypervitamin therapies [ 34 ] has great pedagogical value. The tie-in through the quiz may reinforce the basic biochemistry. What to teach in nutrition. Guidelines on macronutrient recommendations We see the question of what to teach as the most critical problem in introducing or expanding nutrition education in the medical school course. Individual faculty may be resistant to giving up their own interests, but this may depend on how well the case is made for changing to new topics. The original study by Flynn was designed to test physicians' knowledge and expand it to allow them to better implement ATP III recommendations on serum triglycerides. In combination with other questions that we have introduced, the general problem arises as to whether these recommendations or nutritional data should be taught. We feel that official recommendations, such as ATP III, have some inherent contradictions. Given that low fat diets tend to raise triglycerides, the associated recommendations to reduce dietary fat and to raise carbohydrate intake appear somewhat contradictory. The major focus of ATP III, however, is control of cholesterol but again, the literature is not clear-cut. Thus, whereas the association between cholesterol levels and CVD is generally accepted by all but a minority of critics, the effect of diet, especially reduced fat diets, on CVD, or even cholesterol, is far more controversial. The Chapter on "Diet and Coronary Heart Disease (CHD)" in Willett's Nutritional Epidemiology [ 35 ] is 40 pages long with more than 300 references and contains more than one disclaimer on the diet-heart hypothesis, e.g. "Even if a change in dietary lipids influences the incidence of CHD in the direction predicted by its effect on total blood cholesterol level, the quantitative relationship between this dietary change and risk of disease is uncertain because of the possibility of many other potential physiologic effects of this dietary manipulation (p. 422), " or "Although substantial indirect evidence supports the classic diet-heart hypothesis, the magnitude of any association is likely to be modest for ranges of diet found within western culture or attainable by realistic dietary changes if the effects predicted by metabolic studies are correct (p. 443)." Finally, papers have been published by respected authors with such titles as: "Dietary fat is not a major determinant of body fat [ 36 ]," or "Do high carbohydrate diets prevent the development or attenuate the manifestations (or both) of syndrome X? A viewpoint strongly against [ 37 ]" These cautionary reports as well as those of other critics of low fat dietary recommendations [ 8 , 19 , 38 - 40 ] are largely ignored by the ATP III and the body of experts who are making current recommendations. The recent demonstration of a beneficial effect of saturated fat and lower carbohydrate in patients on an overall low fat diet [ 41 ] has been described in an accompanying editorial as an "American paradox." [ 42 ]. The extent to which researchers seek to resolve this paradox remains to be seen. The analysis above also bears on the role of low carbohydrate diets in educating students and physicians. We have previously indicated how such diets can be used to teach basic intermediary metabolism [ 43 ] and whereas we do not recommend any particular diet, we feel that the biochemical rationale of carbohydrate restriction makes it increasingly difficult to justify exclusive recommendations for low fat, high carbohydrate guidelines. In summary, what to teach remains very problematic. There are clear inconsistencies in the dietary recommendations of the ATP III and other professional agencies. This has to make one question whether students and physicians should be educated only in currently recommended practice, or whether we should instead emphasize understanding the underlying data. This is especially true, given the disclaimers in the American Heart statement [ 31 ] that "These recommendations may require modification, based on the results of ongoing and future dietary therapy studies." and that "The available data suggest that it is unlikely that one approach is appropriate for all patients." Of course, presentation of such controversial questions can be introduced into a problem-based learning session but medical students naturally prefer concrete answers and appropriately expect some guidance. The resolution currently depends on individual instructors and departments. It would be good pedagogically to establish the idea that not everything is known about nutrition and that many people consider that a rush to guidelines on insufficient evidence is to be avoided. Conclusions A questionnaire, derived in part from one previously published to assess physician knowledge, can be used to determine medical student awareness of nutritional facts. At the same time, such a quiz can be employed as a teaching device to reinforce earlier material, provide preview of new material, or expose students to factual information that is not easily incorporated into a formal course. One of the areas chosen, the effect of macronutrients on obesity and cardiovascular disease, can lead to discussion and focus on important current issues. The performance of first year medical students as well as the performance of the physicians in the previous study suggest that improvement is needed in imparting knowledge about some basic ideas in nutrition. We believe that the focus should be on these ideas rather than on official recommendations with which the ideas are sometimes in conflict. Finally, the questionnaire is intended as a practical method. The authors would be grateful for any information on the outcome of its use and/or any suggestions for improving the quiz itself. List of Abbreviations HDL: High Density Lipoprotein LDL: Low Density Lipoprotein TAG: Triacylglycerol THF: Tetrahydrofolic acid TLC: Therapeutic Lifestyle Changes Diet ATP III: Third Report of the National Cholesterol Education Program Adult Treatment Program NCEP: National Cholesterol Education Program GI: glycemic index GL: glycemic load CVD: Cardiovascular Disease	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546238.xml
526296	Chemotherapy with cisplatin and vinorelbine for elderly patients with locally advanced or metastatic non-small-cell lung cancer (NSCLC)	Background Although modest improvements in the survival of patients with non-small cell lung cancer (NSCLC) can be achieved with cisplatin-based chemotherapy (CT), its value is disputed in the geriatric setting. In this study, we evaluate the feasibility of vinorelbine/cisplatin CT for elderly NSCLC patients. Methods In this pilot phase I/II trial, all patients received CT with vinorelbine 25 mg/m 2 , on day 1 and 8, and cisplatin on day 1, in 28 days-cycles. After stratification for age (up to 75 years), younger patients were sequentially allocated to moderate cisplatin doses (80 mg/m 2 or 90 mg/m 2 ), and older patients were allocated to lower cisplatin doses (60 mg/m 2 or 70 mg/m 2 ). We recruited patients aged over 70 years with newly diagnosed NSCLC, clinical stage III or IV, Karnofsky performance status ≥ 70%, normal serum creatinine, peripheral neuropathy ≤ grade 1, and no prior cancer therapy. Results Analysis was by intention to treat. Main toxicities (grade 3–4) was as follows: neutropenia, 20%; anemia, 11%; and thrombocytopenia, 2%; alopecia, 55%; fatigue, 11%; and peripheral neurotoxicity, 2%. No grade 3–4 emesis or renal toxicity occurred. Global median time to progression (TTP) and overall survival (OS) were 27.0 (95% CI: 10.1 to 43.7) weeks and 30.1 (95% CI: 24.4 to 35.8) weeks; 1- and 2-year survival rates were 36.3% and 13.2%, respectively. Overall response rate was 50.0% (95% CI: 35.4% to 64.5%), with 1 complete response; no difference on response rate was noticed according to cisplatin dose. Median overall survival was 30.1 weeks, with 1- and 2-year survival rates of 36.3% and 13.2%, respectively. Conclusion Age does not preclude assessment on the role of cisplatin-vinorelbine CT for elderly NSCLC patients with good performance status and adequate bodily functions.	Background Lung cancer is a disease with a great incidence in older people. In Brazil, its incidence rate in male sex was expected to reach 18.8:100.000 in 1999, and aged patients may have accounted for 57% of all lung cancer deaths. Unfortunately, up to two thirds will present at diagnosis with advanced disease, requiring chemotherapy (CT) [ 1 ]. Vinorelbine, a semi synthetic vinca alkaloid, is a highly active drug for NSCLC and its association with cisplatin is worthwhile. European and Southwest Oncology Group trials demonstrated that vinorelbine/cisplatin (VP) offer therapeutic advantage over both drugs alone [ 2 - 4 ], previous cisplatin-based schedules [ 3 ]; comparing to taxane combinations, VP is therapeutically equivalent to carboplatin/paclitaxel [ 5 ] or carboplatin/docetaxel [ 6 ] but inferior to cisplatin/docetaxel [ 6 ]. Although modest improvements in the survival of patients with non-small cell lung cancer (NSCLC) can be achieved with cisplatin-based CT [ 7 ], its value is disputed in the geriatric context. The simultaneous presence of several diseases and homeostenosis, an age-related physiologic process that change the way that the body handles drugs, can shift therapeutic index, allowing harm outweigh any survival gain. On the other hand, underdiagnosis and undertreatment of lung cancer in the elderly is a fact, often explained in terms of ageism in medical oncology staff [ 8 ] and people's beliefs and fears about the disease and its treatment [ 8 , 9 ]. Whether we should treat or not aged patients with cisplatin-based CT surely is an unsolved issue. In this study, we evaluated the feasibleness and activity of vinorelbine/cisplatin CT for elderly NSCLC patients. Methods This study enrolled patients older than 70 years with unresectable locally advanced or metastatic NSCLC. Informed consent was obtained from patients and their relatives, as approved by the Institutional Ethical Committee. Inclusion criteria: all patients had to have histologically confirmed NSCLC; Karnofsky performance status ≥ 70; measurable disease; adequate bone marrow reserve (neutrophils ≥ 2 × 10 9 /L and platelets ≥ 100 × 10 9 /L), bilirubin under 1.25 times upper normal value (UNV), aspartate aminotransferase/alanine aminotransferase (AST/ALT) under 2 times UNV, and renal function (creatinine level under 120 μmol/L); no symptomatic brain metastasis; no prior cancer therapy; no indication for palliative radiation therapy; no previous or concomitant malignancy; and adequate social support. Exclusion criteria were symptomatic peripheral neuropathy and comorbidity regarded as an impediment for CT, such as renal disease, heart failure, coronary heart disease, uncontrolled infection, and cognitive impairment. Baseline work up included a medical history and physical examination; whole blood count (WBC) and biochemistry; chest x-ray; bone scintigraphy scan; chest, abdominal and brain computed axial tomography scans; and electrocardiogram. Although pretreatment bone scan and brain CAT scan are recommended only when signs or symptoms of disease are present, they were added here due to Institutional Protocol Reviewers recommendation. Treatment consisted of vinorelbine 25 mg/m 2 on days 1 and 8, administered intravenously in bolus, followed by intravenous cisplatin over 1 hour. CT was administered every 4 weeks. Prophylactic anti-emetic drugs (intra venous dexamethasone 20 mg and ondansetron 16 mg) and fluid hydration (0.9% saline, 1 L/m 2 ; and magnesium sulfate, 20 mmol) was used to minimize renal toxicity. Dexamethasone 4 mg PO BID plus metoclopramide 10 mg PO QID for 4 days was used to prevent delayed emesis. Patients were time-sequentially assigned to one of two groups, from lower to higher doses, according to age strata. Study doses were defined based on previous reports on renal tolerance of cisplatin in geriatric patients [ 10 , 11 ]. Age stratification was arbitrary. Those aged up to 75 years received cisplatin 60 or 70 mg/m 2 and those aged 70 to 75 received cisplatin 80 or 90 mg/m 2 . Assignation to high dose groups (70 mg/m 2 and 90 mg/m 2 ) occurred after evaluation of toxicity at inferior doses, evaluated according to National Cancer Institute criteria. The protocol required a minimum of 6 and a maximum of 18 patients per group after the 1st cycle for safety analysis. Chemotherapy doses were reduced for haematological, neurological, hepatic and renal toxicities. Toxicity was graded according to National Cancer Institute (NCI) common toxicity criteria guidelines. Changes in dosage were based on WBC results obtained on day 1 of treatment; if neutrophils were <1.5 × 10 9 /L and platelets were <100 × 10 9 /L, treatment was delayed by 1 week. Treament on days 8 had to be cancelled if neutrophil counts were <1.0 × 10 9 /L and platelets were <100 × 10 9 /L. If treatment could not be given after a 2-week interval because of haematological toxicity, it had to be discontinued and the patient withdrawn from the study. Concomitant use of hematopoietic growth factors were not allowed in the first treatment cycle but were administered subsequently on individual basis. Neurological toxicity above grade 2 resulted in suspension of treatment; ototoxicity grade 2 or 3 resulted in a 50% dose reduction of cisplatin. The following dose modifications of vinorelbine were set based on AST/ALT (aspartate aminotransferase/alanine aminotransferase) and bilirubin values on day 1 or day 8 of treatment: if AST/ALT were between 5.1 and 20.0 × UNV or bilirubin was between 1.5 and 3.0 × UNV, dosing was cancelled and the patient was reassessed 1 week later. If AST/ALT were >20.0 × UNV or bilirubin was >3.0 × UNV, vinorelbine was discontinued. If serum creatinine was grade >1, the dose was delayed by 1 week and the test repeated. After a 2-week delay, the patient was taken off the study. WBC and biochemistry were also performed on day + 14 of treatment. We intended to administer a maximum of four CT cycles followed by radiation therapy in responding patients with stage III disease and six CT cycles in patients with wet IIIB or stage IV disease. Notwithstanding radical radiation therapy (RT) should deliver a total dose down to 66 Gy, covering tumor site and regional lymph nodes, and palliative therapy could use doses under to 45 Gy, the final choice of dose, fractioning, irradiated volume, and energies of radiation was at the radiation oncologist's discretion. Treatment interruption was allowed in case of disease progression, severe adverse events, or patient preference. Chest x-ray was performed before each cycle, and CAT scans every two cycle for response evaluation. Tumor response was recorded according to World Health Organization criteria and measured by the same observer (JRP). All responses had to be confirmed 3–4 weeks from initial evaluation. We reported here the best response designation recorded from the start of treatment until disease progression. Patients stopping treatment with an unconfirmed response, or only short stabilisation were considered as inevaluable, unless the response or stabilisation was further confirmed in the absence of any treatment. Patients were monitored for the first month off-study then followed up every 2–3 months. The dose intensity was calculated for both drugs by dividing the actual dose delivered by the length of therapy. Toxic death was defined as death occurring during the chemotherapeutic phase (including four weeks after its end) and due to drug toxicity. Early death was defined as death within four weeks after a chemotherapy cycle without severe toxicity and not related to the malignant disease. Response and survival were calculated by intention-to-treat. Progression was defined in relation to the best response obtained. The time to tumor progression lasted from the first day of treatment to the date of the first observation of progressive disease. Survival was defined as the time elapsed from the beginning of CT until death or last follow-up visit. Time-to-event analysis was performed using the Kaplan-Meier product-limit estimator. All analyses were carried out using a computer program (SPSS version 8.0, Chicago, USA). Results Forty-four patients were recruited from July 1996 to June 1998; twenty-nine aged 70–75 year and fifteen older than 75 years. Cisplatin doses were as follows: in the older cohort, seven patients received 60 mg/m 2 and eight received 70 mg/m 2 ; in the former, fifteen patients received 80 mg/m 2 and fourteen received 90 mg/m 2 . Patient characteristics are given in Table 1 . Most of the patients presented stage III disease (56.8%) and squamous cell carcinoma (52.3%). Treatment results are shown in Table 2 . A total of 125 CT cycles were administered and the median was 3 (range: 1 to 6). No difference was noticed on the dose intensity achieved across the four groups (Kruskal-Wallis test, p = 0.13). Objective response could not be evaluated in six patients due to treatment discontinuation before cycle 2: early death (1), withdrawal of consent (2) and toxicity (3). Twelve out of 25 patients with stage III disease responded to CT and received radical radiation therapy (median delivered dose: 50 Gy; range: 40 Gy to 66 Gy). Fifteen patients (34.0%) received maximum allowed CT cycles; excessive toxicity (n = 8), progressive disease (n = 3), progressive disease after initial response (n = 13), and patient choice (n = 5) were reasons for protocol withdrawal. At a median follow-up time of 77.2 weeks, six patients were alive. Response rate (RR) was 50.0% (95% CI: 35.4% to 64.5%), with 1 complete response. Global median time to progression (TTP) and overall survival (OS) were 27.0 (95% CI: 10.1 to 43.7) weeks and 30.1 (95% CI: 24.4 to 35.8) weeks; 1- and 2-year survival rates were 36.3% and 13.2%, respectively. No significant difference was noticed in RR (p = 0.65), TTP (p = 0.62), and OS (p = 0.44) across study groups. There was no difference according to stage (III vs. IV) in RR (48% vs. 53%, p = 0.76), TTP (32.6 vs. 25.0 weeks, p = 0.73), or OS (31.7 vs. 28.6 weeks, p = 0.33). Likewise there was no difference according to age groups (70–74 vs. ≥ 75 years) in RR (55% vs. 40%, p = 0.34), TTP (31.7 vs. 28.6 weeks, p = 0.33), or OS (30.1 vs. 31.7 weeks, p = 0.76). Toxicity data are presented in Table 3 . Hematological toxicity (grade 3–4) was as follows: neutropenia, 20%; anemia, 11%; and thrombocytopenia, 2%. Common nonhematologic grade 3–4 side effects were alopecia (18%) and fatigue (11%); severe peripheral neurotoxicity occurred in one patient; neither severe emesis nor renal toxicity was noticed. At the highest cisplatin dose (90 mg/m2) there were two early deaths and one toxic death due to neutropenic sepsis. No case of febrile neutropenia was noticed. Discussion Treatment of elderly NSCLC patients with cisplatin-based regimens has been a less contentious matter nowadays but toxicity remains a major issue. In our pilot study, chemotherapy with cisplatin 70–80 mg/m 2 on day 1 plus vinorelbine 25 mg/m 2 on days 1 and 8, repeated each 28 days per in the maximum four cycles, was feasible for elderly NSCLC patients. Neurological and renal tolerance was particularly good. At the time we developed the protocol, no quality-of-life instrument had been validated for use in Brazil. Thus, a drawback in our study is the absence of quality-of-life analysis, which precludes evaluation of key dimensions in geriatric oncology. Cisplatin induces a sensory neuropathy due to axonal damage that is dependent on the total-dose and single-dose intensity [ 12 ], and it is also time-dependent [ 13 ], making histological lesions more common than clinical toxicity. Although in this trial most patients received moderate cisplatin doses, only one had grade 3 peripheral neuropathy. Our data may reflect the low median of cycles actually administered rather than inaccuracy of clinical signs to evidence tissue lesion. Similarly, Ohe et al. [ 14 ] delivered a median of three cycle of cisplatin-containing CT and reported no case of severe neuropathy. Cisplatin renal toxicity has been attributed to drug-protein interactions and the inactivation of specific brush border enzymes, resulting in damage of the loops of Henle, the distal tubules, and collecting ducts. Patients aged above 70 or even 80 are regarded as susceptible to cisplatin-induced renal damage as the younger counterparts [ 15 , 16 ] and current studies have reported a low incidence of renal toxicity in elderly patients [ 14 , 17 ]. No case of severe renal toxicity was noticed in our patients, as estimated by serum creatinine and its clearance (Cockroft-Gault method), but this finding may be artifactual due to small sample sizes, selection bias, and low sensitivity of estimated creatinine clearance to predict actual glomerular filtration rate (GFR) [ 18 ]. The observed response rate here was in the usually range of NSCLC phase II trials, but the absence of external review of radiological data and the widened confidence intervals expected because of small sample sizes in each group limit assertions that could otherwise be drawn. The 1-year survival rate (36%) was good but essentially equivalent to the reported elsewhere [ 19 ] for vinorelbine alone (32%) and inferior to the observed for weekly cisplatin-docetaxel (64%) [ 14 ]. Nonetheless, survival figures should be cautiously considered in this underpowered, heterogeneous, non-randomized pilot study. Vinorelbine is a cytotoxic agent that clearly has expanded the therapeutic options for elderly NSCLC patients [ 20 , 21 ]. The next logical step to improve therapeutic indexes was to combine it with other active drugs. Recently, relevant results emerged from European phase III trials addressing the role of novel drugs in the treatment of elderly NSCLC patients [ 19 , 22 - 25 ]. The Elderly Lung Cancer Vinorelbine Italian Study (ELVIS) [ 19 ], interrupted for slow recruitment, evidenced an improvement of some lung cancer-related symptoms (pain and dyspnea), worsening of toxicity-related symptoms (cognitive function, constipation, and peripheral neuropathy), and a limited survival advantage (28 vs. 21 weeks) for single-agent vinorelbine as compared to supportive care, a survival gain that resembles the benefit reported by meta-analysis for nowadays considered substandard cisplatin-based regimens in advanced NSCLC. Although sequential administration of drugs is an attractive option for aged or frail patients, a setting where minimal treatment-related toxicity should be pursued, research on the role of non-platinum combinations for elderly NSCLC patients aroused attention. The Southern Italy Cooperative Oncology Group (SICOG) evaluated whether the association of vinorelbine and gemcitabine would be better than vinorelbine alone. To date, final results of this trial have not to come. Despite an article focusing on the interim analysis of 120 patients (60 at each group) claimed a survival advantage for the combined arm [ 22 ], further intent-to-treat analysis including 18 patients more reached a less optimistic conclusion: median survival in the combined and single-agent arms were nearly the same (25 weeks and 23 weeks, respectively) and both values were deemed comparable to the observed in the supportive care arm of the ELVIS (21 weeks) [ 23 ]. Since that at least 152 patients was treated in the SICOG trial [ 24 ], a definite report of mature survival data is awaited. In addition, investigators of the Multicenter Italian Lung Cancer in the Elderly Study (MILES) [ 25 ] reported no survival benefit for the combination of vinorelbine plus gemcitabine in comparison to single-agent vinorelbine or gemcitabine in the treatment of elderly NSCLC patients. The question whether or not cisplatin-containing regimens should be used to treat aged patients remains an important, still open, issue. As observed for paclitaxel-carboplatin [ 26 ], gemcitabine-cisplatin [ 27 ], and docetaxel-cisplatin [ 14 ] associations, the role of vinorelbine-cisplatin regimens deserve to be investigated. Until the outcome of large clinical trials addressing this issue proves at least the equivalence of newer drug associations to platinum-based regimens, as seems to be true for the combination of paclitaxel and gemcitabine [ 28 ], there are few reasons to preclude the evaluation of current combined regimens in the chemotherapy of elderly NSCLC patients with normal bodily functions and good performance status. Competing interests This study was supported by Institutional funds only. Authors did hot received reimbursements, fees, funding, or salary from an organization that may in any way gain or lose financially from the publication of this paper in the past five years. Abbreviations (order of appearance) NSCLC: non-small cell lung cancerCT: chemotherapy NCI: National Cancer Institute ELVIS: Elderly Lung Cancer Vinorelbine Italian Study SICOG: Southern Italy Cooperative Oncology Group GFR: glomerular filtration rate Authors' contributions JRP was responsible for study conception, protocol conduction, and results interpretation. SJM carried out the data analysis and results discussion. SMN and FKI were responsible for patient care and data gathering Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526296.xml
545080	Systematic reviews of epidemiology in diabetes: finding the evidence	Background Methodological research to support searching for those doing systematic reviews of epidemiological studies is a relatively neglected area. Our aim was to determine how many databases it is necessary to search to ensure a comprehensive coverage of the literature in diabetes epidemiology, with the aim of examining the efficiency of searching in support of systematic reviews of the epidemiology of diabetes Methods Three approaches were used. First, we defined a set of English language diabetes journals and examined their coverage in bibliographic databases. Second, we searched extensively for diabetes epidemiology articles (in all languages) to determine which are the most useful databases; and third, we analysed the scattering of these articles to determine the core journals in the area. Results The overlap between MEDLINE and Embase for diabetes journals was 59%. A search for diabetes epidemiology articles across both MEDLINE and Embase, showed that MEDLINE alone retrieved about 94% of the total articles. Searching for diabetes epidemiology studies beyond MEDLINE and Embase retrieved no additional English language journal articles. The only diabetes epidemiology studies found by searching beyond MEDLINE and Embase were found in LILACS, and were Spanish or Portuguese language studies from Latin America; no additional English language studies were found. Only 30% of the meeting abstracts were converted to full publication after three years. One third of journal articles were published in just six journals, with Diabetes Care contributing 14.3% of the articles, followed by Diabetic Medicine (5.0%); Diabetes Research & Clinical Practice (4.1%); Diabetologia (4.0%); Diabetes & Metabolism (2.4%) and Diabetes (2.0%). Conclusions Our results show that when searching for articles on diabetes epidemiology, MEDLINE and Embase would suffice for English language papers, with LILACS giving some additional non-English articles from Latin America. Although a MEDLINE-only search will retrieve the vast majority of the relevant literature, Embase and LILACs should also be searched to ensure the search is comprehensive. Searching for meeting abstracts is recommended to alert reviewers to unpublished work. The low rate of full publication of meeting abstracts has the danger of producing bias in reviews. Our findings on scattering show that the core literature in this field is concentrated in just six journals.	Background Review articles have a valuable role in the medical literature, because the volume of journals and articles is such that keeping up to date is very difficult. Reviews are much more valuable if they are systematic reviews done to internationally agreed standards, as non-systematic reviews are known to be subject to bias [ 1 , 2 ]. Dickersin noted that there was a shortage of systematic reviews in epidemiology, and called for more reviews, and more research into the methodology relating to reviews in epidemiology [ 3 ]. A study by Breslow found that more than 60% of epidemiology reviews were not methodologically systematic [ 4 ], and there has been little methodological research relating to their performance. Also, the methods have not been standardised, and the literature searching has not been supported, as done by the Cochrane Collaboration to support systematic reviews of clinical trials. One of the key quality criteria for systematic reviews is the comprehensiveness of the searching, as failure to identify all relevant reports can result in selection bias[ 5 ]. The usual first source for identifying studies for reviews is MEDLINE, which currently indexes 4780 titles [ 6 ] of the estimated 14 000 biomedical titles currently published throughout the world [ 7 ]. In addition to searching MEDLINE, it is recommended that an extensive range of additional sources are searched [ 1 ]. A study, using Ulrich's International Periodicals Directory as a gold standard, found that a MEDLINE search in psychiatry would retrieve only about half the relevant journals [ 8 ]. Similarly, we were interested to investigate the coverage of diabetes journals and diabetes epidemiology articles in medical databases, in order to determine how many databases it is necessary to search to ensure a comprehensive coverage of the literature in diabetes epidemiology. The scattering of the journal literature in a subject area can provide a useful insight into the number and range of journals needed to capture the key literature in a field. Bradford's Law of Scattering states that on any one subject, a small group of 'core' journals (Zone 1) will provide about one third of the articles on that subject, a medium number of less-core journals (Zone 2) will provide another third, and a large number of peripheral journals (Zone 3) will provide the final third of the articles [ 9 , 10 ]. The aim of this study was to examine the efficiency of searching in support of systematic reviews of the incidence and prevalence of diabetes by providing empirical data to answer the following questions: 1. Which databases index diabetes journals (restricted to English language)? 2. Which databases outside MEDLINE and Embase index diabetes epidemiology journal articles and grey literature? 3. How are diabetes epidemiology articles scattered across the journals, and what are the core journals in this area? Accordingly, this study was divided into three parts. (Note: This study is concerned only with searching for the epidemiology of diabetes itself, not with its complications). Methods For the purposes of this study, epidemiology articles were defined as studies of incidence or prevalence of diabetes, or of factors affecting those (thereby excluding studies looking at other epidemiological aspects such as mortality). Basic science studies, e.g. biological mechanisms of disease, were not included. We started from a position that both MEDLINE and Embase should always both be searched, since the overlap between these databases has been estimated to range from 10% to 87% [ 5 , 11 - 17 ]. Also, it is recognised that many relevant studies will appear in non-diabetes journals, and in sources not indexed in MEDLINE or Embase. Hence, a three-part approach was used to investigate literature searching to support systematic reviews of diabetes epidemiology, and address each of our aims in turn. Part one: defining a set of 'diabetes' journals, and determining the databases in which theses journal were indexed Diabetes journals were identified from the 'Medical Sciences – Endocrinology' section of Ulrich's Periodical Directory 2003 (a comprehensive bibliographic database providing detailed information on periodicals published throughout the world) [ 18 ]. This was supplemented with a search of PubList.com [ 19 ] for journals with 'diabet' in title. The inclusion criteria for the journals were: i) the word stem 'diabet' in the title, ii) contains original scientific studies of an academic or scholarly nature, iii) currently in print, and iv) published in English. If inclusion could not be decided on the basis of the information provided in Ulrich's , the contents pages for the past five years, and where necessary, abstracts or the full journal articles, were examined by a diabetes epidemiologist. Journals which fulfilled all the above criteria were then checked against the List of Serials Indexed for Online Users [ 6 ] to see if they were indexed in MEDLINE. To determine if a journal was indexed by Embase, a search was done of 'diabet$' in the Journal Word (JW) field using the Embase OVID search interface, and limiting to publication year 2003. If any journals were not indexed in either MEDLINE or Embase, searches of BIOSIS, BNI, CINAHL, SCI were done to determine if the journals were indexed in any of these databases. Part two: searching databases other than MEDLINE and Embase for journal articles and grey literature on diabetes epidemiology Databases searched were: AMED, ASSIA, BIOSIS (abstracts only), BNI, CINAHL, Conference Papers Index, Dissertation Abstracts US, Health Management Information Consortium (HMIC), Index to Theses UK, ISI Proceedings, PsycINFO, NLM Gateway, LILACS (Latin American and Caribbean Health Sciences), National Research Register (NRR), SIGLE, SCI (abstracts only), SSCI, and Zetoc. The search strategy used was 'diabetes and (epidemiology or incidence or prevalence)' in the Title (TI) field and restricted to publication years 1998 to 2003. The titles (and abstracts when available) of all records were checked by an expert in diabetes epidemiology, in order to determine their potential usefulness for those doing systematic reviews. Part three: investigating the scatter of diabetes epidemiology journal articles found in a search of MEDLINE and Embase, and determining the core journals in this area MEDLINE and Embase were searched using the search strategy: 'Diabetes Mellitus as the major subject heading, with the sub-heading 'Epidemiology' assigned (which includes incidence and prevalence), and restricted to publication years 1998 to 2003'. All languages were included. Duplicates found in both MEDLINE and Embase were removed. The journal titles in which the articles were found were ranked according to the number of articles contributed by each journal. The cumulated numbers of articles and journals were calculated and plotted. This was used to identify Bradford zones; that is, the number of journals needed to cover about one third, two thirds or all the relevant articles in the field. Results Part one: defining a set of 'diabetes' journals, and determining the databases in which theses journal were indexed Searches of Ulrich's Periodicals Directory and PubList.com initially identified four English language journals that were of potential interest but not indexed by MEDLINE or Embase. On closer inspection, three of these, Clinical Diabetes, Journal of Diabetes Nursing , and Diabetes and Primary Care were excluded as they did not appear to contain any primary research; the articles were mainly educational, professional news and views, opinions, or narrative reviews. The fourth journal, The Diabetic Foot , contained primary research, but did not appear to contain studies useful for epidemiological reviews of diabetes itself (as opposed to complications). It is indexed by CINAHL only. As shown in Table 1 , 27 English language diabetes were covered collectively by MEDLINE and Embase in 2003. Seventy-four percent (20) were indexed in MEDLINE, 85% (23) in Embase, and 59% (16) were in both MEDLINE and Embase. Table 1 English language diabetes journals indexed in either MEDLINE or Embase in 2003 Indexed in Acta Diabetologica MEDLINE & Embase Diabetes MEDLINE & Embase Diabetes & Metabolism MEDLINE & Embase Diabetes Care MEDLINE & Embase Diabetes Research MEDLINE & Embase Diabetes Research & linical Practice MEDLINE & Embase Diabetes Technology & Therapeutics MEDLINE & Embase Diabetes, Nutrition & Metabolism – Clinical & Experimental MEDLINE & Embase Diabetes, Obesity & Metabolism MEDLINE & Embase Diabetes/Metabolism Research Reviews. MEDLINE & Embase Diabetic Medicine MEDLINE & Embase Diabetologia MEDLINE & Embase Experimental & Clinical Endocrinology & Diabetes MEDLINE & Embase Experimental Diabesity Research MEDLINE & Embase Journal of Diabetes & its Complications MEDLINE & Embase Pediatric Diabetes MEDLINE & Embase Current Diabetes Reports MEDLINE Diabetes Educator MEDLINE Diabetes Forecast MEDLINE Diabetes Self-Management MEDLINE British Journal of Diabetes & Vascular Disease Embase Canadian Journal of Diabetes Embase Cme Bulletin Endocrinology & Diabetes. Embase Current Opinion in Endocrinology & Diabetes Embase Diabetologia Croatica Embase Journal of Endocrinology, Metabolism & Diabetes of South Africa Embase Practical Diabetes International. Embase The four diabetes journals unique to MEDLINE were all published in the USA. By contrast, only one of the seven journals unique to Embase was published in the USA. Part two results: searching databases other than MEDLINE and Embase for journal articles and grey literature on diabetes epidemiology The results are summarised below: Journal articles (English language) No English language journals articles that were not also indexed in MEDLINE or Embase were identified. Journal articles (non-English language) There were 23 Spanish and Portuguese language articles identified in LILACs. On the basis of the English translation of the titles, they all reported studies done in Latin America. Grey literature We defined grey literature as any literature not published in a peer reviewed journal. After removing duplicates, there were 51 dissertations identified from searches of Dissertations Abstracts US, Index to Theses UK, and SIGLE. The research presented in the vast majority (92%) of the dissertations appeared to have been written up as articles in journals indexed in MEDLINE or Embase. No grey literature studies of any format other than dissertations were retrieved from SIGLE, so there was very little additional information gained by these searches. Research in progress The National Research Register (a database of ongoing and recently completed research projects funded by, or of interest to, the United Kingdom's National Health Service) gave brief details of 18 projects in progress that had not been otherwise identified. Searching the NRR might be useful if unpublished results could be included in the review, but its main value would be to indicate when the review was likely to need updating. Meeting abstracts and conference proceedings The search of the Conference Proceedings Index retrieved 25 articles, none of which appeared to have been published as journal articles after five years. The Zetoc Conference Search found eight articles, of which 50% had been published as full journal articles in MEDLINE or Embase. The search of Science Citation Index (SCI), restricted to meeting abstracts only, found 171 relevant studies. The time to publication of the SCI abstracts was examined by checking how many had subsequently been published as journal articles indexed in MEDLINE or Embase. It was found that 30% had reached full publication after three years. A search of BIOSIS, restricted to meeting abstracts only, retrieved 71 additional relevant abstracts that were not in SCI. Most (65%) of these 71 abstracts came from the supplements of Diabetes and Metabolism and Diabetes Research and Clinical Practice . Of these, 11 (12%) had been published in journals. The average time delay from the date of publication of the abstract to full publication was 1.4 years. Databases searched where no articles not in MEDLINE or Embase were found These included AMED, BNI, HMIC, NLM Gateway meeting abstracts, PsychINFO, and SSCI, In summary, the data indicate that when searching for English language journal articles on diabetes epidemiology, searches of MEDLINE and Embase would suffice. The exception would be for studies from Latin America, where LILACS should also be searched. Searching for meeting abstracts may alert reviewers to forthcoming or unpublished work. Part three: investigation of the scatter of diabetes epidemiology journal articles found in a search of MEDLINE and Embase, and determination of the core journals in this area The searches for diabetes epidemiology articles in MEDLINE and Embase resulted in 2923 articles being found in 696 different journal titles; 39% were found to be in 'diabetes journals' and 14% were non-English language. Figure 1 shows the distribution of all 696 articles retrieved across the journals. Figure 1 Distribution of diabetes epidemiology articles across journals Applying Bradford's Law of Scattering gives three zones, each providing one-third of the articles. Zone 1 The first one-third of articles were from six journals. They are in rank order: Diabetes Care (contained 14.3% of the articles); Diabetic Medicine (5.0%); Diabetes Research & Clinical Practice (4.1%); Diabetologia (4.0%); Diabetes & Metabolism (2.4%), Diabetes (2.0%) These six journals represent 0.9% (of the 696) total journals, and all are indexed in MEDLINE. Zone 2 The second one-third of articles were from 62 journals, representing 9.1% of the total journals. The four journals in Embase only were: Practical Diabetes International ; Diabetologia Polska ; Diabetes und Stoffwechsel and Journal of the Japan Diabetes Society . Hence, 94% of Zone 2 journals are covered by MEDLINE. Zone 3 The final one-third of articles were in 628 journals, representing 90.2% of the total journals. MEDLINE indexed 88% of these journals. Overall, for the three zones, the search of MEDLINE and Embase for diabetes epidemiology articles revealed that MEDLINE indexed 89% of the total journals, and these contained 94% of the articles. Discussion Our results showed that there was an overlap of only 59% in current English language 'diabetes journals' indexed by both MEDLINE and Embase. Also, a search for diabetes epidemiology articles across both MEDLINE and Embase showed that MEDLINE alone retrieved about 94% of the total articles; therefore, both databases should be searched. Embase appears to index more diabetes journals published outside the USA. Therefore, if searching is limited to MEDLINE only (Embase being less accessible and more expensive than MEDLINE, which is free via PubMED) this could potentially introduce a bias. Also, duplication of searching can be useful, as due to differences in indexing practices, a search of one database may retrieve something missed by the other. We also found that despite a wide range of additional databases searched after MEDLINE and Embase, no additional English language journal articles on diabetes epidemiology were identified. The LILACS database was a useful source of Spanish and Portuguese language articles on the epidemiology of diabetes in Latin American countries. Meeting abstracts appeared to be valuable sources of information on forthcoming studies, but their inclusion in systematic reviews is contentious. Some reviewers exclude abstracts on the grounds that the quality of the study cannot be judged because of the inevitably limited detail. However, others include them on the grounds that abstracts provide the most up-to-date information. Nearly all the dissertations identified had been published as journal articles. However, it was found that only 30% of the meeting abstracts were converted to full publication after three years, which is considerably lower than the figure for RCTs, which is 56% [ 20 ]. This has the danger of producing bias in systematic reviews, if failure to publish is based on the size and direction of study results. Scattering of the diabetes epidemiology articles revealed that the 'core' literature in this field is concentrated in just six journals, with Diabetes Care alone containing about 14% of the articles. A similar concentration effect in journals was also shown in a study of 3400 science journals in the SCI database, where just 100 journals accounted for 22% of the published articles and 100 journals also accounted for 44% of cited articles [ 21 ]. This study has a number of limitations. The search to identify diabetes journals was restricted to English language journals only, as we were unable to assess articles in other languages. We did not compare the quality of the articles identified from databases outside MEDLINE and Embase. Also, when searching for articles, we were necessarily limited to the range of databases available to us. Finally, there may be databases inaccessible or unknown to us that cover foreign language and regional journals not indexed in MEDLINE and Embase. Such journals may carry studies of incidence which may seem of primarily local interest, but which may be useful contributions to the international body of evidence because they may show large variations in incidence, or in its relationship to possible aetiological risk factors. It is often useful to study the epidemiology of a disease where it is rare, as well as where it is common. However it is likely that studies which report high incidence are more likely to be published than those which report low incidence. Similarly with risk factors; a study which finds no link between factor x and disease y may be less likely to be published than one which does show a correlation [ 22 , 23 ]. There is a need for further research to see whether our findings apply to searching for epidemiological reviews of other diseases, and on measuring the sensitivity and specificity of various search filters to retrieve epidemiological studies in MEDLINE and Embase. We endorse Dickersin's suggestion of an international collaborative effort to establish an 'epidemiological Cochrane-like database' to identify all relevant studies and to begin systematically reviewing available data for important epidemiological questions [ 3 ]. Conclusions Searching MEDLINE and Embase appears to provide comprehensive coverage of the English language journal literature in diabetes epidemiology. LILACs is a useful source of Spanish and Portuguese language articles on diabetes epidemiology done in Latin American countries and published in regional journals not indexed in MEDLINE and Embase. Searching for meeting abstracts is recommended to alert reviewers to unpublished work. The volume of literature on diabetes epidemiology makes it impossible for one person to read everything. However the provision of systematic reviews makes keeping up with research manageable, and more reviews are needed. Our findings on scattering shows that the core literature in diabetes epidemiology is concentrated in a small number of core journals, and that in the absence of reviews, one can follow the field by reading these journals. It may also be reassuring that a good MEDLINE-only search will retrieve the vast majority of the relevant literature. Competing interests The author(s) declare that they have no competing interests. Authors' contributions PR and NW conceived the study and drafted the initial manuscript. PR and LB collected and analysed the data. All authors contributed to and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545080.xml
524495	The validity of the SF-36 in an Australian National Household Survey: demonstrating the applicability of the Household Income and Labour Dynamics in Australia (HILDA) Survey to examination of health inequalities	Background The SF-36 is one of the most widely used self-completion measures of health status. The inclusion of the SF-36 in the first Australian national household panel survey, the Household, Income and Labour Dynamics in Australia (HILDA) Survey, provides an opportunity to investigate health inequalities. In this analysis we establish the psychometric properties and criterion validity of the SF-36 HILDA Survey data and examine scale profiles across a range of measures of socio-economic circumstance. Methods Data from 13,055 respondents who completed the first wave of the HILDA Survey were analysed to determine the psychometric properties of the SF-36 and the relationship of the SF-36 scales to other measures of health, disability, social functioning and demographic characteristics. Results Results of principle components analysis were similar to previous Australian and international reports. Survey scales demonstrated convergent and divergent validity, and different markers of social status demonstrated unique patterns of outcomes across the scales. Conclusion Results demonstrated the validity of the SF-36 data collected during the first wave of the HILDA Survey and support its use in research examining health inequalities and population health characteristics in Australia.	Background While much health research focuses on objective outcome measures such as mortality or morbidity defined through clinical assessment, there is an increasing emphasis on self-reported measures of health status and health-related quality of life. Self-reported measures of health status have been included in epidemiological and community-based survey research. Their use reflects the importance of considering the patients' point of view and the multidimensional nature of health [ 1 - 3 ]. The Medical Outcomes Study Short Form (SF-36) is one of the most widely used, self-completion measures of health status. It was developed to meet the psychometric standards necessary for group comparisons, to enable profiling of functional health and well-being, and to quantify disease burden [ 3 ]. It comprises 36 items of which all but one are used to measure eight important health concepts that are frequently examined through health surveys. These eight concepts or scales are: Physical Functioning; Role-Physical (interference with work or other daily activities due to physical health); Bodily Pain; General Health; Vitality; Social Functioning (interference with normal social activities); Role-Emotional (interference with work or other daily activities due to emotional problems); and Mental Health (symptoms associated with anxiety and depression and measures of positive affect). In addition, the eight scales yield two summary scales of health, relating to physical (the Physical Component Summary: PCS) and mental (the Mental Component Summary: MCS) functioning and well-being. The SF-36 was first adapted for use in Australia in 1992, as part of the International Quality of Life Assessment (IQOLA) Project [ 4 ]. Previous research has demonstrated the validity of the SF-36 for use by Australian respondents using samples from Canberra and New South Wales [ 2 , 4 ]. This has involved assessment of the psychometric properties of the Australian form of the SF-36, evaluation of internal consistency and reliability, and demonstration of content and construct validity. There are considerable Australian data on the SF-36 from large National samples. In 1995, a subset of National Health Survey respondents (around 18,800 adults) completed the SF-36 and the Australian Bureau of Statistics published Australian population norms [ 5 ]. The SF-36 was also included in the Women's Health Australia survey, with data collected from a sample of around 41,500 women aged 18–22, 45–49, and 70–74 [ 6 ]. In 2001, the SF-36 was included in the first wave of the Household, Income and Labour Dynamics in Australia (HILDA) Survey. This is an Australia-wide survey of approximately 7,680 households, comprising around 14,000 people aged 15 and over. The HILDA Survey is the first longitudinal household survey in Australia and is designed to provide a sound evidence base to support research and analysis of income, labour market and family dynamics. As such, it is a critical resource for social policy development. The inclusion of the SF-36 in the HILDA Survey will enable investigation of the interaction between social, economic and health measures. There is an extensive body of research demonstrating socio-economic inequalities in the distribution of physical and mental health problems [ 7 - 9 ]. The SF-36 has been utilised in research on health inequalities and this research has shown that the SF-36 scales are differentially associated with markers of social-economic circumstance [ 10 - 13 ]. A focus of our research has been on the nature of disadvantage and social exclusion associated with welfare receipt and specifically the association between welfare receipt and mental health problems [ 14 ]. The HILDA Survey provides a valuable dataset with which to investigate this research topic. As such, it is critical to firstly ascertain the validity of the data within the HILDA Survey. Further, demonstrating the validity of the SF-36 scales collected through the survey is critical for other researchers and policy analysts who will utilise the HILDA dataset. The aim of this paper is to evaluate the psychometric properties and criterion validity of the HILDA SF-36 data. We have used the manuscript by Sanson-Fisher and Perkins [ 4 ] as a framework for the analyses reported in the first section. This follows the standard IQOLA validation procedures [ 15 ]. The analysis examines the reliability and validity of the eight SF-36 scales and the PCS and MCS scales. We then compare the SF-36 results obtained from the HILDA Survey with other Australian estimates to assess the representativeness of the data. We also evaluate the criterion validity of the HILDA Survey SF-36 data. We do this by 1) looking for convergent validity, in which scales measuring similar or related constructs demonstrate a positive association, and divergent validity whereby unrelated scales and measures are not associated; and 2) examining the profile of SF-36 results across a range of measures associated with health, disability and social functioning, demographic characteristics as well as a focus on a number of measures of socio-economic circumstances. Methods Data source Data are from the first wave of the HILDA Survey (Release 1.0), a nationally representative household panel survey. The HILDA Survey was funded by the Australian Department of Family and Community Services and managed by a consortium led by the Melbourne Institute of Applied Economic and Social Research at the University of Melbourne. The survey utilised a multi-stage sampling approach (sampling households within Census Collection Districts) and was stratified by State and part-of-State. Four survey instruments were included in Wave 1. A Household Form and Household Questionnaire were completed during a personal interview with one adult member of each household. The Person Questionnaire , also administered during the personal interview, was conducted with all adult household members. Finally, the Self-Completion Questionnaire (SCQ) was provided to all respondents to the Person Questionnaire and was collected at a later date or returned by post. The SF-36 was included as the first element in the SCQ. Fieldwork for wave one of the HILDA Survey was conducted between August 2001 and January 2002. A total of 7,682 households responded to the survey (a household response rate of 66 per cent). Within these households, there were 15,127 eligible adults. Of this group 13,969 (92%) completed the Person Questionnaire and 13,055 (86%) completed and returned an identifiable SCQ. There are some differences between the characteristics of respondents to the HILDA Survey and population estimates from the Australian Bureau of Statistics. However, these discrepancies are not large enough to discredit the data and the differences in rates of response across both sex and location are corrected by applying population weights [ 16 ]. Analysis The first set of analyses evaluated the reliability and validity of the SF-36 scales in the HILDA Survey. Given that these analyses were concerned with the internal structure of the data rather than representativeness, we ignored the clustered and stratified nature of the data and did not take population weights into account. These analyses were conducted using SPSS version 11.7. Item-internal consistency assessed the extent to which each item measures what its associated scale measures. A correlation of 0.40 (corrected for overlap) or greater demonstrates adequate item-internal consistency [ 17 ]. Item-discriminant validity was also assessed. This is demonstrated when an item correlates significantly more strongly with the scale it contributes to rather than with any other scale. We use the multitrait/multi-item correlation matrix approach in which the correlation of each item with each scale is examined [ 18 ]. In this approach, the correlation between each item and its own scale is corrected for overlap: that is, the scale is calculated without the specific item in question to avoid inflating the correlation. The extent to which item-scale correlations within a scale were equal was also assessed, as was the approximate equality of item means and standard deviations. To assess the internal consistency of scale scores, Cronbach's alpha coefficients were assessed, with a criterion of 0.7 used to define adequate internal consistency. Descriptive statistics for the eight SF-36 scales and the two summary scales were calculated, including the mean, median, range, standard deviation, skewness, kurtosis, and the per cent scoring at the lowest value (floor) and the highest value (ceiling). Construct validity of the SF-36 in the HILDA Survey was also evaluated using the principal components method of factor analysis. Results were compared with the factor structure obtained from other analysis of the SF-36 scales. The second set of analyses also examined a form of construct validity, assessing the extent to which scores on the SF-36 scales were associated with other criteria. We included in our analysis a number of measures of health and disability, ratings of satisfaction with life and health, and measures of stressful employment conditions and persistent feelings of loneliness. In addition to examining the profiles across the eight SF-36 scales, we also examined the relationship between these criterion measures and the PCS scale and the MCS scale, which were calculated according to the standard procedure outlined by Ware, Kosinski and Keller [ 19 ]. For these analyses it was critical to take account of the complex survey design. We therefore utilised the svy procedures of STATA to account for the clustered and stratified nature of the data, and to facilitate the use of weights to overcome differential response rates and to replicate Australian population parameters. The HILDA Survey dataset contains person-level weights which, when applied to individual survey respondents, adjusts for the unequal probabilities of selection and completion of the Person Questionnaire. However, these weights do not adjust for the further attrition associated with the SCQ. This is critical as our analysis is focused upon measures from this questionnaire. We therefore, conducted a logistic regression analysis to predict completion of the SCQ using the same predictors used to derive the HILDA person-level weights (geographic location, labour-force status, sex, age, number of adults in household, number of children in household, marital status, English language ability, and dwelling type) [ 20 ]. The probabilities of responding derived from this analysis were used to adjust the person-level weights. Utilising these adjusted weights produced accurate population estimates from the respondents who completed the SCQ (similar to the estimates for Personal Questionnaire respondents using the original person-level weights). We also examine the correspondence between SF-36 results from the HILDA Survey and data from another, large-scale Australian survey to assess the representativeness of the data. We contrast the means for each of the SF-36 scales from the HILDA Survey with those obtained by the Australian Bureau of Statistics through the National Health Survey conducted in 1995 [ 5 ]. To facilitate evaluation of potential differences between means, two sample t-test statistics were calculated using a pooled variance approach. Demographic, socio-economic and health measures Based on previous analysis of the SF-36, we expected results across the eight scales to differ according to subgroups identified by demographic and socio-economic characteristics, including age, sex, marital status, educational attainment, housing tenure and employment status. We also examined differences associated with receipt of welfare payments, a focus of our research endeavours [ 14 ]. The analyses also include a range of measures related to health and life circumstances drawn from other scales and instruments included in the HILDA Survey. Long-term health condition One question from the Person Questionnaire asked respondents if they experienced a disability or health condition that had lasted, or was likely to last 6 months or more, and restricted their everyday activities. A showcard listing examples of health conditions, impairments and disabilities was presented as a prompt for respondents. Responses were either Yes or No . Limited ability to work Those identifying a disability or condition were asked to rate whether this limited the type or amount of work they were able to undertake. Response categories were No , Yes and Can Do Nothing . Satisfaction with health In a series of questions from the Person Questionnaire, respondents were asked to rate their satisfaction with a range of life circumstances using an eleven-point scale with descriptive anchors at either end (0 = totally dissatisfied; 10 = totally satisfied). Responses were categorised as either dissatisfied (if the rating was 4 or less) or satisfied/ mixed (if the rating given was at the midpoint 5 or higher). Satisfaction with life Respondents were also asked, all things considered, how satisfied they were with their life. Responses were categorised as above. Job stress The SCQ included a series of statements about current job characteristics, to which respondents rated their level of agreement or disagreement using a seven-point scale anchored at either end (1 = strongly disagree; 7 = strongly agree). One item asked respondents to rate whether they feared that the amount of job stress would make them physically ill. Respondents were categorised as experiencing significant job stress if they agreed with this statement (a rating of 5 or greater). Social isolation The final measure was also drawn from the SCQ. Using the same scale as the job stress measure, respondents were presented with a series of statements about their level of social support. Respondents were categorised as lonely if they agreed (rating of 5 or above) with the statement " I often feel very lonely ". Based on previous research using the SF-36, we anticipate that increasing age will be associated with poorer physical health, and that women will demonstrate poorer health than men as measured on all SF-36 scales with the possible exception of the General Health scale [ 5 ]. Those respondents who are divorced or separated are expected to demonstrate poorer health than those currently married or in a de facto relationship, with the greatest differences observed on the scales related to social functioning and mental health [ 21 ]. For all of the socio-economic measures, we expect to find a socio-economic gradient for all SF-36 scales, though the pattern may differ for different measures. For example, Sullivan and Karlsson found that educational level was more strongly associated with the measures of physical health whereas employment status was more strongly associated with mental health. Also based on results such as those reported by Sullivan and Karlsson, it was expected that reported experience of a long-term health condition, limited ability to work and satisfaction with health would be more strongly associated with SF-36 scales related to physical health, while job stress, social isolation (loneliness) and satisfaction with life would be more strongly associated with the scales loading on the mental health factor. Results and discussion Sample statistics Of the 13,055 respondents with identifiable SCQs, 53 percent were female. The age of respondents ranged from 15 to age 90 or greater (reported age was capped at 90 in the survey data), with an average age of 43 years (SD = 17.25). A total of 11,264 respondents (86.3%) completed all SF-36 items, with a further 1,326 (10.1%) having 5 or fewer missing items. Overall, the mean number of missing items per person was 0.69, and the median was 0. Details of non-responses/missing values for each item are presented in Figure A1 of the Attached file. The rate of missing values was below 3.5 percent for all items. The highest rate of missing values was evident for items from the Physical Functioning scale, followed by the Role Physical, General Health and Role Emotional scales. Scales were calculated using standard SF-36 scoring procedures, whereby missing values were replaced by scale means where valid responses were available for at least half of the scale items. Therefore, the number of respondents with valid scale scores ranged from 12,686 for the Role Emotional scale to 13,031 for the Social Functioning scale. Analyses were conducted with the maximum number of respondents possible and, therefore, varies across the scales. Tests of scaling assumptions Table 1 demonstrates that the range of item-scale correlations within each scale were moderate to strong (see item-internal consistency column). Indeed, all item-scale correlations were greater than the recommended correlation of 0.40 for adequate item-internal consistency [ 17 ]. The widest range of item-scale correlations was observed for the General Health scale (0.47 – 0.77), with the weakest item-scale correlations evident for items 11a and 11c (see Table A1 in Additional file ). Nonetheless, the item-scale correlations were reasonably similar within each scale. Table 1 Results of item scaling tests and reliability estimates for SF-36 scales. Range of item correlations No. of items Item-internal consistency 1 Item-discriminant validity 2 Scale Reliability 3 SF-36 scale Physical Functioning 10 0.55 – 0.81 0.14 – 0.55 0.93 Role Physical 4 0.76 – 0.81 0.24 – 0.60 0.90 Bodily Pain 2 0.80 0.32 – 0.67 0.87 General Health 5 0.47 – 0.77 0.22 – 0.57 0.82 Vitality 4 0.63 – 0.67 0.24 – 0.59 0.83 Social Functioning 2 0.71 0.42 – 0.60 0.83 Role Emotional 3 0.68 – 0.72 0.27 – 0.54 0.83 Mental Health 5 0.54 – 0.69 0.15 – 0.57 0.82 1 Correlations between items and scale corrected for overlap. 2 Correlations between items and other scales. 3 Internal-consistency reliability (cronbach's alpha). In order to assess item-discriminant validity, items that make up a scale were correlated with other scale constructs. Item-discriminant validity is demonstrated when an item correlates higher with its own scale than with other scales. While there was some overlap in correlations between items from one scale and other scales (compare the range of item correlations for the item-internal consistency and item-discriminant validity columns in Table 1 ), these were relatively minor. At the level of the individual items, it is apparent that all items were more strongly correlated with their own scale than with other scales, and that this difference was statistically significant for all but one item (see Table A.2 in Additional file ). Thus, Table A.3 demonstrates the achievement of 100% scaling success according to accepted SF-36 methods. Reliability Table 1 also shows that all SF-36 scales demonstrated acceptable internal consistency, with Cronbach's alpha ranging from 0.82 (Mental Health and General Health) to 0.93 (Physical Functioning). These reliability scores are similar to those reported in previous Australian research [ 2 , 4 ]. Descriptive statistics for scales In accordance with the standard scoring procedures [ 19 ], the eight scales of the SF-36 were constructed by aggregating 35 of the 36 individual items (the excluded item measures self-reported health transition). Each of the 35 items contributed to one scale, with each scale comprising 2 to 10 items. The range of scores possible on each of the eight scales was from 0 to 100, with 100 representing optimal functioning as measured by the SF-36. The (unweighted) means for the eight scales ranged from 60.87 (Vitality) to 83.19 (Physical Functioning; see Table 2 ). All scales were found to be negatively skewed with the Physical Functioning, Role Emotional, Role Physical and Social Functioning scales moderately skewed. Ceiling effects were high for the Role Emotional (73.3%) and Role Physical (68.8%) scales and moderate for the Social Functioning (49.4%), Physical Functioning (33.8%) and Bodily Pain (32.9%) scales. The greatest prevalence of floor effects was observed for the Role Physical (12.4%) and Role Emotional (9.9%) scales. Minimal ceiling and floor effects were observed for the Vitality, Mental Health and General Health scales. This pattern is similar to that obtained with previous assessment of the SF-36 [ 4 ]. Descriptive statistics for the two summary scores (the Physical and Mental Component Summary scores) are also consistent with previous findings. Table 2 Descriptive statistics for SF-36 scales. SF-36 scale 1 No. of items Mean Median Range SD Skewness Kurtosis Floor (%) Ceiling (%) PF 10 83.19 95.00 0–100 23.11 -1.70 2.10 0.5 33.8 RP 4 79.11 100.00 0–100 35.58 -1.42 0.40 12.4 68.8 BP 2 74.30 84.00 0–100 25.19 -0.81 -0.20 0.9 32.9 GH 5 70.02 72.00 0–100 21.14 -0.80 0.16 0.2 6.1 VT 4 60.87 65.00 0–100 19.84 -0.59 -0.01 0.4 1.1 SF 2 81.96 87.50 0–100 23.55 -1.29 0.86 0.6 49.4 RE 3 82.13 100.00 0–100 33.08 -1.64 1.16 9.9 73.3 MH 5 73.92 76.00 0–100 17.42 -0.98 0.88 0.1 3.3 PCS 21 49.17 52.55 3.61–71.87 10.33 -1.19 0.86 0.0 0.0 MCS 14 49.92 52.74 4.45–73.94 9.95 -1.18 1.24 0.0 0.0 1 PF = Physical Functioning; RP = Role Physical; BP = Bodily Pain; GH = General Health; VT = Vitality; SF = Social Functioning; RE = Role Emotional; MH = Mental Health; PCS = Physical Component Summary; MCS = Mental Component Summary. Principal Components Analysis Principal Components Analysis was conducted to examine the underlying structure in the SF-36. The analysis supported the two factor solution. Two factors had eigenvalues greater than one. The two factor solution was also consistent with the pattern of results evident in the scree plot. The two factors accounted for 69% of the total variance in the 8 SF-36 scales. The first factor accounted for 56% of the variance and the second factor accounted for an additional 13%. The total variance accounted for by the two-factor solution was similar to that found in other studies. For example, in a ten country comparison of the factor structure of the SF-36, Ware and colleagues report that the two factor solution accounted for between 66 and 72 percent of total variance [ 22 ]. Previous Australian analyses also found that the two factors explained around 70 percent of total variance [ 4 , 6 ]. Table 3 shows the correlations between the SF-36 scales and the rotated components. We used the varimax method to obtain orthogonal factors. Across the eight scales, the percentage of variance within each scale explained by the two-factor solution (commonalities) ranged from 0.57 to 0.82. Again, this is consistent with previous international studies [ 22 ]. It is apparent from Table 3 that our two-factor solution is also consistent with the previously reported factor structure, identifying scales that relate to physical and mental health. The scales that correlated strongly with the first factor were Mental Health (0.90), Vitality (0.74), Role Emotional (0.71) and Social Functioning (0.71). The Physical Functioning (0.11), Role Physical (0.28) and Bodily Pain (0.31) scales correlated weakly with this factor. This factor was subsequently labelled mental health. With regard to the second factor, the Physical Functioning (0.84), Role Physical (0.81) and Bodily Pain (0.77) scales correlated strongly, while the Mental Health (0.08) and Role Emotional (0.25) scales correlated weakly. This factor was labelled physical health. Table 3 Rotated principal components associations between the eight SF-36 scales and the mental and physical health components. Commonalities (h 2 ) are also presented. Rotated principal components SF-36 scale Mental Physical h 2 Physical Functioning .106 .842 0.72 Role Physical .278 .806 0.73 Bodily Pain .307 .771 0.69 General Health .475 .627 0.62 Vitality .744 .361 0.68 Social Functioning .709 .465 0.72 Role Emotional .714 .246 0.57 Mental Health .900 .079 0.82 As with most international comparisons (though see [ 23 , 24 ] using Taiwanese and Japanese populations), the factor loadings associated with General Health were stronger for the physical health factor. We also found this pattern of results, though the relationship observed between the General Health scale and the mental health factor was at the upper end of the range of international data. The opposite pattern is generally demonstrated for the Vitality scale, with it loading most strongly on the mental health factor. One previous Australian study [ 4 ] found that the Vitality scale loaded more strongly on the physical health factor. The current results are more consistent with the body of international evidence. Normative comparisons Table 4 presents population estimates of the mean and standard deviation for each of the SF-36 scales and the Physical and Mental Component Summary scores. These results take account of the complex survey design and population weights. Also presented are the corresponding results from the ABS 1995 Health Survey. Table 4 Population estimates of mean SF-36 Scale scores and Component Summary Scores (with N and standard deviation) from the HILDA Survey. Means (and sd) from ABS (1995) are also presented, together with estimate of independent t test to assess differences between estimates. HILDA Population (weighted) Australian population norms 1 Mean (sd) N Mean (sd) N t test Physical Component Summary Score 2 49.23 (10.6) 12 320 Mental Component Summary Score 2 49.79 (10.3) 12 320 Scales Physical Functioning 82.49 (24.0) 12 760 82.6 (23.9) 18 734 0.40 Role Physical 79.06 (36.2) 12 714 79.9 (35.1) 18 710 2.06 Bodily Pain 73.89 (25.6) 12 977 76.8 (25.0) 18 699 10.09 General Health 69.54 (21.6) 12 718 71.6 (20.3) 18 715 8.60 Vitality 60.72 (20.0) 12 938 64.5 (19.8) 18 728 16.64 Social Functioning 81.14 (24.0) 13 031 85.0 (22.5) 18 789 14.64 Role Emotional 81.83 (33.8) 12 686 82.9 (32.3) 18 620 2.82 Mental Health 73.42 (17.7) 12 930 75.9 (17.0) 18 676 12.55 1 From ABS, 1995 2 Physical and mental summary scores were calculated differently in the ABS publication (using factor scores from principle component analysis and stardardised to have a mean of 50 and sd of 10). Therefore, it is not appropriate to directly compare with the HILDA data derived using the standard scoring procedures. The results obtained are broadly consistent with the previous Australian national data. The means are of a similar magnitude. Direct comparison of the means, assessed using independent t tests (see [ 6 ]), indicated that the differences observed were statistically significant (at the .01 level) for six of the eight scales. The mean scores from the HILDA Survey were consistently lower than those obtained in the ABS Health Survey. However, the greatest differences observed on the individual scales (VT and SF) were less than four points and Ware et al. [ 19 ] suggest that a difference of 5 points or more indicates clinical or social meaningfulness. While relatively small, the significant differences in mean SF-36 scale scores from these two national surveys are intriguing. The most marked differences were on scales related to mental health, while the scales most strongly related to physical health (the Physical Functioning and Role Physical scales) were those where no difference between surveys was observed. It may be that the differences observed reflect different methodological approaches adopted in the two surveys, or could be due to non-sampling error. Alternatively, it could reflect a real change in the Australian population between 1995 and 2001, with poorer reported mental health occurring in more recent times. This may be the case, as changes were restricted to the measures related to mental health and therefore are not simply a general change in response bias. There is evidence which corroborates the apparent increased prevalence of mental health problems in the community. Comparison of the 1997 Australian Survey of Mental Health and Wellbeing and the 2001 National Health Survey, both of which included the 10 item Kessler Psychological Distress Scale, showed the rates of substantial psychological distress had increased by over one percent [ 25 ]. In the UK, comparison of the results from the 1993 and 2000 Psychiatric Morbidity Surveys found that, while there was no difference in the overall rate of neurotic disorders amongst adults, there was a significant increase in the prevalence among men [ 26 ]. Further investigation of this finding and possible explanations (e.g. increasing mental disorders, reduced stigma and increased likelihood of disclosure) is warranted. Comparison of groups The data presented in Table 5 show the profile of mean SF-36 scores across a range of demographic and socio-economic characteristics. There were small but significant sex differences across most of the SF-36 scales and summary measures. Females consistently demonstrated slightly poorer health than males on all measures apart from the General Health scale, including the physical and mental summary scores. This pattern of results precisely replicates the pattern found in the previous ABS National Health Survey [ 5 ]. Table 5 Mean SF-36 Scale and Component Summary Scores, by measures of demographic and socio-economic circumstances. The F statistic associated with each comparison is indicated, together with minimum number of respondents in each cell (in brackets). PCS MCS PF RP BP GH VT SF RE MH Gender Male (5817) 49.49 50.45 83.84 80.23 74.96 69.51 62.63 82.21 83.46 74.61 Female (6503) 48.97 49.15 81.18 77.90 72.84 69.57 58.85 80.09 80.22 72.25 F(1,475) 8.19 53.82 44.89 14.56 24.03 0.03 121.39 27.94 30.55 62.19 * * * * ns * * * * Age 18–25 (2135) 53.23 48.26 91.75 88.48 80.26 73.38 62.53 82.65 83.09 71.64 26–39 (3535) 52.16 48.88 89.17 86.40 78.71 73.85 60.97 82.50 83.91 72.80 40–55 (3676) 49.56 49.99 84.28 80.82 73.70 69.49 60.86 81.71 83.35 73.74 56–65 (1452) 44.47 51.51 72.87 69.27 66.23 63.72 60.87 79.87 80.42 74.44 66–75 (992) 41.70 52.92 65.10 61.29 64.59 62.26 59.47 79.61 78.90 76.14 75+ (530) 35.85 51.31 49.49 39.71 57.64 55.57 53.21 70.22 62.76 75.01 F(5,451) 324.72 29.94 302.04 210.84 96.79 85.05 15.69 17.57 25.92 8.15 * * * * * * * * * * Marital status Married/de facto (7928) 48.85 50.73 82.36 79.15 73.68 69.62 61.08 82.97 83.74 74.93 Divorced/separated (1013) 47.01 47.50 77.23 71.74 67.25 65.87 57.02 72.82 73.94 69.29 F(1.475) 18.74 59.43 29.33 28.06 31.20 19.39 25.41 94.83 52.03 60.11 * * * * * * * * * * Educational attainment Yr 12 or above (7975) 50.29 49.92 85.59 81.92 75.86 71.22 61.46 82.40 83.43 74.17 Yr 11 or less (4041) 47.15 49.56 76.64 73.48 70.22 66.32 59.31 78.82 78.74 72.00 F(1.475) 159.19 2.48 235.31 112.92 95.55 102.68 24.48 42.39 40.34 28.45 * ns * * * * * * * * Employment status Employed (7741) 51.77 50.25 89.04 87.14 78.54 73.40 62.57 85.13 86.53 74.87 Unemployed (520) 51.22 47.08 85.07 82.37 76.28 70.67 62.85 77.57 76.39 68.72 F(1.471) 1.58 36.90 17.16 10.00 3.69 6.47 0.10 46.04 36.74 44.94 ns * * ** ns * ns * * * Housing status Own/mortgage (9281) 49.12 50.43 82.58 78.86 74.20 69.99 61.16 82.54 83.10 74.48 Rent (3039) 49.53 47.97 82.24 79.63 73.01 68.26 59.46 77.15 78.18 70.38 F(1.475) 1.77 92.54 0.20 0.75 2.79 6.35 11.43 72.65 36.32 81.86 ns * ns ns ns * * * * * Receipt of welfare payments (working age respondents) Non-recipients (8373) 51.75 50.16 88.93 86.86 78.45 73.25 62.73 84.87 86.09 74.66 Welfare recipients (2091) 46.57 46.16 76.22 68.35 66.22 62.71 55.48 70.35 71.10 66.50 F(1,473) 259.15 174.85 305.32 300.00 223.36 274.19 147.06 431.71 241.05 221.56 * * * * * * * * * * Degrees of freedom for F statistics based on number of clusters (primary sampling units) and number of strata * p < .05, p < .01, * p < .001 Age effects were also consistent with expectations. While those in older age groups reported poorer levels of health across most SF-36 scales, the decline was most pronounced in the measures related to physical health (Physical Health, Bodily Pain). In contrast to this pattern, however, both the Mental Health scale and Mental Component Summary Score showed improved mental health with increasing age, apart from a slight decline evident in the oldest (75+) age group. This is consistent with epidemiological survey results using a variety of mental health measures [ 27 ]. Health differences associated with current marital status (comparing partnered respondents with those who were separated or divorced) were evident across all scales. The greatest differences were observed in the Social Functioning and Role Emotional scales. The analysis showed health was associated with a range of measures of social status, however the profile across the SF-36 scales differed for the different measures. Consideration of educational attainment showed that those respondents who had less than secondary education had significantly poorer physical health, most evident on the Physical Functioning scale. Current unemployment (compared to full or part-time employment) was significantly associated with poorer health on a number of scales, but particularly the measures loading on the mental health factor (Social Functioning, Role Emotional and Mental Health). This too, is consistent with the extensive unemployment literature [ 28 ]. Housing tenure (rental housing vs other) was not related to the SF-36 physical health scales. However, reliance on rental housing was associated with poorer mental health (lower means on Vitality, Social Functioning, Role Emotional, and Mental Health). While Australian welfare payments are universally available, eligibility is highly targeted. Therefore, welfare receipt appears to be a good proxy for poor financial circumstances. In addition, a subset of the welfare population are those with severe disabilities that prevent work. Consistent with this, we found that, amongst working age respondents, receipt of an Australian income support payment was strongly associated with both poorer physical and mental health. The final set of comparisons examined the association between SF-36 scales and respondents' health and social circumstances as measured by other scales and items included in the HILDA Survey (Table 6 ). Consistent with expectations, respondents who reported a long-term health condition or disability demonstrated lower mean scores on all SF-36 scales, but particularly so on the scales loading on physical health (Physical Functioning, Role Physical, Bodily Pain and General Health). A similar pattern of associations was observed when examining the circumstances of those with health conditions that impacted on their ability to work. The greater influence on work ability was associated with poorest physical health as measured by SF-36 scales. Reported satisfaction with health was most strongly associated with General Health and physical health measures more generally, while overall satisfaction with one's life was more highly associated with the SF-36 scales related to mental health. Table 6 Mean SF-36 Scale and Component Summary Scores, by other measures of health and social situation. The F statistic associated with each comparison is indicated, together with minimum number of respondents in each cell (in brackets). PCS MCS PF RP BP GH VT SF RE MH Presence of long term health condition or disability No (9549) 52.16 50.58 88.57 88.09 80.14 74.85 64.19 85.87 86.55 75.34 Yes (2771) 38.89 47.02 61.99 47.91 52.96 51.40 49.10 65.35 65.48 66.94 F(1,475) 1797.99 163.21 1324.59 1367.80 1621.36 1777.15 892.23 858.84 455.68 340.62 * * * * * * * * * * If so, does condition limit ability to work? No (761) 47.69 49.20 80.13 75.08 70.20 65.01 58.66 79.67 78.80 72.31 Yes (1969) 35.76 46.23 56.05 38.30 46.92 46.57 45.72 60.36 60.98 65.08 Can do nothing (41) 28.00 45.10 31.05 22.33 37.13 37.61 41.20 51.72 44.01 61.01 F(2,460) 475.53 15.92 361.10 275.57 272.20 235.87 110.07 159.16 65.12 35.12 * * * * * * * * * * Satisfaction with health Satisfied/mixed (11182) 50.22 50.40 84.65 82.24 76.19 71.88 62.43 83.26 84.14 74.65 Not satisfied (1018) 39.71 43.95 62.67 48.61 52.82 47.63 45.23 61.96 58.97 62.24 F(1,475) 501.04 204.52 391.67 477.58 516.14 672.21 503.83 435.33 250.56 270.29 * * * * * * * * * * Satisfaction with life Satisfied/mixed (11937) 49.40 50.22 82.94 79.90 74.51 70.23 61.42 82.07 82.87 74.21 Dissatisfied (341) 43.99 37.17 69.34 53.93 56.02 49.22 40.10 53.42 51.37 50.20 F(1,475) 48.84 265.85 64.26 84.88 98.41 196.47 275.57 257.92 125.10 364.73 * * * * * * * * * * Report that job stress makes physically ill Not agree, neither (6533) 51.82 51.12 89.11 87.80 79.22 74.59 64.24 86.25 88.23 76.08 Agree (1255) 49.89 44.72 84.90 76.61 71.08 65.19 53.41 74.52 70.55 66.04 F(1,472) 40.41 290.07 39.02 98.10 89.70 174.22 267.47 208.13 200.02 269.57 * * * * * * * * * * Often Lonely Disagree/mixed (9506) 49.74 51.33 84.16 81.58 75.93 71.65 63.04 84.27 85.63 76.17 Agree (2454) 47.45 43.87 77.37 69.90 66.63 61.96 52.26 70.07 67.19 63.14 F(1,475) 63.52 722.01 100.92 147.99 196.06 315.39 486.49 514.90 424.96 763.49 * * * * * * * * * * Degrees of freedom for F statistics based on number of clusters (primary sampling units) and number of strata * p < .05, p < .01, * p < .001 The final two criterion measures (job stress and loneliness) were expected to be most strongly associated with poorer mental health outcomes. Consistent with expectations, those respondents in work who reported that their experience of job stress made them physically ill reported poorest health on the Vitality and Mental Health scales, while those who reported often feeling lonely had low scores on the Mental Health, Social Functioning and Vitality scales. Conclusions This analysis has demonstrated the validity of the SF-36 data collected through the first wave of the HILDA Survey. The eight scales were shown to be psychometrically sound, with good internal consistency, discriminant validity and high reliability. The results supported the underlying two-factor structure, with factors related to physical and mental health. The pattern of factor loadings, variance explained, and the profile and magnitude of the scale means were consistent with previous Australian and international results. There was, however, an indication that the estimated population means on several of the SF-36 scales, particularly on scales loading on the mental health factor, were lower in the HILDA Survey than previous Australian national data collected through the National Health Survey. Self-report measures of health status can be subject to a range of biases, such as reporting bias or cohort effects. They may also be influenced by contextual factors, such as the setting in which the measure is conducted. Further consideration of these different data sources, and an examination of possible methodological and other differences, is warranted. With respect to the relationship between SF-36 scales and external criteria, we demonstrated a pattern of results consistent with expectations and previous research. We examined measures having differential effects on physical and mental health (e.g., disability and health satisfaction vs job stress, loneliness, and life satisfaction). The relationship between demographic characteristics (age, sex, marital status) and SF-36 scales was consistent with previous data. Finally, we focused on health inequalities, using a range of markers of social status and examining their relationship with the SF-36 scales. While each marker of social status was associated with poorer health, the dimensions of health related to each social variable differed. Whereas educational attainment was most strongly associated with poorer physical health, housing tenure and employment status were related to mental health scales. The measure of welfare receipt, which is associated with economic hardship, was strongly associated with both poorer physical and mental health. The application of the probability weights available with the HILDA Survey data ensures the accuracy of the point-estimates of the population parameters, such as the mean, by correcting for non-response and design characteristics. However, it is also important to recognise the impact of the survey design (clustering and stratification) on the calculation of standard errors. As the sampling units in the HILDA Survey are Census Collection Districts and, within that, households, the data contradict assumptions of independence and should not be treated as a simple random sample drawn from the population. We expect more similarity between individuals within Census Collection Districts and households than between those drawn at random from the population and if these characteristics of the survey design are not taken into account the estimates of standard error will overstate accuracy. Calculation of the design effect provides an indication of the increase in error associated with the complex survey design, representing the ratio of the variance for the complex design to that obtained assuming a simple random sample. These measures range from 1.4 (for Role Emotional scale) to 3.1 (for the Physical Functioning scale). Thus, whereas the 95 percent confidence interval around the population estimate for Physical Functioning Scale (84.49) would have been 82.04 to 82.94 had we not taken the survey design into account, our estimate was 81.77 to 83.21. The design effects were greater for the scales assessing physical health than for those loading on mental health. This suggests physical health more strongly clusters within households and areas than does mental health, perhaps as a consequence of the clustering of age (design effect of 3.8) and the strong relationship between age and physical health and disability. Note that the design effects are not evident in the population estimates in Table 4 as these figures were corrected to provide an estimate of the population mean and population standard deviation. Nonetheless, the design effects highlight the need to utilise appropriate statistical techniques in analysis of the HILDA Survey data. This analysis has provided evidence that the SF-36 data collected in the HILDA Survey is valid and supports its use as a general outcome measure of physical and mental health status. It also suggests that the results obtained using the SF-36 in the HILDA Survey can be interpreted by reference to published SF-36 normative data and comparison with previous research findings. Most critically, from our perspective, the analysis provides initial support for the proposition that the SF-36 scales and component summary scores are related in a meaningful and interpretable manner to measures of social status. This supports the use of the HILDA Survey data in our analysis of health inequalities and the effects of social exclusion in Australia. Competing interests The authors declare that they have no competing interests. Authors' contributions PB and TC participated in all aspects of the preparation of this manuscript, including data analysis, preparation of various sections of the manuscript, and commenting on drafts. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 Additional file SF36 HILDA validation.pdf – provides tables and figures used to determine item-discriminant validity and to explore the pattern of missing data across SF-36 items. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524495.xml
546210	Immunolocalization of KATP channel subunits in mouse and rat cardiac myocytes and the coronary vasculature	Background Electrophysiological data suggest that cardiac K ATP channels consist of Kir6.2 and SUR2A subunits, but the distribution of these (and other K ATP channel subunits) is poorly defined. We examined the localization of each of the K ATP channel subunits in the mouse and rat heart. Results Immunohistochemistry of cardiac cryosections demonstrate Kir6.1 protein to be expressed in ventricular myocytes, as well as in the smooth muscle and endothelial cells of coronary resistance vessels. Endothelial capillaries also stained positive for Kir6.1 protein. Kir6.2 protein expression was found predominantly in ventricular myocytes and also in endothelial cells, but not in smooth muscle cells. SUR1 subunits are strongly expressed at the sarcolemmal surface of ventricular myocytes (but not in the coronary vasculature), whereas SUR2 protein was found to be localized predominantly in cardiac myocytes and coronary vessels (mostly in smaller vessels). Immunocytochemistry of isolated ventricular myocytes shows co-localization of Kir6.2 and SUR2 proteins in a striated sarcomeric pattern, suggesting t-tubular expression of these proteins. Both Kir6.1 and SUR1 subunits were found to express strongly at the sarcolemma. The role(s) of these subunits in cardiomyocytes remain to be defined and may require a reassessment of the molecular nature of ventricular K ATP channels. Conclusions Collectively, our data demonstrate unique cellular and subcellular K ATP channel subunit expression patterns in the heart. These results suggest distinct roles for K ATP channel subunits in diverse cardiac structures.	Background ATP-sensitive (K ATP ) channels are widely expressed in both excitable and non-excitable tissue types throughout the body. However, differences exist in the functional and pharmacological properties of various K ATP channels in different tissues. This functional diversity of K ATP channels is also reflected in the cardiovascular system. K ATP channels are abundantly expressed in ventricular myocytes, where they are probably best characterized. These channels have a high unitary conductance, are inhibited by ATP in the micromolar range, are blocked by glibenclamide (but not tolbutamide) and opened by pinacidil (and not by diazoxide). K ATP channels also exist in the coronary vasculature, where they function to maintain basal coronary blood flow [ 1 ]. K ATP channels in the coronary smooth muscle have a low unitary conductance (~30 pS) and are blocked by glibenclamide and activated by K ATP channel openers and adenosine [ 2 ]. K ATP channels exist in the coronary endothelium [ 3 ], but their biophysical properties remain largely unidentified. In addition to this diverse distribution of plasmalemmal K ATP channels in the heart, K ATP channels with unique biophysical and pharmacological profiles are also believed to be expressed in the mitochondrial inner membrane [ 4 ]. K ATP channels are increasingly well characterized at the molecular level. In order to express a functional channel that resembles native K ATP channels in terms of their biophysical and pharmacological properties, a combination of two types of subunits is necessary. It is now understood that Kir6 subunits form a pore-forming structure through which K + ions transverse the membrane whereas SUR subunits assemble with the latter to modulate the channel's function and to confer unique pharmacological properties to the channel complex [ 5 , 6 ]. Two genes each code for the two known Kir6 subfamily members (Kir6.1 and Kir6.2) and for the two known SUR members (SUR1 and SUR2). Alternative splicing of SUR2 gives rise to at least two functionally relevant isoforms (SUR2A and SUR2B) with distinct pharmacological profiles [ 5 ]. It is widely believed that ventricular K ATP channels consist of the specific combination of Kir6.2 and SUR2A subunits and that K ATP channels in vascular smooth muscle consist of Kir6.1 and SUR2B subunits. This view is consistent with results from gene targeting experiments, which demonstrate the absence of functional sarcolemmal K ATP channels in ventricular myocytes from Kir6.2(-/-) mice and the coronary abnormalities that develop in Kir6.1 and SUR2 null mice [ 5 ]. Although they are powerful tools, gene knockout approaches can overemphasize certain important aspects of gene function and may overlook more subtle effects of protein function and interaction. At first sight, these models do not adequately explain the reports of SUR1 mRNA expression in the heart [ 7 ], or the observation that anti-SUR1 antisense oligonucleotides inhibit K ATP channels of ventricular myocytes [ 8 ]. They also do not provide a functional basis for the known expression of Kir6.1 mRNA and protein in cardiac myocytes [ 9 - 12 ]. or explain the molecular composition of the endothelial K ATP channel. The specific cellular and subcellular localization of proteins can be used to predict their function. We therefore used antibodies specific for each of the K ATP channel subunits to determine their cellular and subcellular localization in the mouse and rat heart. Our results suggest distinct roles for each of the K ATP channel subunits in diverse cardiac structures. Results Given the reports of expression of each of the K ATP channel subunits in the heart (see earlier), we performed immunohistochemistry and immunocytochemistry to determine the cellular and subcellular localization of Kir6.1, Kir6.2, SUR1 and SUR2 subunits in mouse and rat ventricle. To this end, we stained frozen sections of cardiac ventricular tissue as well as cardiac myocytes enzymatically isolated from mouse and rat hearts. Where possible, we used different antibodies to the subunits to ensure that the staining pattern observed was specific. Characterization of the antibodies used in this study We performed Western blotting to determine the specificity of the antibodies used in this study. Three different anti-Kir6.1 antibodies (NAF-1, CAF-1 and C-16) all detect a band that migrates with an apparent molecular size of 44 kDa in Western blotting of rat heart membrane fractions (Fig 1 ). A 44 kDa band was also detected by the 78A antibody (not shown). These antibodies did not cross-react with Kir6.2, since they detected only Kir6.1 (and not Kir6.2) in parallel experiments on cells transfected with various K ATP channel subunit combinations (data not shown, see also reference [ 13 ]). Figure 1 Characterization of the antibodies used in this study. Rat heart membrane fractions were separated on 8–12% denaturing PAGE and Western blotting was performed using Kir6.1 (NAF1, CAF-1 or C-16) or Kir6.2 (G-16 or 76A) antibodies. In other experiments, lysates of HEK293 cells stably transfected with Kir6.1/SUR2B, or lysates of COS7L cells transiently transfected with SUR1, SUR2A or SUR2B cDNAs, were separated with PAGE and respectively immunoblotted with anti-SUR2 (C-15) or anti-SUR1 (C-16) antibodies. Molecular size markers are indicted as appropriate. Both the 76A and G-16 anti-Kir6.2 antibodies have previously been characterized and we demonstrated that they specifically detect a ~38 kDa band in Western blotting of Kir6.2 transfected cells and do not detect heterologously expressed Kir6.1 protein [ 13 , 14 ]. Here we show that both of these antibodies also detect Kir6.2 subunits as a ~38 kDa band in Western blotting of heterologously expressed Kir6.2 protein or rat heart membrane fractions (Fig 1 ). The anti-SUR1 antibodies specifically detect SUR1 protein (170 kDa) in cell lysates of COS7L cells transiently transfected with SUR1/Kir6.2 cDNA as well as in membrane fractions obtained from mouse hearts [ 15 ]. In the cell lystates from SUR2B/Kir6.2 transfected cells, the SUR2 antibody recognizes a specific band at 150 kD in transfected cells only (Fig 1 ) and did not detect SUR1 (not shown). Thus, each of the antibodies used in this study detected proteins at the correct molecular size in Western blotting and did not cross-react non-specifically with other proteins. Kir6.1 localization in the murine heart We used an immunohistochemistry approach to identify the localization of Kir6.1 protein in cryostat sections of mouse ventricles. We used three separate antibodies that produced similar results (NAF-1, CAF-1 and C-16). We also used another anti-Kir6.1 antibody (78A) but this antibody did not perform well in these assays. A typical result is shown in Fig 2A where Kir6.1 protein was ubiquitously detected throughout the ventricle. Closer inspection shows that Kir6.1 protein is expressed in a sarcomeric striated pattern in ventricular myocytes. This effect is more pronounced in epicardial myocytes (left hand side of panel A and Fig 2B ). In the midmyocardium, a punctate staining pattern is apparent (Fig 2C ). This was observed in over 50 different cryosections that we examined (also observed with the CAF-1 antibody). Figure 2 Immunohistochemistry demonstrating the regional distribution of Kir6.1 subunits in cryostat sections of the mouse ventricle. A : The low magnification image demonstrates regional expression differences with more prominent myocyte staining apparent in the epicardium (left) than in the mid-myocardium (right). Image width is 750 μm. B and C : Higher magnification of the same slide shown in previous panel showing an epicardial (B) and midmyocardial section (C). The image widths are respectively 102 and 96 μm. D : Double stained section of a midmyocardial section using NAF-1 anti-Kir6.1 antibodies (D1) and ICAM-2 antibodies as a marker of endothelial cells. (D2). The image width is 119 μm. The secondary antibodies used were Cy-3 conjugated donkey anti-rabbit and Cy-5 conjugated donkey anti-rat F(ab')2 IgG fragments. The cylindrical shape of the punctate structures in Fig 2C is reminiscent of coronary blood vessels. We tested various antibodies to find suitable markers for smooth muscle- and endothelial cells in the coronary vasculature. We found antibodies against smooth muscle α-actin to be a good marker for coronary vessels (Fig 3A ) typically having diameters upwards of about 10 μm. We also tested an antibody against the apical endothelial protein ICAM-2 and found the antibody to detect endothelial cells lining the inner layer of coronary arteries (Fig 3B and 3C ). Additionally, the anti-ICAM-2 antibodies stained a large number of smaller vessel-like structures having diameters typically smaller than 15 μm (average around 6–8 μm). These smaller vessels apparently do not have an appreciable amount of vascular smooth muscle cells, as judged by the lack of smooth muscle α-actin staining, suggesting that they may be coronary capillaries (the possibility that some of them may be pre-capillary arterioles with low amounts of smooth muscle cannot be excluded). To examine the localization of Kir6.1 in coronary blood vessels, we co-stained mouse ventricle with NAF-1 anti-Kir6.1 antibodies and anti-ICAM-2 antibodies (a marker for the vascular endothelium). As shown in Figure 2D , there is clear correspondence between Kir6.1 staining and ICAM-2 localization, suggesting high expression levels of Kir6.1 in the coronary vasculature. Figure 3 Immunohistochemistry markers to distinguish between vascular smooth muscle and endothelial cells. A cryosection of mouse heart ventricle was simultaneously stained with FITC-conjugated smooth muscle α-actin antibody (A) and an antibody against the apical endothelial membrane protein, ICAM-2 (B). The secondary antibodies to detect ICAM-2 was Cy-5 conjugated donkey anti-rat (pseudo-colored red for clarity). Panel C is an overlay of the preceding two panels. The image width is 238 μm. Given the high sensitivity of the NAF-1 antibody, we were able to perform a sub-cellular localization study of Kir6.1 protein in the mouse heart (Fig 4 ; this approach was not possible with other antibodies, which produced less sensitive staining). We co-stained a cryostat section with antibodies against smooth muscle α-actin (A1), Kir6.1 (A2) and the endothelial ICAM-2 protein (A3). An overlay of these three images is shown in panel B. A higher magnification of the area roughly represented by the boxed area (panel C) demonstrates that Kir6.1 subunits are ubiquitously expressed and are present in ventricular myocytes, the coronary smooth muscle walls as well as in endothelial cells. Of these cells types, the highest expression levels appear to occur in the vasculature. Figure 4 Triple stain immunohistochemistry of mouse ventricle demonstrating the distribution of Kir6.1 subunits in the coronary vasculature. The cryostat section was stained with antibodies against smooth muscle α-actin ( A1 ), Kir6.1 ( A2 ) and endothelial-specific ICAM-2 ( A3 ). B :Overlay of the three panels shown in panels A. C : Higher magnification of the boxed area highlighted in panel B. Kir6.2 localization in the murine heart We used several antibodies against Kir6.2 subunits to determine their cellular localization. Both the 76A and the G-16 antibodies (Santa Cruz) gave similar results. The W62 antibody developed by us [ 13 ] did not seem to stain more than background and we therefore assumed this antibody not to work in this assay. In terms of staining in the vasculature, we detected Kir6.2 subunit expression mainly in the endothelium (arrows in Fig 5A ) and not in smooth muscle (as judged by a lack of co-localization with smooth muscle α-actin). Evidently, Kir6.2 subunits were expressed in cardiac myocytes as well, as demonstrated in Fig 5B . The staining occurs in a sarcomeric striated pattern in ventricular myocytes (distance interval of roughly 2.2 μm). Although not as apparent as with Kir6.1, there appears to be expression of Kir6.2 subunits in small coronary blood vessels (10 μm or less). These structures are also stained by Kir6.1 antibodies (arrows in Fig 5B and 5C ), suggesting co-localization of Kir6.1 and Kir6.2 subunits in small coronary blood vessels. Figure 5 Immunohistochemistry demonstrating the cellular distribution of the Kir6.2, SUR1 and SUR2 subunits in cryostat sections of the mouse ventricle. A : Co-staining of mouse ventricular section with smooth-muscle α-actin (green) and anti-Kir6.2 antibodies (Santa Cruz). The latter antibody was detected with Cy-5 conjugated donkey anti-goat IgG (blue). B and C : Double-staining with anti-Kir6.2 antibodies (Santa Cruz) and CAF-1 anti-Kir6.1 antibodies (detected with Cy-3 conjugated donkey anti-chicken IgY (red). The image width of panel A is 109 μm and for panels B and C it is 89 μm. D : Co-staining of a midmyocardial section with anti-SUR1 antibodies (Cy3-conjugated donkey anti-goat secondary antibodies were used; shown in red) and FITC-conjugated smooth muscle anti-actin antibodies (shown in green). E and F : Cryosections stained with a pan anti-SUR2 antibody (secondary antibody was Cy-3 conjugated anti goat IgG. The image widths for panels D-F respectively are 238, 238 and 88 μm. Localization of SUR subunits in murine heart Using anti-SUR1 antibodies, we observed staining predominantly in cardiac ventricular myocytes (Fig 5D ). The lack of staining of either large or small coronaries suggests that SUR1 subunits are not expressed in the coronary vasculature of the mouse heart (note the lack of co-localization of SUR1 with smooth muscle α-actin). In contrast, pan-SUR2 antibodies stain both ventricular myocytes in a regular striated sarcomeric pattern as well as small coronary blood vessels (Fig 5E and 5F ). Note the lack of high expression levels in larger coronary arteries (round structure in Fig 5F ), demonstrating that expression of SUR2 subunits occurs predominantly in small coronary vessels (less than 10 μm). In separate experiments, we did not detect a particularly strong co-localization with smooth muscle α-actin (not shown), ruling against the possibility that SUR2 is strongly expressed in the smooth muscle cells of larger coronary vessels. Subcellular localization of K ATP channel subunits in enzymatically isolated ventricular myocytes Fixation and permeabilization procedures can affect the outcome of immunocytochemistry experiments. We therefore compared three different methods. Isolated cardiac myocytes were (a) fixed with paraformaldehyde and permeabilized with Triton X-100, (b) fixed and permeablized in a single step by ice-cold methanol or (c) subjected to a two-step protocol with paraformaldehyde fixation followed by methanol fixation/permeabilization. Typical results obtained with the anti-Kir6.2 (76A) antibody are shown in Fig 6 . In general, we found methanol fixation to preserve membrane staining but to cause a loss of intracellular fluorescence (Fig 6A ). Paraformaldehyde fixation, in contrast, better preserved staining of intracellular structures (Fig 6B ). The two-step fixation protocol in general gave similar results to paraformaldehyde fixation, but the fluorescence intensity was generally much higher and more clearly defined (Fig 6C ). This may be due to better preservation of protein antigenicity and/or improved permeabilization (and hence improved antibody accessibility). Our data are consistent with a recent report describing the two-step protocol to better preserve the in-vivo subcellular localization of proteins [ 16 ]. We consequently used the two-step protocol for all subsequent experiments to examine the subcellular localization of K ATP channel subunits. Figure 6 Effect of different fixation protocols on immunocytochemistry of ventricular myocytes. Enzymatically isolated ventricular myocytes were subjected to fixation protocols with (A) methanol, (B) paraformaldehyde or (C) sequential double fixation with paraformaldehyde followed by methanol. The primary antibody in all cases was against Kir6.2 subunits (76A) and the secondary antibody was Cy-3 conjugated donkey anti-rabbit IgG. The image widths are respectively 137, 107 and 139 μm. The subcellular distribution of Kir6.1 and Kir6.2 subunits in isolated mouse ventricular myocytes is shown in Fig 7A . Kir6.2 subunits are expressed in a regular striated pattern throughout the myocyte (A2). Kir6.1 subunits showed a similar expression pattern (A1), with the exceptions that (a) Kir6.1 expression appears to be more punctate and (b) staining is more prominent at the myocardial surface. Since we used antibodies developed in different species (chicken anti-Kir6.1 antibody and rabbit anti-Kir6.2 antibody) we were able to detect both proteins in the same myocyte with little cross-reactivity of the secondary antibodies. Although there is some degree of overlap in the subcellular expression of Kir6.1 and Kir6.2 subunits (yellow in A3), it is apparent that there are areas within the cell that express Kir6.2 but not Kir6.1 subunits. Thus, the possibility exists that these subunits may have different functional roles within the ventricular myocyte. Similarly, SUR1 and SUR2 subunits also have distinct subcellular localizations. We consistently observed strong surface staining with SUR1 antibodies (Fig 7B ) whereas SUR2 antibodies diffusely stain throughout the width of the cell in a sarcomeric repeating pattern (Fig 7C ). To address the question whether there is co-localization of any of the K ATP channel subunits with mitochondria, we used MitoTracker Red as a mitochondrial marker. We found MitoTracker staining to dissipate during immunocytochemistry protocols and co-labeling with K ATP channel antibodies was therefore not an effective strategy. Nevertheless, the pattern of MitoTracker staining that we observed shortly after fixation (Fig 7D ) was inconsistent with predominant subcellular distribution of any of the K ATP channel subunits and we conclude therefore that K ATP channel subunits are not abundantly expressed in cardiac mitochondria. Figure 7 Immunocytochemistry of isolated mouse ventricular myocytes demonstrating the subcellular localization of Kir6.1, Kir6.2, SUR1 and SUR2 subunits. A : Double staining of a ventricular myocyte with the CAF-1 anti-Kir6.1 antibody (A1) and 76A anti-Kir6.2 antibody (A2). Panel A3 is an overlay of panels A1 and A2. Secondary antibodies used were Cy-3 conjugated donkey anti-chicken IgY (red) and Cy-2 conjugated donkey anti-rabbit IgG (green). Yellow in panel C demonstrates areas of co-localization. The image width is 91 μm. B : Ventricular myocyte probed with anti-SUR1 antibodies and detected with Cy-3 conjugated donkey anti goat secondary antibodies. Image width is 148 μm. C : Staining with a pan-SUR2 antibody (detected with Cy-2 conjugated donkey anti-goat IgG). The image width is 229 μm. D : An isolated myocyte was stained with MitoTracker Red (500 nM) before being paraformaldehyde fixed and viewed with confocal microscopy Image width is 47 μm. Experiments were designed closer to examine the subcellular expression of SUR1 subunits relative to Kir6 pore-forming subunits (Fig 8 ). There was a remarkable degree of co-localization of SUR1 subunits with Kir6.1 subunits (Fig 8A ) but not with Kir6.2 subunits (Fig 8B ). Figure 8 Subcellular localization of Kir6.1, Kir6.2 and SUR1 subunits in ventricular myocytes. A : Co-staining of a mouse ventricular myocyte with antibodies to Kir6.1 (A1: NAF-1) or SUR1 (A2: C-16). The secondary antibodies were Cy2-conjugated donkey anti-rabbit and Cy3-conjugated donkey anti-goat. The degree of co-localization after background elimination is depicted in panel A3 by plotting pixel intensities of the two channels against each other. Pseudo-coloring of pixels indicates degree of co-localization (black is low and white high). B : Co-staining of a mouse cardiac myocyte with anti-Kir6.2 antibodies (76A) and anti SUR1 antibodies. Secondary antibodies used were Cy-2 conjugated donkey anti-rabbit to detect Kir6.2 (green) and Cy-3 conjugated donkey anti-goat IgG to detect SUR1 (red). A 3-D reconstruction was performed (58 stacked images with a voxel size of 0.23 μm 3 ). The image width is 119 μm. The images shown on the bottom and right insets are cut-through projections of the cell width (the regions of cross-sections areas shown by the white lines; the middle of each image is also indicated by a white line for clarity). We finally examined the subcellular expression patterns of K ATP channel subunits in another species. We chose rat ventricular myocytes for this purpose. We essentially obtained the same results as with mouse myocytes. Using different antibodies against Kir6.1 (NAF-1 and C16) we observed strong surface expression of this subunit with a smaller degree of intracellular labeling (Fig 9A and 9B ). SUR1 antibodies also strongly labeled the cell surface (Fig 9C ). In contrast, Kir6.2 and SUR2 antibodies both labeled repetitive patterns at sarcomeric distances (Fig 9D and 9E ). Furthermore, there is strong co-localization of Kir6.2 and SUR2 subunits in ventricular myocytes (yellow in Fig 9F ). Figure 9 Subcellular localization of K ATP channel subunits in isolated rat ventricular myocytes. A : Staining with NAF-1 anti-Kir6.1 antibodies (detection with Cy-3 conjugated donkey anti rabbit). Image width is 128 μm. B : Staining with a different anti-Kir6.1 antibody (C16, Santa Cruz); secondary antibody used was Cy-3 conjugated donkey anti-goat IgG. Image width is 136 μm. C : Staining with an anti-SUR1 antibody. Image width is 128 μm. D : Co-localization of Kir6.2 and SUR2 subunits. The preparation was co-stained with an anti-Kir6.2 antibody (76A; detection with Cy-2 conjugated donkey anti-rabbit IgG; green) and SUR2 antibodies (detected with Cy-5 conjugated donkey anti-goat IgG; pseudo-colored red for clarity). Panel D3 is an overlay of panels D1 and D2; yellow is indicative of co-localization. Image width is 125 μm. Discussion Antibodies used in this study A significant strength of our study is that we extensively characterized the antibodies used. We performed Western blotting with membrane fractions obtained from the heart to ensure that a band of the expected size is detected. Further, we chose antibodies that showed little cross-reactivity with other proteins, as judged by the absence of non-specific bands. In as far as it was possible we used different antibodies to the same K ATP channel subunits in immunostaining experiments to ensure that the same cellular and subcellular distribution staining patterns occurred. Although not shown, we always performed negative controls to ensure that no staining was observed when the primary antibodies were omitted (to eliminate non-specific staining by the secondary antibodies used) or that staining could be blocked by preincubation of antibody with the peptide to which the antibody has been raised. Further, we used the primary antibodies at the lowest concentrations possible to eliminate possible non-specific cross-reactivity with other proteins. Our study is a comprehensive description of the cellular and subcellular expression patterns of K ATP channel subunits in the heart given our stringent criteria and the panel of antibodies available to us. Expression of Kir6.2 and SUR2 subunits in ventricular myocytes Sarcolemmal K ATP channels in ventricular myocytes have been described more than two decades ago [ 17 ]. Cardiac sarcolemmal K ATP channels have been described to consist of hetero-octameric complexes of Kir6.2 and SUR2A subunits [ 5 , 18 , 19 ]. This concept was based on the similarities in the biophysical and pharmacological characteristics when comparing heterologously expressed Kir6.2/SUR2A channels with native cardiac K ATP channels [ 20 ] and also because of the known expression of Kir6.2 and SUR2A mRNA and protein in the heart. Our data demonstrate both Kir6.2 and SUR2A subunits to be expressed in ventricular myocytes. Furthermore, we find that these two subunits co-localize, which is consistent with the biochemical, functional and pharmacological data supporting the concept that they can combine to form a heteromeric channel complex [ 5 ]. Our data are also in agreement with the finding that knockout mice deficient of Kir6.2 or SUR2 subunits lack K ATP channels in the ventricular myocyte [ 21 , 22 ]. It is interesting that Kir6.2/SUR2 subunits are expressed in a regular striated pattern in ventricular myocytes. Furthermore, close inspection of the images shows that both Kir6.2 and SUR2 expression is somewhat punctate. These observations are in complete agreement with previous studies describing the expression of SUR2 isoforms in the t-tubules and sarcolemma [ 23 ] and the subcellular localization of sarcolemmal K ATP channels as determined by functional microscopy. Scanning ion conductance (patch clamp) microscopy data have demonstrated K ATP channels to be organized in small clusters and to be anchored in the Z-grooves (t-tubular openings) of the sarcolemma [ 24 ]. Collectively, these data suggest that K ATP channels are predominantly expressed in the t-tubular system. The implications of K ATP channels present in the t-tubular system are not entirely clear. Since t-tubular ion channels may control action potential propagation into the cardiac myocyte, it may be possible for K ATP channels to have a role in the spread of excitation and action potential duration, particularly during conditions of metabolic impairment when these channels are more prone to opening. A shorter action potential duration in the t-tubular system would imply less Ca 2+ entry at the local control sites of SR Ca 2+ release and hence reduced contractility, which may in part explain the negative inotropic effects observed with K ATP channel openers [ 25 ]. However, the picture may be more complex since both Kir6.1 and SUR1 subunits are also expressed in ventricular myocytes. Expression of Kir6.1 and SUR1 subunits in ventricular myocytes We found clear expression of Kir6.1 and SUR1 subunits in cardiac ventricular myocytes. Interestingly, both of these two subunits are strongly expressed at the sarcolemmal surface, but their functions in the sarcolemma are currently not understood. Since ventricular K ATP channels can be recorded in hearts from knockout mice lacking Kir6.1 subunits [ 9 ], it appears that Kir6.1 subunits are not an absolute requirement for the formation of functional ventricular K ATP channels. It may be possible for Kir6.1 subunits to have a role in the pathophysiological setting, as demonstrated by the upregulated Kir6.1 expression levels during cardiac remodeling after ischemia or hypoxia [ 11 , 26 ]. To our knowledge, cardiac K ATP channels have not been studied in SUR-/- mice. However, SUR1 antisense oligonucleotides inhibit K ATP channels in rat ventricular myocytes [ 8 ], suggesting a functional role for these subunits in ventricular sarcolemmal K ATP channel function. Our data, demonstrating that both Kir6.1 and SUR1 subunits exhibit strong sarcolemmal expression, may require a reassessment of the molecular composition of ventricular K ATP channels during normal and pathophysiological conditions. Expression of Kir6 and SUR subunits in mitochondria The concept has evolved that K ATP channels are expressed in the mitochondrial inner membrane and that these channels are involved in the protection of the heart afforded by ischemic preconditioning [ 27 , 28 ]. The molecular nature of mito-K ATP channels remains to be identified. There are almost as many reports describing the presence of Kir6.0 subunits in cardiac mitochondria [ 23 , 29 , 30 ]. as there are denouncing their existence in these organelles [ 31 , 32 ]. We did not observe strong localization of K ATP channel subunits in ventricular mitochondria. However, the technique of immunocytochemistry does not have sufficient resolution to rule out the existence of K ATP channel subunits in mitochondria and our data therefore do not add significantly to this debate, other than demonstrating that K ATP channel subunits are not abundantly expressed in mitochondria of ventricular myocytes. Expression of K ATP channel subunits in the coronary smooth muscle A tight coupling exists between metabolic status in the heart and coronary blood flow. K ATP channels have been identified in several different vascular tissues, including the coronary vasculature [ 33 ]. K ATP channels in coronary resistance vessels have also been implicated in physiologically important stimuli such as regulation of basal vascular tone, autoregulation of blood flow, hypoxia-induced coronary vasodilation, reactive hyperemia (a clinical index of coronary reserve) and ischemia (reviewed in [ 33 , 34 ]). Molecularly, the identity of coronary smooth muscle K ATP channels has been characterized less extensively than the K ATP channels in cardiomyocytes. A recent study employing in situ hybridization histochemistry examined Kir6.1 and SUR2B mRNA expression in different vascular beds, including the coronary vasculature [ 35 ]. Strong mRNA expression of these two subunits was found in coronary resistance arteries. Interestingly, Kir6.1/SUR2B expression was not found in coronary veins or venules. We found expression of Kir6.1 and SUR2B protein in primary human coronary artery smooth muscle cells and cryosections of human ventricle [ 13 ]. The present study is the first systematic and comparative characterization of K ATP channel subunit expression in the intact coronary vasculature. We find Kir6.1 expression in blood vessels of different sizes, including large vessels such as the aorta and large arteries, but also in small resistance arterioles (as defined by their diameter of larger than 12–15 μm and the presence of a well-defined smooth muscle layer). Without using specific markers, we were not able to distinguish objectively between venules and arterioles, but we did occasionally observe vessels with a thin smooth muscle layer (possibly veins or venules) that only expressed Kir6.1 faintly (if at all). Collectively, our data using various anti-Kir6.1 antibodies generally suggest that Kir6.1 subunits are expressed in coronary arterial smooth muscle, and possibly to a lesser extent in coronary veins. In contrast, we did not observe any staining of the coronary smooth muscle with anti-Kir6.2 antibodies. We found strong staining of smaller coronary resistance vessels with the anti-SUR2 antibodies. We did not have access to suitable SUR2 isoform-specific antibodies, but the staining most probably reflects SUR2B expression. Curiously, we failed to see strong SUR2 expression in larger coronary arteries. This result is in apparent contradiction to the description that SUR2B mRNA expression occurs in larger coronaries [ 35 ] and the lack of K ATP channels in the aortic cells of the SUR2 knockout mouse [ 36 ]. Reasons for this discrepancy are unclear, but may relate to the lack of sensitivity of the anti-SUR2 antibodies used, thus underestimating SUR2 protein expression in other structures. We did not observe SUR1 protein expression in the coronary smooth muscle. Our data are therefore in full support of the notion that K ATP channels in coronary artery smooth muscle (particularly the smaller resistance vessels) are comprised of Kir6.1/SUR2 subunits (most likely the SUR2B isoform). Expression of K ATP channel subunits in the coronary endothelium The evidence that endothelial K ATP channels play a role in regulation of coronary blood flow is compelling. Endothelial K ATP channels, for example, contribute to shear stress-induced endothelial release of the vasodilator nitric oxide in rabbit aorta [ 37 ] and may also mediate vasodilation in response to hyperosmolarity or acidosis in the coronary microvasculature [ 38 , 39 ]. Furthermore, the powerful vasodilatory effect of adenosine may also be mediated (at least in part) by endothelial K ATP channels by stimulating the release of nitric oxide from the endothelium [ 40 ]. In the present study, we used immunohistochemistry approaches and identified Kir6.1, Kir6.2 and SUR2 protein in the endothelium lining coronary vessels (Fig. 4 ) as well as in coronary capillary endothelium (as defined by their small size of less than 10 μm, the presence of the endothelial marker ICAM-2 and the absence of vascular smooth muscle). Our data are supported by previous studies. Using RT-PCR techniques, it has been established that guinea pig coronary endothelial cells express Kir6.1, Kir6.2 and SUR2B subunits [ 41 ]. The presence of Kir6.1 and SUR2B mRNA also detected in the coronary endothelium using in situ hybridization histochemistry techniques [ 35 ]. Recently, using a combination of RT-PCR and Western blotting techniques, we also identified Kir6.1, Kir6.2 and SUR2B mRNA and protein expression in primary human coronary artery endothelial cells [ 13 ]. Importantly, in the latter study we used co-immunoprecipitation approaches to demonstrate that native human coronary endothelial K ATP channels are heteromeric Kir6.1/Kir6.2 complexes in combination with SUR2B subunits [ 13 ]. Thus, the biophysical nature, modes of regulation and functional consequences of these heteromeric K ATP channels in the endothelium may differ fundamentally from homomeric K ATP channels found in other tissues. The investigation of endothelial K ATP channels is currently the subject of some of our ongoing studies. Reservations For this type of study, the specificity of antibodies used is always a concern. To overcome this problem, we used multiple different antibodies where possible and obtained similar staining patterns. However, we only had access to a single antibody to each of the SUR subunits and consequently we have not been able to verify the specificity of these antibodies by comparing different antibodies to each other (as we have done in the case of the Kir6 subunits). Furthermore, we did not have access to antibodies to the various splice variants of SUR1 or SUR2 and our data therefore do not address the possibility of regional expression differences. A definitive study will require the use of tissues obtained from knockout animals (i.e. the immunostaining should be unequivocally absent in tissues from knockout animals). Viable knockout animals for each of the proteins under consideration have been generated, but we have not been able to obtain these animals (or tissues from these animals) for this purpose. Therefore, although we have taken every step possible to minimize non-specificity issues, our results should be interpreted within this limitation. Conclusions Our study is a comprehensive analysis of the various K ATP channel subunits in the heart. We found each of the K ATP channel subunits to be expressed in ventricular myocytes, but with varying expression patterns. The roles of Kir6.1 and SUR1 subunits in ventricular myocytes remain to be elucidated and may require a reassessment of the molecular nature of the cardiac K ATP channel. Coronary smooth muscle expresses predominantly Kir6.1 and SUR2 subunits, whereas the coronary endothelium expresses Kir6.1, Kir6.2 and SUR2 subunits. Thus, there is wide diversity of K ATP channel subunit expression within the heart which determines the functional responses of various cell types to physiological and pathophysiological demands. Methods Immunohistochemistry Adult mice were sacrificed by pentobarbital overdose and the hearts were rapidly removed. All animal experiments were approved by the institutional Animal Care Review Board. Hearts were perfused at 37°C through the aorta (Langendorff mode) with Tyrode's solution (in mM: NaCl 137, KCl 5.4, HEPES 10, CaCl 2 1.8, MgCl 2 1, NaH 2 PO 4 0.33; pH adjusted to 7.4 with NaOH) containing pinacidil (100 μM) to cause maximal vasodilatation as to clear the vasculature of blood. Hearts were fixed by switching the perfusate to paraformaldehyde (4% in phosphate-buffered saline (PBS), pH adjusted to 7.4) for 15 minutes at room temperature. The heart was incubated in 4% paraformaldehyde overnight at 4°C. Following fixation, the tissue was incubated overnight at 4°C in 30% sucrose in PBS. The tissue was then embedded in M1 embedding matrix (Thermo Shandon, Pittsburgh, PA) and placed on dry ice until frozen. The block containing the tissue was sectioned using a cooled (-20°C) cryostat (Microm Cryo-Star HM 560, Kalamazoo MI) at 15 μm thicknesses. The sections were transferred to Superfrost Plus slides (Fisher Scientific) for further processing. Tissue sections were allowed to warm to room temperature. The staining protocol was carried out in a moist chamber to avoid dehydration. Blocking was performed for 60 min with Tris-buffered saline (TBS; in mM 137 NaCl, 50 Tris, 2.7 KCl, pH 7.4) containing 4% goat or donkey serum (depending on the secondary antibody being used) and 0.2% Triton X-100 at room temperature. The slides were incubated overnight at 4°C with primary antibodies (see below) diluted in TBS containing 0.1% serum and 0.2% Triton X-100. Double or triple immunofluorescent studies were carried out by incubating the tissue sections with more than one primary antibody at the same time. Sections were washed three times for 15 min each in TBS, and incubated with fluorescently labeled secondary antibodies (see below) for 1 h at room temperature. The samples were again washed three times for 15 min each in TBS. Sections were drained by blotting with filter paper and a drop of mounting medium (containing an anti-fade reagent) was added to the slides before mounting with a standard coverslip. The mounting medium was allowed to dry before the slides were imaged using a Leica PS2 confocal microscope equipped with an Argon 488 nm gas laser and Helium Neon lasers (543 and 633 lines). Most images were obtained using an emission pin hole of 1.1–1.6 AE with either a 20× (0.7 NA) or a 63× (1.2 NA) oil objective. Isolation and immunocytochemistry of cardiac myocytes Single ventricular myocytes were isolated from adult rats or mice using previously described procedures. Briefly, adult animals were sacrificed and the heart was rapidly removed and perfused in Langendorff mode (at 37°C) for sequential 5 min periods with Tyrode's solution and nominally Ca 2+ -free Tyrode's solution (same composition, but without the addition of CaCl 2 ). Myocytes were dispersed using collagenese (Sigma type I; Sigma-Aldrich Chemical Corp, St. Louis, MO, USA) and proteinase (Sigma type XXIV). The ventricles were removed and chopped into small pieces and digested using Ca 2+ -free Tyrode's solution containing the same enzymes. Single dissociated myocytes were plated onto laminin (10 ug/cm 3 )-coated glass coverslips and incubated at 37°C for 15 min to allow attachment to the coverslips before being fixed. In some experiments, myocytes were incubated with 500 nM MitoTracker Red 580 (Molecular Probes, Eugene, OR) during this period. Three different fixation protocols were employed. Some myocytes were fixed in paraformaldehyde (4%) for 15 min at room temperature, whereas with others fixation and permeabilization was performed in a single step by incubation with ice-cold 100% methanol for 5 minutes at -20°C. However, in the majority of the myocytes presented in this study a two-step fixation protocol was used [ 16 ], in which myocytes were first fixed with paraformaldehyde (as described above) followed by methanol fixation/permeabilization. Irrespective of the fixation method, myocytes were then washed with Ca 2+ and Mg 2+ -free PBS (Invitrogen, Carlsbad, CA). Myocytes were then incubated with 0.1% Triton X-100 (in PBS) for 15 min at room temperature, which permeabilizes surface membranes as well as those of intracellular organelles (this step was omitted when fixing the cells only with methanol but was included in the two-step fixation protocol). Following washing (2 × 5 min) and blocking (5% goat serum in PBS; 2 × 10 min) the cells were incubated with primary antibodies (1 h at room temperature), washed (3 × 10 min in PBS-serum) and incubated with secondary antibodies (45 min at room temperature). Following 4 washes (with PBS; 10 min each) coverslips were mounted and viewed using confocal microscopy. As a negative control, the primary antibody was adsorbed with the peptide against which it was made (when available). Negative staining controls (not shown) also included a null control, in which the primary antibody was omitted, which tested for non-specific staining of the secondary antibody. To avoid background from secondary antibodies alone, we normally pre-blocked the tissue with 5% normal serum from the same host species as the labeled secondary antibody. We used labeled secondary antibodies that have been pre-adsorbed against mouse and human and we titrated the labeled secondary antibody to obtain a maximal signal-to-noise ratio. Transfection of cells The coding regions of Kir6.1 and Kir6.2 (a gift from Dr. S. Seino, Kobe University Graduate School of Medicine, Japan) were subcloned into pcDNA3. HEK-293 or COS-7L cells were cultured in D-MEM (Invitrogen) supplemented with 10% heat-inactivated fetal bovine serum and 20 μg/mL gentamycin. Cells were co-transfected according to the manufacturer's recommendations (Fugene 6, Roche Applied Science, Indianapolis, IN) with Kir6.1 or Kir6.2 cDNA (obtained from Dr S. Seino), SUR1 (Dr J. Bryan, Baylor College of Medicine, Texas) or SUR2A cDNAs (a gift from Dr Seino). Cells were lysed 48 h post-transfection. Preparation of mouse heart membrane fraction Adult Sprague Dawley rats were anesthetized and the hearts were rapidly removed. Membranes were prepared essentially as described before [ 42 ]. The protein content was determined and equal amounts of proteins were subjected to Western blotting. Western blotting Cells were solubilized in lysis buffer [25 mmol/L Tris, 150 mmol/L NaCl, 5 mmol/L EDTA, 1% (v/v) Triton X-100, 0.5% (w/v) deoxycholate, pH 7.5 supplemented with a cocktail of protease inhibitors (Sigma)]. After centrifugation (10 min at 16,000 g), an equal volume of 2 × Laemmli loading buffer was added to the lysate. Proteins were separated by 10% SDS-PAGE and transferred to Immun-blot PVDF membrane (Bio-Rad Laboratories, Hercules, CA). The membrane was blocked and incubated with primary antibody (see below). As secondary antibodies, we used HRP-conjugated anti-rabbit, anti-mouse IgG (Amersham Biosciences, Piscataway, NJ) or anti-goat IgG in TBS-Tween for 1 hour and detected using a chemiluminescent substrate (SuperSignal, Pierce Biotechnology, Rockford, IL). Antibodies Antibodies against Kir6.1 subunits Antibodies (NAF1) were raised against a peptide corresponding to residues 20–31 of the Kir6.1 N-terminus (ENLRKPRIRDRLP). There is a high degree of sequence similarity in Kir6.1 subunits between species in this region. We also raised a Kir6.1 antibody (CAF-1) to this peptide in chickens. In each case, a C-terminal cysteine was added for conjugation purposes. Peptides were synthesized and the antibodies were generated commercially (Quality Controlled Biochemicals, Hopkinton, MA). Each of these antibodies were peptide affinity purified. There is sequence similarity (81% identity) between this peptide and the recently-identified human beta-V spectrin. We tested for possible cross-reactivity of NAF-1 with beta-V spectrin using antibodies generously provided by Dr Jon Morrow (Yale University) and demonstrated that our antibodies had no cross-reactivity with beta-V spectrin (please see online Additional file 1 ). It should further be noted that the antibody epitope has no similarity with mouse beta-V spectrin (see UniGene Cluster Hs.198161). Other anti-Kir6.1 antibodies that we attempted to use with varying degrees of success included the 78A antibody generated in the laboratory of Dr Tinker, which was raised in rabbits against a peptide corresponding to amino acids 399–420 of rat Kir6.1 with a terminal cysteine added for coupling purposes (RRNNSSLMVPKVQFMTPEGNQC) and the goat anti-Kir6.1 R-14 or C-16 C-terminal antibodies (sc-11224 and sc-11225; Santa Cruz Biotechnology, Santa Cruz, CA). In our hands, the 78A and R-14 antibodies failed to detect Kir6.1 protein in Western blotting of cardiac membrane fractions and did not appear to stain above background in immunohistochemistry assays (not shown). Antibodies against Kir6.2 subunits We used a goat anti-Kir6.2 G-16 antibody (sc-11228; Santa Cruz Biotechnology, Santa Cruz, CA) and an antibody (76A) developed in Dr Tinker's laboratory against a peptide (DALTLASSRGPLRKRSC) corresponding to a peptide within the Kir6.2 C-terminus (amino acids 357–372). Antibodies against SUR subunits We used a goat anti-SUR1 C-16 antibody (sc-5789; Santa Cruz Biotechnology) developed to an epitope mapping at the C-terminus and a goat anti-SUR2 C-15 C-terminal antibody (sc-5793; Santa Cruz Biotechnology). This antibody was initially sold as a pan-SUR2 antibody and was able to detect the SUR2A protein in Western blots of cell lysates from Kir6.2/SUR2A transfected cells, although with less sensitivity compared to SUR2B-transfected cells (data not shown). Currently, the antibody with the same catalog number is sold as being specific to SUR2B. We have not tested recent batches of this antibody for isoform specificity. Other antibodies Other antibodies used for immunolocalization included a mouse monoclonal α-actin smooth muscle antibody preconjugated to FITC (Clone 1A4; Sigma-Aldrich Corp, St. Louis, MO) and a rat monoclonal antibody (clone 3C4; BD Biosciences Pharmingen, San Diego, CA) raised against ICAM-2 (CD102), which is a cell surface glycoprotein constitutively expressed on vascular endothelial cells. Secondary antibodies used included Cy3-conjugated donkey anti-rabbit IgG, Cy3-conjgated donkey anti-chicken IgY, Cy5-conjugated F(ab') 2 fragment donkey anti-rat IgG and a Cy3- or Cy5-conjugated donkey anti-goat IgG (all from Jackson ImmunoResearch Laboratories Inc, West Grove, PA). List of abbreviations Kir = Inward rectifying K + channel family SUR = Sulphonylurea receptor K ATP channel = ATP-sensitive K + channel Authors' contributions AM, ER and JL carried out the immunohistochemistry experiments. PDC, SH and GL carried out the immunocytochemistry experiments. TN, AM, LP, HY, XT, JPG and AT participated in characterization of the antibodies. WAC, MA and TN participated in the design of the study. WAC conceived the study, participated in study design and coordination, and manuscript preparation. All authors made substantive intellectual contributions to the study, read and approved the final manuscript. Supplementary Material Additional File 1 NAF-1 antibody does not cross-react with beta V spectrin. COS7L cells have been transfected with Kir6.1 cDNA and cell lystates were subjected to Western blotting with anti-beta V spectrin antibodies or NAF-1 anti-Kir6.1 antibodies. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546210.xml
535573	The Emergence of Complexity: Lessons from DNA	The same molecular qualities that endowed DNA with its capacity to carry hereditary information make it a powerful tool to explore the self-assembly of complex nanostructures	How did life emerge from a soup of chemicals? How do patterns such as schools of fish form from individuals? How do voting patterns emerge? Such a diverse array of problems seems completely unrelated. However, they all involve “emergence of complexity.” When individuals come together, they form patterns, structures, and organizations that cannot be discerned from the individuals alone. The study of the emergence of complexity is one of the most active and important areas of research. It is important not only for understanding nature, but also for technological applications, including the fabrication of large-scale integrated nanocircuits using a bottom-up approach, and the preparation of multifunctional “smart” nanomaterials. DNA evolved to be the primary carrier of genetic information because of its extraordinary chemical properties. These same properties also make DNA an excellent system for the study of self-assembly and self-organization. Two complementary molecules of single-stranded DNA have paired bases that bond with each other and form the well-known double helix structure. Two molecules of double-stranded DNA (duplexes) can further associate together if they have complementary single-stranded overhangs (sticky ends). Intermolecular interactions can be precisely predicted by Watson–Crick basepairing (adenine to thymidine and guanine to cytosine). And, these interactions are structurally understood at the atomic level. Given the diversity of the DNA sequences, we can easily engineer a large number of pairs of DNA duplexes that associate with each other with sequence specificity and in a well-defined fashion. This property is not common among other molecular systems. Small organic and inorganic molecular pairs can interact with each other with specificity and in well-defined structures, but the number of such pairs is limited and their chemistry varies greatly. Protein molecules, such as antibody–antigen pairs, have great diversity and high specificity. However, it is extremely difficult, if not impossible, to predict how proteins interact with each other. In contrast, DNA as a molecular system fulfills all the aforementioned criteria. In nature, DNA occurs predominantly as a linear molecule, and if its conformations were limited to linearity, it would not be very useful for studying self-assembly. Fortunately, branched DNA structures can be engineered. Holliday junctions, for example, are intermediates that occur during genetic recombination. To model Holliday junctions, a stable four-arm junction has been constructed in which, by design, no two strands are fully complementary to each other ( Kallenback et al. 1983 ; Seeman 2003 ) ( Figure 1 A). For example, the 5′ half of strand 2 is complementary to the 3′ half of strand 1, but the 3′ half of strand 2 is complementary to the 5′ half of strand 3 instead of that of strand 1. Combining branched structures and the excellent molecular recognition of DNA, we are ready to engineer complicated DNA nanostructures and use them for studying self-assembly. Figure 1 Basic DNA Structures for Self-Assembly (A) A four-arm junction and (B) its three-dimensional structure; (C) a DNA DX; and (D) a DNA TX. Extensive studies have shown that the four-arm junction adopts an X-shape structure ( Figure 1 B) under physiological conditions, and the angle between its two helical domains can vary widely ( Lilley 2000 ). It is impossible to construct well-defined large structures from flexible components. To overcome this problem, several well-behaved DNA motifs have been engineered. Double crossover (DX) ( Fu and Seeman 1993 ) and triple crossover (TX) ( LaBean et al. 2000 ) molecules are two early examples ( Figure 1 C and 1 D). In such molecules, two or three DNA duplexes lie side by side. Two neighboring duplexes are joined by two crossovers, which prevent any duplex from twisting against its neighbor duplex. Thus, the interhelical angles become fixed at 0°. Other motifs quickly followed, including the paranemic crossover motif ( Shen et al. 2004 ), rhombus/parallelogram motif ( Mao et al. 1999 ), cross motif ( Yan et al. 2003 ), and several triangle motifs ( Chelyapov et al. 2004 ; Ding et al. 2004 ; Liu et al. 2004 ). They all are stable, rigid, and readily designed for self-assembly. One simple example of self-assembly is the formation of two-dimensional (2D) periodic arrays or 2D crystals. This is also one of the greatest successes in the field of DNA self-assembly ( Figure 2 ). The first 2D DNA crystals were assembled from DX motifs ( Winfree et al. 1998 ). In a 2D crystal, each DX molecule contains four sticky ends (A–A′ and B–B′) distributed on its two component duplexes. The complementarity of the sticky ends is designed in such a way that a DX molecule will interact with another four DX molecules through its four sticky ends. Any two DX molecules can interact with each other though only one pair of sticky ends. Any pair of sticky end interactions will position the two DX molecules in a conformation such that no other sticky ends from these two molecules are in sufficient proximity to interact. As a result of this design, regularly ordered 2D arrays have formed ( Figure 2 ). Following similar strategies, others have designed DNA motifs to assemble into 2D arrays, whose symmetries include tetragonal ( Yan et al. 2003 ), pseudohexagonal ( Mao et al. 1999 ; Liu et al. 2004 ), and hexagonal ( Chelyapov el al. 2004 ; Ding et al. 2004 ). Figure 2 Self-Assembly of a 2D DX Array Each rod represents a DNA duplex. The geometric complementarity represents the sequence complementarity of sticky ends. Inspired by early theoretical suggestions ( Winfree 1998 ), experimental exploration of aperiodic self-assembly immediately followed. One study applied algorithmic self-assembly to TX molecules ( Figure 3 ) ( Mao et al. 2000 ). The assembling rule “exclusive OR” (XOR) is encoded in the TX molecules. Consider the value of all inputs and outputs as either 1 or 0. For XOR operations, if two inputs are the same, the output will be 0; otherwise, the output will be 1. If molecules X and Y are the input and output, respectively, Y i molecule takes the input from the X i and Y i−1 molecules. In other words, the values of the X i and Y i−1 molecules determine what Y molecule will be incorporated. There are four different types of Y molecules, whose inputs are (1, 1), (1, 0), (0, 1), and (0, 0). These four, and only four, Y molecules are enough to satisfy any input combination. Two C molecules connect the input and output molecules, which is necessary for the characterization but not essential for the self-assembly process. Sticky ends between the X molecules and the C molecules are longer than those between Y molecules and between Y and X or C molecules. Thus, the C and X molecules assemble first to form the inputs because the association between longer sticky ends is more stable than those between the shorter ones. Then the output Y molecules assemble to the assembled C and X molecules. In that study ( Mao et al. 2000 ), two different input combinations were used, and one of them is shown in Figure 3 . The resulting DNA structures are periodic with respect to the backbones, but they are aperiodic in their sequences. Though the resulting four-byte one-dimensional (1D) structures are quite simple, this study demonstrated that aperiodic structures are achievable through self-assembly. Figure 3 Self-Assembly of a 1D Aperiodic TX Array Based on XOR Operation The value of any input or output is binary, either 1 or 0. If two inputs are the same, the output is 0; otherwise, the output is 1. Winfree and co-workers in this issue of PLoS Biology have extended the algorithmic self-assembly strategy from 1D to 2D ( Rothemund et al. 2004 ). This achievement is certainly a milestone in the field of self-assembly. It overcomes a great challenge, as the structural complexity dramatically increases from 1D to 2D structures. These researchers have applied the same XOR algorithms to DX molecules in their study and achieved fractal structures, Sierpinski triangles ( Figure 4 ). External inputs are in the bottom row. Each row takes inputs from the row immediately below, and sends the operation outputs to the row immediately above. Each position takes two inputs (identical or non-identical) from lower left and lower right positions, and sends identical output to both upper left and upper right positions. The arrows indicate the direction of information flow, or assembly sequences. In their experiment, the rules are encoded in DX molecules. This study is conceptually straightforward, but the experimental challenges are tremendous. One key challenge is assembly fidelity. The right molecules have to compete with partially matched molecules. The concentrations of the competing molecules further complicate the fidelity issue, as some molecules could be rapidly depleted from the solution. In that sense, the current work is quite stunning even though the assembly is far from perfect. Figure 4 Schematic Representation of Self-Assembly of a Sierpinski Triangle Based on XOR Operation The values in the bottom row are the inputs. In principle, a wide range of 2D patterns could be generated with the same set of molecules and the same strategy, changing only the first row of the assembly, which specifies the external inputs. Realization of this goal will critically rely on the elimination of assembly errors, or the introduction of error corrections ( Winfree and Bekbolatov 2004 ). The current work represents a neat approach to understanding the emergence of complexity. It integrates both simulation and wet chemistry. It also provides a plausible approach to nanofabrications. Over the last decade, a variety of methods have been developed, which use biomacromolecules as templates to fabricate nanostructures ( Braun el al. 1998 ; Douglas and Young 1998 ; Mucic et al. 1998 ; Fu et al. 2004 ). Limited by the complexity of the available biomacromolecular templates, simple nanostructures are the usual result: mostly nanowires, nanoparticles, and simple aggregates of nanoparticles. The current work illustrates the possibility of generating more complicated structures and promises unprecedented structural complexity for nanomaterials.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535573.xml
539349	Evaluation of the clinical value of bone metabolic parameters for the screening of osseous metastases compared to bone scintigraphy	Background Bone metastases are common in many types of cancer. As screening methods different imaging modalities are available. A new approach for the screening of osseous metastases represents the measurement of bone metabolic markers. Therefore aim of this study was to evaluate the usefulness of the determination of bone metabolic markers aminoterminal propeptide of type I procollagen (PINP, osteoblastic activity) and the carboxyterminal pyridinoline cross-linked telopeptide of type I collagen (ICTP, osteoclastic activity) for the detection of bone metastases associated with other malignancies. Methods 88 patients aged 21 – 82 years with malignant tumors were prospectively studied. The serum concentrations of PINP and ICTP were measured and compared to the results of bone scintigraphy, radiological bone series, CT, MRI and clinical follow-up. Results Osseous metastases were found in 21 patients. 19 of them were correctly identified by bone scintigraphy (sensitivity: 90%). For bone metabolic markers results were as follows: ICTP sensitivity: 71%, specificity: 42%; PINP sensitivity: 24%, specificity: 96%. Conclusions As markers of bone metabolism PINP and ICTP showed low sensitivity and/or specificity for the detection of osseous metastases. The presented markers did not seem to be sufficient enough to identify patients with bone metastases or to replace established screening methods.	Background Bone metastases are common in advanced cancers of the lung, breast, kidney, prostate and others. In autopsy studies the prevalence ranges from 47–85% [ 1 ]. In patients with osseous metastases early detection is needed, since without effective treatment these bone metastases can cause severe complications leading to considerable morbidity and reduced quality of life. The screening for bone metastases is usually based on bone scintigraphy, confirmed by supplemental radiographic bone surveys, computer tomography, or magnetic resonance imaging. Bone scans are simple to perform and examine the whole skeleton, but although the sensitivity of this method is high its specificity is poor. For this reason, in many cases a positive scan requires confirmation by other imaging modalities, which leads to higher costs and time-consuming investigations. Therefore, an alternative cost-effective screening method with a similar sensitivity and a higher specificity would be very welcome. For the formation of osseous metastases the extracellular matrix consisting of collagens combined with noncollagenous glycoproteins and proteoglycans, including the basement membrane and the interstitial stroma play an important role. Normally, the extracellular matrix serves as a barrier for the attachment and invasion of malignant cells, but the proteolytic activity of tumor cells leads to the destruction of its collagenic components, thus facilitating the local invasion of malignant cells and finally the development of bone metastases [ 2 ]. The major collagen in bone is type I collagen, which is synthesized by osteoblasts and accounts for about 90% of the organic matrix [ 3 ]. Recently, bone metabolic markers have been reported to be useful in diagnosing bone metastases [ 4 , 5 ]. Newly developed methods are able to quantitatively determine concentrations of collagen metabolites. The synthesis of type I collagen can be analyzed by measuring the serum concentration of the aminoterminal propeptide of type I procollagen (PINP) using a specific radioimmunoassay. In addition, bone resorption can also be analyzed by a radioimmunoassay, which measures the serum concentration of the carboxyterminal pyridinoline cross-linked telopeptide of type I collagen (ICTP). Both parameters have been identified as potential candidates for the early detection of bone metastases. Therefore, the purpose of this study was to evaluate the usefulness of PINP and ICTP in patients with newly diagnosed cancer for the screening of osseous metastases compared to current standard protocols. Methods Patients For this study 88 consecutive patients (35 female; 53 male) with malignant tumors (20 lung cancer, 20 breast cancer, 19 head/neck cancer, 5 prostate cancer, 4 thyroid cancer, 3 sarcoma, 2 esophagus cancer, 2 pancreatic cancer, 2 urothel cancer, 1 gastric cancer, 2 plasmocytoma, 2 histiocytosis X, 1 melanoma, 1 rectal cancer, 1 hypernephroma, 2 carcinoma of unknown primary, 1 breast cancer and hypernephroma) between the ages of 21 and 82 were included. None of these patients were under a tumor specific therapy or presented primary bone disease such as osteoporosis or Paget's disease, which could interfere with the results of this study. All patients gave written consent to participate in this prospective study, which was approved by the local ethics committee. Marker assays Blood samples for measuring PINP and ICTP were collected on the same day as bone scintigraphy. Due to higher PINP values at night, all samples were taken early in the morning and stored at -20°C until assayed. Apparent hemolytic serum was excluded. Serum concentrations of ICTP and PINP were measured by using commercially available RIA kits (both: Orion Diagnostica, Espoo, Finland). According to the kit description, the normal range of ICTP was 1.6–5.3 μg/l for females and 1.4–5.2 μg/l for males. For PINP the normal range was 19–102 μg/l for females and 21–78 μg/l for males. Bone scanning Two double-head gamma cameras (ECAM duet and Bodyscan; Siemens, Erlangen, Germany) and a triple-head gamma camera (Multispect III, Siemens, Erlangen, Germany) were used for planar and tomographic (SPECT) bone scans, respectively. Low-energy, high-resolution collimators (1024 × 256 matrix) were used and data acquisition was started 2–4 hrs after intravenous injection of 550–700MBq Tc 99m -DPD (3,3-Diphosphono-1,2-propandicarbonacid). SPECT acquisitions were performed from suspected regions (128 × 128 matrix; 64 steps; 150000–200000 counts/step; Butterworth filter; cut-off level 0.4). The total acquisition time ranged from 100 to 130 min for planar and SPECT scans. The bone-scanning procedure was performed in accordance with procedural guidelines published by the Society of Nuclear Medicine [ 6 ]. Interpretation of bone scintigraphy Two experienced nuclear medicine physicians interpreted planar bone and SPECT images. Initially, neither of the readers knew of the findings of the other or the results of other imaging modalities. Increased tracer-uptake located at joints or on the edge of vertebral bodies adjacent to disk spaces was interpreted as arthritis or osteophytes, respectively. Lesions were classified as fractures when they showed a typical linear and curved pattern e.g. adjacent lesions in the ribs. Multiple lesions of varying size, shape and intensity, elongated rib lesions or photopaenic areas (cold spots) were classified as osseous metastases or suspicious lesions, where further analysis or imaging methods were necessary. In general, interpretation was performed following the criteria described by Krasnow et al. [ 7 ]. Finally, any discrepant interpretations between the two readers were resolved by consensus. Patients were classified as having osseous metastases when other imaging modalities (radiographic, MRI, CT) or histological findings confirmed the diagnosis. Patients were classified as not having bone metastatic disease when no imaging technique or histological finding indicated osseous involvement. To reduce the possibility that bone metastases were not yet visible by the cited imaging methods, a clinical follow-up period of between 9–14 months was used as the gold standard in all patients. Follow-up consisted of a clinical examination, control of tumor markers, imaging techniques (e.g. computed tomography, MRI or plain radiographic), etc. Overall, follow-up was done in compliance with the guidelines for tumor patients published by the cancer society. Data analysis Sensitivity and specificity were calculated. Values are expressed as mean ± standard deviation. Statistical analysis was performed using SPSS-Software version 10.0 (SPSS, Inc.). The Mann-Whitney U test was applied to compare concentrations of PINP and ICTP, respectively. A value of p < 0.05 was considered to be statistically significant. Furthermore, we performed a receiver operating characteristic (ROC) analysis to assess the impact of bone metabolic markers in detecting bone metastases. Results Osseous metastases were found in 21 patients with the following tumors: 8 patients with breast cancer, 3 patients with head/neck cancer and local osseous tumor invasion, 3 prostatic cancer, 2 plasmocytoma, 1 each lung cancer, histiocytosis X, thyroid cancer and sarcoma and 1 patient had breast cancer and hypernephroma. Bone scintigraphy Using planar and tomographic bone scintigraphy, 19 patients were correctly diagnosed as patients with osseous metastases by both investigators independently and confirmed by other imaging modalities or follow-up (sensitivity: 90%). Two patients were false negative, 1 patient with histiocytosis X (diffuse infiltration of the spine and pelvis, identified by computed tomography) and 1 had breast cancer and hypernephroma (multiple osteolytic lesions with diameters up to 1.5 cm in the pelvic bone, which was correctly diagnosed by computed tomography). In 43 patients there was no evidence for osseous metastases by bone scintigraphy or other imaging modalities. In 24 patients, changes in bone scintigraphy were described not typical for osseous metastases, but additional imaging methods were recommended for verification. Most of these lesions were located in the ribs (singular focus) and the spine, typically for fractures, traumatic injuries or osteoporotic changes (compression fracture). Additional verification was recommended especially in those patients where no history of traumatic injuries, degenerative processes or osteoporosis was known. Verification was done in most cases by plain radiography, computed tomography or within the staging by FDG-PET. ICTP Serum ICTP was elevated above the upper reference limit (>5.2 μg/ in males and >5.3 μg/l in females) in 56 patients (31 male, 25 female). In females the mean value was 7.64 ± 4.25 μg/l. In the group with metastases (n = 10) the mean value was 10.71 ± 5.90 μg/l compared to 6.41 ± 2.66 μg/l in the other group (n = 25) without metastases (p = 0.11). In males the mean value was 8.74 ± 9.49. The patients with metastases (n = 11) showed a value of 9.23 ± 7.62 compared to 8.61 ± 9.99 in patients (n = 42) without bone metastases (p = 0.24). Figures 1 and 2 present the values of all patients, separated by male and female, showing the different range of normal values. In females sensitivity was 70% and specificity 32% and in males sensitivity was 73% and specificity 48%. For both groups combined sensitivity was 71% and specificity 42%. Figure 1 Values of ICTP in males (Reference interval: 1.4–5.2 μg/). Red columns indicate patients with osseous metastases. Figure 2 Values of ICTP in females (Reference interval: 1.6–5.3 μg/). Red columns indicate patients with osseous metastases. PINP Serum PINP was elevated above the upper reference limit (>78 μg/ in males and >102 μg/l in females) in 8 patients (5 male, 3 female). In females the mean value was 57.42 ± 38.50 μg/l. In the group of patients with metastases (n = 10) the mean value was 73.29 ± 62.38 μg/l compared to 51.07 ± 22.23 μg/l in the group (n = 25) without metastases (p = 0.95). In males the mean value was 45.59 ± 43.20. The patients with metastases showed a value of 60.70 ± 83.86 compared to 41.63 ± 23.99 in patients without bone metastases (p = 0.70). Figures 3 and 4 show the values of all patients. Sensitivity in females was 30 %, specificity 100 %, in males sensitivity was 18 %, specificity 93 %. Combined sensitivity was 24 % and specificity 96 %. As examples, figures 6 and 7 show patients with non-small-cell-lung cancer as the primary tumor. Figure 3 Values of PINP in males (Reference interval: 21–78 μg/). Red columns indicate patients with osseous metastases. Figure 4 Values of ICTP in females (Reference interval: 19–102 μg/). Red columns indicate patients with osseous metastases. ROC-analysis Neither parameter achieved statistical significance by ROC analysis. According to the ROC analysis, the optimal cut-off level of ICTP that maximizes sensitivity and specificity was 7.5 μg/l in females (60% sensitivity and 88% specificity) and 5.9 μg/l in males (73% sensitivity and 57% specificity), for PINP 85.5 μg/l in females (40% sensitivity and 96% specificity) and 25.8 μg/l in males (46% sensitivity and 74% specificity). Figure 5 shows ROC curves and summarizes the results. Figure 5 left side ROC curves for ICTP and PINP in males Marker AUC ± SE (95%CI) ICTP 0.62 ± 0.1 (0.47–0.75) PINP 0.54 ± 0.1 (0.40–0.68) AUC: area under curve; SE: standard error; 95%CI: 95% confidence interval right side ROC curves for ICTP and PINP in females Marker AUC ± SE (95%CI) ICTP 0.67 ± 0.1 (0.49–0.82) PINP 0.51 ± 0.1 (0.33–0.68) AUC:area under curve; SE: standard error; 95%CI: 95% confidence interval The types of tumors in our study were very heterogeneous, which reflects the normal day-to-day situation seen in a department of nuclear medicine. To obtain additional information concerning special tumor types, we performed a separate examination of two different tumor types with a higher number of patients: breast cancer and head/neck cancer. Breast cancer The group with breast cancer consisted of 20 patients. Subsequently, bone metastases could be verified in eight of these patients. Six cases could be identified clearly by bone scan, in one patient additional plain radiographic analysis was recommended to confirm the diagnosis and one patient was false positive (focal area of increased tracer uptake in the shaft of the femur, identified as local necrosis of the bone by plain radiography and confirmed by follow-up). For bone markers the following results were observed: ICTP: sensitivity: 63%, specificity: 25%. PINP: sensitivity: 25%, specificity: 100%. Figure 8 shows an example of a patient with osseous metastatic disease and normal values for ICTP and PINP. Head and neck cancer 19 patients with head/neck cancer were examined (4 female, 15 male). In these patients determination of whether local osseous structures are involved is essential for the preoperative planning of further treatment. In three patients osseous structures were affected by the tumor and all of these were correctly diagnosed by bone scintigraphy. ICTP and PINP were right positive in two separate patients. Sensitivity for PINP and ICTP was 33%, specificity for ICTP was 56% and for PINP 94%. Discussion Metastatic bone disease is a serious clinical problem. Complications associated with osseous metastases are pain, fractures, spinal cord compression, paralysis, etc., which lead to a significant reduction in the quality of life of tumor patients. Several diagnostic tools are available to detect bone metastases, including plain radiography, computer tomography, magnetic resonance imaging and bone scintigraphy. Of these, radionuclide bone scanning using Tc-99m labeled diphosphonates is the most widely used and accepted method for the detection of bone metastases. Bone scintigraphy can detect osseous metastases several months before changes in plain radiographs can be seen, thus making bone scintigraphy an excellent diagnostic tool. However, this method is expensive, is not always available in every hospital and has the disadvantage of showing a positive reaction even to bone inflammation, degenerative changes and fractures or flare reaction, which leads to a reduced specificity. This is the reason why many authors have described outcomes regarding diagnosis of bone metastases and observation of the clinical course using markers of bone turnover. In recent years, there have been important advances in the field of biochemical markers of bone turnover and new methods have emerged [ 8 ]. Measurement of the metabolites of type I collagen, the predominant collagen in bone, has been reported to be useful for monitoring bone turnover in many different disorders, including diseases with bone metastases [ 9 ]. Yoshida et al. reported on the serum concentration of type I collagen metabolites as a quantitative marker of bone metastases in patients with prostate cancer [ 10 ]. They concluded that, especially in patients with high grade carcinoma cells, the determination of bone metabolic markers should be more useful in evaluating metastatic spread to bone than prostate specific antigen. In our collective, patients with prostate cancer (n = 5) showed the highest correlation between the presence of osseous metastases and elevated markers for PINP and ICTP. Only in one case we found a false positive ICTP. However, due to the small size of our patient groups it prevents any real conclusive statements from being made. For other tumor types we did not observe a similar high correlation. Horiguchi et al. reported on the usefulness of ICTP as a marker for bone metastases in patients with lung cancer [ 11 ]. He suggested that measurement of ICTP is an excellent serological diagnostic method for identifying bone metastases in patients with lung cancer and can also help predict when it might be useful to undertake other examinations like bone scintigraphy. In our study however, we saw a high rate (11 out of 19) of false positive results of increased ICTP levels in this group of patients. One reason for this might be the presence of non-detectable micro-metastases at the time of ICTP measurement. To avoid the possibility of false results, all patients had a follow-up examination between 9 and 14 months, during which micro-metastases would have become apparent. None of the 11 patients developed osseous metastases during this time period so that the likelihood of the presence of such metastases was very low. Another major group of patients in our study were females with breast cancer. The literature on these tumors and the value of bone metabolic markers for detection of osseous metastases is very controversial. Blomqvist et al. reported a positive and significant correlation between ICTP and PICP and the number of bone metastases, plus Wada et al. suggested that ICPT might be a useful marker for screening and monitoring bone metastases in breast cancer [ 12 , 13 ]. In contrast Ulrich et al. showed in a study with 106 patients that the sensitivity for diagnosing bone metastases was 65% [ 14 ]. These results are more similar to our study. However, Ulrich reported a high specificity of 91%, whereas we observed only 25% specificity for ICTP. The types of tumors in our study were very heterogeneous, which seems to reflect the circumstances seen on a daily basis. Due to the fact that tumors can metastasize to bone in different ways (osteoblastic and/or osteolytic) we established parameters for both possibilities. For PINP, the marker for osteoblastic activity, the specificity was high but with a poor sensitivity. For ICTP, the marker for osteolytic activity, both characteristics of sensitivity and specificity were low, such that a general recommendation for the use of this marker as a screening parameter cannot be made. Potential indications for bone metabolic markers might be the therapy control in patients with bone metastases, in which increased parameters had been proven before therapy. Particularly in these patients, the so-called "bone flare reaction", the repair of destroyed bone structures by tumor cells, complicates the assessment of bone scintigraphy. Increased activity in known osseous metastases after or during therapy can be caused by either repair or further tumor growth. Blomqvist and colleagues and Koizumi et al. reported on patients with breast cancer and osseous metastases, where only patients with progressive disease showed an increase in ICTP values during therapy compared to other patients with response to therapy [ 9 , 15 ]. Another indication might be in benign bone disorders like rheumatoid arthritis or Paget's disease. Aman et al. reported a correlation between ICTP values and the disease progression of patients with rheumatoid arthritis [ 16 ]. Conclusions In summary, when a new method is recommended for the diagnostic work up, it is necessary to demonstrate that this new modality is as sensitive and specific as existing routine imaging procedures. The determination of bone metabolic parameters like ICTP or PINP is less expensive than bone scanning, but this prospective study has shown that the results from bone metabolic markers are not yet sufficient enough to demonstrate a bone involvement in different type of malignancies with high sensitivity and specificity. Competing interests The author(s) declare that they have no competing interests. Authors' contributions JS and CE designed the study, performed the statistical analysis and drafted the manuscript. SR and EW recruited patients and carried out the nuclear medicine investigations. All authors read and approved the final manuscript. Figure 6 Bone scan of a 59-year-old female with non-small-cell lung cancer and multiple osseous metastases. Both parameters are increased ICTP: 13 μg/L (1.6–5.3) PINP: 113.8 μg/L (19–102) Figure 7 Bone scan of a 47-year-old male with non small cell lung cancer. Bone scintigraphy and follow-up showed no evidence of osseous metastatic disease. ICTP: 7.3 μg/L (1.4–5.2) and PINP: 102 μg/L (21–78) are above the upper reference limit. Figure 8 55-year-old female with metastatic breast cancer. On the bone scan multiple osseous metastases can be seen especially in the spine, pelvic region and calotte. ICTP: 4.8 μg/L (1.6–5.3) and PINP: 31.5 μg/L (19–102) are within the reference limit. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC539349.xml
524481	Control and maintenance of mammalian cell size	Background Conlon and Raff propose that mammalian cells grow linearly during the division cycle. According to Conlon and Raff, cells growing linearly do not need a size checkpoint to maintain a constant distribution of cell sizes. If there is no cell-size-control system, then exponential growth is not allowed, as exponential growth, according to Conlon and Raff, would require a cell-size-control system. Discussion A reexamination of the model and experiments of Conlon and Raff indicates that exponential growth is fully compatible with cell size maintenance, and that mammalian cells have a system to regulate and maintain cell size that is related to the process of S-phase initiation. Mammalian cell size control and its relationship to growth rate–faster growing cells are larger than slower growing cells–is explained by the initiation of S phase occurring at a relatively constant cell size coupled with relatively invariant S- and G2-phase times as interdivision time varies. Summary This view of the mammalian cell cycle, the continuum model, explains the mass growth pattern during the division cycle, size maintenance, size determination, and the kinetics of cell-size change following a shift-up from slow to rapid growth.	Background Conlon and Raff have described experiments that they claim casts doubt on a basic assumption regarding the way mammalian cell size is maintained during proliferation [ 1 ]. The key question studied by Conlon and Raff asks, "How do cells maintain a constant cell size and cell size distribution during extended cell growth?" In a cell culture growing over many generations, the cell size distribution neither varies nor broadens. Cells do not get progressively larger nor do they get progressively smaller. One formulation of this result is that cell mass increase is regulated during the cell cycle so that there is no disparity between the rate of cell mass increase and the rate of cell number increase. Total cell number and total cell mass increase in parallel during unlimited exponential growth. If there were any disparity or disproportion in the rate of mass and cell number increase, cells would get either larger or smaller during extended growth. In an article accompanying the work by Conlon and Raff [ 2 ], a quote by Robert Brooks (Kings College, London) sums up the problem: "If [cell] growth is exponential, then cells must have a size control over division, since otherwise random differences in size at division would increase continuously from generation to generation. This does not happen. Conversely, if growth is not exponential, then such a size control is not necessary." This quote from Brooks may be thought of in this way. Consider two newborn cells of slightly different size. Exponential growth means that cell mass would be made in proportion to the extant cell mass. The larger cell would increase its mass at a more rapid rate than the smaller cell. When the cells divide, the dividing cell produced by the initially larger cell would be even larger compared to the dividing cell produced by the initially smaller cell. Given equipartition of cell mass at division, the new daughter cells would have an even more disparate size difference. Exponential growth in the next cycle would again lead to larger differences in cell size than in the previous cycle. According to this reasoning, the cell size distribution would grow increasingly broader. Since this is not observed, Conlon and Raff propose that either a cell must grow "linearly," or if a cell grows exponentially the cell must have a cell size control system. This reasoning implies that if cells grow linearly, then no cell size control system is required. The experiments of Conlon and Raff [ 1 ] are presented as supporting linear cell growth. Linear cell growth postulates that there is a constant mass increase during each time interval of the cell cycle. Furthermore, comparing their results on mammalian cells to what is referred to as the "yeast" model of cell size control, Conlon and Raff [ 1 ] conclude that mammalian cells have a different mechanism for cell size control. As Conlon and Raff summarize their experimental conclusion: "We show that proliferating rat Schwann cells do not require a cell-size checkpoint to maintain a constant cell size distribution, as, unlike yeasts, large and small Schwann cells grow at the same rate, which depends on the concentration of extracellular growth factors." A reanalysis of the experiments and reasoning of Conlon and Raff, presented here, leads to a very different view of cell size control and cell size maintenance. It is first shown that there is no problem with either linear or exponential mass increase for size maintenance. Size maintenance does not depend on which pattern of cell mass increase occurs within a cell cycle. The preferred–and experimentally and theoretically supported–pattern of mass increase during the division cycle is exponential growth or exponential mass increase. An exponential growth pattern poses no problem for size maintenance. Constancy of cell size is fully compatible with an exponential pattern of mass increase as well as the hypothetical linear pattern of mass increase. No major difference between the size control systems of yeast, mammalian, or bacterial cells need be postulated to account for the constancy of cell size during the growth of cell cultures. In contrast to the proposed absence of a cell size control system in mammalian cells, it is shown that mammalian cells do have a very simple size control system. The formal elements of this system are similar to that found in the control of the bacterial cell cycle. Discussion Cell size maintenance with exponential and linear mass increase Can cells grow exponentially during the division cycle and maintain a constant cell size? Consider two possible cases of exponential growth for cells with variable cell sizes. For the first case (Fig. 1a ), three cells of the same newborn size have slightly different rates of mass increase. If all three of the cells in Fig. 1a were to have the same interdivision time, the dividing cells would have disparate sizes. But if the interdivision times vary so that cells divide at the same cell size, then cell size is maintained even with exponential growth during the division cycle. A newborn cell that makes mass at a rate slightly faster than average will divide earlier than cells with an average or below average rate of mass increase (Fig. 1a ; arrows indicate division times). Conversely, a newborn cell producing mass at a rate slightly slower than average will divide later than cells with an average or above average rate of mass increase (Fig. 1a ). Variation in interdivision times allows maintenance of constant average cell size even with exponential mass increase. A second case (Fig. 1b ) starts with different sized newborn cells that synthesize mass at the same rate. As in Fig. 1a , the earlier a cell reaches the division size, the earlier the cell will divide and the cell will have a shorter interdivision time. Size constancy is maintained even though mass increases at a constant rate for the three cells with different-sized newborn cells (Fig. 1b ). Mixtures of initial size variation and variation of rates of mass synthesis can be analyzed in the same manner; the analysis is strongly supported by a reanalysis of published experimental data on the variation of mammalian cell interdivision times as determined by time-lapse cinematography [ 3 ]. Figure 1 Exponential and linear growth patterns are both compatible with cell size maintenance. In panel (a) newborn cells of identical size increase mass at slightly different rates with an exponential pattern of mass increase. If cells divide at a constant cell size, here size 2.0, size will be maintained even though the rate of mass increase varies. This occurs as the cells divide at different times (division indicated by the downward arrows) as they reach the same size. In panel (b), exponential growth at identical rates from initial cells of different cell sizes also gives size maintenance as cells divide at the same size because there are different interdivision times for each cell; the larger initial cells have a shorter interdivision time and the smaller initial cells have a longer interdivision time. As shown in panels (c) and (d), linear cell increase (note the different ordinate scale compared to panels (a) and (b)) can also lead to cell size maintenance as cells divide at the same size, 2.0. Linear cell growth during the division cycle (Figs. 1c and 1d ) can also produce size maintenance. Whether cells reach the division size earlier due to a larger initial cell size or due to a more rapid rate of mass increase, the cell size at division can be the same for all cells. Thus size maintenance is also consistent with linear growth. The patterns shown in Figs. 1a,1b,1c,1d show that there is no impediment to size maintenance as long as interdivision times are not invariant. In all four panels in Fig. 1 the interdivision time varies depending on the time required for a newborn cell to reach a particular cell size. To be precise, it is not proposed that cells always divide at "exactly" the same size. There is a statistical variation in mass increase and interdivision times that can lead to variations in cell size at division [ 3 ]. The important point is that when cells deviate from the mean size there is a return to the mean size through compensating interdivision times during the next cell cycle. Large cells will have a relatively shorter interdivision time, leading to a return to the average cell size. Further, it is not proposed that a large, newborn cell "controls" its mass increase to have a slower rate of mass increase (compared to smaller cells) thus compensating for the initial larger cell size. Nor do small cells increase their rate of mass synthesis to compensate for their initial mass deficit. Mass increase variation is postulated to have some inherent statistical variation [ 3 ] but with all cells, no matter what their extant size, having the same relative rate of mass increase. There can be variability in cell mass increase with the rate of mass increase being independent of cell size. A large newborn cell could have a faster than average rate of mass increase. In this case, the interdivision time would be even shorter to compensate for both the larger initial newborn cell size and the greater than average rate of mass increase. The size maintenance pattern is illustrated in Fig. 2 , where the production of large and small cells can arise either by variation in interdivision times or deviations from equipartition. Newborn or baby cells (b) that have a relatively short interdivision time produce small (s) cells while newborn cells that have a relatively long interdivision time produce large (l) cells. Large and small cells may also be produced by deviation from equipartition at division so that an average-sized dividing cell produces one large (l) and one small (s) cell. The return of small and large cells to the average cell size occurs in the next generation by variation in interdivision times so that small cells (s) have a longer (on average) interdivision time than larger (l) cells (on average). Figure 2 Interdivision time variation allows a return of slightly deviant sizes to a constant cell size. In panel (a) newborn "baby" cells (b) grow for slightly different times, producing either large (l) or small (s) newborn cells from large or small dividing cells. Deviation from equipartition for an average sized dividing cell can also produce large and small cells. The resolution of size differences is illustrated in (b) and (c) where the larger cell (l) has a shorter interdivision time dividing at average (a) cell size and the smaller cell (s) has a longer interdivision time also dividing at average cell size. The dividing average sized cells (a) produce newborn baby cells (b) of the original newborn size. It may appear that this simple analysis is merely begging the question by not indicating how the cell "knows" to divide at a particular size. This question will be answered below. But first, two issues should be dealt with. An initial discussion will clarify the relationship of cell mass increase to cell size. This will be followed by a discussion of problems with the proposal of linear growth during the division cycle. What is meant by the proposal that large cells grow faster than small cells? What is meant by the Conlon and Raff proposal that, in yeast culture, large cells grow faster than small cells? And conversely, that in mammalian cells, large and small cells grow at the same rate? There are four different meanings that can be given the notion of the rate of mass increase and its relationship to cell size. These different meanings lead to some verbal confusion that requires clarification. One meaning of the proposal that large cells make mass faster than small cells is that given two cells of disparate sizes, the absolute rate of increase in cell mass is greater in the larger cells. A cell of size 2.0 might add, in some time interval, 0.2 units of cell mass, while a cell of size 1.0, in that same time interval might add only 0.05 units of cell mass. This pattern is a clear and unambiguous difference in the rate of mass increase that is related to cell size. A second meaning of cell size affecting the rate of mass increase considers that a cell of size 2.0 adds 0.2 unit of mass and a cell of size 1.0 adds 0.1 unit of cell mass over the same time interval. Of course, this case could arguably be said to be a constant rate of mass increase, as the rate of mass increase is proportional to the amount of extant mass. This second proposal is equivalent to mass increasing exponentially. This is because as extant mass changes during the cell cycle the absolute rate of mass increase also changes to reflect the newly added cell mass. After the cell of size 1.0 grows to size 1.1, in the next time interval, rather than 0.1 units of mass being added, there are 0.11 units of mass added to the cell mass. Just as interest is compounded in a bank account, and the funds grow exponentially, so mass in this second example increases exponentially. A third meaning of the variation in mass increase with cell size is that the rate of mass increase is determined at birth and continues throughout the cell cycle, unaffected by continued cell size increase. A relatively small newborn cell could have a rate of addition of "X" units per time interval, and this rate would remain constant even as the cell increases its cell mass. The larger cell would add more than "X" units each time interval and not change this rate during the cell cycle. This pattern of increase would be called linear synthesis during the cell cycle. It is interesting to think about these different meaning when considering the theoretical graph drawn by Conlon and Raff [ 1 ] to illustrate the return of cells of disparate sizes to the same cell size. As shown in Fig. 3 (redrawn from Fig. 1 of Conlon and Raff [ 1 ]), consider two cells, one of size 1.0 and one of size 10.0. During one generation of growth 5.5 units of mass are added by the smaller cell to produce a dividing cell of size 6.5, and 5.5 units of mass are also added to the larger cell to produce a dividing cell of size 15.5. As discussed by Conlon and Raff, upon cell division the daughter cells produced by this pattern of growth would be sizes 3.25 and 7.75. Repeating this each generation (5.5 units added to each cell independent of the extant newborn cell mass) leads, according to Conlon and Raff, to a convergence of cell size as shown in Fig 3 . Figure 3 Hypothetical model of Conlon and Raff where constant size increase independent of cell size allows return of deviant cell sizes to a constant cell size over time (This figure is drawn directly from Conlon and Raff (Conlon and Raff, 2003)). The assumption made for this figure is that both large and small cells increase their mass equally over time. Thus, a cell of size 10 and a cell of size 1 increase their mass over one doubling time by 11 units (the sum of the starting masses, 10 and 1). To the large and small cell an increase of 5.5 units of mass is proposed to occur as cells grow. Thus, the large cell grows to size 15.5 and the small cell grows to size 6.5, and at division the daughter cells now have sizes of 7.75 and 3.25 respectively. This continues for a number of generations as the founder cells, originally of disparate sizes, now converge to the same size. But no indication of the length of the division cycles is given in Fig 3 . If the interdivision times are the same for the large and small cells, which is implicit in, and not excluded by, the analysis in Fig 3 (Conlon and Raff's Fig 1), the relative rate of mass increase for the larger cell is 5.5/10.0 or 0.55 and the relative rate of mass increase for the smaller cell is 5.5/1.0 or 5.5. From this point of view, the ratio of the rates of mass increase is a factor of 10, with the smaller cell making mass from its mass at 10 times the rate (relative to extant mass) compared to the larger cell. But if the absolute rates of mass increase were the same, then the smaller cell would have a much longer interdivision time than the larger cell. If, over a unit time, 1.0 unit of cell mass were added to the larger newborn cell, and that cell divided at size 11.0, then the interdivision time would be that unit time. The smaller cell, however, would require 10 time units for its interdivision time, because that is the time required to reach size 11.0 as 1.0 unit of material is added to each small cell during each unit of time. The smaller cell grows for a longer time before division. After this first division the new daughter cells produced by each of the initial cells would be the same size. By allowing interdivision time variation, cell size uniformity is restored in one generation. A similar analysis can be made for exponential mass increase (i.e., mass added proportional to extant mass). If the cell of size 10.0 added 1.0 unit of mass in a unit of time, then the small cell would add 0.1 unit of mass in that same time interval. In this case, there would be even more time required for the small cell to reach the division size of 11.0. In any case, exponential growth coupled with interdivision time variation can allow size maintenance because both the large and the small newborn cells will divide at the same size, as described in Fig. 1 . Of course, the example given by Conlon and Raff (Fig 3 ) as discussed here is unrealistic. Cell sizes do not vary over a factor of 10 in exponential culture. But this re-analysis of Fig. 3 illustrates the power of considering different interdivision times as a factor in maintaining constant cell size. Robert Brooks (personal communication) notes that in some of his experiments cell size is observed as very variable. He states that in experiments with Shields they found that size varied over a range of at least 6-fold. In response it can be pointed out that recent careful measurements of the size variation during the division cycle of cells grown under ideal conditions indicates that size variation is not broad [ 4 - 7 ]. Helmstetter (personal communication) points out that when cells are not grown under optimal conditions there are always some cells of odd or abnormal size. But these cells are cells that are dying or in some way impaired. These abnormally sized cells should not be considered as typical of the cells in a well-maintained, exponentially-growing cell culture. The fourth part of our verbal analysis of mass increase as a function of cell size relates to bacterial cells. As will be seen below, one of the most important results in bacterial physiology is that as growth rate speeds up, cells get larger. As the growth rate of a cell is continuously varied by increasing the richness of the medium, there is a continuous variation in bacterial cell size with the faster growing cells being larger than slower growing cells [ 8 ]. Bacterial cells with an interdivision time of 20 minutes are larger than cells with an interdivision time of 60 minutes. Although it could be said that larger cells make mass faster leading to the shorter interdivision time for larger cells, it is equally possible, and in fact preferred, to reformulate or verbalize this result by saying that faster growth produces larger cells. For a given medium the rate of mass increase is determined for the bacterial cells, and the cell size results from the growth rate. This idea, the fourth way of looking at the relationship of cell size and mass increase, will be illustrated below in the analysis of bacterial patterns of DNA replication and cell size maintenance. As we shall see, in bacterial cells a constant period for DNA replication and a constant time between termination of replication and cell division explains the variation in bacterial cell size as a function of growth rate. This same explanation also applies to mammalian cells: the rate of growth determined by external conditions determines cell size. Rather than taking the results of Conlon and Raff and concluding that larger cells when placed in medium with more serum now grow faster, it is better, as with bacteria, to say that when cells are placed in a condition that provides faster growth (i.e., a shorter interdivision time), the cells grow larger. While this may appear, at first sight, to be a trivial and semantic difference, this distinction actually lies at the heart of the problem and is the key to the solution of size determination and size maintenance. Rather than thinking that cell size produces cells with a particular growth rate (e.g., large cells grow fast), it is preferable to think that a particular growth rate produces cells of a particular size (e.g, fast growing cells are made larger than slower growing cells). What is wrong with linear growth? There are problems inherent and unavoidable in any proposal of linear cell growth during the division cycle. Linear growth means that during the division cycle, as a cell proceeds from size 1.0 to size 2.0, cell mass is added at a constant amount per unit time. If a cell grows linearly, over tenths of a cell-cycle time, a cell increases its size from size 1.0, to 1.1, to 1.2, and so forth. The main problem with linear growth (i.e., constant amounts of cell mass are added at constant time intervals) is that as the cell gets larger, the cytoplasm becomes inefficient. Inefficiency is defined here as producing less mass per extant mass compared to more efficient use of the extant mass. Efficient mass increase would exist when extant cell mass makes new mass as fast as possible. As a cell grows, more cytoplasm is present. With linear growth the extra cytoplasm does not increase the absolute rate of cell mass synthesis. In essence, the new cytoplasm does not work to make new mass. There is a decrease in the relative rate of mass increase (i.e., mass synthesis per extant mass) which means that the ribosomes, after some growth, are not working as efficiently as before there was growth. One mechanistic model explaining the postulated absence of a change in the absolute rate of mass increase to produce linear growth is to propose that the new mass does not enter into active participation in mass synthesis until a cell division; new mass will not be "activated" to enter into mass synthesis until the next division. From this viewpoint, there is a constant rate of mass increase based on the original mass. As a cell approaches division, the efficiency of mass making new mass tends toward half that of the efficiency of the initial, newborn cell mass. An alternative mechanistic proposal to explain linear growth is that during a cell cycle the amount of material able to be taken up by a cell is constant, and only upon cell division is there an "activation" of the new cell surface so that there is an increase in the ability of the cell to take in material. Even more important and troublesome is the result that if a cell grows linearly, at the instant of cell division there must be a sudden saltation or jump in the synthetic activity of the cytoplasm. Toward the end of the cell cycle, 1.9 units of cell mass make 0.1 unit of cell mass to achieve a cell mass of 2.0. Given linear growth, at the instant of division the 2.0 units of cell mass, now apportioned into two daughter cells, must now make, during the next time interval, 0.2 units of cell mass or twice as much as in the previous time interval. When the cell of size 2.0 divides, linear growth implies that the two new daughter cells now immediately activate the "quiescent" cytoplasmic material (or activate the previously inert cell surface uptake capabilities). Irrespective of mechanism, considering the two daughter cells together, linear growth during the division cycle inevitably implies that at division there is a sudden doubling in the rate of mass increase. There is no known biochemical mechanism for these proposals to produce linear cell growth, or the sudden jump in the rate of mass increase. As currently understood, the new cytoplasm joins right in to make new mass. And there is no mechanism known to allow new cell surface to remain inert until a cell division. While the absence of any identification of these mechanisms does not mean that these mechanisms do not exist, there is no need to propose the existence of these mechanisms if cells grow exponentially. The experimental evidence favors exponential mass increase during the cell cycle. In bacteria the evidence for exponential growth is extremely strong [ 9 , 10 ]. Analysis of data on eukaryotic cell size increase also supports exponential growth during the division cycle [ 11 ]. What of the experiments presented by Conlon and Raff [ 1 ] that cell mass increases linearly? Conlon and Raff studied cells cultured in 1% fetal calf serum, forskolin, and aphidicolin. Aphidicolin is an inhibitor of DNA synthesis. While mass increased, there was no concomitant increase in DNA. The cells were incubated for 216 hours (9 days). The cell volume was measured using a Coulter Counter, although in one experiment total protein content was measured. Conlon and Raff realized that it is extremely difficult to distinguish linear from exponential growth over one doubling time. Therefore they measured mass increase over a longer period of time (approximately 3 or more normal interdivision times). The problem with this experiment is that the inhibited cells do not allow an exponential increase in cell number as DNA synthesis is inhibited. Therefore the experiment is subject to the critique that aphidicolin inhibition produced the observed results. The results may not, and very likely do not, reflect the situation in normal, uninhibited, and unperturbed cells. For example, there could have been exponential growth during the first "virtual cell cycle". Then the limitations of DNA content would lead to the observed linearity of growth as measured over the extended period of analysis. But this linearity should not be taken as an indication that during the normal cell cycle the cell mass increases linearly. Even if cells grow linearly during the division cycle, if the rate of mass increase is measured over a number of cell cycles with uninhibited cells, then a priori there should be evidence of an approach to exponential mass increase. If the rate of mass increase during the first cycle is 1.0, during the second cycle it should be 2.0, during the third cycle 4.0, and so on. Thus, even on its own terms, with linear mass increase during the division cycle, the experiments of Conlon and Raff [ 1 ] on the pattern of mass increase are flawed by the presence of an inhibitor of DNA synthesis. An analysis of this idea is presented schematically in Fig. 4 . Figure 4 Approach of cell mass to exponential even if cells had linear synthesis within cell cycle. Panel (a) illustrates cells dividing to produce two, four or eight times the original number of cells (thick line is cell number). The mass (thin line) increases linearly. It is clear that the cell size will not be maintained. In panel (b), even with linear mass growth within the cell cycle (thin line), as cells divide the rate of mass synthesis doubles and then quadruples as cell numbers increase. It is not proposed that mass increases linearly, but merely that even linear synthesis should exhibit, in an uninhibited situation, exponential mass growth. Raff (personal communication) disputes this interpretation of the aphidicolin experiments, proposing that "while the aphidicolin-arrest strategy is certainly artificial, it is not unrealistic...as many cells, including Schwann cells, grow a great deal after they have stopped dividing. Moreover...hepatocytes grow linearly, independent of their size, if a mouse is re-fed after it has been starved for a couple of days." As noted in Fig. 4 , without inhibition, growing cells that grow and divide must, a priori , approach an exponential pattern (i.e., rate of 1, to 2, to 4, to 8 as cells multiply), and therefore the only meaningful discussion of the linear vs. exponential growth pattern relates to growth within the cell cycle. Regarding application of liver growth following starvation and refeeding, this complex situation seems particularly inapplicable to discussions of cell growth in cell culture as there are so many complicating factors. A detailed analysis of the proper systems for cell-cycle analysis has been presented [ 4 ]. The experiments of Conlon and Raff also show some internal inconsistencies that weaken the actual data. A comparison of cell volume increase and protein per cell increase in the same cells over a 96 hour period (Fig.3 of Conlon and Raff) shows that the volume increase was 4.75-fold (~2,000 μm 3 /cell increasing to ~9,500 μm 3 /cell) but the protein increase was only 2.93-fold (~0.16 ng/cell increasing to ~0.47 ng/cell). Until these differences are resolved, it is difficult to accept these experiments as supporting linear cell growth–or any other pattern of cell growth–during the normal division cycle. The discrepancies pointed out here suggest that the quantitative measures of cell size by Coulter Counter may not be able to distinguish different growth patterns. Another problem arises in Conlon and Raff's [ 1 ] analysis of the pulse-chase experiments where cells starved for different times are pulsed and chased to measure protein turnover. They concluded: "...the rate of decrease in radiolabeled protein increased as the cells increased in size." That is, there was a greater release of labeled amino acids from cells that were inhibited with aphidicolin for longer periods of time and which were therefore larger [ 1 ]. But the release data were plotted on rectangular coordinates. This led to the observation that the slope between the 0 hour and 2 hour points in their Fig.4 is steeper for the cells arrested for 72 hours compared to the cells arrested for 48 hours. The 72 hour cells were larger than the 48 hour cells. But considering the actual values, and reading the results from the published graph, the counts for the 72 hour arrested cells went from ~179 to ~121 in two hours, or a ratio of 0.67 for the two hour chase. The 48 hour arrested cells went from ~138 to ~94 for a ratio of 0.68 for the two hour chase. Thus, in contrast to the conclusion of Conlon and Raff [ 1 ] there is no apparent difference in the turnover of proteins as a function of cell size. Robert Brooks (personal communication) has argued against this analysis, noting that the cells starved for 24 hours appeared to show "no turnover" as the line for this graph (Conlon and Raff's Fig.4) was flat. But in the text in the legend to their Fig.4(b) Conlon and Raff state, "The shallowness of the curve for the 24-hour-arrested cells is likely to be the result of the lower than expected value at 0 hours." This explanation comes from the initial counts in Fig.4(a) where it can be seen that there is some apparent error in the zero time value for the 24 hour starved cells in their Fig.4(b). But an even more egregious error in analysis precedes even these technical problems. The cells studied by Conlon and Raff were not synchronized. The cells were not aligned and were in all phases of the cell cycle. Theoretically, it is impossible to determine the pattern of mass synthesis during the cell cycle on cells that are not synchronized. (For complete details see [ 12 ]). This is because of the age distribution of cells in a growing culture. The age distribution for growing cells in culture is given by 2 1-X , where X is the cell age during the cell cycle; X varies between 0.0 and 1.0 (newborn cells are age 0.0 and dividing cells are age 1.0). At age 0.0 the relative number of newborn cells is 2.0 (2 1-0 = 2 1 = 2) while the relative cell number of dividing cells is 1.0 at age 1.0 (2 1-1 = 2 0 = 1). This distribution of cell ages means that any incorporation measurement on asynchronous cells must, and will, yield an exponential pattern of uptake. This is illustrated in Fig. 5 for an idealized case where we imagine cells making all of their cytoplasm only at age 0.5. Because of the age distribution an exponential pattern of incorporation is observed when the entire culture is analyzed (Fig. 5a,5b ). The details of the analysis are presented in the legend to Fig. 5 . If the cells had been synchronized then one would have measured a peaked pattern as illustrated in Fig. 5c . Figure 5 Unsynchronized cells cannot be used to determine cell-cycle pattern of synthesis. Panel (a) shows a series of age distributions starting with the initial age distribution reflecting the pattern Age Distribution = 2 1-X , where X is the cell age going from 0.0 to 1.0. In this Gedanken analysis, it is assumed that cells of age 0.5 (i.e., cells in mid-cycle) are the only cells incorporating amino acid (cross-hatched bars). The asterisk (*) on a bar in each pattern indicates the newborn cells. One reads the cell ages by going from the asterisked bar to the right and then back to the left to finish off the age distribution. The number to the right of each pattern is the relative number of cells incorporating amino acid. Thus, in the uppermost pattern in Panel (a) the relative number is 1.46. After one-tenth of a generation we see that the oldest cells in the first pattern have divided to give double the number of cells and these cells are now the youngest cells in the culture. All of the other cells move up one-tenth of an age so that the cells that were age 0.4 are now age 0.5 (cross-hatched bar) and the rate of synthesis increases to 1.57. This is because there are more cells in the original culture of age 0.4 than there were of age 0.5. Continuing down the patterns in Panel (a) we see that as cells move to age 0.5 there is a continuous, and exponential, increase in the radioactivity. The cells above age 0.5 (in the original topmost diagram) divide and produce two cells each tenth of a cell cycle, so that over one total cell cycle there is an exponential increase in the rate of amino acid incorporation (a measure of cytoplasm increase). The total pattern of incorporation is plotted in panel (b) where the exponential incorporation during one cell cycle is indicated. Panels (a) and (b) thus show that even with a non-exponential pattern of incorporation, if a total culture is studied, the measured incorporation pattern will be exponential. If, however, cells are truly synchronized, as illustrated in Panel (c), a peaked incorporation pattern is observed, accurately reflecting the mid-cycle incorporation of amino acids into the cells at a particular cell-cycle age. Starting with newborn cells at age 0.0 and moving through the cell cycle at one-tenth of an age each pattern in (panel c) the incorporation (noted by the numbers to the right of the diagrams (panel c) shows a peaked pattern. Robert Brooks (personal communication) argues that this critique is incorrect because "they [ 1 ] started with quiescent (G0/G1) cells." Quiescent cells with a G1-phase amount of DNA are not synchronized [ 13 - 15 ]. The reader is referred to these papers for a detailed analysis. Despite the widespread belief and acceptance that cells can be synchronized by growth arrest (i.e., by whole-culture synchronization methods), this idea is incorrect. Cells can only be synchronized by selective methods [ 15 ]. How can one determine whether mass increases exponentially or linearly during a normal, unperturbed, division cycle? To illustrate one approach to determining the pattern of mass increase during the division cycle, consider the following experiment. Grow cells for many generations in a radioactive amino acid (e.g., C-14 labeled amino acid) so that cell protein is totally labeled. Then add a pulse of a counter-labeled amino acid (e.g., H-3 labeled). As shown in Table 1 , if cells grow linearly, the ratio of tritium (H-3) to C-14 should decrease as the cells become larger. With exponential growth the ratio of tritium to C-14 should be constant over the cell cycle. If one now one took such double-labeled cells, fixed them, and spread the cells out on a gradient such that the larger cells were preferentially at the bottom and the smaller cells at the top, if cells grew linearly there would be a decrease in the H-3/C-14 ratio as the larger and larger cells were assayed. If cells grew exponentially there would be a constant radioactivity ratio over the entire set of cell size fractions. The idealized results from Table 1 are illustrated in Fig. 6 . Table 1 Analysis of linear and exponential growth by comparing long-term and short-term isotope incorporation. LINEAR GROWTH EXPONENTIAL GROWTH Cell size Size increase Inc/size Cell age Cell size Size increase Inc/size 1 0.1 0.100 0 1.00 0.07 0.072 1.1 0.1 0.091 0.1 1.07 0.08 0.072 1.2 0.1 0.083 0.2 1.15 0.08 0.072 1.3 0.1 0.077 0.3 1.23 0.09 0.072 1.4 0.1 0.071 0.4 1.32 0.09 0.072 1.5 0.1 0.067 0.5 1.41 0.10 0.072 1.6 0.1 0.063 0.6 1.52 0.11 0.072 1.7 0.1 0.059 0.7 1.62 0.12 0.072 1.8 0.1 0.056 0.8 1.74 0.12 0.072 1.9 0.1 0.053 0.9 1.87 0.13 0.072 2 1 2.00 The center, bold-faced, column lists the cell ages from 0 to 1.0. At the left the linear increase of mass is related to the absolute increase in mass per interval (0.1 each interval for linear increase in mass during division cycle), and the ratio of incorporation per extant cell mass is given in the third column (0.1 to 0.053). Similar results for exponential growth except the mass increase per interval goes from 0.07 at the start of the division cycle to 0.13 at the end. The ratio of incorporation per extant mass in the right-most column is thus constant. Figure 6 Comparison of the ratio of pulse label to total label for exponential and linear patterns of mass increase as described in Table 1. To summarize this critique of the aphidicolin-inhibition results, the experiments of Conlon and Raff do not measure the mass increase during the cell cycle. The experiments using inhibition of DNA replication merely measure the pattern of mass increase in a perturbed experimental situation on cells that are not synchronized. This experiment is not supportive of any particular pattern of mass increase during the normal division cycle. More important, as shown in Fig. 5 , without synchronization of cells, it is impossible to determine the pattern of mass increase during the division cycle. The bacterial cell cycle: Rules, patterns, and regulation This analysis presented here explicitly deals with animal or eukaryotic cells. However, it is relevant to bring to bear on this problem the experience and results obtained regarding cell-size determination in bacteria. In 1968 the rules for the replication of DNA in a simple bacterium ( Escherichia coli ) as well as the relationship of cell size to control of DNA replication were worked out [ 16 - 20 ]. The pattern of DNA replication and cell size are determined by three rules: 1. A round of DNA replication is invariant (40 minutes) over a wide range of growth rates [ 16 - 19 , 21 ]. 2. The time between termination of replication and cell division is invariant (20 minutes) over a wide range of growth rates [ 16 - 19 , 21 , 22 ]. 3. At the time of initiation of replication, the cell mass per origin is a constant [ 16 , 20 , 23 ]. These rules are illustrated in Figs. 7 and 8 . These three rules predict (Fig. 8c ), that cell size should be a logarithmic function of growth rate. Cell size plotted on semi-logarithmic coordinates against the reciprocal of the interdivision time (i.e., the growth rate) gives a straight line. Faster growing cells are larger than slower growing cells. Ten years earlier, in 1958, before the rules predicting the size-growth rate relationship were determined, this experimental result [ 8 ] was clearly obtained in what has been called "the Fundamental Experiment of Bacterial Physiology" (Cooper, 1991)[ 12 ]. An analysis of the history, origins, and meaning of this experiment has been published (Cooper, 1993)[ 40 ]. Figure 7 Diagram of patterns of DNA replication during the division cycle in bacteria. The different patterns go from an infinite interdivision time (i.e., essentially no or extremely slow growth) to cells with 90, 60, 50, 40, 35, 30, 25, and 20 minute interdivision times. In all cases, the rate of replication fork movement is 40 minutes for a round of replication or one-quarter of the genome every 10 minutes. All rounds of replication end 20 minutes before the end of the cell cycle. This is most clearly seen in the 60-minute cells where a newborn cell has one genome, which replicates for 40 minutes ending replication 20 minutes before cell division. The same rules are drawn here for a 90-minute and a very slow growing cell (infinite interdivision time). The large numbers in each pattern at the left indicate the number of origins to be initiated at each time of initiation of replication. Thus, in the 60-minute cells there is one origin in the newborn cell. Consider that the cell mass is given a unit value for each origin to be initiated. Thus, the newborn cell in the 60-minute case is given a size of 1.0 unit of mass. This means that the dividing cell in the 60-minute cells is size 2.0. Mass increases, in the 60-minute case, from 1 to 2. In the 90-minute cells the cell of size 1 is one third of the way through the cell cycle. Since mass increases continuously during the division cycle it is clear that the newborn cell in the 90-minute culture is less than 1.0 in size. Let us say it was something like size 0.7. In this case the newborn cell in the 90-minute cells would be size 0.7 and the dividing cell would be size 1.4. It is clear that the 90-minute cells are, on average, smaller than the 60 minute cells. Similarly, if we consider the very slow cells, the cell of size 1.0 is very near the end of the cell cycle, and the newborn cell is slightly above size 0.5. Since the very slow growing cells (top panel) go from sizes 0.5 to 1.0 and the 60 minutes cells go from size 1.0 to 2.0, the 60 minute cells are twice as large as the very slow growing cells. The 30-minute cells have two origins in the newborn cell and thus the newborn cells can be considered size 2.0 with the dividing cells 4.0. The 20-minute cells have a newborn cell of size 4.0 (four origins in the newborn cell) and a dividing size of 8.0. As one goes from extremely long interdivision time, to 60, to 30 to 20, the relative sizes go from 0.5, to 1, to 2 to 4, with the growth rates expressed as doublings per hour, or 0 (infinite interdivision time), 1 (60 minute interdivision time), 2 (30 minute interdivision time), and 3 (20 minute interdivision time). Cells that initiate DNA replication in the middle of the cycle may be considered as follows. The 40-minute cell has two origins in the middle of the cell cycle so the mid-aged cell is size 2.0. The newborn cell might be some size like 1.5 and the dividing cell something like 3.0. Thus, the 40-minute cell has an average size intermediate between the 60 and the 30-minute cell. Similarly, the 25-minute cell also initiates mid-cycle, but there are 4 origins at the time of initiation. Thus, the mid-aged cell in this case is size 4.0 and the newborn cell may be considered something like size 3.0. The cell sizes go from 3.0 to 6.0, and these cells are larger than the 30-minute cells and smaller than the 20 minute cells. Figure 8 Size determination in bacteria. In panel (a) the rates of growth of cells from infinitely slow (very long interdivision time) minutes to 20 minutes (as illustrated in Fig. 6) are plotted with the relative sizes shown. Thus, the 60 minute cell goes from size 1.0 to 2.0 over 60 minutes. The 30-minute cell (third angled line from top) goes from size 2 to 4 over 30 minutes. And the 20-minute cell (top angled line) goes from size 4 to 8 over 20 minutes. Other rates of growth for 25, 35, 40, 50, 90 and "infinite" interdivision times are also shown. In panel (b) the same results are plotted over relative cell ages from age 0 (newborn) to 1.0 (dividing cell). The open circles indicate when initiation occurs, and corresponds to the numbers in the individual panels. Thus, in Fig. 6 the cells with a 60, 30, and 20 minute interdivision time initiation DNA replication in the newborn cell (age 0.0) at sizes 1, 2 and 4. Besides the cell age at initiation, the open circle also indicates the relative size of the cell at initiation (see numbers in Fig. 7). The cell sizes at age 0.0 for all cells is a measure of the average cell size in the culture. (Given an identical pattern of cell growth during the division cycle the relative cell size of the cells in a culture is precisely proportional to the newborn cell size). These size values are then plotted against the rate of cell growth (the inverse of the interdivision time or doublings per hour) as shown in panel (c). The log of the cell sizes are a straight line when plotted as a function of the rate of cell growth (the inverse of the interdivision time). The important consequence of Figs. 7 and 8 is that we understand how cell size is controlled in bacteria. Cells initiate DNA replication at a certain cell size. This cell size (sometimes referred to as the "initiation mass") is a constant size within experimental limits (Cooper, 1997)[ 23 ]. The cell size at initiation is constant per origin present in the cell. A cell with two origins being initiated is twice as large as a cell with only one origin. The number of origins present at initiation and the cell age during the division cycle at which initiation occurs determines the average cell size of a cells growing in culture. Analysis of size maintenance in animal cells The ideas of the bacterial cycle can be directly applied to animal cells. Cells of different growth rates are shown in Fig. 9a . The different lines, a-g, identify cells of different sizes because they pass through size 1.0 at different cell ages during a cell cycle span. Cell "g" is a faster growing cell than cell "a" with the others of intermediate growth rates. The earlier a cell reaches size 1.0, the larger the cells will be. Thus, in Fig. 9a , the cell "g" is larger than the cell "a" because the cell "g" reaches size 1.0 earlier than the cell "a". As drawn in Fig. 9a , the newborn "g" cell is size 1.0. The mother or dividing cell is size 2.0. We can imagine that the mean size of cells growing at this rate is approximately 1.5. In contrast, the "a" cell varies between newborn size of approximately 0.6 and dividing size of 1.2. The average size of the "a" cells is smaller than the "g" cell, approximately size 0.9. (The precise calculation of the average cell size requires consideration of the age distribution and the actual pattern of mass increase during the division cycle; for purposes of this analysis, these complications are omitted.) Other cells (b-f) may be similarly analyzed to see that faster growing cells are larger than slower growing cells. Figure 9 Mammalian cell size variation as growth rate varies. Panel (a) shows a given mammalian cell growing at different rates and with different sizes. The lines are parallel because the interdivision times are normalized to a relative cell age as cells are born at age 0.0 and divide at age 1.0. All lines are exponentially increasing cell sizes from smallest to largest. Where the lines cross the thick horizontal line indicates a cell of size 1.0. Since the fastest cell (cell g) has a size 1.0 at the start of the cell cycle these cells must go from a newborn sizes of 1.0 to a size at division of 2.0. The slowest cell (cell a) has size 1.0 toward the end of the cell cycle, so the newborn cell is slightly larger than size 0.5 at age 0.0. The size ranges of these cells goes over a factor of 2. In panel (b) the size patterns are re-interpreted in terms of initiation at a particular time during the cell cycle. In this figure the thick, short line on each pattern is the S phase, the thinner line to the right is the G2 phase and the thinner line to the left is the G1 phase. Given that S and G2 are relatively constant in length then the slower cells (e.g., cell "a") have a longer G1 phase than the faster growing cells (e.g., cell "g", which has no measurable G1 phase). This is because the interdivision time is the sum of S+G2+G1. If S and G2 are relatively constant as the interdivision time decreases (i.e., as cells grow at faster growth rates), the G1 phase gets smaller. When the interdivision time equals the sum of S and G2 as in cell "g", there is no G1 phase. Such a situation has been analyzed previously (Cooper, 1979). It is clear from panel (b) that as cells grow faster, the time during the division cycle at which initiation of S phase starts is earlier and earlier. This is illustrated even more directly in panel (c) where the phases are normalized to a unit length. The slowest cell (cell "a") has the shortest fraction of cells with an S or G2 phase and the fastest growing cell (cell "g") has the entire division cycle occupied by S and G2 phases. The topmost line in panel (c) is the fastest cell and it starts S phase early in the cell cycle. Thus we see that the faster a cell grows the earlier in the cell cycle the cell achieves a size of 1.0. This accounts for the result that the slower cell has a smaller cell size than the faster growing cell. As will now be seen, this variation in size is related to, and determined by, the growth rate. It is proposed that mammalian cells initiate DNA replication at some relatively constant cell size. The time for S and G2 phases are relatively constant as the interdivision time varies [ 24 ], so the cell cycle age at initiation of S phase occurs earlier and earlier within the cell cycle as the growth rate increases (or as the interdivision time decreases). This is shown in Fig. 9b , where the interdivision time is varied but S- and G2-phase lengths are constant. In Fig. 9c the cell cycle patterns in Fig. 9b are normalized to a constant length. In Fig. 9c it is clear that the faster cells initiate S phase earlier in the cell cycle. This is because faster growing cells have a relatively short G1 phase. These faster growing cells achieve the initiation mass earlier in the cell cycle and thus these cells will be larger. As in bacteria, faster growth leads to larger average cell sizes . (For a discussion of the case of cells growing so fast as to not have a G1-phase as in cell "g" in Fig. 9b , see [ 24 ]). The rate of cell growth is determined by medium composition. For example, as more and more nutrients are added to a minimal medium, bacterial cells grow at faster and faster rates. The interdivision time shortens as the medium becomes richer. For bacteria the mechanism for growth rate variation with medium composition is, in outline, well understood [ 25 ]. The addition of nutrients to a medium represses the synthesis of enzymes that are not now needed (e.g., addition of leucine stops the synthesis of leucine synthesizing enzymes). This leads to a shift in the synthetic capacity of the cell to the protein synthesizing system (RNA polymerase, ribosomes, related materials, etc.) as these functions are not repressible by external components [ 25 ]. This leads to a more rapid rate of mass increase and thus a shorter mass doubling time [ 24 , 26 ]. Although the details may vary, it is proposed here (and in fact supported by the experiments of Conlon and Raff) that the richer a medium is (e.g., more serum rather than less serum), the faster the cells will grow. The faster a cell grows, the larger it will be (Fig. 9 ). The variation of G1-phase length with interdivision time variation has been analyzed in detail [ 24 , 26 ]. Conlon and Raff [ 1 ] supply evidence for the relationship of cell size and growth rate in their Fig. 7. Cells that have become overcrowded by not being diluted back (their Fig. 7b) decrease their volume (their Fig. 7a). The analysis presented above explains the variation of cell size as function of growth rate as observed by Conlon and Raff (slower growing cells are smaller than faster growing cells). Furthermore, the analysis can also explain the maintenance of cell size, even with exponential mass increase during the division cycle, as shown in Fig. 9 . Larger than average cells will divide sooner as they reach the initiation mass earlier and smaller than average cells will delay initiation until the initiation mass is achieved. Cell division will follow after relatively constant S- and G2/M-phases. This is the underlying and fundamental explanation for the patterns described in Figs. 1 and 2 . Thus, we now have an answer to the question (raised in discussion of Fig. 2 ) "How does the cell 'know' when to divide so that size homeostasis is maintained?" The answer is that initiation of S phase is determined by the cell mass. A relatively large cell initiates S phase earlier than a relatively small cell. This earlier initiation is played out in an earlier cell division after a period equal to the S and G2/M phases. Since the S and G2/M phases are relatively invariant, an earlier initiation produces an earlier cell division. While the analysis in Fig. 1 discussed the size maintenance problem in terms of the cell dividing earlier if a newborn cell was larger and later if a newborn cell was smaller (or if the rate of mass increase was high or low), the deeper analysis presented now proposes that the decision to divide is determined not at the moment of cell division but earlier at the start of S phase. The initiation of S phase is determined by cell size and the faster a cell reaches the S-phase initiation size the earlier the cell will initiate S phase and the earlier it will divide. For reasons not yet understood, there is a relationship between initiation of S phase and cell division such that once S phase is initiated the cell will ineluctably proceed to division. We now see the answer to the problem of cell size at division. Cell size at division is merely a surrogate indicator of cell size at initiation. Further, the time of cell division is a surrogate measure of the time of initiation of S phase. A cell that initiates S phase earlier in the cell cycle will have more time to increase its total mass prior to division. The larger newborn cell, having initiated S phase relatively early compared to its relatively smaller sister and cousin cells, will divide earlier as described in Fig. 1 . Conversely, smaller cells will delay initiation of S phase; that delay will allow more mass increase before the actual cell division because S phase is somewhat delayed and thus division is postponed allowing mass to increase before the ultimate cell division. In this way, the cell size distribution is maintained. Size variation during a shift from slow to fast growth Immediately following the discovery of bacterial cell size variation with growth rate [ 8 ] shift-up experiments of cells from slow growth (relatively small size) to faster growth (relatively large size) were performed [ 27 ]. The phenomenon of "rate maintenance" was discovered in this shift-up experiment. Rate maintenance is the continuation of the rate of cell division for a constant period after the shift-up [ 19 ]. The rate of mass increase changes immediately to the new rate at the instant of shift-up, while the rate of cell division continues for a period of time before abruptly changing to the new rate. The rate maintenance phenomenon occurs over a wide range of shift-ups [ 19 ]. The continuation of the original, slower rate of cell number increase, combined with an immediate transition to the new rate of mass increase, leads to an increase in cell size over the period of rate maintenance (Fig. 10a ). Rate maintenance is now understood to result from the constant S and G2 periods (C and D periods in bacteria) that do not allow new divisions to occur until the newly inserted replication forks pass through the S (i.e., C) period and the G2 (i.e., D) period. Without going into details here (see [ 12 ] for a complete analysis and explanation), suffice it to say that the rate maintenance phenomenon leads to the observed variation in bacterial cell sizes as the rate of cell growth varies over a wide range. Figure 10 Comparison of shift-up of bacterial cells and mammalian cells. In the panel (a), after a shift of bacterial cells from slow-growth medium to fast growth medium there is an immediate change in the rate of mass synthesis to the new rate while the rate of cell division continues at the old rate for a fixed period of time (rate maintenance). At the end of this "rate maintenance" period, there is a sudden shift in the rate of cell number increase to the new rate. The thick line in panel (a) shows the change in cell size following the shift-up. In contrast, in panel (b) a slower and more gradual change in the rate of mass synthesis, concomitant with the cell number pattern also changing slowly over a period of time, will give a longer period of change in cell size. Conlon and Raff observed this slow pattern of mammalian cell size change. Conlon and Raff [ 1 ] studied mammalian cells during a shift-up from slow to rapid growth and small to large cell size. Upon shifting slow cells to faster medium (e.g., shifting cells from low serum to high serum) there is a concomitant increase in cell size (Fig. 6e of Conlon and Raff [ 1 ]). One major difference from the bacterial shift-up result is that with animal cells the time for cell size to increase took a much longer time, between 6 and 9 days. To explain the difference between the bacterial shift-up result (Fig. 10a ) and the mammalian cell shift-up result (Fig. 10b ) one can postulate that for reasons unrelated to the cell cycle but merely related to cellular metabolism occurring continuously throughout the cell cycle, the change in external conditions does not immediately lead to the new rate of mass increase (Fig. 10b ). The rate of mass increase is predicted to change relatively slowly as mammalian cells are shifted from serum-free (slow growth) medium to serum-containing (fast growth) medium. Of course, this is just the result reported by Conlon and Raff (their Fig. 6e). This view of the change in cell size following a shift from slow to rapid growth is quite different from the description Conlon and Raff present for the case of yeast cells switched from a nutrient-poor to a nutrient-rich medium. They write [ 1 ], "When switched from a nutrient poor medium to a nutrient-rich medium, the cell cycle arrests and resumes only when the cells have reached the appropriate size for the new condition, which occurs within one cell cycle...Thus, the cells can adjust their size threshold rapidly in response to changing external conditions." The bacterial model of the shift-up allows a rapid change in cell size within one cell cycle without postulating any "arrest" of passage through the cell cycle. Rather than postulate a mechanism that slows or actively shuts down the cell cycle, it is proposed that no change in cell division occurs until the increased initiations of S phases pass through the S and G2/M phases, as in the bacterial model. No additional mechanism need be proposed to "stop" some event of the cell cycle until cell size has increased. The age-size distribution summarizes size control One way to consider a growing culture is to see that every cell in a growing culture has an age and a size. The age/size structure of a population is a representation of each of the cells in the culture and its age and size. If there is no statistical variation, and cells move through the cell cycle with a perfectly precise exponential growth pattern, then the age/size distribution is seen in Fig. 11a . The projections of the dots in this panel to the age axis (bottom, abscissa) and the size axis (left, ordinate) indicate that when cells are growing exponentially, there are a greater number of smaller cells relative to the population than there are younger cells. This reflects the age distribution discussed above. Figure 11 Age-size structure of a growing culture. Panel (a) is the age-size structure for a perfectly deterministic population growing exponentially in mass during the division cycle. The dots on the exponentially increasing line are placed at equal age intervals shown by their representation at the bottom of the panel. The representation of the dots at the left of panel (a) indicates that there is a greater concentration of smaller cells than younger cells. In panel (b) the age-size structure for a population with variation in size and interdivision times is illustrated. The cloud of points (indicated by a few points as representative of the population) is one possible age-size structure. In panel (c) the newborn cells are indicated by the filled circles, the dividing cells by open circles, and the cells in the act of initiation of DNA synthesis by + signs. It can be seen that the larger cells at birth will, on average, reach the size required for initiation of DNA replication more quickly than smaller cells. This is because the larger cells are closer to the initiation size (represented by I on the right side of panel (c)). The B and D distributions at the right of panel (c) indicate the size distributions of newborn (B) cells and dividing (D) cells. The B, D, and I distributions at the top of panel (c) illustrate the age distributions for newborn, dividing, and initiating cells. The size distribution of initiating cells is drawn with a narrower distribution. Variations in mass increase during the period after initiation lead to the widening of the size distribution at division. Panel (d) is a replotting of the pattern in panel (c) with the bottom time scale defined by setting the time of initiation of DNA synthesisas age 0.0. Cells before initiation have a negative age value, and cells after initiation have a positive age value. Initiation takes place, by definition in this panel, at age 0.0. There is some variation in the size of cells at initiation, but it is proposed that this variation is less than the variation at other events of the cell cycle. The narrowing of the age-size structure at the time of initiation is a graphic representation of the size-homeostasis mechanism. No matter what size cells are present at birth or division, these cells are returned to their proper age-size relationship at the instant of initiation of DNA synthesis. Larger cells at division produce larger newborn cells which then reach initiation size earlier than smaller cells which were produced by the division of smaller dividing cells. This is a restatement of the idea that larger cells get to initiation earlier because larger cells have less of a negative age value at cell birth. At the top and right panels of (c) and (d) are representation of the presumed variation of the sizes and ages of cells at particular events. The size at birth is always a little more widely distributed than the size at division due to a slight inequality of partition of mass at division. The size at initiation of DNA replication is drawn with a relatively small variability. There are, however, statistical variations in cell sizes at a given age and cells of a given age may have different sizes. The precise statistical distribution is not known, but one view of the possible result is shown in Fig. 11b . The shaded area indicates a cloud of points preferentially collected around the middle of the shaded area, with fewer cells at the outer edges. If this were a three-dimensional graph, there would be a peaked "ridge" up the center of the shaded area indicating that more cells reside with a particular age/size distribution than those at the edges of the age/size distribution. It is possible to indicate the cells at particular times during the division cycle such as birth, division and initiation of DNA synthesis, as shown in Fig. 11c . There is no distribution in age at birth, since by definition age at birth is 0.0. The graphs at the upper and right sides of Fig. 11c are representations of the spread of the various distributions. There is variability in the age at initiation of DNA replication (I) and the age at division (D). The age at cell birth (B) is defined as constant (i.e., 0.0) but the size at birth (B) is not constant. It is expected that the size at division will be slightly less variable than the size at birth. This is attributable to the probability that any deviation from equipartition of dividing cells leads to a broadening of the size distribution in newborn cells. From Fig. 11c it can also be seen that larger newborn cells (distribution B) will reach the size at initiation (I) earlier than smaller newborn cells. This explains the size maintenance pattern drawn in Fig. 2 . Note that in Fig. 11c the size at initiation is relatively narrow. It is this narrowing of size at initiation that leads to the slightly narrower size at division relative to the wider size distribution at birth. An instructive way of looking at the age/size distribution is to replot cell ages using the age at initiation of DNA synthesis (I) as a starting point (Fig. 11d ). By defining age at initiation as 0.0 one gets negative ages for cells before initiation and positive ages for cells after initiation. Fig. 11d shows that there is no distribution in the age at initiation (since by definition the age at initiation is 0.0 for all cells) but there is now a variation in the "age" of newborn cells. Smaller cells have a more negative age and take longer to reach initiation size than larger newborn cells; that is why there is a distribution of cell ages of newborn cells with respect to the time until initiation. The "bottleneck" at initiation of DNA synthesis enables cells born of different sizes to retain size homeostasis. Since all cells to the left of initiation must pass through the bottleneck of initiation on the way to division, all cells, of any newborn size, are realigned and assigned a new age and a new size as they pass through the act of initiation. Summary Understanding mammalian cell size control This analysis explains how cell size is maintained through a combination of interdivision time variation and cell growth rate variation. Exponential growth is possible and allowed during the division cycle, in contrast to the proposal of Conlon and Raff [ 1 ]. The ideas presented here are a fresh way to look at the cell cycle and cell growth in eukaryotic cells, even though the ideas have been around for over three decades due to work on size determination in bacteria (Cooper, 1979; Helmstetter, 1969). The model of the cell cycle presented here explains many experimental results without postulating checkpoints, G1-phase events, restriction points, or similar phenomena. Experimental support for these ideas [ 28 , 29 ] and the application of these ideas to other problems of cell growth and differentiation [ 3 , 13 - 15 , 26 , 28 - 35 ] have been published. These ideas have also been reviewed [ 36 - 39 ]. It may be best to summarize these two contrasting views of size maintenance by looking at cell growth in a simple manner, and asking how the rate of mass increase is related to the passage of the cell through the cell cycle. The model of Conlon and Raff looks at the events of passage through the cell cycle as occurring independently of mass increase. The problem then remains as to how mass increase fits into, or coordinates with, the pattern or timing of passage through the cell cycle. It is as though the cell moves through the cell cycle without considering the mass problem, and then the mass of the cell looks at the cell cycle and says "I must grow at some rate so that I do not get too big or too small." In the Conlon/Raff model, a control exists that coordinates mass increase with the rate of cell division. The model presented here–in contrast to the model of Conlon and Raff–situates mass as the driving force of the cell cycle. Mass increases at some rate that is determined by external conditions (medium, growth factors, pH, etc.). As the mass increases, the accumulation of mass starts or regulates passage through the cell cycle. A cell cannot grow to an abnormal larger size because at a certain cell size or cell mass the S phase is initiated and this event starts a sequence of events leading to mitosis and cytokinesis. A cell cannot get too small because if mass grows slowly (or even stops growing) then the later events of the cell cycle (S-, G2-, and M-phases) are delayed (or do not occur) until mass increases sufficiently to start S phase. A cell cannot get too large because at a certain size the cell initiates S phase leading to the relatively early cell division. A cell cannot get too small because if mass accumulation is inhibited then S phase initiations are also inhibited. Thus, there is no problem relating mass increase and the cell cycle. Cell mass growth and cell cycle passage cannot be dissociated because one (mass increase) is the determinant of the other (S-phase initiation). For this reason one needs neither checkpoints nor control elements outside of mass increase. The time for mass to double in a particular situation determines the doubling time of a culture. This is because initiations of S phase occur every mass doubling time, and cell divisions similarly occur every mass doubling time. Thus total mass increases at the same rate as total cell number. The model presented here explains size determination, size maintenance, and the relationship of mass increase and cell number increase in a growing, exponential, unperturbed, mammalian cell cultures.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524481.xml
514542	The p68 and p72 DEAD box RNA helicases interact with HDAC1 and repress transcription in a promoter-specific manner	Background p68 (Ddx5) and p72 (Ddx17) are highly related members of the DEAD box family and are established RNA helicases. They have been implicated in growth regulation and have been shown to be involved in both pre-mRNA and pre-rRNA processing. More recently, however, these proteins have been reported to act as transcriptional co-activators for estrogen-receptor alpha (ERα). Furthermore these proteins were shown to interact with co-activators p300/CBP and the RNA polymerase II holoenzyme. Taken together these reports suggest a role for p68 and p72 in transcriptional activation. Results In this report we show that p68 and p72 can, in some contexts, act as transcriptional repressors. Targeting of p68 or p72 to constitutive promoters leads to repression of transcription; this repression is promoter-specific. Moreover both p68 and p72 associate with histone deacetylase 1 (HDAC1), a well-established transcriptional repression protein. Conclusions It is therefore clear that p68 and p72 are important transcriptional regulators, functioning as co-activators and/or co-repressors depending on the context of the promoter and the transcriptional complex in which they exist.	Background The DEAD/H box family of RNA helicases has been demonstrated to be involved in virtually all processes that require manipulation of RNA including transcription, pre-mRNA and pre-rRNA processing, RNA export, ribosome assembly and translation [ 1 ]. Although, in vitro , several members of this family have been shown to unwind RNA duplexes, relatively few appear to be true processive helicases and it is clear that, in the cell, many are likely to be involved in unwinding of short base paired regions of RNA or in the modulation of RNA-protein interactions. DNA helicases belong to a superfamily of proteins that is distantly related to DEAD/H box RNA helicases and includes the Werner syndrome protein (WRN) [ 2 ] and the Xeroderma pigmentosum XPB and XPD proteins [ 3 ], which have well established roles in transcription. Although the functions of DEAD/H box RNA helicases in other cellular processes, such as pre-mRNA processing and translation have been well studied, their role in transcriptional regulation is only now emerging. Examples of DEAD/H box RNA helicases involved in transcription include RNA helicase II (RHII/Gu) and RNA helicase A (RHA/NDHII). RHII/Gu was demonstrated to be a cofactor for c-Jun-activated transcription [ 4 ] and was shown to translocate from the nucleolus to the nucleoplasm after UV or anisomycin treatment (which activates JNK signalling). Although RHII/Gu was found to associate with phosphorylated c-Jun in a non-phosphorylated state, this association was observed to increase after anisomycin treatment, implying a stronger interaction when c-Jun is phosphorylated [ 4 ]. RHA is a homologue of the Drosophila maleless (MLE) gene product [ 5 ] and is thought to be important for gene dosage compensation on the X-chromosome [ 6 ]. RHA has been shown to be required for complex formation between the transcriptional co-activator, CREB binding protein (CBP), and RNA polymerase II [ 7 ]. Furthermore, different regions of the RNA helicase protein were found to interact with both CBP and RNA polymerase II. The association of RHA with RNA polymerase II was further investigated, and narrowed down to a 50 amino acid stretch, outwith the conserved helicase motifs [ 7 ]; this study also showed that RHA could regulate CREB-dependent transcription either through recruitment of Pol II or by ATP-dependent mechanisms. A later study reported that RHA acts as a bridging molecule between the breast tumour specific transcriptional activator, BRCA1 and the RNA polymerase II holoenzyme complex [ 8 ]. These reports thus provide clear evidence of a role for RNA helicases as transcription factors. p68 is a prototypic member of the DEAD box family of proteins [ 9 ] and an established RNA helicase [ 10 ]. The subsequent discovery of p72 [ 11 ] and the finding that p68 and p72 share remarkable homology (90% over the central conserved core and 60% and 30% at the N- and C-terminal extensions respectively) suggests that these proteins may form a specific sub-group of DEAD box proteins and may have similar, but perhaps subtly different, functions in the cell, perhaps through interaction with different RNA substrates or proteins. In vitro , both proteins exhibit the RNA-dependent ATPase and RNA helicase activities characteristic of members of the DEAD box family [ 10 - 14 ] and have also been reported to catalyse rearrangement of RNA structure via branch migration [ 13 ]. Moreover p68 and p72 can interact with each other, as well as self-associate, and appear to preferentially form heterodimers in cells [ 15 ]. This provides the potential for a wide range of functions for p68 and p72 with the possibility of their co-operation in some contexts. More recently p68 and p72 have been shown to be involved in a range of processes in the cell, including pre-mRNA and pre-rRNA processing and alternative splicing [ 16 - 18 ]. p68 and p72 have also been shown to be growth- and developmentally-regulated [ 19 - 22 ] and, furthermore, p68 appears to be over-expressed and poly-ubiquitylated in colorectal tumours [ 23 ]. Interestingly p68 has been shown to act as a transcriptional co-activator, specific for the activation function 1 (AF-1) domain of estrogen receptor alpha (ERα) [ 24 ]. This interaction was dependent upon phosphorylation of AF-1 at serine 118 , a residue phosphorylated by mitogen-activated protein kinase (MAPK). Interestingly the RNA helicase function of p68 appeared to be dispensable for this activity as a mutant p68 (Lys 144 to Arg) in the ATP binding site (conserved motif I) retained the ability to co-activate ERα, although in a later study RNA binding appeared to be required [ 25 ]. p72 was subsequently shown to share this property and both proteins were shown to interact with the activation domain 2 (AD2) of p160 co-activators [ 25 ]. Furthermore p68 was also found to interact with the CBP co-activator and RNA polymerase II [ 26 ]. More recently p68 has been shown to be recruited to the promoter of the ERα target gene pS2 [ 27 ], suggesting a direct involvement in transcriptional regulation. In this report we explore further potential mechanisms through which p68 and p72 may contribute to transcriptional regulation. We find that, in some contexts, p68 and p72 can also act as transcriptional repressors and that these proteins exhibit clear promoter specificity in this function. By directing GAL4-tagged p68/p72 (GAL4 DNA binding domain aa1-147) to promoters containing GAL4 binding sites, we show that both p68 and p72 can repress transcription from the herpes virus thymidine kinase (TK) promoter but not the simian virus 40 promoter/enhancer. Moreover, while p72 can repress transcription from the Adenovirus major late promoter, p68 appears to have no effect, suggesting that these proteins do not behave in an identical way in all contexts. Furthermore we show, by co-immunoprecipitation, that both p68 and p72 interact with histone deacetylase 1 (HDAC1) and that HDAC1/p68/p72 co-elute by gel filtration, indicating that they exist in the same complex in the cell and suggesting a possible mechanism by which these proteins may exert their repressive effect. Results p68 and p72 differentially repress constitutively active promoters/enhancers In order to determine whether, in addition to their reported role as co-activator proteins [ 25 ], p68 and p72 helicases have any intrinsic transcriptional activity we generated plasmids to express p68-/p72-GAL4 DNA binding domain (aa1-147) fusion proteins (p68G4 and p72G4). U2OS cells were then co-transfected with p68G4, p72G4, or GAL4-tagged pcDNA3 (pcG4) plasmid as a control, and a Herpes virus thymidine kinase promoter/chloramphenicol acetyl transferase (CAT) reporter plasmid (TK-CAT) bearing 5 copies of the GAL4 binding site, and CAT activity was measured in standard assays. The amount of TK-CAT used had previously been titrated to give a basal level of activity within the linear range of measurement in the CAT assay system used (data not shown). Interestingly, we observed a marked decrease in CAT activity for both p68G4 and p72G4 (Figure 1a ). This was confirmed using other cell lines, including MCF-7 and 293 HEK (data not shown). As many transcription factors have been shown to act differentially depending upon promoter context [ 28 ], we also tested the transcriptional activity of p68 and p72 with the Adenovirus major late promoter (MLP-CAT), and the SV40 promoter/enhancer (SV40-CAT). In each case we used the same amount of p68G4/p72G4 DNA as previously and appropriate amounts of the reporter constructs that would give similar basal levels of CAT activity using the GAL4 control plasmid (Figure 1a ). The amounts of reporter plasmid DNA transfected had again been titrated previously to give similar basal levels, which were within the linear range (data not shown). We also confirmed that p68 and p72 were not limiting under these conditions (see Figure 2c and data not shown). Surprisingly p68 and p72 acted differentially with the MLP promoter, with p72 acting as a repressor, while p68 had no significant effect on CAT activity. Furthermore, neither p68 nor p72 repressed transcription from the simian virus 40 promoter/enhancer (SV40-CAT-Figure 1a ). Western blotting, using an antibody against GAL4 confirmed that p68 and p72 were expressed at similar levels in these cells (Figure 1b ). Taken together these results reveal not only that p68 and p72 appear to have a previously unknown transcriptional repressive ability, but also that this activity is variable depending on the promoter context. Figure 1 Effect of GAL4-tagged p68 and p72 on transcriptional activity as measured by CAT assays using the TK-CAT, MLP-CAT and SV40-CAT promoter-reporter plasmids, each harbouring 5 copies of a GAL4 binding site fused to the promoter. The pcDNA3-GAL4 expression vector (pcG4) was used as a control. In each case U2OS cells were co-transfected with pcG4 or plasmids expressing GAL4-tagged p68/p72 (p68G4/p72G4) and the appropriate promoter-reporter construct. The amounts of DNA transfected were: pcG4-, p68G4-/p72G4- 1 μg; TK-CAT- 2.5 μg; MLP-CAT- 9 μg; SV40-CAT- 0.5 μg. The amount of DNA used had been previously titrated to achieve appropriate levels of baseline CAT activity. a) Transcriptional repression by p68 and p72 as measured by CAT activity, which is shown as % conversion of 14 C-labelled chloramphenicol to acetylated forms. The data represent results from 5 independent assays, which were each performed in triplicate. b) Western blot, using a GAL4-specific antibody, showing expression levels of the pcG4, p68G4 and p72G4 plasmids. Figure 2 Control CAT assays to examine repression of the TK-CAT promoter-reporter in U2OS cells by p68/p72. The amounts of DNA transfected in each assay are indicated below and in all cases the % conversion of 14 C-labelled chloramphenicol to acetylated forms is shown as an average of three independent experiments. a) Effect of untagged p68/p72 on TK-CAT transcription. 7.5 μg of control pcDNA3 vector, pcDNA3-p68 (p68) or pcDNA3-p72 (p72) were co-transfected with 2.5 μg of TK-CAT. b) Effect of GAL4-tagged p68 and p72 on transcriptional activity of a TK-CAT promoter-reporter which incorporated a 1.6 kb DNA 'spacer' between the GAL4 binding sites and the promoter (TK-S-CAT). 1 μg of pcDNA3-GAL4 (pcG4) or GAL4-tagged p68/p72 (p68G4/p72G4) were co-transfected with 5 μg of TK-S-CAT. The amount of TK-S-CAT had previously been titrated to achieve an appropriate baseline level of CAT activity. c) Titre of repression of TK-CAT activity by GAL4-tagged p68/p72. 2.5 μg of TK-CAT were co-transfected with different amounts of pcG4 vector, p68G4 and p72G4 as indicated. d) Effect of p300 and CBP on repression of TK-CAT transcription by GAL4-tagged p68/p72. 2.5 μg of TK-CAT were co-transfected with 1 μg of pcG4 vector, p68G4 or p72G4 together with 6.5 μg of either bluescript (as control) or p300/CBP. The repression of transcription is an active process To confirm that the repressive effect observed is not due to an artefact of the assay conditions we initially repeated the experiment with the TK-CAT reporter plasmid, using non-tagged p68 or p72 or the pcDNA3 expression plasmid vector alone (Figure 2a ). In this experiment neither p68 nor p72 significantly reduce transcription of the TK-CAT reporter, suggesting that the repression observed with the GAL4-tagged p68/p72 plasmids (Figure 1 ) is not due to competing out of an essential factor required for TK-CAT transcription but, instead, implies an active mechanism of repression in which the p68/p72 proteins are required to be directed to the TK-CAT promoter via the GAL4 tag. The possibility still remained that p68 and p72 were repressing transcription by direct interference with the transcriptional machinery of the TK-CAT promoter (e.g. perhaps by physically blocking the promoter). To rule out this possibility, we used a similar TK-CAT promoter construct, which had, however, a 1.6 kb DNA 'spacer' between the GAL4 binding sites and the promoter (the inserted DNA is reported to have no effect on transcription [ 29 ]). Both GAL4-p68 and -p72 retained the ability to repress transcription, almost to the same degree as previously (Figure 2b ). Furthermore, a titration using the GAL4- p68/p72 fusion proteins clearly shows a dose-dependent concentration curve (Figure 2c ) as would be expected for active repression. In addition, p68 and p72 have been observed to interact with p300/CBP co-activators [ 24 , 26 ]. Therefore it remained possible that the observed transcriptional repression was due to competition for, or interference with, p300/CBP. If this were the case co-expression of p300/CBP would be expected to relieve repression by p68/p72. As shown in Figure 2d , no such relief of expression was observed. Thus our findings that transcriptional repression by p68/p72 was dose-dependent and not due to steric blocking of the promoter or competition/interference with p300/CBP, suggest that it is an active process. Deletion experiments do not identify a distinct repression domain for p68 or p72 but reveal an activation domain To determine whether specific regions or domains of p68 and p72 act as transcriptional repressors, we performed CAT assays, again using the TK-CAT reporter, but with a range of deletion derivatives of p68/p72 covering the entire coding region. Deletion derivatives encompassing domains between amino acids 1–478 and 1–474 for p68 and p72 respectively repressed transcription in this assay while the C-terminal region for both proteins (aa 477–614 for p68 and 468–650 for p72) acted as a strong transcriptional activator (Figure 3a,3b ). Residues 1–478 of p68 and 1–474 of p72 include all the conserved motifs that characterise the DEAD box family of proteins (Figure 3c ). Within this conserved core we have shown that there are three domains, which can independently repress transcription (Figure 3c ), while the complete region (aa 1–478/474) can repress as well as the full-length respective proteins. Additionally, the finding that the C-terminal regions of p68 and p72 activate transcription is consistent with earlier reports of transcriptional activation by p68/p72 [ 25 , 26 ]. Interestingly, ATPase inactive mutants of p68 and p72 (in which the DEAD motif had been mutated to NEAD) as well as the more recently identified p82 (a derivative of p72 which uses an alternative non-AUG upstream translation initiation codon [ 30 ]) also repress transcription (Figure 3b ). Thus ATPase and helicase activity appear to be dispensable for transcriptional repression suggesting that this function of p68/p72 may not specifically require RNA unwinding; again this is consistent with reports that helicase activity is not required for co-activation of ERα transcriptional activity [ 24 , 25 ] although another report suggested that p68 helicase activity is required for synergism with the transcriptional co-activators CBP/p300 [ 26 ]. Figure 3 Deletion mapping of potential repression/activation domains in a) p68 and b) p72 as observed in CAT assays using the TK-CAT promoter-reporter plasmid. The pcDNA3-GAL4 expression vector (pcG4) and full-length GAL4-tagged p68/p72 were used as controls. All p68/72 deletion derivatives were expressed as GAL4-tagged fusion proteins in pcG4 and included the amino acids indicated. Additional proteins tested in this assay included the ATPase/helicase GAL4-tagged inactive mutants of p68/p72 (p68N/p72N) and the alternative upstream initiation product of the p72 gene (p82). The amounts of DNA used in the transfections were; TK-CAT- 2.5 μg; pcG4 and all p68/p72 constructs- 1 μg. The % conversion of 14 C-labelled chloramphenicol to acetylated forms is shown as an average of five independent experiments. c) Diagram correlating the deletion end-points to the position of the motifs conserved in the DEAD box family of proteins. p72 immunoprecipitates a HDAC activity Many studies have implicated histone deacetlyase (HDAC) proteins in active repression of transcription, and many transcriptional repressors have been shown to associate with HDACs [ 31 ]. To test whether the observed transcriptional repression by p68 and p72 was dependent on HDAC activity, CAT activity assays were performed as before, using the TK-CAT and MLP-CAT reporter plasmids, in the presence and absence of the HDAC inhibitor trichostatin A (TSA). The relief of repression by TSA was then determined for p68-/p72-GAL4 compared with that observed for the GAL4 vector alone since TSA will also increase basal levels of transcription. No effect was observed with the TK-CAT promoter (data not shown). For the MLP-CAT promoter repression was relieved two-fold in the case of p72 while no effect was seen with p68 (Figure 4a ). This is not surprising since p68 does not repress transcription from this promoter (Figure 1 ). These findings therefore suggest that HDAC activity appears to be important for transcriptional repression of the MLP promoter by p72 and imply that repression of the TK-CAT may employ a different mechanism [ 32 ]. Figure 4 The involvement of HDAC activity in transcriptional repression by p68/p72. a) Relief of p68/p72 repression of MLP-CAT transcription by TSA. 1 μg of pcDNA3-GAL4 (pcG4) or GAL4-tagged p68/p72 (p68G4/p72G4) were co-transfected with 9 μg of MLP-CAT and TSA was added 16 hr after transfection, at a final concentration of 300 nM. The values for p68G4 and p72G4 are given relative to the baseline value for the pcG4 vector control, which was set at 1, and represent the average from three experiments. b) Immunoprecipitation/western blotting of myc-tagged p68 and p72 from 293 cells expressing these proteins. Myc-tagged proteins were immunoprecipitated with the anti-myc epitope antibody, 9E10, and western blotted with the same antibody to detect the presence of p68-myc and p72-myc fusion proteins. A myc-tagged pSG5 vector control is included. pSG5, p68 and p72 all refer to myc-tagged versions. H denotes cross reaction with the antibody heavy chain. Molecular weight markers (in kDa) are indicated. Equal amounts of these immunoprecipitated proteins were used in the HDAC activity assay shown in c. c) HDAC activity assay of immunoprecipitated p68 and p72 (see b). HDAC activity in the presence and absence of TSA is shown relative to that of the myc-tagged pSG5 vector control, which was set at 1, and represent the average from three experiments. To examine further the involvement of HDACs in transcriptional repression by p68/p72 we determined whether p68 and/or p72 co-immunoprecipitate a HDAC activity in cells. Plasmids expressing myc-tagged p68 and p72 were expressed in 293 cells (with myc tag plasmid vector as a negative control) and the tagged proteins were immunoprecipitated from cell lysates using the myc epitope antibody, 9E10. Equal amounts of immunoprecipitated protein bound to the antibody, as confirmed by western blotting (Figure 4b ), were then used in an HDAC activity assay (Biomol) in the presence and absence of TSA. In this assay (Figure 4c ) p72 was found to co-immunoprecipitate a HDAC activity, which is abolished by TSA, while p68 did not. These findings are consistent with p72 interacting with a HDAC and repressing transcription in a HDAC-dependent manner. p68 and p72 associate with HDAC1 in cells Three classes of HDACs have been described. Class I HDACs, which include HDAC1, 2 and 8, are expressed in the nucleus and have been shown to bind several transcription factors and to mediate transcriptional repression [ 31 ]. Since HDAC1 is the prototypical member (in mammalian cells) and has been well studied we decided to examine whether p68 and/or p72 associate with HDAC1 in cells. We had previously shown that a large proportion of p68 and p72 co-elute by gel filtration [ 15 ]. Therefore we examined fractions from a DNAse/RNAse-treated 293 cell lysate, which had been separated by gel filtration, by western blotting using antibodies against p68, p72 and HDAC1, and showed that a significant proportion of HDAC1 co-elutes with the majority of p72 and a substantial proportion of p68 in the cell in complexes that are of a size consistent with p68/p72 interacting with HDAC1 (Figure 5 ). Since co-elution does not, in itself, indicate an interaction, we went on to examine whether p68 and or p72 co-immunoprecipitate with HDAC1 from cell lysates. For p68, nuclear extracts were prepared from U2OS cells and HDAC1 was immunoprecipitated with an HDAC1-specific antibody. Immunopreciptiated proteins were then separated by SDS-PAGE and the presence of p68 was detected by western blotting with a p68-specific antibody (Figure 6a ). As currently available antibodies against p72 cross react with other nuclear proteins [ 15 ] 293 cells were transfected with myc-tagged p72 and interactions between the myc-tagged p72 and HDAC1 were examined. In this case, therefore, nuclear extracts were prepared from transfected cells, HDAC1 was immunoprecipitated as before and associated myc-tagged p72 was detected by western blotting using the myc epitope-specific antibody, 9E10 (Figure 6b ). In each case, the presence of HDAC1 in the immunoprecipitate was confirmed by western blotting with the HDAC1-specific antibody (Figure 6a,6b ). As an additional control we carried out a reciprocal co-immunoprecipitation experiment in which nuclear extracts from 293 cells transfected with myc-tagged p68/p72 were prepared, the myc-tagged p68/p72 proteins were immunoprecipitated using the myc epitope-specific antibody and associated HDAC1 was detected by western blotting using the HDAC1-specific antibody (Figure 7a ). Cells that had not been transfected were used as a control. Additionally, since deletion derivatives encompassing residues 1–478 of p68 and 1–474 of p72 can repress transcription as well as the respective full-length proteins, we examined whether these deletions could co-immunoprecipitate with HDAC1. We therefore transfected 293 cells with GAL4-tagged proteins containing residues 1–478 of p68 and 1–474 of p72, prepared nuclear extracts, immunoprecipitated HDAC1 and western blotted for associated GAL4-tagged p68/p72 with a GAL4-specific antibody. As shown in Figure 7b , these deletion derivatives co-immunoprecipitate efficiently with HDAC1, suggesting that this region of p68/72 interacts with HDAC1. These findings thus indicate that p68 and p72 associate with HDAC1 in cells and that the interaction appears to be mediated by the regions which, in our system, are responsible for the transcriptional repression activity of p68 and p72. Moreover, since the immunoprecipitations were performed with extracts that had been treated with DNase/RNase (see Materials and Methods), the interactions of p68/p72 and HDAC1 are not mediated by nucleic acid. Figure 5 Western blots showing gel filtration elution profiles of p68, p72 and HDAC1. p68, p72 and HDAC1 in the fractions were detected by western blotting using appropriate antibodies. Note that the antibody raised against p72 also recognises p82 and cross-reacts with NFAR-2 [15]. All lysates had been treated with DNase and RNase prior to gel filtration. The void volume and elution position of the Pharmacia FPLC size markers are indicated, as are molecular weight markers (in kDa). Figure 6 Co-immunoprecipitation of a) p68 and b) p72 with HDAC1. a) HDAC1 was immunoprecipitated from U2OS nuclear extracts using an HDAC1-specific antibody. Immunoprecipitated proteins were separated by SDS-PAGE and the presence of HDAC1 and associated p68 was detected by western blotting with p68- and HDAC1-specific antibodies. b) HDAC1 was immunoprecipitated from nuclear extracts of 293 cells expressing myc-tagged p72 using an HDAC1-specific antibody. Immunoprecipitated proteins were separated by SDS-PAGE and the presence of HDAC1 and associated myc-tagged p72 was detected by western blotting with HDAC1- and myc epitope- specific antibodies. In both experiments a control immunoprecipitation (IP) was performed using an irrelevant rabbit IgG. An aliquot of nuclear extract (NE) was also included in the western blots (west.) as an additional control. H denotes cross reaction with the antibody heavy chain. Molecular weight markers (in kDa) are indicated. Figure 7 a) Reciprocal co-immunoprecipitation of p68 and p72 with HDAC1. Myc-tagged p68 and p72 were immunoprecipitated from nuclear extracts of 293 cells expressing these proteins using a myc epitope-specific antibody and associated HDAC1 was detected by western blotting with an HDAC1-specific antibody. 293 cells, which had not been transfected, were used as control. b) Co-immunopreciptiation of p68/p72 deletion derivatives with HDAC1. HDAC1 was immunoprecipitated from nuclear extracts of 293 cells expressing GAL4-tagged p68 and p72 deletion derivatives, which encompass residues 1–478 and 1–474 of p68 and p72 respectively. Associated p68/p72 were detected by western blotting with a GAL4-specific antibody. 293 cells, which had not been transfected, were used as control. NE-nuclear extract, IP-immunopreciptiation. Molecular weight markers (in kDa) are indicated. Discussion We have shown that the highly related DEAD box RNA helicases p68 and p72 act as repressors of transcription in a promoter-context manner. When targeted to the TK-CAT promoter-reporter construct they both strongly repress transcription (Figure 1 ). Furthermore, this transcriptional repression does not appear to be due to either squelching or physical blocking of the transcription apparatus (Figure 2 ), implying an active transcriptional mechanism. Moreover repression of TK-CAT was observed in several cell lines (U2OS, 293, MCF-7) suggesting that it is not cell line dependent. In order to determine whether this repression activity exhibited any promoter specificity we also tested the ability of p68 and p72 to repress transcription of other constitutively active promoter-reporter constructs with high basal levels of transcription, namely MLP-CAT and SV40-CAT. Interestingly MLP-CAT revealed a difference in the ability of p68 and p72 to repress transcription, with p72 strongly repressing transcription of this promoter-reporter and p68 having no effect (Figure 1 ). This observation suggests that, although highly homologous (70% overall identity at the amino acid level [ 11 ]) p68 and p72 proteins may act differently in some contexts, perhaps through the association with different protein partners. Neither p68 nor p72 repressed transcription of SV40-CAT (Figure 1 ) suggesting that the repression by p68 and p72 is promoter context-dependent, an observation that has been reported for other transcription factors [ 33 ]. These findings thus are consistent with the observed repression activity of p68/p72 being an active process. Interestingly, in this context, another DEAD box protein DP103 (Ddx20) has been found to act as a co-repressor of the Ets repressor METS/PE1 [ 34 ]. Using a series of deletion derivatives of p68 and p72 we identified three domains, within the core conserved among the DEAD box family of proteins, which can independently repress transcription (Figure 3c ). Moreover regions encompassing residues 1–478 of p68 and 1–474 of p72, which contain the complete conserved core (Figure 3c ), can repress transcription as well as the respective full-length proteins (Figure 3a,3b ). In contrast, the C-terminal extension of both proteins acts as a transcriptional activator in this context (Figure 3a , 5 ) consistent with earlier reports of these proteins acting as transcriptional co-activators [ 24 , 25 ]. Thus, using this system, we have shown that there are separable transcriptional repression and activation domains within p68 and p72. Since HDAC proteins have been extensively implicated in the repression of transcription, it was important to examine whether these proteins are likely to play a role in transcriptional repression by p68 and/or p72. Firstly, the ability of the HDAC inhibitor, TSA, to relieve repression was examined. No effect was observed on repression of TK-CAT (data not shown) implying the involvement of a HDAC-independent mechanism in repression of the TK promoter. In contrast, repression of MLP-CAT by p72 was relieved two-fold compared with the vector control (Figure 4a ) suggesting the involvement of HDACs in this process. (Since p68 did not repress MLP-CAT transcription, the lack of effect by TSA is not surprising.) Supporting these data, p72 was found to co-immunoprecipitate a HDAC activity which was abolished by the addition of TSA, while p68 did not (Figure 4b ). We chose to investigate whether p68 and/or p72 associate with HDAC1, since it is a well-studied example of Class I HDACs. Both p68 and p72 co-immunoprecipitate with HDAC1 (Figures 6 and 7 ); furthermore HDAC1, p68 and p72 co-elute in similar sized complexes by gel-filtration, which are of an appropriate size (Figure 5 ) supporting the idea of interactions between p68/p72 and HDAC1 in cells. Moreover, the finding that these proteins co-immunoprecipitate and co-elute from extracts which had been treated with DNase/RNase suggest that these represent protein-protein interactions rather than merely interactions via nucleic acid. While an interaction between HDAC1 and p68 is not supported by the results of the HDAC assay, or the TSA experiment, it is possible that, in some instances, p68 does recruit HDAC1 and that this mechanism is not being triggered in the MLP-CAT assay or HDAC assay. Alternatively, it is possible that the observed co-immunoprecipitation of p68 and HDAC1 is occurring through the interaction between p68 and p72 [ 15 ] or that HDAC1 associated with p68 may have other, possibly non-transcriptional, roles [ 35 , 36 ]. However, the data are entirely consistent with p72 recruiting HDAC1 to achieve active repression of transcription. Future investigations should also reveal whether the differential ability of p68 and p72 to recruit active HDAC proteins is responsible for the difference observed upon the MLP promoter. Our attempts at correlating the different repressive functions of p68 and p72 to specific domains of the respective proteins, using deletion derivatives, were unsuccessful, as the equivalent regions of both either caused transcriptional repression or activation. While this might suggest that both helicases repress transcription in the same manner, it is more likely that the HDAC recruitment by p72 may be an additional mechanism of repression, used at specific promoters. We also cannot rule out the recruitment of other repression complexes at this stage. Our findings that p68 and p72 differ in their ability to repress the MLP promoter and to recruit HDAC activity suggest that, at least in some contexts, p68 and p72 repress transcription by different mechanisms. In summary, it is clear that p68 and p72 act to repress transcription in a differential manner dependent upon promoter context. It will be important to determine which endogenous promoters are subject to repression by p68/p72 in a physiological context. However, until the signal transduction pathways, which target these proteins to the appropriate promoters, are elucidated it will be necessary to use a targeting system (such as GAL4) to undertake a molecular analysis of transcriptional repression by p68 and p72. Since it is now clear that p68/p72 can act both to activate and repress transcription future work will involve dissection of the transcriptional activation/repression complexes in which p68 and p72 are involved, as well as characterisation of the molecular 'switch' which determines whether these proteins will be part of transcriptional activation or repression complexes. Conclusions We have shown that the highly related RNA helicases p68 and p72 can repress transcription in a promoter context-dependent manner. Both proteins associate with HDAC1, a well-established transcriptional repressor protein. Strikingly, however, p68 and p72 behave differently in their ability to repress transcription from different promoters and in their ability to recruit HDAC activity suggesting that they may, at least in some contexts, repress transcription by different mechanisms. Methods Cell culture U2OS human osteosarcoma cells and 293 human embryo kidney cells were maintained at 5% CO 2 at 37°C in DMEM with 10% FBS, 2 mM L-glutamine, 100 units/ml penicillin and 100 μg/ml streptomycin (all supplied by Invitrogen). Plasmids A pcDNA3-GAL4 expression plasmid was used to express full-length and deletion derivatives of p68/p72/p82 tagged at the N-terminus with the DNA binding domain (amino acids 1–147) of GAL4 [ 37 ]. The majority of deletion derivatives were created by PCR and inserted, as Bam HI/ Eco RI fragments, in frame with the GAL4 tag. The CAT reporter constructs have been described previously: TK-CAT [ 38 ], TK-Spacer-CAT [ 29 ], MLP-CAT and SV40-CAT [ 39 ]. The MLP-CAT and SV40-CAT plasmids were a kind gift from Douglas Dean (Washington University, St Louis, USA), and the TK-spacer-CAT plasmid was kindly provided by Dr. Alain Nepveu (McGill University, Canada). Untagged p300, CBP, p68, p72 were expressed from pcDNA3. A myc-tagged derivative of pSG5 (Stratagene) [ 15 ] was used to express the myc-tagged p68/p72 or the myc epitope alone as negative control. Antibodies p68: The antibodies used were the mouse monoclonal antibody PAb 204 and the rabbit polyclonal antibody 2906, generated against the C-terminal 15 amino acids of p68 [ 19 ]. PAb 204 was originally generated against the SV40 large T antigen but it cross-reacts with p68 [ 9 ]. It is specific for p68 in cells that are not infected or transformed by SV40. p72: A rabbit anti-peptide polyclonal antibody generated against amino acids 624 to 638 [ 15 ] was used to detect p72/p82 in fractions from gel filtration. Myc epitope: A mouse monoclonal antibody (9E10) was used to detect proteins tagged with the myc epitope (MRQKLISEEDL). HDAC1: A rabbit polyclonal antibody (Oncogene Research Products) was used both for immunoprecipitation and western blotting. GAL4- a mouse monoclonal antibody (Santa Cruz) was used to detect GAL4-tagged proteins. A negative control rabbit IgG antibody for immunoprecipitation was obtained from R&D systems and appropriate anti-mouse and anti-rabbit secondary antibodies were obtained from DAKO. Transient transfections and chloramphenicol acetyl-transferase (CAT) assays 3×10 5 U2OS cells were seeded for transfections, which were performed as previously described [ 38 ]. Cells were co-transfected with the appropriate CAT reporter construct and GAL4-tagged p68/p72, and in each case the DNA was made up to a total of 10 μg with Bluescript DNA (Stratagene). CAT activity was determined 48 hr after transfection using 100 μg of total protein from cleared whole cell lysates. Typically each experiment was performed in triplicate within each assay, and each assay was repeated 3 to 5 times. Where applicable, the HDAC inhibitor Trichostatin A (TSA) (Upstate Biotechnology) was added, at a final concentration of 300 nM, 16 hr post transfection. Nuclear extract preparation and co-immunoprecipitation Nuclear extracts were prepared from U2OS and 293 cells essentially as described in [ 40 ] except that the NaCl concentration was reduced to 330 mM NaCl, then diluted to 150 mM NaCl (after nuclear lysis) and treated with RNAse/DNAse. The extract was pre-cleared in buffer D [20 mM Hepes (pH7.9), 150 mM NaCl, 0.5 mM DTT, 20% (v/v) glycerol, 10 mM NaF and protease inhibitor cocktail (Roche)] with protein G sepharose beads for 30 min. at 4°C. Immunoprecipitations were carried out in the presence of appropriate antibodies and protein G sepharose beads for one hour at 4°C. After washing in buffer D plus 0.1% Igepal (Sigma), immunoprecipitated proteins were separated by SDS-PAGE and western blotted using standard conditions and appropriate primary and secondary antibodies. Immunoreactive proteins were detected using the ECL method (Amersham). Gel filtration 293 cell lysates were prepared in RIPA buffer [50 mM Tris-HCl (pH8.0), 150 mM NaCl, 1% Igepal, 0.1% SDS, 1% Na deoxycholate, 1 mM EDTA and protease inhibitor cocktail (Roche)]. After treatment with RNase/DNase, lysates were fractionated on a Pharmacia Superose 6 HR column in 50 mM Tris-HCl (pH7.5), 150 mM NaCl, 10% glycerol and 1 mM benzamidine, using a Pharmacia AKTA FPLC system. 0.5 ml fractions were collected and alternate fractions were analysed by western blotting. Molecular weight standards from Pharmacia were used to calibrate the column. HDAC assay 293 cells were transfected with either pSG5-myc vector, pSG5-p68-myc, or pSG5-p72-myc. Cells were lysed in buffer B 48 hr after transfection. The lysate was then diluted in buffer A and myc-tagged proteins immunoprecipitated with 9E10 (myc epitope) antibody as described above. The HDAC assay employs Fleur de lys substrate, which contains an acetylated lysine side chain (Biomol) and was performed according to manufacturers instructions. Antibody-bound beads were washed in HDAC assay buffer prior to being added to the 96-well plate, to remove immunoprecipitation buffer. Reactions were incubated for 30 min at 37°C with or without the addition of 1 μM TSA. Samples were excited at 360 nm and emitted at 460 nm and were read in a fluorometer. Typically each assay was performed 3 times. Authors' contributions BJW carried out the transcriptional and HDAC activity assays, the deletion mapping and the gel filtration, GJB and SMN carried out the co-immunoprecipitation experiments, DJG carried out the original experiments showing that p68 and p72 could act as transcriptional repressors, and NDP and FFP participated in the design of this study. FFP co-ordinated the study and prepared the final draft of the manuscript. All authors read and approved the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514542.xml
535567	Spectrum of clinical disease in a series of 135 hospitalised HIV-infected patients from north India	Background Literature on the spectrum of opportunistic disease in human immunodeficiency virus (HIV)-infected patients from developing countries is sparse. The objective of this study was to document the spectrum and determine the frequency of various opportunistic infections (OIs) and non-infectious opportunistic diseases, in hospitalised HIV-infected patients from north India. Methods One hundred and thirty five consecutive, HIV-infected patients (age 34 ± 10 years, females 17%) admitted to a tertiary care hospital in north India, for the evaluation and management of an OI or HIV-related disorder between January 2000 and July 2003, were studied. Results Fever (71%) and weight loss (65%) were the commonest presenting symptoms. Heterosexual transmission was the commonest mode of HIV-acquisition. Tuberculosis (TB) was the commonest OI (71%) followed by candidiasis (39.3%), Pneumocystis jiroveci pneumonia (PCP) (7.4%), cryptococcal meningitis and cerebral toxoplasmosis (3.7% each). Most of the cases of TB were disseminated (64%). Apart from other well-recognised OIs, two patients had visceral leishmaniasis. Two cases of HIV-associated lymphoma were encountered. CD4+ cell counts were done in 109 patients. Majority of the patients (82.6%) had CD4+ counts <200 cells/μL. Fifty patients (46%) had CD4+ counts <50 cells/μL. Only 50 patients (37%) received antiretroviral therapy. Twenty one patients (16%) died during hospital stay. All but one deaths were due to TB (16 patients; 76%) and PCP (4 patients; 19%). Conclusions A wide spectrum of disease, including both OIs and non-infectious opportunistic diseases, is seen in hospitalised HIV-infected patients from north India. Tuberculosis remains the most common OI and is the commonest cause of death in these patients.	Background The first case of human immunodeficiency virus/acquired immunodeficiency syndrome (HIV/AIDS) in India was detected in 1986 in the state of Tamilnadu [ 1 ] and since then the spread of HIV/AIDS across the nation has been relentless. Cases have been reported from all states and union territories of India. Though the overall prevalence of HIV infection is low (<1%), the total number of cases is high. As per estimates, by the end of 2002, about 4.58 million adults were infected with HIV [ 2 ]. Even more ominous has been the shift of the epidemic from high-risk groups such as injectable drug users (IDU) and patients with other sexually transmitted diseases, to low-risk groups like married, monogamous women [ 3 ]. Though the majority of HIV-infected population lives in developing nations, there is a paucity of data on natural history, pattern of disease and survival of hospitalised patients with HIV/AIDS from these regions, especially India. It is well established that manifestations of AIDS are influenced by factors such as endemic infections and malnutrition that are widely prevalent in these regions [ 4 ]. Conventional disease staging criteria, which were developed in western populations may not hold good in these settings [ 4 , 5 ]. Added to this, resource constraints prohibit evaluation and decision-making based on cost and labour-intensive methods such as CD4+ cell counts and viral RNA load estimation. Timely initiation of prophylaxis for opportunistic infections (OIs) and their prompt recognition and treatment are the only economically viable options [ 6 ]. In this scenario, knowledge regarding the pattern of OIs, will be useful. This study was conducted to elucidate the frequency of various OIs and non-infectious opportunistic conditions in hospitalised HIV-infected patients, from north India. Methods We describe a series of 135 consecutive patients with HIV/AIDS, aged 13 years and above, admitted to the All India Institute of Medical Sciences (A.I.I.M.S.) hospital, New Delhi during the period of January 2000 through July 2003. A.I.I.M.S. is a large tertiary level teaching hospital and referral centre located in north India, attending to HIV-infected patients apart from others. The catchment area of this hospital includes the New Delhi national capital territory and neighbouring states mainly, Uttar Pradesh, Bihar, Uttharanchal, Haryana, Punjab and Himachal Pradesh. All patients were under the care of a single unit, of the three units existing in the Department of Medicine. Decision to admit was taken by the treating physician and all patients were hospitalised for the evaluation and treatment of a suspected OI or HIV-related disorder. We have earlier reported the determinants of in-hospital mortality in this series [ 7 ]. Diagnosis of HIV infection was made using methods described previously [ 8 ]. All patients referred to A.I.I.M.S. as 'HIV-positive' underwent retesting. Apart from this, all patients treated for tuberculosis (TB), those reporting risk factors for HIV infection and patients with clinical presentation suggestive of underlying immunosuppression, were offered HIV testing. All except one patient were infected with HIV-1, the latter with HIV-2. CD4+ lymphocyte counts were done by flowcytometry, using a FACSCalibur flowcytometer (Becton Dickinson, USA) in patients who could afford for the same. Patients were evaluated using a predesigned instrument regarding the demographic characteristics, risk factors for HIV infection, presenting symptoms, physical findings and laboratory parameters. Various laboratory abnormalities were defined using cut-off values as follows: anaemia – haemoglobin <100 g/L; leucopenia – total leucocyte count (TLC) <4 × 10 9 /L; lymphocytopenia – absolute lymphocyte count (ALC) <1.2 × 10 9 /L; decreased CD4+ – <500 cells/μL; thrombocytopenia – platelet count <150 × 10 9 /L; elevated erythrocyte sedimentation rate (ESR) – >20 in the first hour; hypoalbuminaemia – albumin <35 g/L; hyperglobulinaemia – globulin >40 g/L; hyperbilirubinaemia – serum total bilirubin >20.5 μmol/L; elevated liver enzymes – >2 times the upper limits of normal [aspartate aminotransferase (AST) – >100 IU/L, alanine aminotransferase (ALT) – >100 IU/L, alkaline phosphatase (ALP) – >560 IU/L]; hypoxaemia – PaO 2 <12 kPa; increased arterial-alveolar oxygen gradient (A-a DO 2 ) – >3.3 kPa, while breathing ambient air (F i O 2 = 0.21). While genuine efforts were made to establish the diagnosis of an OI, patients at times were too sick to undergo invasive diagnostic procedures such as bronchoscopy or tissue biopsy and hence the following criteria were used to define an OI concerned. Pulmonary tuberculosis (PTB): clinical features suggestive of TB with radiological features compatible with TB on chest radiograph or computed tomographic scan (CT) and/or demonstration of acid-fast bacilli (AFB) in sputum smears or growth of Mycobacterium tuberculosis in sputum culture; disseminated tuberculosis (DTB): clinical features suggestive of TB with concurrent involvement of at least two non-contiguous organs, in the presence of bacteriological and/or histopathological evidence of TB and improvement with anti-tuberculosis therapy; miliary tuberculosis (MTB): clinical presentation consistent with TB and bacteriological and/or histopathological evidence of TB and demonstration of bilateral miliary infiltrates on chest radiograph or high-resolution CT; tuberculosis meningitis (TBM): clinical and cerebrospinal fluid (CSF) features consistent with tuberculosis meningitis, with bacteriologic demonstration of M. tuberculosis in CSF or after exclusion of other aetiologies such as cryptococcus and syphilis, with evidence of tuberculosis elsewhere and/or improvement with anti-tuberculosis therapy; cryptococcal meningitis: in CSF, demonstration of Cryptococcus sp. yeast cells by India ink or antigen by latex agglutination or growth in culture; cerebral toxoplasmosis: demonstration of multiple ring-enhancing cerebral parenchymal lesions on contrast-enhanced CT or magnetic resonance imaging (MRI), in the presence of anti-toxoplasma antibody in serum and clinical response to anti-toxoplasma therapy; progressive multifocal leucoencephalopathy (PMLE): compatible clinical presentation with demonstration of characteristic cerebral white matter changes by MRI; Pneumocystis jiroveci pneumonia (PCP): bilateral, diffuse interstitial infiltrates on chest radiograph or high-resolution CT, with hypoxaemia (PaO 2 <12 kPa) and sputum smears/cultures negative for aerobic bacteria and AFB and/or demonstration of Pneumocystis jiroveci in induced sputum [ 9 ]. Patients received primary chemoprophylaxis for PCP and toxoplasmosis and received treatment for specific OIs followed by secondary prophylaxis, as per recommendations [ 10 ]. For patients with PCP, whenever hypoxaemia was severe (PaO 2 <9.3 kPa), corticosteroids were given in addition to oral co-trimoxazole. None of these patients received assisted ventilation. All patients were counselled and offered antiretroviral therapy, whenever indicated. Upon discharge from the hospital, patients were advised to follow-up as outpatients, on a regular and as needed basis. Statistical analysis Entry and analysis of all data were done using a statistical software package (SPSS for Windows, Version 10.0, SPSS Inc., Chicago, IL). Entered data were double-checked for discrepancies. Data are presented as mean ± standard deviation (SD) when distributed normally and as median with interquartile range (IQR), if the distribution was skewed. Frequency of various clinical and laboratory findings and the frequencies of individual OIs are expressed as proportions (%). The relation between ALC and CD4+ count was analysed by Pearson's product moment correlation. Paired-samples t-test was applied to assess the improvement in CD4+ cell counts following antiretroviral therapy. Results Over a period of about three and a half years, 135 patients with HIV/AIDS were admitted and all these patients were included in the study. Mean age of the study group was 34 ± 10 (range: 13–73) years. Females constituted less than one fourth of the study group (17%). Details of other demographic characteristics and transmission categories are presented in Table- 1 . More than two-thirds of patients (76%) were in the second and third decades of their lives. Labourers were the commonest occupational group (40%) and long-haul truck drivers constituted a considerable proportion of the study group. Professionals were the least common occupational group. An identifiable risk factor for HIV infection was present in 59% of patients, extramarital heterosexual contact being the commonest (41.5%). None of the patients reported homosexual practices. About 40% of patients denied any risk factor for the acquisition of HIV infection. A large number of patients (44.4%) were from states other than Delhi. Three symptoms namely fever, weight loss and diarrhoea were the common symptoms at presentation (70.4%, 65.2% and 23.7% respectively). Productive cough and dyspnoea were present in about a fourth of patients (Table- 2 ). Two-thirds of patients (66.7%) were malnourished (body mass index <19 kg/m 2 ) and generalised lymphadenopathy was present in a considerable proportion of patients (16.3%). Poor performance on mini mental status examination (MMSE score <23) was not uncommon (20.7%). Altered sensorium and focal neurologic deficit were encountered occasionally. Anaemia was present in about half of the patients (50.5%) and among those who where anaemic, 17 (21.8%) patients were on zidovudine. A considerable number of patients were leucopenic and in 22% of patients ALC was less than 1200/μL (Table- 3 ). CD4+ cell counts were done in 109 patients. The distribution of CD4+ cell counts is shown in Figure- 1 . Most of the patients (n = 90; 82.6%) had CD4+ counts less than 200 cells/μL. Fifty patients (46%) had CD4+ counts less than 50 cells/μL. The correlation of ALC with CD4+ count was not significant (r = 0.14; P = 0.18). HIV viral load estimation was done in only four patients (range 33752–289176 RNA copies/mL). Twenty patients (14.8%) had hypoxaemia. Of these, five patients had PCP (Figure 2-A ), ten had DTB, three had extensive PTB and one patient had massive unilateral tuberculosis pleural effusion. The mean number of OI was 1.4 per patient. The commonest OI was TB (71.1%), followed by oral candidiasis (39.3%). Most of the cases of TB were disseminated (64%; Figure 2-B ) and PTB was seen in only a small number of patients (8.1%). Eleven patients had MTB (Table- 4 ). Of the 53 patients with oral candidiasis, five patients had concomitant oesophageal involvement. Cryptosporidium sp. and Giardia sp. infection causing diarrhoea was found in three and two patients respectively. Of the 13 cases of meningitis, eight were tuberculosis and the rest due to Cryptococcus sp.; no case of acute, pyogenic meningitis was seen. Cerebral toxoplasmosis (Figure 2-C ) and one case of PMLE were the other central nervous system infections seen. Of the five cases of cerebral toxoplasmosis, one patient presented as recent onset dilated cardiomyopathy with no features to suggest a central nervous system pathology. On further investigation, the patient was found to be seropositive for anti-toxoplasma antibodies and contrast enhanced CT revealed multiple ring enhancing cerebral lesions. The cerebral lesions cleared and cardiac ejection fraction improved with anti-toxoplasma treatment. The cardiac dysfunction was probably due to toxoplasma myocarditis. Ten patients presented with PCP (Figure 2-A ) and three patients had cytomegalovirus (CMV) retinitis (Figure 2-D ). One patient was diagnosed as disseminated histoplasmosis. This patient presented as fever of unknown origin. Bone marrow biopsy revealed caseating granulomas and subsequently fungal culture grew Histoplasma capsulatum . Non-infectious AIDS-defining conditions were not common. One case of B-cell non-Hodgkin's lymphoma presenting as unilateral maxillary swelling (Figure 2-E ) and another case of Hodgkin's lymphoma that presented as generalised lymphadenopathy with hepatosplenomegaly, were seen. Only 50 patients (37%) received antiretroviral therapy, of which three patients were taking protease inhibitor (PI) based regimens. All other patients received non-nucleoside reverse transcriptase inhibitor (NNRTI) based (nevirapine – 46 patients, efavirenz – one patient) triple-drug regimens. Follow-up CD4+ counts were done in 16 patients after three to six months of antiretroviral therapy. The mean CD4+ count increased to 224 ± 178 cells/μL from a baseline mean of 131 ± 122 cells/μL, which was statistically significant (P = 0.003). Myelosuppression during antiretroviral therapy developed in four patients, who were taking regimens containing zidovudine. One patient who was on didanosine developed acute pancreatitis (Figure 2-F ). Another patient presented with severe, uncompensated, wide anion-gap metabolic acidosis with no apparent cause. This was probably due to lactic acidosis caused by stavudine, which he had been taking. Twenty one patients (15.6%) died in hospital, most of them due to TB (16 patients) and PCP (4 patients). All patients who died in hospital except for one, had CD4+ counts less than 200 cells/μL (Figure- 1 ). Discussion Opportunistic infections (OIs) are the major cause of morbidity and mortality in patients with HIV infection. In resource-limited settings, knowledge regarding the prevalence of various OIs might aid in making decisions regarding empirical treatment and would help to prioritise limited resources. To this end, we studied the frequency of various opportunistic conditions in this series of 135 HIV-infected patients. Nearly a half of these patients came from places other than New Delhi, being referred by medical practitioners or approached on their own. In India, HIV-infected patients are usually treated in tertiary level centres, due to lack of expertise and facilities in primary and district level hospitals. Moreover, for reasons of privacy patients and their kin prefer treatment at places far off from their home towns. This series comprises predominantly of patients from the northern states of India. Hence the findings of the present study may not apply to south India and north-east Indian states. Moreover, being a hospital based study, patients with severe illness causing rapid death and patients with milder symptoms who prefer treatment from local health facilities may not have been proportionately represented in the study group. Most of the patients were in the age group 21–40 years and males were predominantly affected. This is similar to nation-level statistics in which, of the 57781 cases of HIV/AIDS reported to the National AIDS Control Organisation (NACO), 89% of the cases were in the age group 15–44 years and 74% were males [ 11 ]. This section of the population is more affected because they are sexually more active and the social structure is patriarchal. Unfortunately, these patients also happen to be in the economically most productive years of their lives. The male preponderance might have been due to the fact that in the existing social milieu, females do not seek medical care fearing ostracism and loss of family support. Interestingly, a large number of patients (41%) did not give a history of any of the known risk factors for the acquisition of HIV infection and homosexual practice was singularly not seen. It is likely that these patients were reluctant to reveal, due to the prevailing social values, which discourage polygamy and homosexuality. For the same reason, despite a history of prior blood transfusion in a considerable proportion of patients, it may not be possible to ascribe causation; not all these patients would have acquired HIV infection through blood transfusion. Injectable drug users (IDU) constituted only a minority of the study group as has been observed in other parts of India except for the north-eastern states where IDU is widely prevalent [ 12 ]. The three symptoms, namely fever, weight loss and diarrhoea were common presenting features. In this series of patients, all of whom had some opportunistic condition at presentation indicating advanced HIV disease, wasting is anticipated. The association of severe weight loss at presentation with TB in HIV-infected patients has been reported before [ 13 ]. TB was very common in this series and was the commonest OI. Previous studies also confirm the high prevalence of TB in HIV-infected individuals in India [ 4 , 9 , 14 , 15 ]. Most of these cases of TB were either disseminated or extra-pulmonary. We have previously reported the increased prevalence of HIV seropositivity among cases of extrapulmonary TB as compared to PTB and as such, HIV co-infection is now on the rising trend among cases of TB, in this population [ 16 ]. This trend is in consonance with findings from across the globe. The cases of TB attributable to HIV are increasing and have reached alarming proportions in certain countries. An example of such a nation is the United States of America, where almost 26% of cases of TB are now attributable to HIV infection [ 17 ]. Although these figures are from countries where the incidence of TB has otherwise been low, it is bound to be similar in countries where TB is endemic. PCP was seen in only a small number of patients. This is in sharp contrast with western populations where PCP is the commonest AIDS-defining illness [ 18 ]. Earlier studies from India also have reported a low prevalence (6–7%) of PCP [ 9 , 14 ]. The low prevalence of PCP in developing countries is well known and the incidence is probably increasing [ 19 ]. The frequencies of various OIs as observed in the present study are similar to those reported earlier from other parts of India [ 4 , 9 , 14 , 15 ]. Some of the opportunistic conditions were conspicuous by their absence. Disseminated infection caused by the fungus Penicillium marneffei has been reported to be the third commonest OI in southeast Asia [ 20 ] and is known to be highly prevalent in the north-east Indian state, Manipur [ 21 ] which shares common geographic, ecologic and climatic conditions with the south-east Asian regions of P. marneffei endemicity. No case of P. marneffei infection was found in the present study. Kaposi's sarcoma, atypical mycobacterial infection and disseminated CMV disease were not seen, as has been observed by a previous study from south India [ 14 ]. It appears that, TB being an endemic infection in this population, takes its toll before other less virulent pathogens could come into play. Development of TB per se might have afforded some protection against atypical mycobacterial disease [ 22 ]. It is also possible that TB masked the recognition of other OIs, which might have been present since it is well known that a large number of potentially fatal OIs are not diagnosed antemortem [ 23 , 24 ]. In India, Kaposi's sarcoma occurring in the setting of HIV infection and other immunocompromised conditions is a rarity, though not undescribed [ 25 - 27 ]. As such the prevalence of human herpes virus-8 (HHV-8) infection among the healthy population in India is low when compared with African nations and this is also likely to be responsible for the rarity of Kaposi's sarcoma [ 28 ]. Apart from the well-recognised OIs, infection unique to the local population namely, visceral leishmaniasis occurs in association with HIV and this has to be borne in mind. Overall in-hospital mortality in this series is considerable and is probably reflective of the advanced nature of disease at presentation. This is evidenced by the fact that more than 80% of patients had CD4+ counts less than 200 cells/μL and all but one deaths occurred in this group. However, as we had reported earlier, CD4+ cell counts do not determine the immediate outcome of patients with an OI, whereas the latter does [ 7 ]. The proximate causes of death in these patients are OIs, which are amenable to effective treatment. Many of these patients were from poor socioeconomic strata and were not able to afford for antiretroviral therapy. Cost constraints directly or indirectly, probably also resulted in poor compliance with scheduled follow-up visits. With the implementation of World Health Organisation sponsored '3 by 5' initiative, it is hoped that antiretroviral therapy will be within the reach for a large number these patients and also compliance is likely to become better. Conclusions A wide spectrum of OIs is seen in hospitalised patients with HIV/AIDS from north India. Tuberculosis is the most common OI in this group of patients. Though uncommon, non-infectious opportunistic conditions also occur. Many patients suffer from more than one OI. Hospital mortality is significant, with TB and PCP being responsible for most of these deaths. Whenever a definite diagnosis is not established quickly, empirical treatment should be considered. Such an approach is likely to improve immediate outcome in hospitalised patients with HIV/AIDS. Competing interests The author(s) declare that they have no competing interests. Authors' contributions SKS conceived of the study, designed and coordinated the study and revised the manuscript critically for important intellectual content. TK participated in the design of the study and statistical analysis and drafted the manuscript. AB participated in the design of the study, performed the statistical analysis and revised the manuscript for important intellectual content. TG participated in the design of the study, collected the clinical data and contributed significantly to the content of the manuscript. IB participated in the design of the study, collected the clinical data and participated in drafting the manuscript. PKS participated in the design of the study and coordination and also contributed significantly to the content of the manuscript. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535567.xml
546204	Prolactin daily rhythm in suckling male rabbits	Background This study describes the 24-h changes in plasma prolactin levels, and dopamine (DA), serotonin (5HT), gamma-aminobutyric acid (GABA) and taurine concentration in median eminence and adenohypophysis of newborn male rabbits. Methods Animals were kept under controlled light-dark cycles (LD 16:8, lights on at 08:00 h), housed in individual metal cages, and fed ad libitum with free access to tap water. On day 1 after parturition, litter size was standardized to 8–9 to assure similar lactation conditions during the experiment. Groups of 6–7 suckling male rabbits were killed by decapitation on day 11 of life at six different time points during a 24-h period. Results Plasma prolactin levels changed significantly throughout the day, showing a peak at the beginning of the active phase (at 01:00 h) and a second maximum during the first part of the resting phase (at 13:00 h). Median eminence DA concentration also changed significantly during the day, peaking at the same time intervals as plasma prolactin. A single maximum (at 13:00 h) was found for adenohypophysial DA concentration. Individual adenohypophysial DA concentrations correlated significantly with their respective plasma prolactin levels. A maximum in median eminence 5HT concentration occurred at 21:00 h whereas adenohypophysial 5HT peaked at 13:00 h. Median eminence 5HT concentration and circulating prolactin correlated inversely. In the median eminence, GABA concentration attained maximal values at 21:00 h, whereas it reached a maximum at 13:00 h in the pituitary gland. Median eminence GABA concentration correlated inversely with circulating prolactin. In the median eminence, taurine values varied in a bimodal way showing two maxima, at the second half of the rest span and of the activity phase, respectively. In the adenohypophysis, minimal taurine levels coincided with the major plasma prolactin peak (at 01:00 h). Circulating prolactin and adenohypophysial taurine levels correlated inversely. Conclusion The correlations among the changes in the neurotransmitters analyzed and circulating prolactin levels explain the circadian secretory pattern of the hormone in newborn male rabbits.	Background The mechanisms that regulate prolactin secretion are complex [ 1 ]. Two major regulatory inhibitory inputs for prolactin secretion are dopamine [ 2 ] and gamma-aminobutyric acid (GABA) [ 3 - 6 ]. In addition, many other neuromodulators have been implicated in the control of prolactin secretion, among them, vasoactive intestinal peptide, thyrotropin releasing hormone and serotonin (5HT) [ 1 ]. More recently, taurine has also been implicated in the regulation of prolactin secretion [ 1 ]. It is well known that basal secretion of prolactin varies throughout the day, describing a characteristic pattern with maximal values close to the light-dark transition [ 7 , 8 ]. Such a circadian pattern has been described not only in rodents (rat and mouse) but also in many other species [ 1 ]. In the rat, we previously demonstrated changes of the secretory pattern of prolactin along the year [ 7 - 11 ], as well as a function of aging [ 12 , 13 ]. The rat is very immature at birth, so that newborn and suckling rats are very sensitive to manipulations that can affect adulthood [ 14 - 18 ]. Circadian rhythms of developing mammals seem to be entrained by the rhythmicity of their mother [ 19 , 20 ], and several studies have indicated that maternal melatonin is necessary to entrain the circadian rhythms in the newborn [ 21 , 22 ]. The rabbit is probably the best-studied laboratory animal in the wild, due to its abundance, size and importance as an agricultural pest [ 23 , 24 ]. Wild and laboratory rabbits are essentially nocturnal and display a clear daily pattern of activity [ 25 ]. The rabbit possesses a number of behavioral specializations that make it uniquely suited for circadian studies. Female rabbits visit their altricial young only for a few minutes once every 24 h to nurse, and survival of the young depends on the tight circadian-controlled synchronization in behavior and physiology with the mother. This unusual pattern of maternal care and the demands it places on the litter provide an excellent opportunity to analyze circadian rhythms during early development [ 25 ]. In contrast to the large amount of information available on circadian rhythms in adult mammals, studies on circadian phenomena in neonates are few [ 26 , 27 ]. For example, in 21 day-old male rats the daily circadian pattern of prolactin secretion seen in adults is absent [ 18 ]. Considering that no information on circadian rhythmicity of prolactin secretion in neonatal male rabbits is available, we undertook the present study to analyze whether neonatal male rabbits show defined 24-h changes in plasma prolactin levels and whether neonatal male rabbits show circadian changes in DA, 5HT, GABA and taurine concentration in median eminence and the adenohypophysis, all of which are well known modulators of prolactin secretion. Methods Animals This study was performed using 24 multiparous, lactating Californian × New Zealand White crossbreed doe rabbits. Animals were housed in research facilities of the Animal Production Department. They were maintained under controlled light-dark cycles (LD 16:8, light on at 08:00 h), housed in individual metal cages, fed at libitum using a commercial pellet diet (Lab Rabbit Chow, Purina Mills, Torrejón de Ardoz, Madrid, Spain) with free access to tap water. On day 1 after parturition, litter size was standardized to 8–9 by adding or removing kits to assure similar lactation conditions during the experiment. This study was performed according to the CEE Council Directive (86/609, 1986) for the care of experimental animals. Groups of 6–7 suckling male rabbits were killed by decapitation on day 11 of life at six different time points throughout a 24-hour cycle. The brains were quickly removed, and the median eminence and the anterior pituitary were taken out. Anterior pituitaries were weighed and homogenized in chilled (0–1°C) 2 M acetic acid. After centrifugation (at 15000 × g for 30 min, at 5°C), the samples were either analyzed for DA and 5HT or boiled for 10 min and further centrifuged at 14000 rpm for 20 min to measure GABA and taurine. Hormone assay Plasma prolactin levels were measured by a specific homologous RIA method [ 28 ] using AFP-991086 antibody supplied by the National Institutes of Health (NIH, Bethesda, MD, USA) and Dr. A. F. Parlow (Harbour-UCLA Medical Center, CA, USA). The titer of antibody used was 1:62,500. The PRL standard used was RbPR L -RP-1. Hormone was labeled with 125 I by the chloroamine-T method [ 29 ]. The volume of plasma for PRL determinations was 10 μl. Staphylococcus aureus (prepared by the Department of Plant Physiology, U.A.M., Madrid, Spain) was used to precipitate the bound fraction [ 28 ]. All samples were measured in the same assay run to avoid inter-assay variations. The sensitivity of the assay for PRL was 0.125 ng/ml and the intra-assay coefficient of variation was < 5%. The intra-assay coefficient of variation was calculated using a pool of plasma measured ten times in the same assay; mean (± S.E.M.) concentration was 106.9 ± 4.1 ng/ml. Catecholamine and indoleamine analysis DA and 5HT concentration was measured by high pressure liquid chromatography (HPLC) using electrochemical detection (Coulochem, 5100A, ESA; USA), as described elsewhere [ 12 ]. A C-18 reverse phase column eluted with a mobile phase (pH 4. 0.1 M sodium acetate, 0.1 M citric acid, 0.7 mM sodium octylsulphate and 0.57 mM EDTA containing 10% methanol, v/v) was employed. Flow rate was 1 ml/min, at a pressure of 2200 psi. Fixed potentials against H 2 /H + reference electrode were: conditioning electrode: -0.4 V; preoxidation electrode: +0.10; working electrode: +0.35 V. Indoleamine and catecholamine concentration was calculated from the chromatographic peak heights by using external standards and was expressed as pg/μg protein. The linearity of the detector response for DA and 5HT was tested within the concentration ranges found in median eminence and adenohypophysial supernatants. Amino acid analysis Amino acids were isolated and analyzed by HPLC with fluorescence detection after precolumn derivatization with O-phthalaldehyde (OPA) as described elsewhere [ 30 ]. An aliquot of the tissue supernatant containing homoserine as an internal standard was neutralized with 4 M NaOH and was then incubated at room temperature with OPA reagent (4 mM OPA, 10% methanol, 2.56 mM 2-mercaptoethanol, in 1.6 M potassium borate buffer, pH 9.5) for 1 min. The reaction was stopped by adding acetic acid (0.5 % v/v). Samples were immediately loaded through a Rheodyne (Model 7125) injector system (50 μl loop) to reach a C-18 reverse-phase column (4.6 mm ID × 150 mm, Nucleosil 5, 100A). Elution was achieved by means of a mobile phase consisting of 0.1 M sodium acetate buffer (pH 6.5) containing 35 % methanol, at a flow rate of 1 mL/min and a pressure of 140 Bars. The column was subsequently washed with the same buffer containing 70 % methanol and re-equilibrated with the elution buffer before re-use. The filter fluorometer was set at the following wavelengths: excitation: 340 nm, emission: 455 nm. The procedure allowed a distinct separation and resolution of the amino acids measured. Amino acid content was calculated from the chromatographic peak heights by using standard curves and the internal standard. The linearity of the detector response was tested within the concentration ranges found in median eminence and adenohypophysial extracts. Statistics Statistical analysis of results was performed by a one-way analysis of variance (ANOVA) followed by post-hoc Tukey-Kramer's multiple comparisons tests. Curve estimation in regression analysis was made by using SPSS software, version 10.1 (SPSS Inc., Chicago, ILL). P values lower than 0.05 were considered evidence for statistical significance. Results Figure 1 shows the levels of prolactin throughout the day in suckling male pups. Plasma prolactin levels changed significantly throughout the day (F = 21.1; p < 0.0001), showing two maxima, a major one at the beginning of the active phase (at 01:00 h) and a second one during the first part of the resting phase (at 13:00 h). Figure 1 24-h changes in plasma prolactin levels of 11 days old male rabbit pups. Groups of 6–7 pups were killed by decapitation at 6 different time intervals throughout a 24 h cycle. Bar indicates scotophase duration. Results are the means ± SEM. a p < 0.01 vs. all time points. b p < 0.01 vs. 01:00 h, 05:00 h and 13:00 h, Tukey-Kramer's multiple comparisons test. For further statistical analysis, see text. Figures 2 , 3 , 4 , 5 depict the changes in median eminence and adenohypophysial concentration of DA, 5-HT, GABA and taurine. Mean plasma prolactin concentration is plotted as a reference in every case. Figure 2 24-h changes in median eminence and adenohypophysial DA concentration in 11 days old male rabbit pups. Groups of 6–7 pups were killed by decapitation at 6 different time intervals throughout a 24 h cycle. Bar indicates scotophase duration. Results are the means ± SEM. Circulating prolactin levels are shown in shaded line. Letters indicate the existence of significant differences between time points within each tissue after a Tukey-Kramer's multiple comparisons test, as follows: a p < 0.01 vs. all time points. b p < 0.01 vs. 05:00 h, 09:00 h and 21:00 h. c p < 0.01 vs. 05:00 and 21:00 h. For further statistical analysis, see text. Figure 3 24-h changes in median eminence and adenohypophysial 5HT concentration in 11 days old male rabbit pups. Groups of 6–7 pups were killed by decapitation at 6 different time intervals throughout a 24 h cycle. Bar indicates scotophase duration. Results are the means ± SEM. Circulating prolactin levels are shown in shaded line. Letters indicate the existence of significant differences between time points within each tissue after a Tukey-Kramer's multiple comparisons test, as follows: a p < 0.01 vs. all time points. b p < 0.01 vs. 01:00 h, 09:00 h, 17:00 and 21:00 h. For further statistical analysis, see text. Figure 4 24-h changes in median eminence and adenohypophysial GABA concentration in 11 days old male rabbit pups. Groups of 6–7 pups were killed by decapitation at 6 different time intervals throughout a 24 h cycle. Bar indicates scotophase duration. Results are the means ± SEM. Circulating prolactin levels are shown in shaded line. Letters indicate the existence of significant differences between time points within each tissue after a Tukey-Kramer's multiple comparisons test, as follows: a p < 0.01 vs. all time points. b p < 0.01 vs. 01:00 h, 05:00 h and 09:00 h, p < 0.05 vs. 17:00 h. For further statistical analysis, see text. Figure 5 24-h changes in median eminence and adenohypophysial taurine concentration in 11 days old male rabbit pups. Groups of 6–7 pups were killed by decapitation at 6 different time intervals throughout a 24 h cycle. Bar indicates scotophase duration. Results are the means ± SEM. Circulating prolactin levels are shown in shaded line. Letters indicate the existence of significant differences between time points within each tissue after a Tukey-Kramer's multiple comparisons test, as follows: a p < 0.01 vs. all time points. b p < 0.01 vs. 01:00 h, 09:00 h and 13:00 h. For further statistical analysis, see text. Median eminence DA concentration changed in a bimodal way as a function of time of day, showing two maxima, coinciding with those of plasma prolactin at the active and resting phase of the diurnal cycle (F = 14.1; p < 0.0001, Figure 2 ). In the case of adenohypophysial DA concentration, a single maximum occurred during the first half of the rest phase (at 13:00 h) (F = 29.9; p < 0.0001). Only in the adenohypophysis, plasma prolactin and DA concentration correlated in a direct way. This correlation was best described by a log model with r 2 = 0.16, b 0 = -123.7 and b 1 = 18.1 (F = 4.69, p= 0.04). As shown in Figure 3 , a maximum in median eminence 5HT concentration occurred at the second half of the rest span (F = 64.1; p < 0.0001) whereas a maximum in adenohypophysial 5HT levels was found at the first half of rest span. Circulating prolactin and median eminence 5HT concentration correlated inversely in a linear way (r 2 = 0.18, b 0 = 677.6 and b 1 = -4.9, F = 5.3, p < 0.03). Figure 4 shows the changes in median eminence and adenohypophysial GABA concentration. In the median eminence, GABA concentration attained maximal values at the rest phase, with a peak at late evening (i.e. at 21:00 h, F = 11.1, p < 0.0001). In the anterior pituitary, GABA concentration reached a maximum at 13:00 h (F = 21.6, p < 0.0001). Circulating prolactin and median eminence GABA concentration correlated inversely in a linear way (r 2 = 0.21, b 0 = 25.7 and b 1 = -0.22, F = 6.6, p < 0.01). Figure 5 depicts the 24-h changes in taurine concentration. In the median eminence, taurine values varied in a bimodal way showing a peak at the second half of the rest period, a nadir at the early activity span (coinciding with the prolactin peak) and a second maximum late in the activity phase (at 05:00 h, F = 32.9, p < 0.0001). Likewise, in the adenohypophysis, taurine levels exhibited minimal values at the time of the prolactin peak (i.e., at 13:00 h, F = 21.6, p < 0.0001). Circulating prolactin and adenohypophysial taurine levels correlated inversely in a linear way (r 2 = 0.42, b 0 = 11.6 and b 1 = -0.11, F = 17.4, p < 0.0001). Discussion The present study, performed in neonatal male rabbit pups sacrificed at 6 different time intervals during a 24-h cycle, describes for the first time significant changes in plasma prolactin levels throughout the day. In concomitant measurements of median eminence and adenohypophysial concentration of DA, 5HT, GABA and taurine, a clear daily pattern was found in almost every case. Contrasting with neonatal rats that did not display any circadian pattern of plasma prolactin [ 18 ], a daily rhythm of plasma prolactin occurred in neonatal male rabbits, with a maximal value attained 1 h after lights-off (at 01:00 h) and a secondary peak found during the first part of the resting phase (at 13:00 h). In adult rabbits, daily patterns of prolactin secretion depend on light/dark phases [ 25 ]. The present results indicate that, already on day 11 of life, male rabbit pups display daily changes in plasma prolactin levels, remarkably similar to those described in adult male rats (e.g., the maximum displayed 1 h after the dark onset) [ 7 - 10 ]. The activity of several nuclei of rabbit hypothalamus increases with age and with experience of anticipatory arousal [ 27 ]. However, no study has been published on the regulatory mechanism of prolactin in rabbits. Considering that DA is the major inhibitory input for prolactin secretion [ 1 , 32 ], the present study indicating that DA concentration in median eminence of rabbit pups is high during the rest phase of the day (when plasma prolactin levels are low), and decreases at day-night transition (coinciding with the increase in circulating prolactin), may support a cause-effect relationship. The afternoon decrease in median eminence DA concentration could be a prerequisite for prolactin release in neonatal male rabbits [ 2 ]. However, median eminence DA concentration of male rabbit pups also presents a peak during the activity phase (01:00 h) associated with the highest prolactin levels. Therefore, the data suggest that the inhibitory regulatory influence of DA on prolactin secretion is exerted mainly during the light phase of the photoperiod, whereas during the dark phase other hypothalamic neuromodulators could be operative, as it was previously described in rats [ 13 ]. These hypotheses must be tested rigorously (e.g., by using pharmacological blocking agents) before a definitive conclusion can be made. Among other possible neuromodulators of prolactin secretion, the arcuate nucleus receives a dense serotonergic innervation consisting of a population of brainstem neurons arising mainly from the midbrain raphe nuclei [ 33 ] and from fibers originated in 5HT cell bodies located within the hypothalamus. There is a close proximity of 5HT fibers to dopaminergic cell bodies in the arcuate nucleus [ 34 ]. Therefore, an indirect effect of 5HT on prolactin release could be linked to the modulation of the inhibitory dopaminergic inputs to the pituitary. Our foregoing results agree with this hypothesis since 5HT concentration in median eminence changes diurnally in an opposite way to that of plasma prolactin levels, albeit without a significant correlation between them. Indeed, previous experiments in rats indicated that 5HT could probably modulate directly the secretion of prolactin [ 13 ]. Taurine has also been implicated in the regulation of prolactin release [ 5 , 13 , 35 , 36 ]. The foregoing results indicate that in median eminence and anterior pituitary of male rabbit pups taurine concentration varies inversely to plasma prolactin levels, displaying a mirror pattern. In the adenohypophysis a negative correlation between plasma prolactin and taurine levels was found, similarly to previous data obtained in rats [ 13 ]. Therefore, taurine may play a role in prolactin regulation in newborn rabbits. A relatively dense innervation of GABA terminals exists in the external layer of the median eminence [ 37 ], and the ability of median eminence neurons to release GABA in portal blood has been demonstrated [ 38 ]. We previously demonstrated a possibly inhibitory control of GABA on prolactin secretion during the activity phase in male rats [ 3 - 6 ]. Results obtained in the present study in suckling male rabbits support such an inhibitory effect of GABA on plasma prolactin levels exerted mainly during the dark phase of daily photoperiod. The data indicate that GABA concentration in median eminence decreased during the day-night transition, while plasma prolactin levels were increasing. Actually, in median eminence a negative correlation between GABA concentration and plasma prolactin was found, thus suggesting an inhibitory effect of GABA on prolactin secretion. GABA acting on specific receptors in the anterior pituitary has been reported to suppress prolactin secretion [ 39 , 40 ], although whether this effect was physiological has been questioned [ 40 ]. Data from literature suggest that the role of GABA on prolactin release is quite complex [ 41 ]. In some conditions, such as aging [ 13 ] or hyperprolactinemia [ 6 ], the inhibitory role of GABA becomes more pronounced whereas the inhibitory control exerted by DA diminishes. Our results in male rabbit pups indicated that, although no correlation between plasma prolactin and pituitary GABA concentration was found, the pattern may confirm the main role of this amino acid in the control of prolactin secretion during the dark phase of the photoperiod that was developed later. Again, all these hypotheses must be tested. e.g. pharmacologically, before a definitive conclusion on this matter can be drawn. Conclusions In suckling male rabbits plasma prolactin and median eminence and anterior pituitary concentration of several neuromodulators change on a daily basis. The existence of significant correlations among several of the neurotransmitters analyzed and plasma prolactin levels may explain the circadian secretory pattern of prolactin at this age in suckling rabbits. Collectively, the present results differ from the reported absence of circadian rhythmicity of prolactin and median eminence and adenohypophysial neuromodulators in rats at a comparable age. Competing Interests The author(s) declare that they have no competing interests. Authors' Contributions MPA and PC carried out the experiment and the immunoassays and the analysis of catecholamines, indoleamines and amino acids. DPC and AIE designed the experiments. Also, DPC performed the statistical analysis. PR took care of the experimental animals. AIE supervised its technical implementation and drafted the manuscript. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546204.xml
517936	The role of circulating anti-p53 antibodies in patients with advanced non-small cell lung cancer and their correlation to clinical parameters and survival	Background Lung cancer causes approximately one million deaths each year worldwide and protein p53 has been shown to be involved in the intricate processes regulating response to radiation and/or chemotherapeutic treatment. Consequently, since antibodies against p53 (anti-p53 antibodies) are associated with mutations within the p53 gene it seems likely that these antibodies could, hypothetically, be correlated with prognosis. Methods Serum samples from patients with non-small cell lung cancer (NSCLC) admitted to the Department of Oncology, University Hospital, Uppsala, Sweden, during 1983–1996 were studied. Anti-p53 abs were measured using a sandwich ELISA (Dianova, Hamburg, Germany). Results The present study included 84 patients with stage IIIA-IV (advanced NSCLC). At least three serum samples from each patient were collected and altogether 529 serum samples were analysed for the presence of anti-p53 antibodies. The median value of anti-p53 antibodies was 0.06 (range 0 – 139.8). Seventeen percent of investigated NSCLC first serum samples (n = 84) expressed elevated levels of anti-p53 antibodies. Anti-p53 antibodies were not correlated to tumour volume or platelets. Survival analysis showed that anti-p53 antibodies were not associated with survival as revealed by univariate analysis (p = 0.29). However, patients with adenocarcinoma had a significantly poorer survival if they expressed anti-p53 antibodies (p = 0.01), whereas this was not found for patients with squamous cell carcinoma (p = 0.13). In patients where the blood samples were collected during radiation therapy, a statistically significant correlation towards poorer survival was found (p = 0.05) when elevated anti-p53 antibodies levels were present. No correlations to survival were found for serum samples collected prior to radiation therapy, during chemotherapy, or during follow-up. When anti-p53 antibodies were measured continuously, no increase in median anti-p53 values was observed the closer the individual patient come to death. Conclusion The result of the present retrospective study indicates that anti-p53 antibodies are not suitable for predictions concerning selection of patients with a more favourable outcome. Further prospective studies are, though, needed to fully elucidate this issue.	Background Lung cancer causes approximately one million deaths each year worldwide [ 1 ]. Treatment of these patients is based on surgery, but at diagnosis approximately 80 % of NSCLC patients are inoperable [ 2 ]. These inoperable patients are treated with radiotherapy and/or chemotherapy. Several studies have tried to improve survival through introducing new chemotherapeutic treatment combinations [ 3 ] or applying different radiation fractionation schedules [ 4 ], resulting in modest improvements in survival. The continuous progress in the field of lung cancer biology has resulted in gradually increased insights into the intricate processes resulting in development and progression of malignancy. One of the first proteins whose occurrence was thought to affect prognosis was p53. This protein was identified during the late 1970s [ 5 ] and the corresponding gene was localized on the short arm of chromosome 17 [ 6 ]. Endogenous p53 protein is maintained at very low levels within the cell, but when the cell is exposed to hypothermia, oncogene activation, hypoxia or DNA damage, a rapid elevation of p53 levels is found, resulting in cell cycle arrest, DNA repair or induction of apoptosis [ 7 ]. If mutations are present in the p53 gene, these functions might be disturbed. In patients with NSCLC, p53 mutations are common, with mutation frequencies between 45–75% [ 8 , 9 ]. The question if mutations within the p53 gene are correlated to radiosensitivity is not defined. In an in vitro study from our group, we found that p53 mutations located in exon 7 were associated with significantly higher radiosensitivity than in those cell lines expressing p53 mutations in other exons [ 10 ]. In a clinical study performed on breast cancer patients with axillary lymph node metastases, the authors found that irradiation prolonged life in patients with p53 mutations compared with patients with wild type p53 [ 11 ]. Antibodies against p53 can be detected in sera from patients with cancer and a correlation exists between mutations in the p53 gene and antibodies against p53 in sera [ 12 ]. In a study from our group, investigating 67 patients with NSCLC, we found that the presence of anti-p53 antibodies prior to radiation therapy predicted increased survival (p = 0.025) [ 13 ]. In the present study, we included 84 patients with stage III-IV, donating 529 serum samples with the intention to investigate if predictions concerning outcome can be made and to investigate whether or not increased amounts of anti-p53 antibodies develop during disease progression, a question that, according to our knowledge, has not been explored previously. Methods Serum samples from patients with NSCLC admitted to the Department of Oncology, University Hospital, Uppsala, Sweden, during 1983–1996 were studied. All patients gave informed consent prior to the collection of blood samples and the samples were stored at -70°C until analyzed. The study was reviewed and approved by the research ethics committee, Uppsala University, Uppsala, Sweden. The inclusion criteria were: a verified histology of NSCLC, advanced NSCLC defined as stage IIIA-IV according to TNM and a minimum of three serum samples donated from each patient during progression of the disease. The patients included were all followed from admittance until death. The first serum samples from 67 patients included in this study had previously been investigated for the presence of antibodies against p53 [ 13 ]. In the present study, only patients with stages III and IV were included and the number of serum samples was enlarged with serum samples obtained during follow-up. The clinical parameters investigated were: age, gender, histology, performance status, weight reduction, smoking, tumour volume (according to RECIST criteria), objective response, subjective response (defined from the clinical charts as either: feeling better after treatment or feeling worse after treatment) and the presence of anti-p53 antibodies. Complete information concerning clinical parameters could not be obtained due to inadequate information in the clinical charts. Anti-p53 antibody investigation Blood was collected in 7 ml serum tubes without additive (367609, Becton Dickinson, Rutherford, NJ). Anti-p53 abs were measured using a sandwich ELISA (Dianova, Hamburg, Germany). Human recombinant p53 was bound to microtiter plates. Standards and samples were pipetted into the wells. After incubation and washing, a horseradish peroxidase conjugated polyclonal goat anti-human IgG was added. After renewed incubation and washing, a chromogenic substrate was added and the colour intensity was measured at 450 nm in a Titertek Multiskan. A relative index for patient sera was calculated as follows: E450 (sample) - E450 (low control)/E450 (high control) - E450 (low control). The ELISA assays were performed without knowledge of clinical data. According to the manufacturer's instructions, serum samples with an anti-p53 antibody index >1.1 were considered positive, whereas a serum sample with an index <0.9 was considered negative. Serum samples with an index between 0.9–1.1 were considered intermediately positive. Statistics Survival was estimated using the Kaplan-Meier product limit method, where univariate analysis was performed using a log-rank test. Cox regression analysis was performed to investigate if certain continuous factors had a significant effect on survival or to perform multivariate survival analyses. Spearman's rank order correlation was utilised for tests of associations between factors. The survival analysis together with the descriptive statistics is based on the first serum sample collected for each patient, whereas the correlation analyses were performed using all sera samples. In order to investigate if the levels of anti-p53 antibodies increased during the progression of the disease a statistical model was designed. In the model, time zero was set to be the date of pathological diagnosis and time one was set to time of death, i.e. the time to death was standardized for all patients. Using a fixed effect leased square estimator to allow for individually different starting values of anti-p53 antibodies, the effect on anti-p53 antibodies from diagnosis to death was calculated. In the descriptive statistics, range was defined as the minimum and maximum. Throughout the paper a 5% significant level was used. Results The median value of anti-p53 antibodies was 0.06 (range 0 – 139.8). Seventeen percent of the investigated NSCLC first serum samples expressed elevated levels of anti-p53 antibodies according to the manufacturer's instructions. From the time of pathological diagnosis until the time of death, no statistically significant effect on levels of anti-p53 antibodies was found (p = 0.8). Anti-p53 antibodies were not correlated to tumour volume (p = 0.19) or platelets (p = 0.27). A numerically higher median value of anti-p53 antibodies was found for adenocarcinoma patients (index level: 0.12) than for squamous cell carcinoma patients (index level: 0.08). The median anti-p53 antibody level, prior to therapy, was statistically significantly elevated in comparison with serum samples were collected during follow-up (p < 0.001) (Fig 1 ). No statistically significant difference was found when the other groups were compared. Descriptive data including survival, median and range for anti-p53 antibodies, as well as univariate analysis (based on the first serum sample for each patient), are shown in Table 1 . Analysis concerning histology and the presence of anti-p53 antibodies showed that patients with adenocarcinoma had a significantly poorer survival if they expressed anti-p53 antibodies (n = 23, p = 0.01). This was not found for patients with squamous cell carcinoma (n = 59, p = 0.13). Survival analysis based on when the first serum sample was collected in relation to therapy revealed that anti-p53 antibodies collected prior to radiation therapy (continuous variable) were not associated with survival (n = 53, p = 0.14). In patients from whom blood samples were collected during radiation therapy, a statistically significant correlation towards poorer survival was found (n = 13, p = 0.05). However, no correlations to survival were found when the first serum sample was taken during chemotherapy (5 patients), or during follow-up (n = 12, p = 0.68). Survival analysis showed that increased amounts of anti-p53 antibodies were not associated with survival according to univariate analysis (p = 0.29) (Fig 2 ). Discussion In the present study, anti-p53 antibodies have been investigated and correlated to various clinical parameters with the intention of studying if predictions concerning a more favourable outcome can be made for NSCLC patients, based on the presence or absence of these antibodies. Correlation analysis showed that anti-p53 antibodies were not correlated with tumour volume (measured according to RECIST criteria). These results support results of a previous study from our group, in which we neither did find a correlation between anti-p53 antibodies and tumour volume in patients with NSCLC prior to thoracic surgery [ 14 ]. The clinical utility of anti-p53 antibodies in NSCLC patients with advanced disease has not yet been defined. In a previous study by our group, the presence of anti-p53 antibodies did not correlate to survival but the presence of anti-p53 antibodies prior to radiation therapy resulted in increased survival for anti-p53 positive patients [ 13 ]. In the present study, patients from the previous study were included and the number of patients, as well as the amount of investigated serum samples, were increased. According to our knowledge, the present study is currently one of the largest studies in patients with advanced NSCLC concerning the amounts of investigated anti-p53 antibodies in sera. We were unable to find a correlation between survival and anti-p53 antibodies (p = 0.29) but differences in survival were observed between the different histologies and the presence of anti-p53 antibodies. Patients with adenocarcinoma had poorer survival if expressing anti-p53 (p = 0.01), whereas this was not found for squamous cell carcinoma patients. In a study by Gao et al., examining p53 mutations through exons 5–8 in patients with squamous cell carcinoma and adenocarcinoma of the lung [ 15 ], it was shown that patients with adenocarcinoma expressed mutational hotspots at codons 248 and 249, whereas patients with squamous cell carcinoma had mutational events spread throughout exons 5–8. Further, the authors concluded that mutations in the p53 gene in adenocarcinoma patients resulted in production of a more rigid p53 protein. In the present study, although, no significant survival differences were found between these two histological subgroups, it may be speculated that immunogenic differences exist between adenocarcinoma and squamous cell carcinoma histologies, differences that might explain our results. In the present study, the presence of anti-p53 antibodies was not correlated to survival when analyzed prior to radiation therapy, possibly because only patients with stages IIIA-IV were included, whereas our previous study also included patients with stages I-II. In patients where the first serum sample was collected during radiation therapy, a correlation to poorer survival was found (p = 0.05). Naturally, since the number of investigated patients was small, the obtained results should be interpreted with caution. The issue of using anti-p53 antibodies in monitoring NSCLC patients has not yet been determined. Zalcman et al. used anti-p53 antibodies in NSCLC patients during follow-up in order to detect relapse [ 16 ], and concluded that anti-p53 antibodies in sera might be of clinical usefulness during follow-up of NSCLC patients. In the present study, serum samples were collected prior to, during, and after treatment. The distribution of serum samples was not homogeneously distributed during the different treatments, as seen in Fig 1 . A change in antibody titer was observed during treatment and the levels of anti-p53 antibodies seemed to be higher during chemotherapy. These results should be interpreted with caution since the number of investigated serum samples was small. However, these results are intriguing, since anti-p53 antibodies have been correlated with the half-life of IgG1 and IgG2 [ 12 ]. Since patients receiving chemotherapy as well as radiation therapy often develop immunogenic deprivation, a subsequent reduction of anti-p53 antibodies seems reasonable, but this was not found. The results might be explained by tumour lysis causing increased exposure to the p53 antigen and, consequently, increased production of anti-p53 antibodies. Another hypothesis is that this elevation in the levels of anti-p53 antibodies during chemotherapy might be due to chemotherapy effects on normal tissues, thus explaining the difference between serum samples collected during radiation treatment and serum samples collected during chemotherapy treatment. A statistically significant difference was found between mean values of anti-p53 antibodies, prior to therapy and during follow-up (p < 0.001). However, since the investigated number of serum samples differs between the groups, no general conclusions can be made. Conclusion The result of the present retrospective study indicates that anti-p53 antibodies are not suitable for predictions concerning selection of patients with a more favourable outcome. Further prospective studies are, though, needed to fully elucidate this issue. Authors' contributions The authors have fulfilled the criteria for authorship according to rules stated and used in medical journals. Competing interests None declared. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517936.xml
517934	TM4SF10 gene sequencing in XLMR patients identifies common polymorphisms but no disease-associated mutation	Background The TM4SF10 gene encodes a putative four-transmembrane domains protein of unknown function termed Brain Cell Membrane Protein 1 (BCMP1), and is abundantly expressed in the brain. This gene is located on the short arm of human chromosome X at p21.1. The hypothesis that mutations in the TM4SF10 gene are associated with impaired brain function was investigated by sequencing the gene in individuals with hereditary X-linked mental retardation (XLMR). Methods The coding region (543 bp) of TM4SF10, including intronic junctions, and the long 3' untranslated region (3 233 bp), that has been conserved during evolution, were sequenced in 16 male XLMR patients from 14 unrelated families with definite, or suggestive, linkage to the TM4SF10 gene locus, and in 5 normal males. Results Five sequence changes were identified but none was found to be associated with the disease. Two of these changes correspond to previously known SNPs, while three other were novel SNPs in the TM4SF10 gene. Conclusion We have investigated the majority of the known MRX families linked to the TM4SF10 gene region. In the absence of mutations detected, our study indicates that alterations of TM4SF10 are not a frequent cause of XLMR.	Background Brain Cell Membrane Protein 1 (BCMP1) cDNA was fortuitously isolated from a thyroid cDNA library [ 1 ]. It encodes a 181aa-long putative four-transmembrane domain protein which appears related to both Peripheral Myelin Protein 22 / Epithelial Membrane Proteins and Claudins protein families, and exhibits significant similarities to the Caenorhabditis elegans VAB-9 protein, a protein that has recently been shown to be involved in the control of cell adhesion and epidermal morphology [ 2 ]. The protein sequence itself has been extremely well conserved during evolution, as exemplified by the observation that human and canine sequences are identical and differ from the mouse sequence at only 2 positions. The corresponding gene has now been renamed TM4SF10 in man and mouse, and is located on the X chromosome in both species, as well as in the other mammalian species investigated to date [ 1 ]. Initial Northern blot analysis of TM4SF10/BCMP1 gene transcripts distribution in adult dog tissues revealed very high expression in the brain, and lower but clearly detectable levels of expression in most of the tissues examined [ 1 ]. Data mining in the SAGEmap database [ 3 ] confirmed this observation in man, as elevated tag counts have been reported in brain astrocytoma (SAGE H127 library), brain ependymoma (SAGE ependymoma 353 and 582 libraries) and normal spinal cord (SAGE normal spinal cord library) as compared to other tissues. Together with its localization on the X chromosome, the high expression level detected in the brain and the putative role of the encoded protein in specific cell contacts raised the possibility that the TM4SF10 gene may be involved in X-linked mental retardation (XLMR) in man. Initially, the TM4SF10 gene was assigned to Xp11.4 [ 1 ]. As the integration between human cytogenetic and DNA sequence-based maps is still evolving, the gene has been reassigned to band p21.1. It is noteworthy that TM4SF2, another gene encoding a four-transmembrane domain protein, is located at the p11.4-p21.1 border on human chromosome X, in the very close vicinity of TM4SF10, and constitutes a known XLMR gene [ 4 , 5 ]. Recent compilations of XLMR families [ 6 - 8 ] mention several conditions mapped to the Xp11.4-p21.1 region. We report the result of mutation screening of TM4SF10 in a cohort of XLMR patients whose gene was mapped to this region of the X chromosome and does not correspond to TM4SF2. Methods Blood genomic DNA was collected from 16 patients (14 unrelated) and 5 unrelated healthy volunteers using a standard procedure [ 9 ]. The patients were affected males from families with definite, or possible, linkage to the region at Xp11.4-p21.1. Patients belonged to the following published MRX(S) families: MRX9 [ 10 ], MRX10 [ 11 ], MRX11 [ 11 ], MRX12 [ 11 ], MRX18 [ 12 ], MRX56 [ 6 ] and MRXS10 [ 13 ]. Additional patients were from an XLMR family with epilepsy [ 14 ] and 4 other XLMR families (C.S., F.K., J.G., unpublished), a MRXS family with macrocephaly and large ears (C.S., unpublished), and another MRXS family (J.G., unpublished). Chromosomal linkage data and major phenotypic traits are described in table 1 . All samples were studied anonymously and all procedures met the standards of our institutional ethics committee. Table 1 Description of the patients included in the study Patient Family Linkage data Phenotype P1, P2 XLMR family (F.K., now MRX79) chromosome X X-linked mental retardation P3, P4 MRXS10 Xp11.21-Xp11.4 Mental retardation, choreoathetosis, abnormal behavior P5 XLMR family (F.K.) Chromosome X, pericentromeric Non-syndromic mental retardation P6 MRX9 Xp11.22-Xp11.4 Non-syndromic mental retardation P7 MRX10 Xp11.3-Xp21.2 Non-syndromic mental retardation P8 MRX11 Xp11.3-Xp21.2 Non-syndromic mental retardation P9 MRX12 Xp11.21-Xp21.2 Non-syndromic mental retardation P10 MRX18 Xp11.3-Xp21.2 Non-syndromic mental retardation P11 XLMR family with epilepsy Xp11.23-Xp22.22 Non-syndromic mental retardation, epilepsy P12 MRXS family (J.G.) Xp21.3-Xq21.3 Non-syndromic mental retardation P13 XLMR family with macrocephaly (J.G.) Xp11.4-Xq13.1 Macrocephaly, moderate to profound mental retardation P14 XLMR family (C.S.) Xp11.3-Xp21.1 Seizures, ataxia, aggressive and hyperactive behavior, speech delay, mild to moderate mental retardation P15 MRXS (C.S.) Xp22.22-Xq24 Macrocephaly, prominent ears and moderate to severe mental retardation P16 MRX56 Xp11.21-Xp21.1 Non-syndromic mental retardation PCR reactions were performed in a final volume of 100 μl containing 200 ng of genomic DNA, 1 μg of each primer (see table 2 for primer sequences), 200 μM of each dNTP, 1X PCR buffer (QIAGEN) and 2.5 units of Taq polymerase (QIAGEN). Additionally, 10% DMSO or 20% Q solution (QIAGEN) were also included in some reactions (see table 2 ). After an initial denaturation step (93°C, 45 sec.), 35 cycles were conducted as follows: 93°C, 45 sec.; annealing temp. (see table 2 ), 45 sec.; 72°C, 45 sec. (amplicons Ex1–Ex3) or 1 min. (amplicons 3'UTR F1–F4). A final extension step (72°C, 3 min.) was done at the end of the reaction. PCR products were purified using the QIAquick PCR purification kit (QIAGEN) before sequencing. Purified PCR fragments were roughly quantitated by agarose gel electrophoresis using SmartLadder molecular weight marker (EUROGENTEC) as a quantitative reference. Table 2 PCR conditions and primer sequences used to amplify TM4SF10 gene fragments from total genomic DNA Amplicon Size Primer sequence PCR conditions Ex1 442 bp fw: AGAGCCCCGAGGGAGCGA, rev: GGGGACAGGCGGTGACTG T anneal = 55°C, 10% DMSO Ex2 447 bp fw: AAATCCTAGCAAACCCCTGG, rev: TCTGCATAGGAAAGGAAGATGG T anneal = 50°C Ex3 447 bp fw: CCATCTAGAACAAGCCATCTTTAA, rev: TAAATCAACTGAGCAAACTGCTTG T anneal = 50°C 3'UTR F1 959 bp fw: GGCCTGGGGTGCAACTATAT, rev: TAGGCAAATGTATGTGGAGGGT T anneal = 55°C, 10% DMSO 3'UTR F2 1101 bp fw: ATTGGTGCCTCAGCCCTATCTA, rev: GCAACCATTCTTAAGACAAGCT T anneal = 57°C, 20% Q solution 3'UTR F3 1130 bp fw : CAGTATGTTCTGGTTTTGGCCC, rev : TATCTAACAATGGGTTTGTGGC T anneal = 57°C, 20% Q solution 3'UTR F4 1097 bp fw : CCTTCTCAGCAAAGAGCCCTAC, rev : AAGGATCTTGGGAGATAATTTG T anneal = 57°C, 20% Q solution About 50–100 ng of purified PCR fragment was used in a DNA sequencing reaction using a nested, internal primer. DNA sequencing was performed using ABI PRISM BigDye Terminator Cycle Sequencing Ready Reaction kit (Applied Biosystems) on an Applied Biosystems 3100 automatic DNA sequencer. The sequences of the primers are given in table 3 . Table 3 Sequences of the primers used for sequencing of TM4SF10 gene fragments Amplicon Primer sequence Ex1 fw: CCGAGGGAGCGAGTCCCC Ex2 fw: CACATCTGTTGAGCCACTGC Ex3 rev: GATGCTCCACAAGTGTTTTAGA 3'UTR F1 fw : TGCCTGAACCCTAAGAACTATG, rev : GGAGGGTTAGGGAACAACTTAT 3'UTR F2 fw : CTGCATGAGTTGCTTTTGTACC, rev : GCAACCATTCTTAAGACAAGCT 3'UTR F3 fw : TCTGTTAAGAGCAGGACCACAT, rev1: ACTCGAGATGTGATGATATTGG, rev2 : TATCTAACAATGGGTTTGTGGC 3'UTR F4 fw : AACATGAAAATTGTTGCTTCTC, rev : AAGGATCTTGGGAGATAATTTG Results and discussion Sequencing of TM4SF10 coding region The human TM4SF10 gene is composed of 3 exons and produces a 4 kb-long mRNA. The short coding region (543 bp) is interrupted by 2 introns and the last and the largest exon also contains a 3 233 bp-long 3'UTR (see Figure 1 ). Initially, we sequenced the coding region and exon-intron junctions of the gene in the DNAs from 16 XLMR patients and from one normal male (amplicons Ex1–Ex3, Figure 1 ). No mutations were identified. One silent polymorphism was observed at position 186 in the cDNA sequence (clone DKFZp761J17121; GenBank accession number AL136550), corresponding to the 3 rd base of the codon Arg59, where a C residue was present in half of the sequences and a G residue in the other half. Individual single nucleotide polymorphism (SNP) data are shown in table 4 . It is noteworthy that patients P3 and P4 who belong to the same family exhibit a difference in their TM4SF10 gene sequence at this level. This observation argues against a causal role of the gene in this family. The 186C>G polymorphism in the TM4SF10 gene had been previously reported [ 15 ]. Figure 1 Schematic of the TM4SF10 gene and single nucleotide polymorphisms identified in the study. The structure of the gene is outlined with exons represented as light-blue boxes and the coding region as dark-blue areas within the boxes. The 3' end of the gene is also enlarged (bottom). Amplified regions are delineated and locations of sequence primers (arrows) used in this study are depicted. Identified SNPs are indicated (dotted lines) with reference to genomic contig (GenBank accession number NT_011757) sequence coordinates and cDNA (clone DKFZp761J17121; GenBank accession numberAL136550) sequence coordinates (italics) when relevant. Numbers in parentheses indicate the occurrences of the nucleotide in the individual sequences. The 3 novel SNPs identified in the work appear in red. Table 4 Description of the individual TM4SF10 gene SNP haplotypes determined in this study Exon 1: 186 (Arg59), [21589434] 3' UTR: 705, [21562747] 3' UTR: 2870, [21560583] 3' UTR: 2907, [21560546] 3' intergenic: [21559611] N1 C C G T C N2 (n.d.) C C C C N3 (n.d.) C C T C N4 (n.d.) T C T T N5 (n.d.) C C T C P1 C T C T T P2 C T C T T P3 G C C T C P4 C C C T C P5 G C C T T P6 G C C T T P7 G C C T C P8 G C C T C P9 G C C T C P10 C C C T T P11 C T C T T P12 G C C T C P13 G C C T C P14 C C C T T P15 C T C T T P16 G C C T C The first row identifies the source of the DNA, either a normal individual (Nx) or a patient (Px; see also table 1 for the detailed description of the patients). The individual bases found at each polymorphic position in the DNA sequence (identified by the nucleotide position in the cDNA sequence and/or in the genomic contig sequence [in square brackets]; see also figure 1) are given in the following rows. (n.d.): not determined. Sequencing of the 3'UTR The long 3'UTR sequence of the TM4SF10/BCMP1 transcript is highly conserved, with an overall score of 72% when human, dog, mouse and rat sequences are compared. As the 3'UTR of mRNAs may contain regulatory sequences that participate in the control of gene expression, we decided to screen this part of the gene as well. The entire region, including the sequences around the polyadenylation site, was subdivided into four overlapping fragments of approximately 1 kb in length (3'UTR F1–F4, Figure 1 ) and sequenced from both ends. In fragment 3' UTR F3 the presence of a stretch of 12 consecutive A residues on the sense strand resulted in difficulties in proper reading of the sequences located downstream of this motif. In order to overcome this problem, an additional sequence primer (rev2) was used to obtain overlaps between the 3 separate sequences for each individual fragment. In the cDNA sequence AL136550 the motif is composed of 13 A residues, which is a likely sequence artefact. TM4SF10 sequence was obtained from 16 patients and 5 controls. Four SNPs were identified in the non-coding part of the gene: 3 of them were located in the 3'UTR of the mRNA while the fourth one was located downstream of the polyadenylation site (Figure 1 ). Only this last one (C21559611T) had been previously reported in the SNP database [ 15 ], the other three representing novel SNPs in the TM4SF10 gene. The individual SNP haplotypes determined here are described in table 4 . It is also noteworthy that during the course of our investigation, the genetic defect of one of the unpublished XLMR family included in the study (see top of table 1 ) has been identified and mapped to Xq28, within the MECP2 gene [ 16 ]. Conclusions In this study, we have investigated the majority of the known MRX families linked to the TM4SF10 gene region. In the absence of mutations detected, our results indicate that alterations in the transcribed region of TM4SF10 are not a frequent cause of XLMR. Although the gene promoter has not been identified and screened yet, it appears very unlikely that all mutations would be there. This work also identified three novel SNPs in the TM4SF10 gene, which adds to our knowledge of SNP occurrence within this gene. Competing interests None declared. Authors' contributions CCH performed/managed the PCR and sequencing reactions, and analyzed the DNA sequences. FK, JG, MJA, EHF and CS provided the DNA samples. These authors also participated in the writing of the manuscript. DC conceived and supervised the study, and drafted the manuscript. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517934.xml
533880	Validating the Time and Change test to screen for dementia in elderly Koreans	Background We assessed the applicability of the T&C test as an accurate and convenient means to screen for dementia in primary care and community settings. Methods The study group comprised 59 patients and 405 community participants, all of who were aged 65 years and over. The time component of the T&C test evaluated the ability of a subject to comprehend clock hands that indicated a time of 11:10, while the change component of the T&C test evaluated the ability of a subject to make 1,000 Won from a group of coins with smaller denominations (one 500, seven 100, and seven 50 Won coins). Results The T&C test had a sensitivity and specificity of 73.0 and 90.9%, respectively, and positive and negative predictive values of 93.1, and 66.7%, respectively. The test-retest and interobserver agreement rates were both 95% (κ = 0.91) (time interval, 24 hours). The association between the T&C test and K-MMSE test was modest, while significant (r = 0.422, p < 0.001). The T&C test scores were not influenced by educational status. Conclusions We conclude that the T&C test is useful as supplemental testing of important domains (e.g., calculation, conceptualization, visuospatial) to traditional measures such as the MMSE. However, because T&C test is simple, rapid, and easy to use, it can be applied conveniently to elderly subjects by non-specialist personnel who receive training.	Background Dementia, an acquired persistent impairment of cognitive functioning is an increasingly common problem in Korea, and is associated with increased morbidity and mortality, functional loss, caregiver burden, and institutionalization [ 1 ]. Nevertheless, most patients with dementia are not detected by family members or clinicians at an early stage, even though such patients have memory problems [ 2 , 3 ]. Screening for dementia has been recommended to increase detection of dementia. The Mini-Mental State Examination (MMSE) is a brief screening test that quantitatively assesses the cognitive status of elderly people [ 4 , 5 ]. Although it has shown good validity, the MMSE has been found to be influenced largely by racial differences and educational status [ 6 - 10 ]. Korean elderly show very low levels of educational background, and it is necessary to develop appropriate cognitive assessment method that is not influenced by education. The Time and Change (T&C) test was developed by Inouye et al. [ 11 ] to assess (i) how well a patient understands time and (ii) the ability of the patient to calculate using money. Several studies have found that the T&C test allows direct assessment of two important activities of daily living and supplements the testing of calculation, conceptualization, and visuospatial cognitive domains irrespective of race and education [ 11 - 13 ]. We have therefore introduced a T&C test, and assessed the applicability of the T&C test as a convenient and accurate means to detect early stage dementia in primary care and community settings. Methods Subjects One study group initially comprised 60 participants who either visited or were admitted to a hospital that was located in an urban area, namely Gwangju city (referred to forthwith as area A) between November and December 2001. Of these 60 participants, 37 were diagnosed with Alzheimer's disease or vascular dementia, 11 had mental illness such as schizophrenia or depression, 7 had organic brain disease such as cerebral apoplexy or Parkinson's disease, and 5 were alcoholics. Another study group initially comprised 412 participants who were recruited from all residents of Jangseong-county, Jeonnam province, South Korea, aged 65 and over in 2002(referred to forthwith as area B). The area consists of 11 towns and had an estimated population 54,528 of whom about 16.5% were aged 65 and over. The subjects were selected from each stratum (town) using cluster sampling. All subjects in whom vision or hearing was impaired were excluded from the study. We selected 59 (of 60) and 405 (of 412) participants from area A and B, respectively. All participants gave informed consent, and when participants with dementia cannot provide informed consent, caregivers were asked to provide it. The Time and Change test The 'time' component of the T&C test evaluated the ability of a subject to comprehend that the hands of a clock indicated 11:10. The diameter of the clock was 15 cm, and the distance between the clock and the participant was 25–35 cm (in consideration of the visual deficits of the older subjects). When the subject responded incorrectly on the first attempt, the interviewer posed the same question, i.e., a second attempt was permitted. The time (in seconds) that it took for the participant to respond correctly was recorded. A time limit of 60 s for a response was imposed. In the 'change' component of the T&C test, the participants were required to make 1,000 Won from a group of coins of several smaller denominations, namely one 500, seven 100, and seven 50 Won coins, that were placed on a table in front of the participant. The participants were given 120 s to complete the task, and an additional 120 s was granted if the subject failed to complete the task at the first attempt. The time (in seconds) that it took for the participant to complete the task was recorded by the interviewer. The participants who completed both of the aforementioned tasks of the T&C test successfully were determined to be negative for suspected dementia, whereas participants who failed to complete either or both of the tasks were determined to be positive for suspected dementia. The T&C test was performed prior to the other cognitive tests by an interviewer who was unaware of the results of the other tests. Measurement In the participants from area A, the T&C test was conducted by a nurse, while a physician who were blinded to the result of the T&C test performed a clinical examination and neuropsychiatric inventory. The diagnostic criteria for dementia were based on those of the Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) [ 14 ]. To detect the causative diseases of dementia, the following were carried out: general blood and chemical tests (including VDRL, Vitamin B12, T3, T4, TSH); a chest radiograph; an electrocardiogram; a brain computerized tomography scan; and psychometric tests including the K-MMSE [ 15 ]. Alzheimer's dementia was diagnosed based on the guidelines of the National Institute of Neurological and Communicative Disorders and Stroke and Alzheimer's disease and Related Disorders Association (NINCDS-ADRDA) [ 16 ]. Vascular dementia was diagnosed based on the guidelines of the National Institute of Neurological Disorders and Stroke and the Association Internationale pour la Recherche et l'Enseignement en Neurosciences (NINDS-AIREN) [ 17 ]. The CDR developed by Hughes et al. [ 18 ] was used to determine the progression of dementia. The participants from area B were interviewed by interviewers who had undergone sufficient training to be able to conduct the K-MMSE and T&C test. All participants were interviewed for data on social and demographic factors such as address, age, sex, educational status, and the number of family members living together. Assessment of the validity of the T&C test To assess the validity of the T&C test as a method to screen for dementia in the participants from area A, we compared the results of the test to a reference standard (diagnosed by a physician), and evaluated sensitivity, specificity, and positive and negative predictive values. To assess the test-retest reliability, the one type of interviewer (specifically, physicians) conducted the same test twice at an interval of 24 h (n = 22 participants). To assess the interobserver reliability, two different types of interviewer (specifically, physicians and psychometrists) conducted the same test, at an interval of 24 h (n = 22 participants). To assess the applicability of the T&C test as a method to screen for dementia in the participants from area B, the association between the T&C and K-MMSE test scores was evaluated. Statistical analysis The sensitivity, specificity, and positive and negative predictive values and 95% Confidence interval of the T&C test were analyzed. After classifying the participants from area B into a group in which dementia was suspected, and another group in which dementia was not suspected based on the results from the T&C test, a Student's t-test was used to compare the total K-MMSE score with the scores for each of the components of the K-MMSE test, and Spearman's rank correlation was used to analyze the correlation between the T&C and K-MMSE test scores. SPSS for Windows (version 10.0; SPSS Inc., Chicago, IL, USA) and Stata Software 6.0 (Stata Corporation, College Station, Texas) were used to analyze data and statistics. Results Characteristics of subjects For the participants from area A, the average age was 73.2 ± 7.9 years, 55.9% of the group was female, the average number of years of education was 4.2 ± 5.4, and 62.7% lived alone. For the participants from area B, the average age was 73.1 ± 6.1 years, 58.5% of the group was female, the average number of years of education was 3.1 ± 4.0, and 29.0% lived alone (Table 1 ). Table 1 Characteristics of the study groups. Characteristic Hospital* (n = 59) Community † (n = 405) Age: years, mean ± SD 73.2 ± 7.9 73.2 ± (range) (58–90) (65–97) Gender: female, n (%) 33 (55.9) 238 (58.5) Education: years, mean ± SD 4.2 ± 5.4 3.1 ± 4.0 (range) (0–16) (0–16) Living alone: n (%) 32 (62.7) 118 (29.0) SD = standard deviation. Referred to in main text as area A (see Methods). † Referred to in main text as area B (see Methods). Validity of the T&C Test Table 2 shows the results of an analysis of the validity of the T&C test. In a comparison of the T&C test and a diagnosis of dementia by physicians (which served as a reference), 27 of 37 patients that had been diagnosed as demented were classified as positive for dementia according to the T&C test, and the sensitivity was 73.0%. Of the subjects who were diagnosed as not demented, 20 of 22 were classified as negative according to the T&C test, and the specificity was 90.9%. The positive predictive value that indicated the probability of dementia, as determined by a positive classification according to the T&C test, was 93.1%, while the negative predictive value that indicated the probability of not being demented, as determined by a negative classification according to the T&C test, was 66.7%. Table 2 Concurrent validity of the Time and Change (T&C) test in a sample of elderly hospital patients. Reference standard Total Dementia No dementia T&C + 27 2 29 assessment - 10 20 30 Total 37 22 59 Sensitivity = 27/37 (73.0%) (CI = 61.6–84.3%) Specificity = 20/22 (90.9%) (CI = 83.6–98.2%) Positive predictive value = 27/29 (93.1%) (CI = 86.6–99.6%) Negative predictive value = 20/30 (66.7%) (CI = 54.6–78.7%) Accuracy = 47/59 (79.7%) CI = 95% confidence interval for sensitivity and specificity values. Diagnosis of dementia made by a clinician. +, positive for dementia according to T&C test. -, negative for dementia according to T&C test. Reliability of the T&C test The rate of agreement of the test-rest & interobserver variability was analyzed to assess the reliability of the T&C test. The rate of agreement of both the test-retest and interobserver was 95% (κ = 0.91) (time interval, 24 hours). Comparison of the T&C and K-MMSE test There was a significant difference between the total K-MMSE score and the scores for each of the components of the K-MMSE test between participants that were classified as positive and those that were classified as negative for dementia according to the T&C test (Table 3 ). The association between the T&C test and K-MMSE test revealed modest, while significant (r = 0.422, p < 0.001) (Table 4 ). Table 3 Korean Mini-Mental State Examination (K-MMSE) scores according to T&C test performance in a sample of elderly community population. K-MMSE components K-MMSE score (overall) (n = 30) Orientation (n = 10) Registration (n = 3) Concentration & Calculation (n = 5) Recall (n = 3) Language & Diagram (n = 9) T&C test + † (n = 65) 14.8 ± 5.8 6.1 ± 2.8 2.1 ± 1.1 0.5 ± 1.4 0.8 ± 1.1 5.4 ± 1.7 - (n = 340) 22.4 ± 5.0 8.9 ± 1.8 2.8 ± 0.6 2.2 ± 1.9 1.6 ± 1.1 7.0 ± 1.6 Time test* + ‡ (n = 53) 14.3 ± 5.3 6.0 ± 2.6 2.0 ± 1.1 0.4 ± 1.2 0.7 ± 1.1 5.2 ± 1.6 - (n = 352) 22.3 ± 5.1 8.8 ± 1.9 2.8 ± 0.6 2.2 ± 1.9 1.6 ± 1.2 7.0 ± 1.6 Change test* + § (n = 31) 14.6 ± 6.9 6.0 ± 3.0 2.0 ± 1.2 0.8 ± 1.7 0.5 ± 1.0 5.3 ± 2.0 - (n = 374) 21.8 ± 5.4 8.6 ± 2.0 2.7 ± 0.6 2.0 ± 1.9 1.5 ± 1.2 6.8 ± 1.7 () : number mean ± SD. p < 0.01 for K-MMSE score (or a component thereof) versus T&C test score (Student's t-test). +, positive for dementia according to T&C test. -, negative for dementia according to T&C test. † Incorrect response for either or both the time and change task of the T&C test (see Methods for details). ‡ Incorrect response for the time task. § Incorrect response for the change task. Table 4 Convergent validity of the T&C test in a sample of elderly community population. Correlation coefficient (r) K-MMSE (overall score) 0.422 K-MMSE component scores: Orientation 0.431 Registration 0.331 Concentration & Calculation 0.358 Recall 0.258 Language & Diagram 0.314 *p < 0.001 for all values (Spearman's rank correlation analysis). Association between the T&C test and educational status The results of a logistic regression in which the T&C test as a dichotomous and dependent variable was performed with education status adjusting for age and sex revealed no association between them. The odds ratio for T&C test associated with educational status is 0.877 (95% CI = 0.766–1.004). Response times in the T&C test For the time task in the T&C test, 75.8% of the participants produced a correct response on the first attempt after 6.3 ± 6.7 s, and 45 participants (11.1%) produced a correct response on the second attempt. For the change task in the T&C test, 81.2% of the participants produced a correct response on the first attempt after 12.7 ± 14.2 s, and 43 participants (10.6%) produced a correct response on the second attempt. 34 participants(8.4%) were tested twice for both the time and change task. None of the subjects refused to respond during the tests. Discussion Interracial variability in both the etiology of dementia and the accuracy of cognitive testing suggests that there is an urgent need to develop racially appropriate methods of cognitive assessment. The rate of vascular dementia due to cerebrovascular disease is much higher in Koreans than in other races; this is due to insufficient prevention, diagnosis, and treatment of hypertension, diabetes, and hyperlipidemia in Korea. Vascular dementia, unlike Alzheimer's disease, can be prevented by appropriate treatment and prevention for the risk factors of cerebrovascular diseases, can be treated to improve symptoms and inhibit progression of the disease. In the present study, the original T&C test of Inouye et al. [ 11 ] was modified to apply specifically to Korean elderly sample, and was used to screen for dementia. We found that the sensitivity of the T&C test was 73.0%, specificity was 90.9%, and the positive and negative predictive values were 93.1% and 66.7%, respectively. The disparity between the sensitivity and the specificity might mean the insensitivity to mild stage or pre-dementia. We advocate using the T&C test sequentially with other dementia measures (with increased sensitivity). When compared to the results of Kawamato [ 19 ], in which sensitivity and specificity was 49.1 and 95.2%, respectively; the sensitivity of the T&C test in the present study was remarkably high, while specificity was relatively low. In a study by Inouye et al. [ 8 ] in which the MMSE and modified Blessed Dementia Rating Scale (mBDRS) served as standards, sensitivity was 86%, specificity was 71%, and the negative predictive value was 97% when the subjects were hospital patients [ 8 ], while sensitivity was 63%, specificity was 96%, and the negative predictive value was 93%, for outpatients [ 12 ]. This discrepancy might be due to the characteristics of different samples or the diagnostic standards for dementia. The rate of dementia among the participants of the present study (62.7%) was higher than in the studies by Inouye and colleagues (14 and 16% for admitted and outpatients, respectively). Therefore, there are limitations when it comes to comparing the predictive values of screening tests, which would appear to be influenced by the prevalence. In elderly patients, the assessment of cognitive function is affected by psychological factors and by the circumstances under which the tests are conducted. In our measurement of the reliability of the T&C test, the test-retest and interobserver agreement rates were both remarkably high (95% for each, κ = 0.91). In a study by Inouye et al. [ 8 ] the test-retest agreement rate was 88%, and the interobserver agreement rate was 78% when the subjects were hospital patients [ 8 ]. However in the present study test-rest was conducted in the short interval, therefore it might to be influenced by carry-over effects and learning effects. In the present study, no association was observed between T&C test scores and educational status. This result may be explained by the fact that interpreting the hands of a clock and calculation using change are behaviors that are common to all people during daily life, irrespective of race and education. In addition, unlike the language-focused MMSE, the questions in the T&C test cannot be misinterpreted or misunderstood, and are effective for assessing calculation ability and attention. Clock test (similar to T&C test) is less likely to be confounded by educational attainment [ 20 ], and measures various facets of cognitive functioning at varying levels of difficulty [ 21 ]. This study has several important limitations. First, while the hospital patients underwent an evaluation of their medical history as well as neurological and physical examinations before a diagnosis of dementia was made, the elderly community sample were evaluated using only the K-MMSE as a reference. Second, the hospital patients would appear to be a rather unusual sample in that there was a high prevalence of psychiatric morbidity in the non-demented subjects. Third, dementia in the present study was not classified as vascular or Alzheimer's-type, nor did we consider the severity of symptoms. Finally, the T&C test has limitations in assessment of a wide range of deficits associated with dementia. However, because T&C test is simple, rapid, and easy to use, the T&C test may pose particular advantages in primary care and community settings where frequent assessment of cognitive functioning is required. The T&C test adds supplemental testing of important domains (e.g., calculation, conceptualization, visuospatial) to traditional measures such as MMSE. In addition, because the T&C test is less influenced by educational status, it may be particularly useful in populations with diverse educational and cultural backgrounds. Conclusions We conclude that the T&C test is useful as a supplemental testing of important domains (e.g., calculation, conceptualization, visuospatial) to traditional measures such as the MMSE, because sensitivity of T&C test is not great and the association between the T&C test and MMSE is modest. However, because T&C test is simple, rapid, and easy to use, it can be applied conveniently to elderly subjects by non-specialist personnel who receive training. In addition, the T&C test is less influenced by educational status Funding This work was supported by grants from Chonnam National University Hospital (CUHRI-U-200241). Competing interests The author(s) declare that they have no competing interests. Authors' contributions JAR conceived of the study, collected data and drafted the manuscript. EKC participated in its design, performed data collection, and reviewed the manuscript. MHS participated in data analysis and reviewed the manuscript. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC533880.xml
535565	Endotracheal tube cuff pressure in three hospitals, and the volume required to produce an appropriate cuff pressure	Background Cuff pressure in endotracheal (ET) tubes should be in the range of 20–30 cm H 2 O. We tested the hypothesis that the tube cuff is inadequately inflated when manometers are not used. Methods With IRB approval, we studied 93 patients under general anesthesia with an ET tube in place in one teaching and two private hospitals. Anesthetists were blinded to study purpose. Cuff pressure in tube sizes 7.0 to 8.5 mm was evaluated 60 min after induction of general anesthesia using a manometer connected to the cuff pilot balloon. Nitrous oxide was disallowed. After deflating the cuff, we reinflated it in 0.5-ml increments until pressure was 20 cmH 2 O. Results Neither patient morphometrics, institution, experience of anesthesia provider, nor tube size influenced measured cuff pressure (35.3 ± 21.6 cmH 2 O). Only 27% of pressures were within 20–30 cmH 2 O; 27% exceeded 40 cmH 2 O. Although it varied considerably, the amount of air required to achieve a cuff pressure of 20 cmH 2 O was similar with each tube size. Conclusion We recommend that ET cuff pressure be set and monitored with a manometer.	Background A critical function of the endotracheal tube cuff is to seal the airway, thus preventing aspiration of pharyngeal contents into the trachea and to ensure that there are no leaks past the cuff during positive pressure ventilation. However, complications have been associated with insufficient cuff inflation. Consequences of micro-aspiration of oropharyngeal secretions include nosocomial pulmonary infections [ 1 ]. Conventional high-volume, low-pressure cuffs may not prevent micro-aspiration even at cuff pressures up to 60 cm H 2 O [ 2 ], although some studies suggest that only 25 cm H 2 O is sufficient [ 3 ]. In contrast, newer ultra-thin cuff membranes made from polyurethane effectively prevent liquid flow around cuffs inflated only to 15 cm H 2 O [ 2 ]. In the absence of clear guidelines, many clinicians consider 20 cm H 2 O a reasonable lower limit for cuff pressure in adults. Catastrophic consequences of endotracheal tube cuff over-inflation such as rupture of the trachea [ 4 - 6 ], tracheo-carotid artery erosion [ 7 ], and tracheal innominate artery fistulas are rare now that low-pressure, high-volume cuffs are used routinely. However, post-intubation sore throat is a common side effect of general anesthetic and may partly result from ischemia of the oropharyngeal and tracheal mucosa [ 8 - 10 ], and the most common etiology of non-malignant tracheoesophageal fistula remains cuff-related tracheal injury [ 11 , 12 ]. In addition, acquired laryngeal stenosis may be caused by mechanical abrasion or pressure necrosis of the laryngeal mucosa secondary to high cuff pressure [ 13 , 14 ]. Animal data indicate that a cuff pressure of only 20 cm H 2 O may significantly reduce tracheal blood flow with normal blood pressure and critically reduces it during severe hypotension [ 15 ]. Similarly, inflation of endotracheal tube cuffs to 20 cm H 2 O for just four hours produces serious ciliary damage that persists for at least three days [ 16 ]. Thus, appropriate inflation of endotracheal tube cuff is obviously important. Lomholt et al. recommended selecting a cuff pressure of 25 cmH 2 O as a safe minimum cuff pressure to prevent aspiration and leaks past the cuff [ 17 ]; Bernhard et al. supported this recommendation [ 18 ]. On the other hand, Nordin et al. studied the relationship between cuff pressure and capillary perfusion of the rabbit tracheal mucosa and recommended that cuff pressure be kept below 27 cm H 2 O (20 mmHg) [ 19 ]. Seegobin and Hasselt reached similar conclusions in an in vitro study and recommended cuff inflation pressure not exceed 30 cm H 2 O [ 20 ]. It is thus essential to maintain cuff pressures in the range of 20–30 cm of H 2 O Adequacy of cuff inflation is conventionally determined by palpation of the external balloon. Previous studies suggest that this approach is unreliable [ 21 , 22 ]. One study, for instance, found that cuff pressure exceeded 40 cm H 2 O in 40-to-90% of tested patients [ 22 ]. However, increased awareness of over-inflation risks may have improved recent clinical practice. Our first goal was thus to determine if cuff pressure was within the recommended range of 20–30 cmH 2 O, when inflated using the palpation method. Because cuff inflation practices are likely to differ among clinical environments, we evaluated cuff pressure in three different practice settings: an academic university hospital and two private hospitals. It is also likely that cuff inflation practices differ among providers. We therefore also evaluated cuff pressure during anesthesia provided by certified registered nurse anesthetists (CRNAs), anesthesia residents, and anesthesia faculty. Cuff pressure can be easily measured with a small aneroid manometer [ 23 ], but this device is not widely available in the United States. It would thus be helpful for clinicians to know how much air must be injected into the cuff to produce the minimum adequate pressure. We designed this study to observe the practices of anesthesia providers and then determine the volume of air required to optimize the cuff pressure to 20 cmH 2 O for various sizes of endotracheal tubes. Taking another approach to the same question, we also determined compliance of the cuff-trachea system in vivo by plotting measured cuff pressure against cuff volume. Methods With approval of the University of Louisville Human Studies Committee and informed consent, we recruited 93 patients (42 men and 51 women) undergoing elective surgery with general endotracheal anesthesia from three hospitals in Louisville, Kentucky: 41 patients from University Hospital (an academic centre), 32 from Jewish Hospital (a private hospital), and 20 from Norton Hospital (also a private hospital). Patients with emergency intubations, difficult intubations, or intubation performed by non-anesthesiology staff; pregnant women; patients with higher risk for aspiration (e.g., full stomach, history of reflux, etc.); and patients with known anatomical laryngeo-tracheal abnormalities were excluded from this study. The Human Studies Committee did not require consent from participating anesthesia providers. However, no data were recorded that would link the study results to specific providers. Protocol Independent anesthesia groups at the three participating hospitals provided anesthesia to the participating patients. Because one purpose of our study was to measure pressure in the endotracheal tube cuff during routine practice, anesthesia providers were blinded to the nature of the study. They were only informed about the second purpose of the study: determining the relationship between cuff volume and pressure. Ninety-three patients were randomly assigned to the study. The groups were not equal for the three different types of practitioners; however, determining differences of practice between different anesthesia providers was not the primary purpose of our study. General anesthesia was induced by intravenous bolus of induction agents, and paralysis was achieved with succinylcholine or a non-depolarizing muscle relaxant. Male patients were intubated with an 8 or 8.5 mm internal diameter endotracheal tube, and female patients were intubated with a 7 or 7.5 mm internal diameter endotracheal tube. This is a standard practice at these hospitals. Patients who were intubated with sizes other than these were excluded from the study. Anesthesia was maintained with a volatile aesthetic in a combination of air and oxygen; nitrous oxide was not used during the study period. At the University of Louisville Hospital, at least 10 patients were evaluated with each endotracheal tube size (7, 7.5, 8, or 8.5 mm inner diameter [Intermediate Hi-Lo ® Tracheal Tube, Mallinckrodt, St. Louis, MO]); at Jewish Hospital, at least 10 patients each were evaluated with size 7, 7.5, and 8 mm Mallinckrodt Intermediate Hi-Lo ® Tracheal Tubes; and at Norton Hospital, 10 patients each were evaluated with size 7 and 8-mm Mallinckrodt Intermediate Hi-Lo ® Tracheal Tubes. Consecutive available patients were enrolled until we had recruited at least 10 patients for each endotracheal tube size at each participating hospital. All tubes had high-volume, low-pressure cuffs. Measurements We recorded endotracheal tube size and morphometric characteristics including age, sex, height, and weight. An anesthesia provider inserted the endotracheal tubes, and the intubator or the circulating registered nurse inflated the cuff. This is the routine practice in all three hospitals. Adequacy is generally checked by palpation of the pilot balloon and sometimes readjusted by the intubator by inflating just enough to stop an audible leak. Investigators measured the cuff pressure at 60 minutes after induction of anesthesia using a manometer (VBM, Sulz, Germany) that was connected to the pilot balloon of the endotracheal tube cuff via a three-way stopcock. This type of aneroid manometer is nearly as accurate as a mercury manometer, but easier to use [ 23 ]. Pressure was recorded at end-expiration after ensuring that the patient was paralyzed. The cuff pressure was measured once in each patient at 60 minutes after intubation. We did not collect data on the readjustment by the providers after intubation during this hour. A syringe attached to the third limb of the stopcock was then used to completely deflate the cuff, and the volume of air removed was recorded. The cuff was considered empty when no more air could be removed on aspiration with a syringe. The cuff was then progressively inflated by injecting air in 0.5-ml increments until a cuff pressure of 20 cmH 2 O was achieved. The entire process required about a minute. Data analysis Our primary outcomes were 1) measured endotracheal tube cuff pressures as a function of tube size, provider, and hospital; and 2) the volume of air required to produce a cuff pressure of 20 cmH 2 O as a function of tube size. Outcomes were compared by tube size, provider, and hospital with either an ANOVA (if the values were normally distributed) or the Kruskal-Wallis statistic (if the values were skewed). Compliance of the cuff system was evaluated by linear regression of measured cuff pressure vs . measured cuff volume. Data are presented as means (SD) or medians [interquartile ranges] unless otherwise noted; P < 0.05 was considered statistically significant. Results Morphometric and demographic characteristics of the patients were similar at each participating hospital (Table 1 ). Measured cuff pressures averaged 35.3(21.6)cmH 2 O; only 27% of the patients had measured pressures within the recommended range of 20–30 cmH 2 O. Fifty percent of the values exceeded 30 cmH 2 O, and 27% of the measured pressures exceeded 40 cmH 2 O. Thus, 23% of the measured cuff pressures were less than 20 mmHg. Measured cuff volume averaged 4.4 ± 1.8 ml. Neither measured cuff pressure nor measured cuff volume differed among the hospitals (Table 2 ). Table 1 Demographic and Morphometric Characteristics as a Function of Hospital. University of Louisville Hospital Norton Hospital Jewish Hospital P N 41 20 32 --- Tube size (N) 7 / 7.5 / 8 / 8.5 10 / 11 / 10 / 10 10 / 0 / 10 / 0 10 / 10 / 12 / 0 --- Age (yr) 41 (14) 50 (14) 51 (16) 0.006 Weight (kg) 88 (27) 100 (40) 78 (17) 0.020 Height (m) 1.7 (0.1) 1.7 (0.1) 1.7 (0.1) 0.669 Results presented as number or mean (SD). Table 2 Principal Results as a Function Hospital University of Louisville Hospital Norton Hospital Jewish Hospital P Measured Cuff Pressure (cmH 2 O) 26 [18, 38] 34.5 [20, 50.5] 33 [22.5, 48.5] 0.469 Measured Cuff Volume (ml) 4.0 [3.0, 5.0] 4.3 [3.0, 6.0] 4.5 [3.2, 5.5] 0.646 Volume Required for 20 cmH 2 O (ml) 2.7 [2.1, 3.8] 2.5 [2.3, 3.3] 2.9 [2.2, 3.7] 0.792 Results presented as median [interquartile range]. There was no correlation between the measured cuff pressure and the age, sex, height, or weight of the patients. Nor did measured cuff pressure differ as a function of endotracheal tube size. Measured cuff volumes were also similar with each tube size. Interestingly, the amount of air required to achieve a cuff pressure of 20 cmH 2 O was similar with each tube size (Table 3 ). However, there was considerable variability in the amount of air required. Table 3 Principal Results as a Function of Tube Size. 7.0 mm 7.5 mm 8.0 mm 8.5 mm P Patients with measured cuff pressure >30 cmH 2 O (%) 57 57 47 30 0.444 Measured Cuff Pressure (mmHg) 32 [22, 52] 38 [24, 49] 30 [16, 38] 24 [21, 40] 0.467 Measured Cuff Volume (ml) 3.9 [3.0, 5.0] 4.5 [2.7, 5.0] 4.6 [3.5, 6.1] 3.8 [3.0, 5.0] 0.616 Volume Required for 20 cmH 2 O (ml) 2.6 [2.0, 3.1] 2.5 [1.8, 3.0] 3.0 [2.5. 4.1] 3.3 [2.0, 3.9] 0.143 Results presented as medians [interquartile range] or percent. CRNAs (n = 72), anesthesia residents (n = 15), and anesthesia faculty (n = 6) performed the intubations. There were no statistically significant differences in measured cuff pressures among these three practitioner groups ( P = 0.847). The compliance of the tube was determined from the measured cuff pressure (cmH 2 O) and the volume of air (ml) retrieved at complete deflation of the cuff; this showed a linear pressure-volume relationship: Pressure= 7.5. Volume+2.7, r 2 = 0.39 (Fig. 1 ). Figure 1 The relationship between measured cuff pressure and volume of air in the cuff. There was a linear relationship between measured cuff pressure (cmH 2 O) and volume (ml) of air removed from the cuff: Pressure = 7.5. Volume + 2.7, r 2 = 0.39. Discussion Previous studies suggest that the cuff pressure is usually under-estimated by manual palpation. For example, Braz et al . [ 22 ] observed cuff pressure exceeding 40 cm H 2 O in 91% of PACU patients after anesthesia with nitrous oxide, 55% of ICU patients, and 45% of PACU patients after anesthesia without nitrous oxide. In an experimental study, Fernandez et al . [ 21 ] observed that when the cuff was inflated randomly to 10, 20, or 30 cmH 2 O, participating physicians and ICU nurses were able to identify the group in 69% of the high-pressure cases, 58% of the normal pressure cases, and 73% of the low pressure cases. Our results are consistent in that measured cuff pressure exceeded 30 cmH 2 O in 50% of patients and were less than 20 cmH 2 O in 23% of patients. Cuff pressures were thus less likely to be within the recommended range (20–30 cmH 2 O) than outside the range. It was nonetheless encouraging that we observed relatively few extremely high values, at least many fewer than reported in previous studies [ 22 ]. This result suggests that clinicians are now making reasonable efforts to avoid grossly excessive cuff inflation. Measured cuff inflation pressures were virtually identical at the three study sites: one academic center and two private hospitals. These data suggest that management of cuff pressure was similar in these two disparate settings. Interestingly, there was also no significant or important difference as a function of provider – measured cuff pressures were virtually identical whether filled by CRNAs, residents, or attending anesthesiologists. Our results thus fail to support the theory that increased training improves cuff management. We evaluated three different types of anesthesia provider in three different practice settings. Although we were unable to identify any statistically significant or clinically important differences among the sites or providers, our results apply only to the specific sites and providers we evaluated. While it is likely that these results are fairly representative, it is obvious that results would not be identical elsewhere because of regional practice differences. Fernandez et al . [ 21 ] found that the volume of air required to inflate the endotracheal tube cuff varies as a function of tube size and type. But interestingly, the volume required to inflate the cuff to a particular pressure was much smaller when the cuff was inflated inside an artificial trachea; furthermore, the difference among tube sizes was minimal under those conditions. We similarly found that the volume of air required to inflate the cuffs to 20 cmH 2 O did not differ significantly as a function of endotracheal tube size. These data suggest that tube size is not an important determinant of appropriate cuff inflation volume. A caveat, though, is that tube sizes were chosen by clinicians in our study and presumably matched patient size; results may well have differed if tube size had been randomly assigned. We intentionally avoided this approach since our purpose was to evaluate cuff pressures and associated volumes in three routine clinical settings. A limitation of this study is that cuff pressure was evaluated just once 60 minutes after induction of anesthesia. Because nitrous oxide was not used, it is unlikely that the cuff pressures varied much during the first hour of the study cases. We recognize that people other than the anesthesia provider who actually conducted the case often inflated the cuffs. However, a full hour was plenty of time for the provider to have checked and adjusted cuff pressure to a suitable level. We observed a linear relationship between the measured cuff pressure and the volume of air retrieved from the cuff. The regression equation indicated that injected volumes between 2 and 4 ml usually produce cuff pressures between 20 and 30 cmH 2 O independent of tube size for the same type of tube. However, there was considerable patient-to-patient variability in the required air volume. Measuring actual cuff pressure thus appears preferable to injecting a given volume of air. Conclusions Cuff pressure should be measured with a manometer and, if necessary, corrected. Competing interests The author(s) declare that they have no competing interests. Authors' contributions SP oversaw day-to-day study mechanics, collected data on many of the patients, and wrote an initial draft of manuscript. DIS contributed to study design, data analysis, and manuscript preparation. PM, SW, and AV recruited patients and performed many of the measurements. JD conceived of the study and participated in its design. AW contributed to protocol development, patient recruitment, and manuscript preparation. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535565.xml
533882	The impact of two multiple-choice question formats on the problem-solving strategies used by novices and experts	Background Pencil-and-paper examination formats, and specifically the standard, five-option multiple-choice question, have often been questioned as a means for assessing higher-order clinical reasoning or problem solving. This study firstly investigated whether two paper formats with differing number of alternatives (standard five-option and extended-matching questions) can test problem-solving abilities. Secondly, the impact of the alternatives number on psychometrics and problem-solving strategies was examined. Methods Think-aloud protocols were collected to determine the problem-solving strategy used by experts and non-experts in answering Gastroenterology questions, across the two pencil-and-paper formats. Results The two formats demonstrated equal ability in testing problem-solving abilities, while the number of alternatives did not significantly impact psychometrics or problem-solving strategies utilized. Conclusions These results support the notion that well-constructed multiple-choice questions can in fact test higher order clinical reasoning. Furthermore, it can be concluded that in testing clinical reasoning, the question stem, or content, remains more important than the number of alternatives.	Background The assessment of problem-solving skills, and specifically diagnostic skills, was once reserved for examination formats such as free-response questions, patient management problems (PMPs) or oral examinations. These evaluation methods, however, are all resource-intensive, thus making it difficult to provide the representative sampling of problems necessary to circumvent the problem of case specificity, which predicts that success in solving one clinical presentation does not predict success in another [ 1 ]. As a consequence of case specificity, reliability and content validity of an examination are dependent on a broad sampling of problems. Such extensive sampling is more easily done with pencil-and-paper type of tests. This study will examine two pencil-and-paper formats specifically in regards to their relative problem-solving testing abilities. Previous literature has demonstrated that altering item stems tends to determine clinical challenge, while psychometric properties such as discrimination and difficulty tend to be affected by the number of answer options [ 2 ], hereby referred to as 'number of alternatives'. The central focus of this paper surrounds whether altering the number of alternatives within a pencil-and-paper format alters diagnostic higher order thinking and/or format psychometric properties. Two formats were studied, both with a stem consisting of a long vignette with distracters, but with different number of alternatives. The format presenting five options to the examinee will henceforth be referred to as the "multiple-choice question" or MCQ format, while the second format, presenting greater than ten options to the examinee, will be referred to as "extended-matching" or EMQ format. The first examination format studied is the five-option MCQ (see Appendix A for example). Although MCQs have always been considered an efficient and reliable testing tool, they have not always been perceived as ideal for the evaluation of higher-order thinking skills such as problem solving. Prevailing perceptions that MCQs assess lower levels of knowledge such as recall of isolated facts, and/or encourage trivialization, do exist in the medical education community [ 3 ]. To the extent that some clinicians question whether MCQs can test problem-solving skills, suggests that this format may have low validity [ 4 ]. However, as discussed by Case and Swanson [ 5 ] well constructed MCQs could challenge students to problem solve. Maguire et al also recognized that MCQs could yield valid information of clinical reasoning skills, providing that stems and alternatives are well constructed [ 6 ]. Evidence does exist that MCQs have predictive value for more recognized problem-solving tasks [ 7 ] and can elicit higher order problem solving such as forward reasoning [ 8 ]. The second examination format the EMQ format, initially designed in response to some of the criticisms of the MCQs. EMQs (see Appendix A for example) were introduced in the 1990s in both the NBME and USMLE, amongst others. Case and Swanson [ 5 ] have been instrumental in the development of these questions, which are defined as any matching format with more than the five alternatives traditionally used by MCQs. From its conception, the question preparation of the EMQs has been very careful in designing stems that test higher cognitive levels such as problem solving. The first study that examined the psychometric features of Extended-matching [ 5 ] questions showed that Extended-matching items were more difficult, more discriminating, had higher reliability and needed significantly less testing time to achieve reproducible scores than traditional MCQs. Other studies have shown that EMQs, by increasing the number of alternatives used, increased mean item difficulty as well as, perhaps by reducing guessing, provided improvement in item discrimination over the five-option MCQ [ 9 ]. By increasing item discrimination, EMQs offer comparable levels of reproducibility with 30% fewer items than the MCQ with five options [ 9 ]. Reliability coefficients were also markedly higher with Extended-matching [ 5 ]. Positive psychometric outcomes have been found in other studies using the format [ 10 - 14 ]. These studies have focused on psychometrics, whereas potential benefits, and possible reasons for such benefits, of the EMQ format over standard MCQs in eliciting higher order problem solving remain unclear. No study has formally used think-aloud protocols to assess whether a well-written MCQ differs from EMQs in challenging examinees to problem solve. There is little doubt that poorly written MCQs can encourage students to learn isolated facts by rote. In fact, all available evaluation methods potentially yield information on clinical reasoning if the content is appropriate, suggesting that content is more important than question type [ 15 ]. The two examination formats will be tested for their ability to elicit the three different diagnostic reasoning strategies generally available to learners: hypothetico-deductive reasoning, pattern recognition, and scheme-inductive reasoning. Deductive reasoning (hypothetico-deductive) [ 16 ] is a "to-and-fro" strategy of problem solving, also termed "backward reasoning". The method is generally used by novices or experienced diagnosticians to include or exclude a single diagnosis, when faced with a particularly complex problem, or as a fallback strategy when faced with clinical problems that are outside their domains of expertise. Pattern recognition has been identified by other research as a very successful approach used by experts to solve clinical problems [ 17 - 19 ]. Before becoming more expert in problem solving, learners progress through several transitional stages characterized by different knowledge structures: elaborated causal networks, abridged networks, illness scripts, and instance scripts [ 18 ]. Extensive experience eventually leads to acquisition of a repertoire of problems common to the domain of expertise termed "illness scripts". This repertoire permits problem resolution by recognition of new problems as ones that are similar or identical to old ones already solved, and the solutions are recalled. The third strategy is scheme-inductive reasoning. "Schemes" are defined as a mental categorization of knowledge that includes a particular organized way of understanding and responding to a complex situation. They are drawn on paper like "inductive trees" or "road maps" to recreate the major divisions (or chunks) used by expert clinicians for both storage of knowledge in memory and its retrieval for solving problems [ 19 , 20 ] (see Figure 1 for an example of the scheme for "dysphagia"). Decisions are explicitly at the forks in the road or branching of the tree. The organizational structure, or "scheme", proceeds from alternative causes in a forward direction, through crucial "tests", to exclusion of some alternative causes and adoption of what is left. These tests may be based on an evaluation of symptoms, signs, or results of investigations, singly or in any combination. Scheme-inductive reasoning is a strategy used by experts when pattern recognition is not possible [ 21 ]. This type of problem solving represents the "climbing of a conditional inductive tree" [ 22 ]. Figure 1 Example of the scheme for "dysphagia". By directly comparing the problem-solving strategies elicited by the two pencil-and-paper formats, using the think-aloud method previously described, two major questions will be addressed. The first question is whether pencil-and-paper formats such as EMQ and MCQ can in fact assess problem-solving skills. The examination formats' capacity to evoke more 'expert' methods of problem solving, such as scheme-inductive reasoning or pattern recognition, will be taken as evidence of their ability to assess problem-solving skills. The second question relates to the impact of the alternatives number on psychometric properties and diagnostic higher order thinking, considering that a shift to hypothetico-deductive reasoning could conceivably occur with the shorter alternatives lists of the MCQ format. A corollary to these questions is whether in testing problem solving, it is the construction of question stems that is important, as opposed to the number of alternatives or examination format. Methods Examination construction An examination for four clinical presentations, each representing a different domain in gastrointestinal medicine, was constructed: dysphagia, chronic diarrhea, nausea and vomiting, and elevated liver enzymes. The examination consisted of eight pencil-and-paper questions, with two questions, one of the MCQ type and another of the EMQ type, created for each of the four clinical presentations. While completing the questions, the examinees were permitted to write notes. The two question stems written for each of the clinical presentations were long vignettes created with a problem-solving task in mind. Furthermore, the stems within each clinical presentation (see Appendix A for the two stems for clinical presentation 'diarrhea') were designed to be as similar as possible in length, difficulty, and the presence of distracters. The stems differed only in the presence of a few key different pieces of information that led to a different diagnosis. The stems were then randomly assigned to one of the examination formats described above, MCQ or EMQ. The alternatives list included the correct diagnosis, and two plausible 'competing alternatives' to the correct answer. Subjects The examination was administered to twenty experts in Gastroenterology in two centers, Calgary (15) and Ottawa (5), as well as twenty non-experts, final-year medical students at the University of Calgary. Candidates were considered experts if they were specialists who spent more than 80% of their clinical time in the practice of Gastroenterology. Data collection The subjects were first asked to answer the eight questions. The examinees were not given a time constraint to complete the examination, though most completed it in 45 minutes. After the completion of the eight questions, the subjects, with the examination paper in hand and any written notes made during the examination, were asked to explain how they arrived at each diagnosis. A panel of two judges (experts in the Gastroenterology presentations being tested and in the recognition of the diagnostic reasoning process) interviewed the examinees. With as little prompting as possible, the examinees were asked to think-aloud [ 23 ] and describe how each diagnosis was derived. Based on the examinees' verbal discourse for that question, the two judges assigned a discrete 'Process Score' of 1, 2, or 3, depending on the predominant diagnostic process used. Once the score was assigned, the examinee was encouraged to proceed to the next question, until a diagnostic process score had been assigned for all eight questions. A 'Process Score' of 3 was assigned if pattern recognition was used. Determination that "pattern recognition" was used occurred when the subject directly reached a single diagnosis with only perfunctory attention to the alternatives. A 'Process Score' of 2 was assigned if a well-structured and accurate scheme was predominantly used to guide the inductive inquiry. Determination that a scheme-directed diagnostic reasoning strategy was used occurred by analysis of the verbal discourse using modified propositional analysis [ 24 ]. A proposition is defined as "the smallest unit of meaning that underlies the surface structure of a text" [ 25 ]. This analysis consisted of searching the examinees' discourse for key predetermined propositions that linked categories and thus provided evidence for chunking (i.e. scheme use). These key chunking propositions were determined by the authors based on information from texts, databases, consultation with experts not participating in the study, and personal experience. A recall method was utilized and felt appropriate given that of major interest to the present study was global description of representations in memory, as opposed to exact numbers of recall or specific inferences made from recalled texts [ 26 ]. The key propositions are shown in Table 1 . Table 1 Propositions demonstrating evidence of chunking. Clinical presentation Key chunking propositions Dysphagia - Oropharyngeal vs. esophageal - Mechanical vs. motility Elevated liver enzymes - Hepatocellular vs. cholestatic - Intra vs. extrahepatic cholestasis Nausea and vomiting - GI vs. non-GI causes - GI vs. metabolic vs. CNS vs. drugs Diarrhea - Small bowel vs. large bowel - Steatorrhea (malabsorption) vs. non-steatorrhea - Osmotic vs. secretory vs. inflammatory vs. motility A 'Process Score' of 1 was assigned if the examinee relied on hypothetico-deductive reasoning exclusively or predominantly. It was determined that hypothetico-deductive reasoning was the diagnostic strategy utilized when the subjects analyzed one by one each alternative diagnosis presented with the clinical vignettes prior to selecting the most likely diagnosis. The interviews were audio taped or videotaped for later review. Such reviews were required infrequently, but were found necessary when the two judges identified different reasoning strategies. The most frequent cause for differences in identification of diagnostic reasoning strategy was examinees' use of more than one strategy. For example, the candidate might initiate the diagnostic reasoning process using scheme-inductive inquiry, but resort to deductive reasoning immediately after. Disagreement between the two judges was resolved by discussion until concurrence about the diagnostic reasoning strategy was reached. The final assigned mark reflected the predominant diagnostic reasoning strategy utilized. A dichotomous score (0 for incorrect answer, 1 for correct answer) was assigned in order to compute the format psychometric properties. Data analysis Reliability of the process scores and formats was estimated using Cronbach's alpha coefficient. Item statistics were generated for each item including a discrimination index. Inter-rater reliability of diagnostic reasoning scores was estimated by a Pearson correlation coefficient. Effects of expertise, examination format, and clinical presentations on diagnostic reasoning or 'process score' A logistic regression analysis was used to determine which of the three independent variables being studied (examination format, expertise, and clinical presentation) had an impact on diagnostic reasoning or 'process score' (the dependent variable). Specifically, the analysis will model the odds of using an 'expert' method of problem-solving, that is scheme-inductive or pattern recognition (in other words, odds of not using hypothetico-deductive reasoning) in relation to the three independent variables of format, expertise and clinical presentation. An expertise effect, which would be expected, will lend evidence of construct validity to the 'process score'. Analysis was carried out using the Stata software system [ 27 ]. Results A. Reliability of 'Process Score' The two judges found it easy to agree on the broad type of strategy used by the subjects (hypothetico-deductive, scheme-directed, and pattern recognition). However, there was less agreement when the same subject used more than one diagnostic strategy. The initial diagnostic reasoning scores resulted in an agreement between the two judges of 0.84. B. Reliability and discrimination of examination formats Both formats demonstrated quite acceptable reliability and discrimination, as per Table 2 . Table 2 Cronbach alpha reliabilities and discrimination indices based on question format over all subjects. Question format Alpha coefficient Average disc. index Multiple-choice 0.76 0.63 Extended-matching 0.66 0.58 C. Relationship of examination format to cognitive process The results of the logistic regression analysis are as follows, in Table 3 . There was no difference in the odds of using 'expert' methods of problem solving (scheme-inductive or pattern recognition) across the two examination formats (MCQ or EMQ). As would be expected, experts had approximately threefold higher odds of using either of these two problem-solving methods as opposed to novices (p 0.00). There was a negative odd of using scheme-inductive and pattern recognition (-1.55) within the diarrhea and nausea/vomiting clinical presentations (i.e. more likely to use hypothetico-deductive) as opposed to the elevated liver enzymes presentation. Explanation for this lies in the fact that the diarrhea questions were the most complex for both novices and experts (in which case experts and non-experts resorted to hypothetico-deductive reasoning, as has been described in the literature [ 28 ]), while the nausea and vomiting questions were complex for the experts especially, given that the experts were gastroenterologists, but the diagnoses for this clinical presentation were 'metabolic' causes of nausea and vomiting. Table 3 Logistic regression of the odds of using an 'expert' process (either pattern recognition or scheme-inductive) Independent variable Baseline Level OR (95% CI) p value Examination format Extended-matching Multiple-choice -0.59 (-1.73, 0.56) 0.31 Expertise Expert group Non-expert group 2.69 (1.64, 3.75) 0.00 Clinical presentation Nausea and vomiting Liver enzymes -1.55 (-2.67, -0.43) 0.01 Diarrhea Liver enzymes -1.55 (-2.67, -0.43) 0.01 Dysphagia Liver enzymes 1.11 (-1.21, 1.21) 1.00 D. Ability of the two formats to evoke higher-order thinking Table 4 and Table 5 are frequency tables for the Expert and Non-expert Process Scores, across the two examination formats and four Clinical Presentations. They demonstrate that experts utilized either scheme-inductive or pattern recognition more than 90% of the time for both pencil-and-paper examination formats, while non-experts utilized these two reasoning strategies less often than experts, but still greater than 50% of the time for both formats. Table 4 Frequency table for the expert (n = 20) process scores, across two formats and four clinical presentations Question format Process score Liver enzymes Nausea and vomiting Diarrhea Dysphagia Total Multiple-choice 1: Hypothetico-deductive 0 3 2 0 5 2: Scheme 9 4 10 13 36 3: Pattern recognition 11 13 8 7 39 Total 20 20 20 20 80 Extended-matching 1: Hypothetico-deductive 1 2 1 1 5 2: Scheme-inductive 8 3 5 14 30 3: Pattern recognition 11 15 14 5 45 Total 20 20 20 20 80 Table 5 Frequency table for the non-expert (n = 20) process scores, across two formats and four clinical presentations Question Format Process score Liver enzymes Nausea and vomiting Diarrhea Dysphagia Total Multiple-choice 1: Hypothetico-deductive 6 12 13 6 37 2: Scheme-inductive 10 2 6 10 28 3: Pattern recognition 4 6 1 4 15 Total 20 20 20 20 80 Extended-matching 1: Hypothetico-deductive 8 4 14 6 32 2: Scheme inductive 8 1 4 13 26 3: Pattern recognition 4 15 2 1 22 Total 20 20 20 20 80 Discussion The present study had two major goals. The first was to determine whether the two pencil-and-paper formats studies, the MCQ and EMQ, could in fact assess problem-solving skills. In Table 4 and Table 5 , the two pencil-and-paper formats demonstrated high preponderance of scheme-inductive and pattern recognition utilization, in both experts and non-experts, thus suggesting that these question types can potentially elicit higher order clinical reasoning strategies. Another aim was to assess, by using think-aloud protocols, the impact of the alternatives number on psychometric properties and reasoning strategies employed. The logistic regression analysis shown in Table 3 demonstrates that the number of alternatives, in the form of the two examination formats used (MCQ and EMQ), did not exert an independent effect on reasoning strategy utilized. Table 2 demonstrates that both formats had good and comparable psychometric properties. The first research question of this paper was to investigate whether the examination formats used in this study, the standard five-option Multiple-choice and Extended-matching questions, were capable of testing problem-solving abilities. The observation from the data is that the two formats can potentially evoke more 'expert' methods of diagnostic reasoning processes such as scheme utilization or pattern recognition. Table 4 and Table 5 , demonstrate preponderance in both experts (greater than 90%) and non-experts (greater than 50%) of scheme-inductive and pattern recognition utilization in answering the questions. It can be concluded that by evoking these 'expert' methods of clinical reasoning, the two pencil-and-paper formats used in this study have the capability to assess diagnostic higher order thinking, assuming the question stems are constructed with a problem-solving task in mind, as was done in this study. In regards to the second main research question, the two question formats, with different number of alternatives, did not exert an independent effect on diagnostic reasoning strategy, as shown in the logistic regression analysis (Table 3 ). Shortening the number of alternatives in the MCQ format to five did not lead to an examinee 'shift' of relying on hypothetico-deductive reasoning. Explanation of this result may be found in the view raised by several authors [ 6 , 15 ] that it is not the examination format, or the number of alternatives, that dictates the cognitive level of the testing, but rather the specific construction of the question stems. We have demonstrated that a well-constructed Multiple-choice question, designed specifically to target problem solving, can achieve the purpose of testing higher order cognitive reasoning. Critics of the Multiple-choice format, who believe that it only tests recall of isolated facts, need to consider altering the construction of the stems rather than the format. That no difference exists between MCQs' and EMQs' relative ability to test for problem-solving may lie in the notion that a person's problem-solving strategy is a trait attribute and not dependent on the item format or number of alternatives. In other words, a given diagnostician will use a given strategy, such as scheme-inductive, on all questions which ask for a problem-solving task (i.e. diagnosis), regardless of the format. A well-created problem-solving question stem will challenge the examinee to use, in many cases, 'expert' (scheme-inductive, pattern recognition) diagnostic reasoning strategies to arrive at an answer, prior to looking at the alternatives. This minimizes the impact of alternatives number on the diagnostic reasoning strategy utilized, and specifically minimizes any shift to hypothetico-deductive reasoning that could have been feared occurring with the smaller number of alternatives presented in the MCQ format. The key is to create the stem with a problem-solving task in mind, and not looking for rote memorization of facts. While the findings presented do support ongoing use of the MCQ format, there is no denying that the EMQ format has demonstrated superior psychometric properties over the MCQ format in a number of studies mentioned earlier in this paper. Furthermore, in our own study, several non-expert and some expert examinees did comment that Extended-matching questions made it more difficult to go through the list of alternatives prior to answering the question. For examinees relying on hypothetico-deductive reasoning, the Extended-matching format, because of the inherent difficulty of reading through an extended alternatives list, may, at least subjectively, provide a better challenge than the Multiple-choice format. A significant limitation to the study is the manner in which the cognitive problem-solving process selected by the subjects was ascertained. Thinking aloud was used. After the completion of the examination, subjects were asked to verbally report their thinking method to two judges. The two judges independently noted the cognitive problem solving process the subjects had used in arriving at a diagnosis. Although agreement between the two judges on the process selected was identical in more than 85% of the think-aloud interviews, in the remaining 15% there was disagreement. The cognitive process was then decided by reviewing audiotapes and videotapes, so that 100% agreement could result. In other words, consensus and not initial judgements were used. Conclusions This is the first study that has used this type of think-aloud analysis to directly assess the ability of pencil-and-paper examination formats to test higher order problem solving. The results failed to show a significant difference between the two formats used, but did show that both formats can potentially evoke higher order diagnostic thinking. The results have several potential implications for medical education. Firstly, the results are important to examination construction, by demonstrating direct evidence that problem solving can be tested by pencil-and-paper formats, and specifically change some of the presented misperceptions about the standard five-option MCQ format. Secondly, demonstrating that the two formats can evoke scheme utilization is important. There is evidence [ 29 ] that the odds of diagnostic success are greater in examinees using scheme-inductive (and pattern recognition) as opposed to hypothetico-deductive reasoning. However, over and above their potential advantage in problem solving, schemes can be a very powerful tool for knowledge organization in an undergraduate curriculum. In this light, showing that MCQs and EMQs can test for scheme utilization is an important step for medical schools planning to include schemes as a teaching tool in their curricula. Lastly, this study demonstrates that testing higher order problem solving requires careful attention to question stem rather than question format or number of alternatives. A well-constructed stem will challenge examinees to choose the correct response, potentially using more expert reasoning strategies, prior to examining the alternatives. This has great potential impact on examination writers, who need not feel obliged to provide more than five alternatives, once they have carefully constructed a stem with a problem-solving task in mind. Competing interests The author(s) declare that they have no competing interests. Authors' contributions SC conceived of the study, participated in its design and coordination, and drafted the final manuscript. HM participated in the study conception, design, and revised the initial manuscript draft. PH participated in the study design and performed the statistical analysis. GF participated in the statistical analyses. All authors read and approved the final manuscript. Appendix A: The two examination formats Format: Multiple-choice question A 35 year-old woman presents with a one year history of diarrhea. She describes her stools are 10 – 12 profuse, watery, non-bloody bowel movements per day. She is eating well but has lost 7 kg over the last year. She has no abdominal pain. She is unsure if her stools are oily, but they are difficult to flush. She is otherwise perfectly well, with no previous surgeries. She smokes 1/2 pack a day but does not drink alcohol. She has never traveled, camped or drank well water. Her family history reveals an aunt with ulcerative colitis. Examination is unremarkable except for pallor. Stool C & S, O & P and C. difficile are all negative. Laboratory work shows a microcytic anemia (Hb 95, MCV 63), with low ferritin (4), but normal B12 and folate levels. 1) What is the most likely diagnosis for this patient? A) Celiac disease B) Crohn's colitis C) Villous adenoma of rectum D) Pancreatic insufficiency E) Bacterial overgrowth ANS:__________ Format: Extended-matching question A 33 year-old woman presents with a one year history of diarrhea. She describes her stools as 10 – 12 profuse, watery, non-bloody bowel movements per day. She is eating well, but has lost 9 kg over the last year. She has no abdominal pain. She sometimes sees oil droplets in her stool, and they are very difficult to flush. She had surgery for stomach ulcers at age 20, and had repeat surgery five years later for "bile gastritis". She is otherwise healthy. She smokes 1/2 pack per day but does not drink alcohol. She has not drank well water, and has not traveled or gone camping recently. Her family history is significant for two cousins with Crohn's disease. Examination is unremarkable. Stool C & S, O & P and C. difficile are all negative. Her CBC shows a macrocytic anemia (Hb 108, MCV 110) with a normal ferritin, but low B12 and elevated folate levels. Select the most likely diagnosis from the list below: __________________ A) Bacterial overgrowth B) Celiac disease C) Collagenous colitis D) Crohn's colitis E) Crohn's ileitis F) Colonic carcinoma G) Factitious diarrhea H) Giardiasis I) Ischemic colitis J) Irritable bowel syndrome K) Lactose intolerance L) Pancreatic insufficiency M) Shigella dysentery N) Villous adenoma of rectum O) Viral gastroenteritis Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC533882.xml
545082	A power law global error model for the identification of differentially expressed genes in microarray data	Background High-density oligonucleotide microarray technology enables the discovery of genes that are transcriptionally modulated in different biological samples due to physiology, disease or intervention. Methods for the identification of these so-called "differentially expressed genes" (DEG) would largely benefit from a deeper knowledge of the intrinsic measurement variability. Though it is clear that variance of repeated measures is highly dependent on the average expression level of a given gene, there is still a lack of consensus on how signal reproducibility is linked to signal intensity. The aim of this study was to empirically model the variance versus mean dependence in microarray data to improve the performance of existing methods for identifying DEG. Results In the present work we used data generated by our lab as well as publicly available data sets to show that dispersion of repeated measures depends on location of the measures themselves following a power law. This enables us to construct a power law global error model (PLGEM) that is applicable to various Affymetrix GeneChip data sets. A new DEG identification method is therefore proposed, consisting of a statistic designed to make explicit use of model-derived measurement spread estimates and a resampling-based hypothesis testing algorithm. Conclusions The new method provides a control of the false positive rate, a good sensitivity vs. specificity trade-off and consistent results with varying number of replicates and even using single samples.	Background DNA microarrays have become common tools for monitoring genome-wide expression in biological samples harvested under different physiological, pathological or pharmacological conditions. One of the most challenging problems in microarray data analysis is probably the identification of differentially expressed genes (DEG) when comparing distinct experimental conditions. In spite of its biological relevance, there is still no commonly accepted way to answer this question. An ideal DEG identification method should limit both false positives, i.e. genes wrongly called significant (type 1 errors), and false negatives, i.e. genes wrongly called not significant (type 2 errors). To this end, understanding how gene expression values measured in replicated experiments are spread around the true expression level of each gene, would help to distinguish biologically relevant gene expression changes from fluctuations due to different sources of variability that are unrelated to the biological phenomenon under investigation. Measurement error estimates can be obtained in two ways: either by empirically inferring noise from highly replicated data or by deducing noise from a theoretical error model [ 1 ]. Especially when the experimental design requires the investigation of a high number of conditions, the former strategy is not always feasible, because of the high cost of these experiments or due to the availability of biological material. In addition, there is still a lack of consensus on how gene expression values from replicated experiments should be theoretically distributed, which restricts the application also of the latter strategy. The most widely used methods for identifying DEG range from purely empirical filtering techniques (e.g. selecting genes that show a fold change higher than a fixed threshold) to more sophisticated statistical tests such as the signal-to-noise ratio described by Golub et al . [ 2 ] or the Significance Analysis of Microarrays (SAM) method by Tusher et al . [ 3 ]. While empirical filtering techniques rely on arbitrarily chosen thresholds and are unable to provide any type of control on the significance of the results, the more sophisticated statistical tests usually need a high degree of replication in the data to accurately measure gene-specific variability. In the past years various authors have proposed competing error models for microarray data from which discordant implications for the variance versus mean dependence can be deduced. Chen et al . [ 4 ] first proposed a simple Gaussian model, more recently Ideker et al . [ 5 ] and Li and Wong [ 6 ] introduced two-component models containing a multiplicative and an additive error term. All of these models implicitly or explicitly assume a constant coefficient of variation (CV), implying that standard deviation should vary proportionally with the mean. More recently, Rocke and Durbin [ 7 ] proposed a variation of the two-component model from which they derived that variance of repeated microarray measures is a quadratic function of the mean. Dealing specifically with spotted cDNA microarray technology Baggerly et al . [ 1 ] proposed a beta-binomial model, from which it can be derived that variance is a second-order polynomial function of the mean. Unfortunately, most of these models are based on theoretical assumptions that have been verified on simulated data or on data sets consisting of small numbers of replicates. More recently, Tu et al . [ 8 ] empirically modeled the variance versus mean dependence from a data set consisting of ten replicated oligonucleotide microarray experiments. According to the authors, the variance of the genes should decay exponentially with the mean, but only for moderately expression values. Taken together, all these aspects could limit the applicability of these error models. Independently from the choice of the error model, another point that remains to be faced is on how to estimate residual error. A discussed by Wright et al . [ 9 ] the possibilities range between two extremes: either obtaining a single variance estimate across all genes or obtaining a gene-specific residual variance. In the same paper a hybrid approach is proposed in which information from all genes is used to fit a single linear model from which the gene-specific variance estimates can be deduced. In the present work we chose to follow an approach similar to the latter. The aim of this study was to use highly replicated microarray data to empirically determine the true variance versus mean dependence that exists in this type of data. This knowledge enabled the proposal of PLGEM as a simple but powerful error model. We fitted the proposed model on various data sets without pre-filtering the data, deriving an improved test statistics and identifying DEG even in data sets with very little number of replicates. Results Variance versus mean dependence The relationship between measurement variability and average expression values was investigated by means of scatter plots where different measures of spread were displayed against different measures of location. For each gene absolute or relative standard deviation was plotted against the mean expression value in either linear or log-log plots using data from the 16iDC data set (Figure 1 ). Independently from the choice of standard deviation or inter-quartile range as the estimate of spread and of mean or median as the location estimate we obtained qualitatively similar plots (data not shown). Log-log plots of both absolute and relative spread estimates revealed a strikingly linear dependency, indicating that measurement spread could depend on signal location following a power law. Figure 1 Relationship between measurement variability and mean expression level . For each of the 12488 probe sets displayed on Affymetrix MG-U74Av2 chips the standard deviation (st. dev.) or the coefficient of variation (CV = st. dev. / mean) is plotted against the mean in either linear or log-log plots based on absolute expression values derived from the 16iDC data set. The power law global error model Based on the previous observation, we chose to empirically model measurement noise through linear regressions: where s and respectively represent standard deviation and mean of repeated measures. Error term ε is the realization of a random variable E that we will show later to be normally distributed as assumed when fitting a linear model. Inspired by the previous experimental observations we propose the following power law global error model (PLGEM): Model parameters α and β can be estimated from linear regression coefficients in 1 in a straightforward way: α = e c eqn . 3 β = k eqn . 4 PLGEM fitting method Instead of performing a simple linear fit through the whole set of points, we preferred to implement a method that could provide improved model robustness by partitioning the data to gain local estimates of spread as in Mutch et al . [ 9 ]. Most importantly, this method should also provide the possibility to choose different levels of confidence when modeling the spread of the data. Note that Mutch et al . [ 9 ] proposed to model within-replicates fold changes as a function of average expression using a model that was very different from PLGEM. Therefore, the following algorithm was applied: • Rank genes according to their ln( ) value and subdivide the overall expression range into a given number p of partitions containing an equal number of ranked genes. • Choose a "modeling quantile" q and determine for all the genes contained in each partition a single "modeling point" with median of ln( ) values as the x-coordinate and q-th quantile of ln( s ) values as the y-coordinate. • Finally, find a linear fit through the set of p modeling points using the least-squares method and obtain a slope k and an intercept c of the resulting regression function. Thus, for all possible combinations of p and q a slope k p,q , an intercept c p,q and a correlation coefficient r 2 p,q can be obtained. Performance of this modeling method was tested also using different combinations of partitions, in the range between 5 and 500, and quantiles, ranging from 0.01 to 0.99 (Supplementary Table 1 ). For all 77 analyzed combinations of p and q regression lines gave good fit with the modeling points, with an adjusted r 2 that was always very close to 0.99. In addition, all regression lines were strikingly parallel as judged by their slopes: 0.81 ± 0.068 (mean ± sd). The reason for not considering p > 500 was that above this number we empirically revealed a poorer modeling quality in terms of correlation coefficients (data not shown), most likely due to the decrease of the number of data points contained in each partition. Table 1 Effect of the presence of DEG when applying the classic permutation strategy to the PLGEM-STN statistic. FPR vs. significance level estimated vs. observed FDR including DEG excluding DEG including DEG excluding DEG # of genes in data set slope intercept adj. R^2 slope intercept adj. R^2 slope intercept adj. R^2 slope intercept adj. R^2 22300 1.690 3.070 0.856 0.871 -0.114 0.938 0.187 -0.383 0.314 1.090 -0.692 0.826 10000 1.710 2.470 0.881 0.888 -0.083 0.931 0.224 -0.247 0.433 1.040 -0.555 0.824 5000 1.460 1.270 0.880 0.877 -0.147 0.934 0.093 -0.228 0.067 1.080 -0.425 0.836 2500 1.590 1.180 0.888 0.864 -0.155 0.939 0.092 -0.176 0.135 1.110 -0.348 0.853 1500 1.670 0.935 0.908 0.876 -0.166 0.948 0.078 -0.122 0.170 1.130 -0.182 0.880 1000 1.720 0.689 0.874 0.944 0.002 0.956 0.038 -0.122 0.082 0.991 -0.226 0.897 500 1.900 0.263 0.864 0.857 -0.255 0.956 0.062 -0.030 0.307 1.160 0.038 0.909 200 2.490 -0.059 0.875 0.905 -0.378 0.946 0.064 -0.012 0.426 1.050 0.233 0.914 Eight data sets with different percentages of DEG were constructed from the Latin Square data set by keeping the 62 known spiked-in probe sets, but randomly removing increasing amounts of the remaining probe sets reaching the total indicated in the first column. The null distributions of the PLGEM-STN statistics were evaluated through the classic permutation strategy either including or excluding DEG. A wide range of significance levels was used to select DEG and correlation between the FPR and the significance level or between estimated and observed FDR was evaluated through linear regressions in log-log plots. Table reports slopes, intercepts and adjusted r 2 of linear models. See text for details on estimation of FDR. As a straightforward application of this modeling method, PLGEM could be fitted at the 50 th -percentile to obtain a central tendency of standard deviation to be used for improving test statistics (see next section). Another application of this method could be to fit PLGEM at the 5 th - and at the 95 th -percentile of standard deviation, to consequently find the limits of the corresponding 90% empirical confidence interval of standard deviation. In order to verify the feasibility of the former application, fitting of PLGEM on real-life data as well as distribution properties of the random variable E were investigated by analyzing the residuals of the model, i.e. differences between observed and expected values: Figure 2A–D shows distribution of residuals ε g computed from the 16iDC data set and its dependency on the rank of mean expression values. Figure 2E–L summarizes model validation on other two completely unrelated high-density oligonucleotide microarray data sets, the HG-U133A Leigh syndrome data set (Figure 2E–H ) and the HG-U95Av2 Muscle biopsies data set (Figure 2I–L ). For each tested data set an individual model was fitted and a distinct set of parameters α and β was determined. In all of the three independent data sets measurement variability could be accurately modeled through equation 2 , with a power coefficient β that was always between 0 and 1 and a random variable E that appeared to be normally distributed with zero-mean and constant standard deviation over the whole range of expression values. Of course, these findings were eventually expected only for q = 0.5, and their occurrence demonstrated a goodness of fit of PLGEM on a series of unrelated real-life data sets. Figure 2 Analysis of residuals of PLGEM fitted on three different real-life data sets . (A) PLGEM is fitted on the 16iDC data set (MG-U74Av2) following the method described in the text, setting p = 10 and q = 0.5. (B) Model residuals are plotted as a function of the rank of the mean absolute expression level. (C) Distribution of pooled residuals. (D) The quantiles of the distribution of pooled residuals are plotted against the quantiles of a standard normal distribution. The same procedure is applied to the Leigh syndrome data set (HG-U133A, panels E-H) and to the Muscle biopsies data set (HG-U95Av2, panels I-L). Improved test-statistics for detecting differential expression In order to identify DEG, we implemented the following general algorithm derived from the framework of statistical hypothesis testing, in which we test against the null hypothesis of non-differential expression. First of all, we chose to implement as the test statistic the signal-to-noise ratio (STN) already used by Golub et al . [ 2 ], because it explicitly takes unequal variances into account and because it penalizes genes that have higher variance in each class more than those genes that have a high variance in one class and a low variance in another [ 11 ]: where in the original version and represent, respectively, the mean of the replicated expression measures of gene g in condition 1 and 2, whereas and are the corresponding standard deviations. Instead, we propose to use model-derived standard deviation estimates predicted by PLGEM in equation 2 for the corresponding signal mean, rather than data-derived standard deviation values calculated independently from the few data points that are usually available for every single gene. The improvement of the test statistic in ranking DEG was evaluated as done by Broberg [ 12 ] through receiver operator characteristic (ROC) plots on the HG-U133A Latin Square data set, where there is an a priori knowledge on the truly differentially expressed transcripts. ROC plots investigate the relationship between false positive rates (FPR) and false negative rates (FNR) at different significance levels; in this way the performance of the PLGEM-derived STN statistic (PLGEM-STN) has been compared with the original STN statistic (CLASSIC-STN) and the statistic implemented in the commonly accepted Significance Analysis of Microarrays (SAM) DEG identification method (SAM-STAT). To this purpose Exp01 of the Latin Square was taken as the baseline to which the remaining 13 experiments were compared. For each comparison absolute values of each statistic were ranked in decreasing order and first n genes selected (where n ranged from 5 to 200). Figure 3 summarizes results only for the most informative comparisons, but in each tested comparison analysis PLGEM-STN was at least as good as the other two statistics for each tested value of n (data not shown). In addition, the ROC curve of PLGEM-STN always had the shortest distance from origin, indicating that it resulted in the best trade-off between sensitivity and specificity. Interestingly, improved sensitivity was observed especially when the nominal fold change was particularly low (see Exp02 vs. Exp01 and Exp14 vs. Exp01). Figure 3 Performance of PLGEM-STN in ranking differentially expressed genes . ROC plots were used to compare the sensitivity vs. specificity trade-off of the following three statistics: PLGEM-STN (black), CLASSIC-STN (blue) and SAM-STAT (red). SAM-STAT values were obtained using the R package "siggenes" [21]. Absolute values of the corresponding statistics were sorted in decreasing order, first n genes were selected (where n ranged from 5 to 200) and false positive and false negative rates were evaluated on the HG-U133A Latin Square dataset. Note that, while the transcripts in Exp02 (Exp14) are spiked-in at twice (half) the concentration than in Exp01, in both Exp06 vs. Exp01 and in Exp10 vs. Exp01 comparisons the nominal fold-change of spiked-in transcripts ranged from 32 to 512. Apart from discriminating between significant and not significant gene expression changes, an optimal test-statistic should additionally provide an accurate quantification of the actual degree of differential expression. Figure 4 shows that PLGEM-STN outperforms the competing statistics in correlating the value of the statistic with the nominal concentration variation of the known Latin Square DEG; this was particularly true for the most extreme variations. Figure 4 Correlation between the value of PLGEM-STN and the nominal concentration variation in comparison to competing statistics . Exp01 of the Latin Square data set was taken as the baseline to which the remaining 13 experiments were compared. The observed values of the indicated statistics are plotted against the nominal log ratio, deduced from the known spiked-in concentrations (left panels). A nominal log ratio of 0 is assumed for the remaining transcripts and a box-plot of their corresponding values of the indicated statistics is superimposed to the plot. For those cases where one of the two known spiked-in concentrations is 0, the value of the statistic is instead plotted against the non-null concentration (right panels). Red and black dots represent transcripts that are present in Exp01 or in the remaining 13 experiments, respectively. Identification of differentially expressed genes A resampling-based method for estimating the null distribution Though ranking of genes based on the absolute value of their test-statistic has been proven to be an effective method for selecting DEG, an even more useful way would be to compare the observed statistic with its null distribution (the distribution of values of the statistic that are expected by chance for a not differentially expressed gene), in order to control the FPR. A classic approach to empirically obtain the null distribution of a test-statistic is running a series of random permutations of the chip indexes of the full data set and re-computing the test-statistics at each permutation. Permutated test-statistics can then be pooled and significance thresholds (i.e. expected false positive rates) are found as specific quantiles of the null distribution. Nevertheless, we can foresee that the classic permutation strategy may not be optimal for estimating the actual FPR when the test-statistic makes use of a global error model such as PLGEM. We can in fact hypothesize that measurement spread of DEG may not be accurately described by means of a global error model that was designed to describe signal variability in absence of differential expression. To test this hypothesis we compared the correlation between the expected significance level and the observed FPR using PLGEM-STN and the classic permutation strategy either including or excluding DEG during the permutation step. To this end, data sets containing different percentages of DEG were obtained by merging the 62 known DEG of the Latin Square data set with differently sized random samples of not DEG extracted from the same data set. As predicted, the presence of DEG during the permutation step caused the significance level to be less correlated with the observed FPR and this correlation worsened with increasing percentages of DEG (Table 1 ). This lack of correlation was dramatically amplified when expected and observed numbers of false positives were divided by the number of selected genes to obtain an oversimplified estimate of the false discovery rate (FDR) and the observed FDR. Conversely, when DEG were omitted during the permutation step the correlation between estimated and observed FPR or estimated and observed FDR was sensibly higher for each tested percentage of DEG. We hereby by no means claim that this FDR estimate is the most accurate. A more appropriate relationship between FPR and FDR can be found in the paper by Storey and Tibshirani [ 13 ]. Nevertheless, the explicit control of the FDR goes beyond the scope of the present paper. Since in real-life data sets true DEG are unknown in advance, we propose the following resampling-based method to obtain the null distribution of not DEG when comparing n 1 replicates of condition A with n 2 replicates of condition B: • Artificial condition A* is obtained by randomly sampling with replacement n 1 indexes corresponding to the replicates of only one experimental condition. If available, chose the condition with the highest number of replicates; • Similarly sample n 2 values from the same set to obtain indexes of artificial condition B; • Compute resampled test-statistics between A and B* at each cycle. The previous resampling should be repeated a sufficiently large number of times – as large as possible compared to the total number of possible combinations and compatibly with available computational resources – and the resampled test-statistics finally pooled. In our opinion resampling the expression values from only one experimental condition, rather than permutating indexes of both conditions, makes more sense with this particular statistic, because in this way we avoid merging true and false null hypothesis. Note that when more than one condition (all with the same number of replicates) are to be compared to a common baseline, the distribution of resampled test-statistics needs to be determined only once, obviously providing a computational advantage. As a test of substantial equivalence between this resampling method and the classic permutation strategy (excluding DEG), we compared the distribution of the permutated and of the resampled PLGEM-STN test-statistics in Q-Q plots. The distribution of the PLGEM-STN resampled from Exp01 of the Latin Square data set was almost identical with the distributions of permutated PLGEM-STN obtained with the classic strategy from each comparison with the remaining 13 experimental conditions (data not shown). Figure 5 shows that the quantiles of the resampled PLGEM-STN values have a good concordance with the mean quantiles of the classically permutated statistics averaged over the 13 comparisons, implying that no differences are expected also in the gene selection step. Figure 5 Comparison of two methods for inferring the null distribution of the PLGEM-STN statistic . The classic permutation strategy (excluding DEG) was performed for each comparison in the Latin Square data set and the quantiles of the distribution of PLGEM-STN values were averaged over the 13 comparisons. The mean quantiles of the permutated statistics are plotted against the quantiles of the distribution of PLGEM-STN values obtained through the proposed resampling approach performed on the same data set but including DEG. In accordance with the previous observations, the ROC curve of the resampling method applied to the PLGEM-STN statistic was not significantly different from the ROC curve of the classic permutation strategy (excluding DEG) applied to the same statistic on the Latin Square data set (data not shown). Conversely, ROC curves of the classic permutation strategy (including DEG) applied to the CLASSIC-STN statistic and of the SAM method gave poorer performance similarly to the results in Figure 3 (data not shown). Increased robustness to varying number of replicates Another appealing feature of an optimal DEG identification method is that it should provide consistent results when different replicates of a same data set or different numbers thereof are analyzed. We therefore compared the performance of our resampling approach applied to the PLGEM-STN statistic (method 1) with SAM (method 2) and with the classic permutation strategy applied to the CLASSIC-STN statistic (method 3). The number of available replicates for each experimental condition in the Latin Square data set was unfortunately too small to investigate this particular task. We therefore took advantage of the 16iDC+LPS data set, where the first sixteen columns can be considered as the baseline condition for the remaining four experimental replicates. We then constructed a series of reduced data sets in which the baseline columns were kept constant while all possible combinations of 1, 2 or 3 replicates of LPS-stimulated DC were systematically deleted from the 16iDC+LPS data set, reaching a total of fifteen distinct data sets including the original one. Since methods 2 and 3 are not applicable on the four reduced data sets containing single samples for the LPS experimental condition, only the eleven data sets with at least two replicates were used for comparison purposes. Since the sixteen baseline columns are identical in each reduced data set, PLGEM parameters were determined only once on this common baseline condition. Significance levels used by each method in all eleven data sets were empirically selected in order to achieve a similar number of significant genes (ca. 500 probe sets) in the full data set, i.e. the one containing all available replicates. Thus, for each method eleven lists of identified DEG were obtained and the consistency between these lists was evaluated by counting the number of times each probe set was selected, giving a probes set count between 1 and 11. In Figure 6 we compared the three distinct cumulative frequency curves for each method, which show the percentage of identified DEG that were selected at least a given number times. While method 2 and 3 gave similar results, the method proposed in the present work identified a larger number of probe sets in a larger number of lists. Figure 6 Consistency of findings when different replicates or numbers thereof are analyzed . Ten reduced data sets were constructed by removing all possible combinations of 1 or 2 replicates of LPS-stimulated DC from the data set. The plot shows the cumulative frequency at which the probe sets are consistently selected in the 16iDC+LPS and in the ten reduced data sets by the following methods: the resampling approach applied to PLGEM-STN (filled squares), the permutation strategy applied to CLASSIC-STN (filled circles) and the SAM method (open triangles). In case of PLGEM, a single model was fitted on the common 16iDC baseline data set. The cumulative frequencies are normalized with respect to the total number of probe sets identified by the corresponding method. We finally evaluated the possibility of applying our method also to data sets where one of the experimental conditions was investigated only with a single sample without replication. To this end, we used the remaining four reduced data sets that could not be used in the previous comparison. In this case, the same PLGEM parameters derived from the sixteen baseline columns were applied to each of the single LPS-treated DC sample to obtain an estimate of standard deviation associated to each gene expression value, treated here as if it was a mean value from a larger group of values. Interestingly, when results obtained through this procedure were compared to the previously described results a comparable number of DEG was identified and only one probe set was newly detected in comparison to the previously identified ones (data not shown), arguing for a good consistency of results. Discussion PLGEM accurately describes GeneChip data variability In the present work we described a new global error model for microarray gene expression data that describes measurement variability with the same degree of accuracy over the whole dynamic range of values and that can be fitted at any desired quantile of spread. PLGEM has proven to correctly model signal standard deviation, in spite of the presence of different sources of variability, e.g. biological variability as well as the use of different target preparation protocols or of different chips. Moreover, PLGEM has shown to be able to deal with the great variability that exists at low expression levels while at the same time considering the significant relative reproducibility of highly expressed genes. Previously proposed error models assumed that measurement spread depended on signal location following different mathematical relationships, but none of them was based on a power law thus far. Analysis of the residuals showed a good fit of PLGEM to a number of high-density oligonucleotide microarray data sets, with model parameters being very similar to each other even when dealing with RNA samples coming from completely different biological sources and analyzed on different array layouts. This suggests that PLGEM could represent a general Affymetrix GeneChip measurement noise model. Even though scaled MAS5 Signals gave satisfactory modeling results, a further improvement could be achieved by using other emerging gene expression indices [ 6 , 14 ] or more sophisticated normalization techniques, e.g. quantile normalization [ 15 ]. Interestingly, if the same evaluation of sensitivity vs. specificity using ROC plots on the Latin Square data set was done using GCRMA expression values [ 16 ], the results were even more striking than using MAS5 Signals (data not shown). Further studies will be needed to assess if PLGEM is also able to deal with data coming from microarray technologies others than Affymetrix GeneChips. Interestingly, model parameter β was found to be quite stable and comprised between 0 and 1 in all analyzed data sets. It is noteworthy that for β ∈ (0:1) absolute variability increases with growing expression values, while relative variability decreases (compare panel B with panel D of Figure 1 ). On the other hand, none of the models mentioned in the background section seem to agree with these experimental observations. Formal statistical reasoning could unravel the underlying theoretical error model that leads to the power law relationship that was observed to be at the basis of the variance versus mean dependence in replicated microarray data. A PLGEM-based method successfully detects differential expression In spite of the lack of a theoretical statistical model, the empirical model presented here has proven its applicability in the identification of DEG, providing improved results under a wide range of different testing conditions. In comparison to other commonly used DEG identification methods, the proposed approach demonstrated improved specificity and sensitivity on the Latin Square data set and robustness to decreasing number of replicates on the 16iDC+LPS data set. The good performance of our proposed method is reasonably due to the fact that it relies on a global error model. As an example, when the classic permutation strategy is applied to the CLASSIC-STN statistic or when the SAM method is used, the selected genes are apparently more dependent on the number and identity of the replicates than when our proposed approach is used. We hypothesize that, when no error model is assumed and a small number of replicates is present in the data set, the probability of observing for some genes coincidently very similar (or very dissimilar) values increases, thus leading to an underestimation (or overestimation) of the standard deviation and a consequent overestimation (or underestimation) of the test statistic, finally leading to false positives (or false negatives). Interestingly, when the performance of our method was compared on a data set of DC stimulated for 24 hours with LPS, SAM showed a decreased sensitivity in identifying down-regulated genes when the number of LPS replicates was low (data not shown). Under these experimental conditions DC undergo a process known as maturation, which is a specialized form of cellular differentiation, for which both up- and down-regulation of gene expression is expected [ 17 , 18 ]. We speculate that SAM did not select these genes, because of the combination of two effects. First of all, down-regulated genes are expected to have lower and therefore intrinsically more variable expression values in the four LPS replicates than in the sixteen replicates of immature DC. When, in addition, the number of LPS replicates becomes too low, SAM filters these genes out to control the FDR. In agreement with this hypothesis SAM was perfectly able to identify down-regulation when the full data set was used (data not shown). The gene selection method proposed in the present work does not provide a direct control on the FDR, but the significance level has been proven to be a direct estimate of the FPR. Thus, if a significance level of 0.001 is used and 12488 probe sets are displayed on the MG-U74Av2 chip, 12–13 genes are expected to be selected by chance in cases where all genes are in fact not differentially expressed. Therefore, a researcher can test how many genes would be selected over a range of different significance levels and chose the one that results in the most acceptable compromise between number of selected genes and estimated FPR. Conclusions The proposed DEG identification method provides a direct control of the FPR and an indirect control of the FDR. Moreover, as tested on the Latin Square data set, our method improved the specificity vs. sensitivity trade-off in comparison to other commonly applied DEG selection techniques. It finally showed an increased robustness when different replicates or numbers thereof are analyzed, giving consistent results even in data sets containing single samples. In conclusion, the global error model presented here may facilitate the analysis of microarray gene expression data by discriminating information from noise, and thus possibly helping the formulation of new hypothesis concerning gene functions. Methods Data sets 16iDC RNA was harvested from ten biological samples of unstimulated immature mouse dendritic cells (DC), each extracted from an independent batch of cells. One operator prepared the biotin-labeled cRNA for hybridization from three of the ten RNA samples, a second operator prepared the remaining seven. While operator 1 applied the total RNA protocol to all of its three samples, operator 2 applied the purified mRNA protocol to five of its seven samples and the total RNA protocol to the remaining two. Two of the three cRNA samples prepared by operator 1 and four of the seven cRNA samples prepared by operator 2 have been hybridized twice; therefore, a total of 16 MG-U74Av2 GeneChips (Affymetrix, Santa Clara, CA) have been employed. Leigh syndrome Eight RNA samples were harvested from human fibroblast cell lines each deriving from a distinct Leigh syndrome patient [ 19 , 20 ] and individually hybridized on HG-U133A GeneChips (Affymetrix). Muscle biopsies Four individual and two pooled RNA samples from human muscle biopsies of sixteen healthy young male donors were hybridized on six HG-U95Av2 GeneChips (Affymetrix). This data set was downloaded from [ 21 ], experiment code: GSE80 [ 22 ]. Latin Square This data set consists of 3 technical replicates of 14 separate hybridizations (named Exp01–14) of 42 spiked transcripts in a complex human background at concentrations ranging from 0.125 pM to 512 pM. Thirty of the spikes are isolated from a human cell line, four spikes are bacterial controls, and eight spikes are artificially engineered sequences believed to be unique in the human genome. Further details on the design of the Latin Square data set can be found at [ 23 ]. Considering the redundancy of some probe sets, there are a total of 62 distinct probe sets designed to match the 42 spiked transcripts. 16iDC+LPS This data set consists of the same samples of the 16iDC data set, but includes additional four samples as a second experimental condition. To this end dendritic cells were stimulated to mature with lipopolysaccharide (LPS) for 24 hours. Two independent biological samples were harvested and individually processed by the same two operators that prepared the samples for the 16iDC data set: one applied the total RNA protocol, the other one applied the purified mRNA protocol. Each cRNA sample was hybridized twice, thus using a total of four Affymetrix MG-U74Av2 chips. Software All chips mentioned in the present study were hybridized and scanned following Affymetrix recommendations and MicroArray Suite 5.0 (MAS5) was used as the image acquisition and analysis software. All data sets used passed quality control tests and probe set signals were scaled so that the 4%-trimmed mean of all expression values of each chip was equal to a predefined reference intensity (called TGT) following manufacturer's recommendations: TGT = 100 for MG-U74Av2 and HG-U133A chips and TGT = 500 for HG-U95Av2 chips. All procedures for fitting PLGEM, for calculating observed PLGEM-based signal-to-noise ratios (STN), for obtaining expected PLGEM-STN through the resampling-based approach and for comparing observed with expected STN values have been implemented as R functions [ 24 ] and will be soon submitted for integration into the Bioconductor project [ 25 ]. Authors' contributions NP conceived the study and drafted the manuscript. MP wrote the software, participated in the design of the study and in the editing of the manuscript. CV performed the microarray experiments, participated in the design of the study and the editing of the manuscript. MC participated in the microarray experiments, AS participated in the design of the algorithms, FG and PRC coordinated the study. All authors read and approved the final manuscript. Supplementary Material Additional File 1 Performance of modeling method using different combinations of parameters p and q . The modeling method described in this study was tested on the 16iDC data set using different combinations of partitions (5, 10, 20, 50, 100, 200 and 500), and quantiles (0.01, 0.02, 0.05, 0.1, 0.2, 0.5, 0.8, 0.9, 0.95, 0.98 and 0.99). For all 77 analyzed combinations of p and q regression lines were fitted to the data as described in the text. Goodness of fit was evaluated from the resulting slope (panel A), intercept (panel B) and adjusted r 2 (panel C). Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545082.xml
526257	Autoimmune inflammatory disorders, systemic corticosteroids and pneumocystis pneumonia: A strategy for prevention	Background Pneumocystis pneumonia (PCP) is an increasing problem amongst patients on immunosuppression with autoimmune inflammatory disorders (AID). The disease presents acutely and its diagnosis requires bronchoalveolar lavage in most cases. Despite treatment with intravenous antibiotics, PCP carries a worse prognosis in AID patients than HIV positive patients. The overall incidence of PCP in patients with AID remains low, although patients with Wegener's granulomatosis are at particular risk. Discussion In adults with AID, the risk of PCP is related to treatment with systemic steroid, ill-defined individual variation in steroid sensitivity and CD4+ lymphocyte count. Rather than opting for PCP prophylaxis on the basis of disease or treatment with cyclophosphamide, we argue the case for carrying out CD4+ lymphocyte counts on selected patients as a means of identifying individuals who are most likely to benefit from PCP prophylaxis. Summary Corticosteroids, lymphopenia and a low CD4+ count in particular, have been identified as risk factors for the development of PCP in adults with AID. Trimethoprim-sulfamethoxazole (co-trimoxazole) is an effective prophylactic agent, but indications for its use remain ill-defined. Further prospective trials are required to validate our proposed prevention strategy.	Background Pneumocystis carinii was identified a hundred years ago by Chagas [ 1 ] and recognised as a pathogen in marasmic children at the end of World War II [ 2 ]. The organism came to the fore again in the early 1980s when apparently healthy homosexual men developed PCP and heralded the acquired immunodeficiency syndrome (AIDS) epidemic [ 3 ]. With highly active antiretroviral therapy (HAART) and prophylactic antibiotics, attention has turned to PCP in human immunodeficiency virus (HIV) negative individuals. We recently treated a young woman with steroid-based immunosuppression for dermatomyositis. Four months after diagnosis, she was admitted acutely breathless to the intensive care unit with a presumptive diagnosis of PCP. Although the final diagnosis was rapidly progressive alveolitis related to dematomyositis, it prompted us to consider whether we should use PCP prophylaxis for selected patients receiving systemic steroids for AID. In this article we explore the background of PCP in HIV negative patients, consider the incidence of PCP in AID, discuss predisposing factors and propose a strategy for prevention. Epidemiology Pneumocystis pneumonia is caused in humans by the recently renamed Pneumocystis jiroveci (Frenkel 1999), previously known as Pneumocystis carinii and now thought to be related to fungi on the basis of DNA analysis, despite morphological similarities to protozoa [ 4 ]. The organism shows significant genetic divergence and is host specific with no cross-species infectivity. There has been considerable debate about the nature of the relationship between humans and P. jiroveci. At one time it was hypothesised that infection occurred through reactivation of colonisation acquired in childhood, as specific antibody was found in seven out of eight normal adults [ 5 ]. This suggestion has been refuted by the absence of detectable P. jiroveci in bronchoalveolar lavage specimens from healthy volunteers, despite amplification by the polymerase chain reaction [ 6 ]. In addition, genotypic analysis of P. jiroveci from infected adults identified strains found in patients' place of residence, rather than their place of birth [ 7 ]. Investigation of apparent clusters has shown different genetic strains affecting most cases, indicating that transmission from affected cases to susceptible persons does not account for the majority of infections [ 8 , 9 ], despite the recognised transmission of Pneumocystis DNA from affected patients to their immunocompetent contact health care workers to produce colonisation [ 10 ]. As P. jiroveci has been isolated in samples of air [ 11 ] and pond water [ 12 ], it is likely that the environment represents the main source of infection for most patients. Nevertheless, isolation of known cases of PCP is advisable. Pneumocystis pneumonia is almost always exclusive to immunocompromised hosts. Two thirds of cases occur in HIV positive patients and constitutes the initial manifestation of AIDS in 46% of these patients [ 13 ]. A third of cases arise in HIV negative patients [ 14 ], a group consisting of organ transplant recipients (0–75%), haematological malignancies (9–58%), solid organ tumours (4–17.5%) and AID, usually on immunosuppressive treatment. The latter group accounts for 13–36% of cases in HIV negative patients [ 15 - 19 ]. It is this group we wish to consider in more detail. Pathogenesis P. jiroveci trophozoites proliferate and attach to type I alveolar pneumocytes causing desquamation, leading to a foamy eosinophilic exudate visible on hematoxylin-eosin staining and a "honey comb" appearance of lung tissue. Both antibody and cell mediated immunity have been postulated as being involved in host protection [ 20 ]. Clinical presentation HIV-negative patients with PCP are older (48–61 years) than HIV positive patients. Males are affected more often than females (male to female ratio 1:1.4). The clinical course is typically more acute, with an average duration of 6–13 days. The most frequent clinical symptoms are dyspnoea (63–100%), cough (55–74%), weakness (47%), loss of appetite (38%) expectoration (22–25%), sweating (19%), weight loss (13%), haemoptysis (9%) and thoracic pain (9%). Physical findings include fever >38°C (63–85%), râles (55–66%), tachycardia (25%) and tachypnoea (22%). Chest radiography may show bilateral abnormalities (68–88%), interstitial opacities (64–80%) and alveolar opacities (31–47%), but may be normal (5–7%) [ 14 - 16 , 18 , 21 ]. The clinical picture can be altered by the use of aerosolised pentamidine prophylaxis, resulting in extra-pulmonary disease [ 20 ]. P. jiroveci in HIV negative patients is associated with concurrent pulmonary infections in over 50% of cases. The implicated pathogens include Cytomegalovirus (35%), Candida (18%) and Mycobacterium tuberculosis, which contribute significantly to the mortality of PCP [ 17 , 18 , 22 ]. Diagnosis P. jiroveci infection is diagnosed in most cases by bronchoalveolar lavage, which, depending on the staining method used, has been reported to have a sensitivity of 81 to 90% and a specificity of 90 to 100% [ 23 ]. Other methods used include hypertonic saline induced sputum production, which is less sensitive, and definitive open lung biopsy. Cysts or trophozoites are morphologically identified by methenamine-silver nitrate or giemsa stains respectively. Immunospecific stains are now available and have increased the sensitivity of detection in sputum and bronchoalveolar lavage fluid [ 24 ]. The number of organisms in diagnostic specimens is high in HIV positive patients, but often low in HIV negative cases. An induced sputum sample may therefore be insufficient in the latter situation and bronchoalveolar lavage and/or transbronchial biopsy is the preferred method of investigation [ 20 ]. In patients who are at risk of PCP it is essential to have a high index of suspicion and low threshold for investigation, to allow early diagnosis and treatment. This is particularly important in those with AID, as PCP may mimic pulmonary involvement by the underlying condition, such as dermatomyositis, as exemplified by our patient who prompted this review. In SLE, 33% of patients die from infections, with 62.5% of all fatal infections being opportunistic, but only 10% of opportunistic pathogens are detected ante-mortem [ 25 ]. Management Two antibiotic regimes have been shown to work, co-trimoxazole and parenteral pentamidine isethionate. Efficacies are comparable, although side-effects are particularly common with co-trimoxazole in HIV associated PCP and affect up to 60% of such patients [ 26 ]. These include rash, neutropenia, gastrointestinal upset and liver enzyme disturbance. Although side-effects are less common with parenteral pentamidine, toxicity tends to be more severe and includes pancreatitis, hypoglycaemia, neutropenia, thrombocytopenia and orthostatic hypotension. Treatment with co-trimoxazole is usually given for 21 days in HIV cases and 14 days in HIV negative PCP. Co-trimoxazole is used in 90% of cases of PCP [ 15 ] and is associated with fewer reactions (15%) in HIV negative patients [ 22 ]. Prognosis The prognosis of HIV negative PCP is worse than for those HIV positive, with intensive care admission in 31–60%; mechanical ventilation in 14–64%; overall mortality 19–47%, rising to 50–71% on intensive care, although some series suggest an improved mortality in recent years [ 14 - 18 , 27 ]. Poor prognosis has been associated with tachypnoea, tachycardia, elevated C-reactive protein, raised lactate dehydrogenase, mechanical ventilation, and some studies suggest a correlation with previous mean steroid dose and treatment with cyclophosphamide [ 17 , 18 ]. The mortality rate of patients with underlying AID appears to be worse than for other HIV negative patients [ 15 ] and varies according to underlying pathology: 63% in Wegener's granulomatosis; 58% in inflammatory myopathy; 48% in polyarteritis nodosa; 31% in rheumatoid arthritis and 17% in systemic sclerosis [ 27 ]. Several case reports suggest that PCP in inflammatory myopathy may follow a fulminant course [ 28 ]. While PCP in HIV positive patients has been associated with a high incidence of relapse after successful therapy, this has not been seen in patients with AID, even without secondary prophylaxis and despite ongoing immunosuppressive treatment. This is thought to be related to better clearance of organism, confirmed by repeat bronchoalveolar lavage [ 22 ]. Incidence There is no specific surveillance system in the United Kingdom (UK) for PCP, other than in HIV infection. Although laboratories are invited to report isolates of P. jiroveci to the Communicable Disease Surveillance Centre, it is estimated that only a fifth of clinically diagnosed cases of PCP are reported [ 29 ]. As the laboratory-reported number of cases was 0.36 per million in 1999, we would estimate the true incidence of PCP to be approximately 1.8 per million in that year. While Pneumocystis is the second commonest reported invasive mycosis, this estimate suggests that PCP is still a rare pathogen in the UK. The overall incidence of PCP from 1990 to 1999 has declined in the UK [ 29 ], as a result of the advent of HAART for AIDS. However, probably as a result of the use of ever more immunosuppressive therapy, the number of cases of PCP diagnosed in HIV negative patients increased throughout the 1980s and 1990s [ 14 - 16 , 18 , 27 ]. These two opposing trends have resulted in HIV negative patients constituting an ever increasing proportion of the total number of cases of PCP. Attempts have been made to characterise the incidence of PCP amongst patients with AID, usually by retrospective analysis of case records. Ward & Donald's review of 223 cases of PCP in AID is the largest series and covered 2.6 million hospitalisations over the period 1983 to 1994 [ 27 ]. They found the underlying AID in this group to be SLE (42%), rheumatoid arthritis (18%), Wegener's granulomatosis (14%), inflammatory myopathy (12%), polyarteritis nodosa (9%) and systemic sclerosis (5%). An estimate of the incidence of PCP in a particular AID was derived by determining the number of cases of PCP in a particular AID per 1,000 hospitalisations with the said AID per year. The results were: 8.9/1,000 hospitalisations /year for Wegener's granulomatosis; 6.5 for polyarteritis nodosa; 2.7 for inflammatory myopathy; 1.2 for SLE; 0.8 for systemic sclerosis and 0.2 for rheumatoid arthritis. Clearly the denominator in these frequencies reflects hospital admissions per year with a particular AID. For an AID where the average annual rate of admission is less than once per year, the true incidence for that condition will be less than the rate quoted and vice versa. These frequencies, with the exception of polyarteritis nodosa, are broadly comparable with findings in other studies which report a long-term risk of PCP of 6–12% for Wegener's granulomatosis and less than 2% for other AID [ 14 , 22 , 30 ]. There are only a few isolated reports of PCP in dermatoses treated with medium-term systemic steroids such as pemphigus, pemphigoid, cutaneous necrotizing vasculitis and Behçet's syndrome [ 22 , 31 ]. Discussion Pneumocystis pneumonia in AID is unusual in the absence of steroid treatment. Corticosteroids have been recently administered in over 90% of cases in most series [ 14 , 17 , 18 , 22 ] and were the sole immunosuppressant in 17–28% of AID patients [ 18 , 22 ]. The median duration of treatment prior to the diagnosis of PCP is three to four months. Occurrence within a month of starting treatment is uncommon, with the exception of inflammatory myopathies [ 28 , 32 ]. Most cases have taken prednisolone in excess of 15 mg per day, or equivalent doses of corticosteroid. Notable is the profound inter-subject variation in response to standard steroid doses as measured by in vitro inhibition of lymphocyte proliferation [ 33 ], indicating that host factors are likely to have a significant, but as yet ill-defined role. Several mechanisms have been postulated to explain the role of steroids in promoting the development of P. jiroveci including CD4+ lymphocyte depletion and immune dysfunction [ 17 , 30 ]. Porges et al. found an association between the risk of PCP and the dose of prednisone used in SLE [ 32 ]. Similarly, Hellman et al. found an association between prednisolone dosage and risk of fatal opportunistic infection in SLE, the commonest cause of which was P. jiroveci [ 26 ]. However, other studies have failed to show an association between cumulative steroid dose and risk of PCP [ 34 ]. A number of cytotoxic and other immunosuppressive agents commonly used in the treatment of AID are frequently associated with PCP, including cyclophosphamide, azathioprine, methotrexate and ciclosporin [ 14 ]. Cyclophosphamide is routinely used in the treatment of Wegener's granulomatosis and has transformed the previous one year survival figure of 20% [ 35 ] into the present eight year survival of 80% [ 36 ]. Godeau et al. [ 34 ] showed a significant association between cyclophosphamide cumulative dose and the risk of PCP. However, this was not an independent factor in multivariate analysis when lymphopenia was taken into account [ 19 ]. In one series involving 180 patients with Wegener's granulomatosis, no cases of PCP were identified amongst patients on cytotoxic therapy alone (although the authors did not specify the numbers involved), suggesting a permissive role for corticosteroids [ 30 ]. Data on PCP associated with AID indicates lymphocytopenia (<1,000 cells/mm 3 ) is almost a prerequisite, with 91% of patients exhibiting a low lymphocyte count. Fifty percent of such PCP patients have total lymphocyte counts of <400 cells/mm 3 [ 22 ]. The pre-treatment lymphocyte count and lymphocyte counts during the first three months of immunosuppressive treatment in Wegener's granulomatosis have been shown to be predictive for PCP in multivariate analysis. A total lymphocyte count <600 cells/mm 3 was recorded in ten (83%) of 12 patients with PCP, but such a low lymphocyte count was recorded in 11 (34%) of 32 with Wegener's unaffected by PCP [ 34 ]. A similar association was found in a prospective study involving patients with SLE [ 37 ]. Porges et al [ 32 ] proposed a cut off of total lymphocyte count of <350 cells/mm 3 which captured 4 out of 6 cases with PCP and SLE, but only 1 of 20 patients with SLE unaffected by PCP. Information on CD4+ counts, which have been shown to be highly predictive of the risk of PCP in HIV infected individuals [ 38 ], is less well documented in AID patients. The issue was addressed by Mansharamani et al [ 19 ] who prospectively observed 171 patients in various risk categories for PCP, including 22 patients with active PCP. They found that patients who were at high risk of PCP had significantly lower CD4+ counts than patients at low risk. They noted that 91% of cases of PCP had CD4+ counts <300 cells/mm 3 at the time of diagnosis. Their findings are echoed by an increased risk of respiratory colonisation by P.jiroveci in HIV negative patients with CD4+ counts of <400 cells/ mm 3 [ 39 ]. Kadoya et al [ 37 ] in their study on the occurrence of PCP in 75 patients with inflammatory myopathy and SLE, noted a significant association between radiological interstitial pulmonary fibrosis (IPF) and the risk of PCP (8.8% IPF in non-PCP vs 100% IPF in PCP, p < 0.001). In contrast, PCP has only rarely been reported in idiopathic pulmonary fibrosis [ 15 ], indicating that more than systemic steroids and pulmonary fibrosis are required to put patients at excessive risk of PCP. Prevention Co-trimoxazole is commonly used for PCP prophylaxis in Wegener's granulomatosis when CD4+ counts are <300 cells/mm 3 [ 15 ] or even with normal counts in some centres [ 21 ]. This combination of antibiotics has been shown to be effective prophylaxis when used daily or thrice weekly at a dose of 960 mg in HIV positive patients [ 40 ]. Adverse effects occur in less than 20% of patients, usually manifesting as a rash, which resolves on temporary discontinuation and often does not recur on re-challenge [ 41 , 42 ]. Apart from Wegener's granulomatosis, identifying patients with AID who are at risk of PCP has proved a challenge, as the overall incidence is low. Nevertheless it remains an important issue, as AID patients contribute a considerable proportion of cases of PCP in HIV negative patients (up to 36%) and have a particularly poor prognosis, as discussed earlier. Any method used to select patients for prophylactic treatment needs to be assessed against set standards and have a high sensitivity and specificity. The standards should address the percentage of PCP cases captured by the selection criteria (ideally 100% but in practice >80%) and the risk of the condition in the selected group (which should be significant). In HIV patients who meet the criteria for PCP prophylaxis as set out by the US Public Health Service [ 43 ], the annual risk of PCP is 18% [ 38 ]. As the mortality from PCP in HIV negative patients is approximately double that of HIV-positive patients [ 14 - 18 , 27 ], we would suggest that an annual risk >9% of PCP would be sufficient to justify prophylaxis. In the study by Mansharamani et al, their proposed cut off of <300 CD4+ cells/mm 3 would capture 91% of cases of PCP in all HIV negative patients, but would also include 39–46% of patients on systemic steroids, most of whom would be unaffected by PCP. Administering prophylaxis to such large numbers of patients would unnecessarily expose patients to drug side-effects and potentially encourage drug resistance. However, analysis of their data reveals that the subgroup of patients with AID who developed PCP had CD4+ counts of <250 cells/mm 3 and six out of eight had counts <200 cells/mm 3 [ 19 ]. Given the laboratory costs, we would argue in favour of performing CD4+ counts after one month's immunosuppression only on patients who satisfy the following three screening criteria: • Steroid dosage >15 mg prednisolone or equivalent/day • >three months corticosteroid treatment proposed • total lymphocyte count <600 cells/mm 3 A CD4+ count <200 cells/mm 3 might then warrant the use of prophylactic co-trimoxazole, if the annual risk of PCP in these patients is greater than 9%. Most cases of PCP in patients with AID would be captured by these criteria, according to published series. Clearly, further prospective investigation is required to gather sufficient data to validate any selection method. To justify our proposed threshold for prophylaxis we would need to know the risk of PCP for patients on steroid-based immunosuppression for AID with CD4+ counts of <200 cells/mm 3 , information which is currently unavailable. Summary P. jiroveci infection in HIV negative patients presents more acutely and has a worse prognosis. The incidence of PCP in patients with AID is low, although these patients still represent a considerable proportion of all HIV negative cases. It is important to have a high index of suspicion of PCP when treating AID with steroid based immunosuppressive regimes, as early treatment could improve prognosis. The risk of infection is related to treatment with systemic steroid, ill-defined individual variation in steroid sensitivity and CD4+ lymphocyte count. Effective and relatively safe prophylaxis is available. Rather than opting for PCP prophylaxis on the basis of disease or treatment with cyclophosphamide, we argue the case for carrying out CD4+ lymphocyte counts on selected patients as a means of identifying individuals who are most likely to benefit from PCP prophylaxis. Further prospective trials are required to validate our proposed prevention strategy. Competing interests The authors declare that they have no competing interests. Authors' contributions AJ Carmichael initiated the discussion, appraised results, lead departmental debates and helped revise the manuscript. ES carried out the literature search, presented at meetings and wrote the original manuscript. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526257.xml
534109	Use of interrupter technique in assessment of bronchial responsiveness in normal subjects	Background A number of subjects, especially the very young and the elderly, are unable to cooperate and to perform forced expiratory manoeuvres in the evaluation of bronchial hyperresponsiveness (BHR). The objective of our study was to investigate the use of the interrupter technique as a method to measure the response to provocation and to compare it with the conventional PD 20 FEV 1 . Methods We studied 170 normal subjects, 100 male and 70 female (mean ± SD age, 38 ± 8.5 and 35 ± 7.5 years, respectively), non-smoking from healthy families. These subjects had no respiratory symptoms, rhinitis or atopic history. A dosimetric cumulative inhalation of methacholine was used and the response was measured by the dose which increases baseline end interruption resistance by 100% (PD 100 Rint, EI) as well as by percent dose response ratio (DRR). Results BHR at a cut-off level of 0.8 mg methacholine exhibited 31 (18%) of the subjects (specificity 81.2%), 21 male and 10 female, while 3% showed a response in the asthmatic range. The method was reproducible and showed good correlation with PD 20 FEV 1 (r = 0.76, p < 0.005), with relatively narrow limits of agreement at -1.39 μmol and 1.27 μmol methacholine, respectively, but the interrupter methodology proved more sensitive than FEV 1 in terms of reactivity (DRR). Conclusions Interrupter methodology is clinically useful and may be used to evaluate bronchial responsiveness in normal subjects and in situations when forced expirations cannot be performed.	Background It is known that assessment of bronchial responsiveness incorporating measurements of forced expiration can be problematic because of limited co-operation and fatigue and dizziness due to repeated forced expiratory manoeuvres. In addition, a deep inspiration, as it is required during an FEV 1 (Forced Expiratory Volume in 1 sec) procedure, causes transient bronchodilatation particularly in normals during challenge with pharmaceutical substances, resulting in interpretation difficulties [ 1 ]. Determining bronchial reactivity using a technique which measures airways resistance is less influenced by inspiratory and expiratory efforts. Furthermore, it is more sensitive to small changes in bronchoconstriction [ 2 ] and hence more suitable for studies in normal subjects, in whom the response to bronchoconstrictors is, limited [ 3 ]. The interrupter method has been shown to be a simple and non-invasive technique of measuring airway mechanics in children or patients with limited co-operation [ 4 ]. It is also suitable for diagnostic purposes in the detection and exclusion of asthma [ 5 ] and in obtaining valid rhinomanometric measurements in various groups of patients [ 6 ]. An official statement by the ATS (American Thoracic Society) on methacholine provocation indicated that the interrupter method may be useful in testing patients who cannot perform acceptable spirometry manoeuvres but its use should be restricted to laboratories with expertise in their application and interpretation [ 7 ]. Furthermore, concerns have been raised about pressure equilibration during flow interruption [ 8 ] and when small increases in resistance are used as provocation thresholds, the repeatability of the method was found unacceptably low and unsuited for clinical and research purposes [ 9 ]. In addition, the studies performed so far with this technique were done on too small numbers of subjects to allow firm scientific conclusions. We hypothesized that, since normals present with lower levels of airway obstruction during challenge, the interrupter technique in this case might be suitable and comparable with the reference PD 20 FEV 1 method, and therefore clinically useful. Methods Subjects The study was conducted in a tertiary referral centre for respiratory disease and 198 subjects were initially enrolled. All subjects were healthy with a negative history and physical examination, normal blood counts, chemistries, chest radiography and spirometry. One individual from this sample reacted to the diluent control solution (0.6%), defined as a resistance difference of > 30% baseline [ 9 ] and was excluded together with non evaluable data from 27 subjects. The final data of 170 normals were finally included, consisting of 100 (59%) males and 70 (41%) females. Predicted values for spirometry were obtained according to the European Community Coal and Steel (ECCS) [ 10 ]. All participants were given detailed information of the purpose of the study, which was approved by the hospital ethics committee, and signed a consent form. They were asked to come in the next morning, avoiding all factors listed in the ATS guidelines [ 7 ] that might cause a false negative test. Methods Routine spirometry was performed according to standardized guidelines [ 11 ]. Interrupter resistance was measured at end interruption (Rint, EI) using the technique by Phagoo et al.[ 12 ], who showed that the Rint, EI reflected changes in lung mechanics more sensitively, than interrupter resistance measured at mid or begining of interruption. The Rint, EI is calculated from the airway opening pressure (time function) signal airway opening pressure Pao(t) as follows: based on the assumption that, during a brief (100 ms) airflow interruption there is equilibration between alveolar pressure (PA) and Pao, the Rint, EI is obtained by dividing the change in pressure by the immediately preceding flow. In this study we used the alternative method of opening the interrupter [ 13 ], which calculates Rint, EI from the Pao signal using a calibration resistance. The airflow interruptions were performed using the Bronchoscreen system (Jaeger, Würzburg, Germany) [ 8 ], a computerized apparatus with a combined nebulizer-shutter head, which allows the changes in resistance of the respiratory system Rint, EI) to be measured with each breath. During quiet breathing, the opening interrupter Rint, EI was calculated. The seated subject (with noseclip in place and the cheeks partly supported by a rubber mouthpiece) breathed in a relaxed manner (in order to avoid glottic artifacts) ambient air to get accustomed to the apparatus. The shutter closed within 15 ms. The time of complete airflow interruption was 100 ms. It was triggered 150 ms after the onset of expiration. The dead space of the apparatus was 0.35 L. The pressure transducer (Honeywell 142 PC 01G; Chesham, Bucks, UK) was connected via a side port directly to the mouthpiece at a distance of 18 cm from the airway opening. Rint, EI was calculated by the formula: Rint, EI = (PA/Pm) × Rref, where PA is the end interruption mouth pressure, Pm the pressure generated during free flow and Rref is a fixed serial resistance. The triggering volume was determined by integrating the signal from a low resistance Lilly Pneumotach, which had a linearity of ± 2% at a flow below 12 L/s. Before each challenge the interrupter was calibrated. A vent produced an airflow of 105 L/min, which was led through the shutter and a calibrating resistance (0.10 kPa/L/s) and the determined Rint,EI had to be within ± 10% of the reference resistance. The above method has been found valid in the presence of mild to moderate bronchoconstriction [ 14 ], conditions that are normally met during bronchial challenge. Bronchial responsiveness was measured by a rapid methacholine provocation dosimetric test, as previously described by our group with the same apparatus but using histamine instead [ 15 ]. Briefly, 1% methacholine in saline (Lofarma, Italy) was inhaled in doubling doses starting from 200 μg, until FEV 1 had fallen = 20% compared with FEV 1 after an initial saline inhalation. The bronchial aerosol provocation system (APS Jaeger, Wurzburg, Germany) was used in this procedure. The nebulizer was calibrated to draw 5 μL of solution per automatic actuation lasting 0.6 seconds. The 100 μL of aerosol bolus had a mass median aerodynamic diameter of 1.9 μm with 80% of the droplets being less than 5.5 μm at a set pressure of 1.6 bar (22.8 psi). The subjects inhaled methacholine by slow inspiratory capacity manoeuvres guided by the green colour of light emission diodes (so that inspiratory flow was <0.5l/s) and the response was assessed 1 min after each inhaled dose. Data was assessed by using two different estimates: 1) provocation dose which increases Rint, EI by 100% (PD 100 Rint, EI), calculated by interpolation from the last two points of the cumulative semilogarithmic dose-response diagram, and 2) the percent slope (dose-response ratio-DRR) of a line extending from the origin to the last point of the curve (DRR) [ 16 ]. Plateau response was defined as difference in Rint, EI <40% after the delivery of three consecutive doubling doses and/or DRR<40% after a total cumulative dose of 4 mg or a PD 100 Rint, EI >4 mg. The DRR data were analyzed from the whole sample. The 10 day reproducibility of the PD 100 Rint, EI was investigated by randomly asking 39 subjects to come again after one week for a second examination. During the second visit we compared Rint, EI with FEV 1 as measurements of response to provocation, the latter determined 30 s after the assessment of Rint, EI [ 9 , 17 ]. At least two technically correct forced expiratory manoeuvres with an FEV 1 variation within ± 5% were received and the highest value was used for calculating the dose producing a 20% fall in FEV 1 (PD 20 FEV 1 ). Statistical analysis Regression analysis and correlation, Shapiro-Wilk test for normality and the non-parametric Mann-Whitney-U-/Wilcoxon Rank Sum test with normal approximation and the x 2 test, were used for statistical analysis. The relative duplicate error was used to assess test-retest reproducibility of the PD 100 Rint, EI (assuming a normal distribution), defined as a standard deviation of the differences divided by the v2 after log transformation (approximates coefficient of variation)[ 18 ]. Agreement between PD100Rint, EI and PD 20 FEV 1 was defined and calculated according to Bland and Altman [ 19 ]. Normal bronchial responsiveness was defined at a cut-off level of > 0.8 mg methacholine [ 20 ], while negative non-response reactions were those > 2.0 mg (8.8 μmoL)[ 21 ]. Results Subjects' anthropometric data and baseline spirometry are shown in Table 1 . Mean values of vital capacity (VC), FEV 1 and maximal expiratory flow when 50% of the forced vital capacity (FVC) remains to be exhaled (Vmax 50 ), were higher in males by 16.3%, 14, 7% and 2.5% than in females. While Rint, EI was higher in females, possibly reflecting smaller airway size, but these differences were not statistically significant. The distribution of PD 100 Rint, EI is shown in Figure 1 . Table 1 Characteristics of the study population stratified by gender. Variables Men Women n = 100 n = 70 Age, mean (range) yr 38 (18–60) 35 (18–55) Height, mean (cm), SEM 174 (0.78) 160 (0.71) Weight, mean (kg), SEM 79 (0.9) 63 (0.9) Rint, EI, mean (kPa/l/s), SEM 0.24 (0.069) 0.29 (0.074) VCin, mean (%pred), (range) 111.5 (83–144) 95.2 (76–129) FEV 1 , mean(%pred), (range) 107.9 (75–125) 93.2 (78–110) FEV 1 % mean (range) 83 (77–92) 82 (75–90) Vmax 50 , mean (%pred), (range) 83.5 (70–155) 81 (65–145) Abbreviations : VC in: inspiratory vital capacity, FEV 1 : forced expiratory volume in 1 sec, FEV 1 %: ratio of forced expiratory volume in 1 sec to forced vital capacity, Vmax 50 : maximum flow at 50% of forced vital capacity, Rint, EI:Interrupter Resistance at End Interruption Figure 1 Analysis of the distribution of PD 100 Rint, EI (threshold dose) in males and females. Values >4 mg are derived by extrapolation. Twenty one males and ten females (18%) of our subjects exhibited bronchial hyperesponsiveness. These values were normally distributed (W = 0.93, p = 0.12), with no gender related difference (x 2 = 1.48, p = 0.22, odds = 1.79). Furthermore, 5 of these 31 subjects (3 men and 2 women, 3% of total) were found to show moderate bronchial hyperesponsiveness (PD 100 Rint, EI < 0.4 mg or < 1.66 μmol methacholine), as frequently found in current symptomatic asthmatics [ 21 ]. No correlation was found of PD 100 Rint, EI to baseline post-saline Rint, EI and the respective DRR. Subjects with negative reactions (> 8.8 μmol) showed DRRs that were ten times lesser compared to those with BHR (mean ± SD = 67.52 ± 10.66 vs 690 ± 390, p < 0.001). Plateau response was exhibited by 66 (38%) of the subjects, (36 males) without gender related statistical difference (x 2 = 0.81, p= 0.36). They had DRRs that were 2.5 times smaller compared to the subjects with normal but measurable reactions (30.1 ± 9.8 vs. 75.0 ± 49.9, p = 0.024). PD100Rint, EI was found reproducible with a duplicate error of 8.3% or 0.65 doubling doses (within 140 μg). A close correlation was found between PD 100 Rint, EI and PD 20 FEV 1 (r = 0.76, 95% CI 0.53-0.88) with relatively narrow limits of agreement (Figure 2 .) Stratification of data according to BHR status is shown in Table 2 . The interrupter method showed DRRs that were more reactive in comparison to the respective DRRs of FEV 1 (approximately seven-fold). Figure 2 Bland and Altman plot of the differences between two methods against their mean value. The limits of agreement ( -2s and +2s) are -0.334 mg (-1.39 μmol) and 0.306 mg (1.27 μmol) respectively. The 95% confidence intervals are -0.364 to -0.303 mg and 0.275 to 0.336 mg, respectively. Table 2 Comparison of the methods described in the text in terms of threshold dose (sensitivity) and dose-response ratio (reactivity), stratified according to BHR status. NS: p value not statistically significance between the two methods. The greater reactivity of the interrupter method is shown. Methods (× ± SD) Subjects showing BHR PD 100 Rint, EI (mgs) 0.57 ± 0.20 Dose-response ratio (%/mg) 690 ± 390 PD 20 FEV 1 (mgs) 0.72 ± 0.66 Dose-response ratio (%mg) 98 ± 90 NS P < 0.05 Subjects with normal measurable reactions (> 0.8 mgs) PD 100 Rint, EI (mgs) 3.42 ± 3.10 Dose-response ratio (%/mg) 74.86 ± 49.87 PD 20 FEV 1 (mgs) 3.13 ± 2.65 Dose-response ratio (%/mg) 20 ± 3.82 NS P < 0.05 Abbreviations : BHR :Bronchial Hyperresponsiveness, Rint, EI:Interrupter Resistance at End Interruption, PD 20 FEV 1 :Dose Producing a 20% fall in FEV 1 . Discussion In this study we have shown that the interrupter technique, and specifically PD 100 Rint, EI, is comparable to the conventional PD 20 FEV 1 method for evaluation of BHR in a large sample of normal subjects. We have also found that this technique has a specificity of 81.2% for normal subjects and its dose response ratio is 7-fold more sensitive than the conventional FEV 1 method. It is known that specific airway conductance is four times more sensitive than FEV 1 as a measure of response to provocation but the use of a body plethysmograph makes assessment of bronchial challenge expensive and time consuming. The present methodology is particularly useful in children and in the elderly, since it is non-invasive, sensitive to changes in airway calibre and requires no subject co-operation. The opening interrupter technique offers the additional advantage of simplicity and ease of application, being particularly useful in subjects unable to perform forced manoeuvres. The Rint, EI was measured during expiration above forced residual capacity (FRC), because resistance hardly changes above this level and since subjects performed relaxed tidal flow maoeuvres, measurements were not affected by variations in breathing. Furthermore, we did not correct Rint, EI by lung volumes because the variability formed in the FRC can reduce the benefit of standardization of Rint, EI and the correlation between respiratory resistance and FRC is not significant over the limited FRC range of healthy subjects [ 21 ]. Problems in repeatability have been reported when one uses provocative concentration causing a 30% increase in Rint, EI (PC 30 Rint, EI), so we assessed the PD 100 Rint, EI threshold, which is above the 95 % confidence interval at one tail direction (1.96SD) observed in our normal sample at baseline (Table 1 ). Furthermore, although a correlation of the PC 40 Rint, EI with the classical provocative concentration causing a 20% fall in FEV 1 (PC 20 FEV 1 ) has been reported (17), data on agreement are presented for the first time in this study. If a cut-off value is set at 0,4 mg PD 100 Rint, EI methacholine, which defines severe and moderate hyperresponsiveness compatible with asthma [ 22 ] then 3% of the studied normal population was found to be in this area. This is similar to the percentage found by Malo et al. [ 23 ] working with PC 20 FEV 1 as well as to 2.5%, which represents the proportion of subjects beyond the 2 SD of the mean on one side of a normal distribution. If the PD 100 Rint, EI threshold is set at 0,8 mg methacholine [ 18 ], which includes mild BHR, then 18% of the total subjects studied had some degree of hypersensitivity. Contradictory results have been previously reported regarding the clinical significance of asymptomatic BHR. Some studies, using even stricter definitions of BHR, showed that asymptomatic BHR is of no significance [ 24 ]. In contrast other studies have reported an increased rate of decline of lung function in an asymptomatic population with BHR [ 25 ]. A plateau with a low maximal response was exhibited by 38% of the subjects, thus representing the least reactive part of the sample. Seppala et al.[ 26 ], using a method incorporating a deep inhalation (FEV 1 ) showed that 50% of normals had no calculable PC 20 FEV 1 . In this study, by direct comparison of two methods, one involving a deep inhalation, the relative effects of maximal expiratory manoeuvres on airway calibre can be indirectly assessed. The data in Table 2 show that while threshold doses between the two methods are essentially similar and not statistically significant, there is a significant difference in the DRRs between the two methods, being more pronounced in normal subjects showing BHR. Recently, Sundblad et al [ 27 ] reported a significant correlation between dose response slopes of FEV 1 and airway conductance in a large sample of subjects but not all were normals. It is known that in bronchial challenge the dose-response curve is expressed mainly by the threshold dose indicating hyperresponsiveness and the rate and magnitude of the response (hyperreactivity, DRR). The less reactive DRRs of the FEV 1 method lend support to the perturbed actomyosin equilibrium hypothesis recently described, in that with stretching there is a decrease in myosin duty cycle and the magnitude of the contractile response becomes functionally disengaged from the level of the contractile stimulus [ 28 ]. Furthemore, since there is indirect evidence of a lack of airway inflammation or remodelling that could prevent smooth muscle from stretching, our data are in agreement with those of Kolnaar et al. [ 24 ]. The fact that airway elastic recoil decreases (increase in hysteresis) when smooth muscle is contracted [ 1 ], explains the greater difference in DRRs exhibited by subjects showing BHR (the prevailing distending force of the lung allows the airway to dilate more after deep inhalation). No correlation was found in this study between BHR and baseline airway calibre, although Malo et al.[ 23 ] found a weak correlation by using a more sensitive parameter i.e. the PC 6 FEV 1 . The limits of agreement between PD 100 Rint, EI and the classical PD 20 FEV 1 were found relatively small at -1.39 μmol and 1.27 μmol respectively. This implies that this method may be used as an alternative to FEV 1 during provocation, as it is simple and easy to perform and requires no patient co-operation. Gender differences in BHR were explored because of the smaller lung size in women. Our data are in agreement with recent studies [ 29 ] that in non smoking women, lung size has no effect on bronchial sensitivity. Since RintL was found more sensitive than RintEI, a study comparing the two methods with the classical method would be interesting [ 30 ]. Conclusions In summary, the interrupter technique as an index of response to provocation has been shown to be useful to assess bronchial responsiveness in normal subjects, when maximal efforts cannot be performed. We recommend threshold doses of 100% baseline, because they show reliable agreement with the classical PD20FEV1 method. List of abbreviations ATS:American Thoracic Society, BHR:Bronchial Hyperresponsiveness, DRR:Dose Response Ratio, FEV 1: Forced Expiratory Volume in 1 sec, FVC:Forced Vital Capacity, FRC:Functional Residual Capacity, PA:Alveolar Pressure, Pao: airway opening Pressure, PC 20 FEV 1 :Provocative Concentration causing a 20% fall in FEV 1 , PC 30 Rint, EI:Provocative Concentration causing a 30% increase in Rint, EI, PD 100 Rint, EI: Provocation dose which increases Rint, EI by 100%, PD 20 FEV 1 :Provocation Dose producing a 20% fall in FEV 1 , Rint, EI:Interrupter Resistance at End Interruption, VC:Vital Capacity, Competing interests The authors declare that they have no competing interests. Authors' contributions PP, IK, AT, SA, DB were involved with the study conception. PP, AT, SA performed the interrupter technique and collected the data. PP did the statistical analysis. PP, AT, DB prepared the manuscript. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC534109.xml
539361	α1-antitrypsin and its C-terminal fragment attenuate effects of degranulated neutrophil-conditioned medium on lung cancer HCC cells, in vitro	Background Tumor microenvironment, which is largely affected by inflammatory cells, is a crucial participant in the neoplastic process through promotion of cell proliferation, survival and migration. We measured the effects of polymorphonuclear neutrophil (PMN) conditioned medium alone, and supplemented with serine proteinase inhibitor α-1 antitrypsin (AAT) or its C-terminal fragment (C-36 peptide), on cultured lung cancer cells. Methods Lung cancer HCC cells were grown in a regular medium or in a PMN-conditioned medium in the presence or absence of AAT (0.5 mg/ml) or its C-36 peptide (0.06 mg/ml) for 24 h. Cell proliferation, invasiveness and release of IL-8 and VEGF were analyzed by [ 3 H]-thymidine incorporation, Matrigel invasion and ELISA methods, respectively. Results Cells exposed to PMN-conditioned medium show decreased proliferation and IL-8 release by 3.9-fold, p < 0.001 and 1.3-fold, p < 0.05, respectively, and increased invasiveness by 2-fold (p < 0.001) compared to non-treated controls. In the presence of AAT, PMN-conditioned medium loses its effects on cell proliferation, invasiveness and IL-8 release, whereas VEGF is up-regulated by 3.7-fold (p < 0.001) compared to controls. Similarly, C-36 peptide abolishes the effects of PMN-conditioned medium on cell invasiveness, but does not alter its effects on cell proliferation, IL-8 and VEGF release. Direct HCC cell exposure to AAT enhances VEGF, but inhibits IL-8 release by 1.7-fold (p < 0.001) and 1.4-fold (p < 0.01) respectively, and reduces proliferation 2.5-fold (p < 0.01). In contrast, C-36 peptide alone did not affect these parameters, but inhibited cell invasiveness by 51.4% (p < 0.001), when compared with non-treated controls. Conclusions Our data provide evidence that neutrophil derived factors decrease lung cancer HCC cell proliferation and IL-8 release, but increase cell invasiveness. These effects were found to be modulated by exogenously present serine proteinase inhibitor, AAT, and its C-terminal fragment, which points to a complexity of the relationships between tumor cell biological activities and local microenvironment.	Background The relationship between inflammation, innate immunity and cancer is widely accepted. Early in the neoplastic process, inflammatory cells and their released molecular species influence the growth, migration and differentiation of all cell types in the tumor microenvironment, whereas later in the tumorigenic process neoplastic cells also divert inflammatory mechanisms, such as metalloproteinase production, and chemokine/cytokine functions to favour tumor spread and metastasis [ 1 - 3 ]. Human polymorphonuclear neutrophils (PMN) comprise 50–70% of circulating leukocytes and induce inflammatory reactions that can be either cytotoxic for tumour cells or promote tumour growth and metastasis [ 4 , 5 ]. For example, animal studies show that circulating neutrophils isolated from tumor-bearing animals reduce the number of metastatic foci in the lung. Other studies reveal that neutrophils stimulate tumour cell attachment to endothelial monolayers [ 6 , 7 ] and neutrophil-released cytokines, growth and angiogenic factors enhance tumour growth and spread [ 8 ]. Most of the neutrophil-induced tumour-promoting effects are attributed to their abilities to release proteases. Neutrophil degranulation results in the release of serine proteases, such as elastase, cathepsin G and protease-3, which may contribute to the activation of matrix metalloproteinases (MMPs) that mediate tumor cell invasiveness [ 9 - 12 ]. Recently, Schwartz and co-workers found that neutrophil-secreted factors can activate pro-MMP-2, which is important in ECM degradation and tumor cell invasion [ 14 ]. It is now also apparent that tumor cells can be sources of proteolytic enzymes themselves. For example, some cells possess immunoreactive neutrophil elastase [ 15 ], which is suggested to play a role in increasing tumor cell invasiveness of surrounding tissue [ 16 ]. Levels of immunoreactive neutrophil elastase in tumour extracts are reported to be an independent prognostic factor for patients with breast cancer and non-small cell lung cancer [ 17 - 19 ]. A large number of studies support the notion that proteases play an important role in the progression of malignant tumors. Therefore, the expression of proteinase inhibitors is considered to be an antimalignant event [ 20 ]. Alpha-1 antitrypsin (AAT), a major inhibitor of human serine proteases in serum, is produced mainly by the liver, but also by extrahepatic cells, including neutrophils and certain cancer cells [ 21 , 22 ]. AAT is an acute phase protein and its concentration rises up to 3–4-fold above normal during acute inflammation [ 23 ]. Several types of cancer, including non-small cell lung adenocarcinoma, have been associated with increased serum levels of AAT [ 20 , 24 ]. Clinical studies have shown that high circulating levels of AAT directly correlate with tumour progression [ 24 - 26 ]. Recently we also found that plasma levels of AAT are significantly elevated in lung cancer patients and, particularly in cases with metastases [ 27 ]. The role of AAT is poorly understood in tumor diseases although it is suggested that it can inhibit tumor cell growth and invasion, in vitro [ 28 ]. Moreover, not only the native, inhibitory form of AAT, but also conformationally modified, non-inhibitory forms are suggested to play a role in modulating tumour growth and invasiveness. For example, Kataoka and co-workers have shown that the C-terminal fragment of AAT can enhance the growth and invasiveness of human pancreas adenocarcinoma cells, in vivo [ 29 ]. Our in vitro studies revealed that the C-terminal fragment of AAT, corresponding to amino acid sequence 358–396 (C-36 peptide), induces breast tumour cell proliferation and invasiveness [ 30 ]. Recently we found that C-36 peptide induces neutrophil chemotaxis, adhesion, degranulation and superoxide generation, in vitro [ 31 ], which we propose may indirectly affect tumor cell biological activities, in vivo . Together these findings prompted us to design an experimental model in vitro that would allow us to evaluate how neutrophil-derived products themselves and in combination with exogenously added native AAT or its C-36 peptide influence lung cancer HCC cell growth, invasion and metastatic properties. Results HCC cell proliferation To examine the effect of the degranulated PMN-conditioned media alone or supplemented with exogenously added native AAT (0.5 mg/ml) or its C-36 peptide (0.06 mg/ml) on lung cancer HCC cell proliferation, we measured DNA biosynthesis using a [ 3 H]-thymidine incorporation assay (Fig. 1 ). HCC cells cultured in PMN-conditioned medium for 24 h decrease proliferation by 3.9-fold, (p < 0.001), while under the same experimental conditions cells exposed to PMN-conditioned medium supplemented with native AAT show no changes in proliferation compared to control cells cultured in a regular media. HCC cells exposed directly to native AAT decrease [ 3 H] thymidine incorporation by 2.5-fold, (p < 0.01) compared to control cells cultured in a regular medium. Thus, our data show that both PMN-conditioned medium and AAT loose their inhibitory effects on cancer cell proliferation when used in combination. Under the same experimental conditions, the C-36 peptide of AAT added either directly to HCC cells or as a supplement with PMN-conditioned medium did not exert any significant effects on HCC cell proliferation relative to controls (Fig. 1 ). Figure 1 [ 3 H]Thymidine incorporation assay. The HCC cells were exposed to AAT (0.5 mg/ml) or its C-36 peptide (0.06 mg/ml) in a regular medium or in a PMN-conditioned medium for 20 h following addition of [ 3 H] thymidine for 4 h. Each bar represents mean ± standard deviation from three separate experiments with three repeats in each. * p < 0.001, p < 0.01. HCC cell invasiveness To characterize lung cancer cell properties after their exposure to PMN-conditioned medium with and without added AAT or its C-36 peptide, we performed cell invasiveness assays. As illustrated in Fig. 2 and Table 1 , cells cultured in a PMN-conditioned medium show markedly increased cell invasiveness (by 56.8%, p < 0.001) compared to controls. When PMN-conditioned medium is supplemented with native AAT or its C-36 peptide, its effects on cancer cell invasiveness are diminished by 41.5%, and 77%, (p < 0.001), respectively, compared to PMN medium alone. It must be pointed out that HCC cells cultured in a regular medium in the presence of C-36 peptide decreased cell invasion by 51.4%, (p < 0.001), while native AAT showed slight, but not significant up-regulation of cell invasiveness, compared to controls. Together these data show that native AAT as well as its C-36 peptide abolish the capacity of PMN-conditioned medium to stimulate HCC cell invasiveness. Figure 2 HCC cell invasiveness after exposure to PMN-conditioned media, AAT or C-36 peptide separately and in combinations. HCC cells cultured in a regular medium (I) and in a PMN-conditioned medium (II) alone or supplemented with AAT or C-36 peptide (this figure represents one of three independent experiments). Table 1 Quantitative evaluation of HCC cell invasiveness (migrated cells/per view, in ten randomly selected fields). Stimulus HCC cells in a regular medium HCC cells in PMN-conditioned medium mean ♠) SEM mean SEM Control (medium) 27.8 ± 0.85 43.6* ± 1.6 AAT (0.5 mg/ml) 32.7 ± 2.1 25.5* ± 1.99 C-36 (0.06 mg/ml) 13.5* ± 1.5 10.0* ± 0.86 ♠ mean and standard error of 3 independent experiments; * -p < 0.001 Release of Vascular Endothelial Growth Factor (VEGF) and interleukin-8 (IL-8) Vascular endothelial growth factor (VEGF) and chemokine IL-8 are factors produced by tumor cells as well as by neutrophils [ 32 - 34 ], and are known to correlate with lung cancer angiogenesis in vivo [ 35 ]. We analysed for VEGF and IL-8 released by HCC cells alone or cultured in PMN-conditioned medium with and without added AAT or C-36 peptide for 24 h (Fig. 3 and 4 ). As shown in Fig 3 , AAT increases VEGF release by 1.7-fold, p < 0.001, and inhibits IL-8 by 1.4-fold, p < 0.01, relative to control cells. Under the same experimental conditions, cancer cells cultured in a PMN-conditioned medium supplemented with AAT increase VEGF and IL-8 release by 3.7-fold, p < 0.001 and 1.6-fold, p < 0.01, respectively, compared to PMN-condition medium alone. In addition, HCC cells in a PMN-conditioned medium show a slight (~27%), but not significant, decrease in VEGF levels compared to cells grown in regular medium. It must be noted that C-36 peptide had no influence on VEGF and IL-8 levels under any experimental conditions used (data not shown). Figure 3 VEGF release from HCC cells alone or exposed to PMN-conditioned medium with and without addition of AAT. Each bar represents the mean ± standard deviation of six repeats from two separate experiments with three repeats in each. * p< 0.001 Figure 4 IL-8 release from HCC cells alone or exposed to PMN-conditioned medium with and without addition of AAT. Each bar represents the mean ± standard deviation of six repeats from two separate experiments with three repeats in each. ** p < 0.01, * p < 0.05. Enzymatic activity in PMN-conditioned medium Neutrophil degranulation results in a large release of proteases, and to verify this we monitored proteolytic activity in PMN-conditioned medium by zymography, an electrophoretic technique used for the qualitative evaluation of proteases. As illustrated in Fig. 5 and 6 line-1, PMN-conditioned medium reveals a specific pattern of gelatinolytic and caseinolytic activity at a position between 72 and 180 kDa, which most likely represents neutrophil collagenase (MMP-8) (75 kDa), gelatinase A (MMP-2) (72 kDa), gelatinase B (MMP-9) (92 kDa) and several MT-MMPs, such as MT-MMP-2, -5 (72–75 kDa). Lung cancer cells grown in regular medium (Fig. 5 and 6 , line-7) or treated with AAT and C-36 peptide show no gelatinolytic (Fig. 5 , lines 6, 5) and caseinolytic activity (Fig. 6 , lines-5, 3). PMN-conditioned medium incubated with HCC cells for 24 h manifested a decrease in certain enzymatic activities (Fig. 5 , line 4, Fig. 6 , line 6) compared to PMN media alone (Fig 5 and 6 , line-1). However, no further changes in enzymatic activity profiles were observed when HCC cells were grown in PMN-conditioned medium supplemented with native AAT (Fig. 5 , line 3 and 6 line-5) or C-36 peptide (Fig 5 and 6 , line-2). Figure 5 Effects of PMN-conditioned medium alone and supplemented with AAT or C-36 peptide on HCC cell gelatinolytic activity. Lane 1-PMN-conditioned medium alone; lanes 2 and 3, PMN-conditioned medium supplemented with C-36 peptide and AAT, respectively; and incubated with cancer cells for 24 h, lane 4-PMN-conditioned medium incubated with HCC cells alone; lanes 5 and 6 – HCC cells exposed to C-36 peptide and AAT, respectively, in a regular medium; 7-HCC cells alone. This figure represents 1 of 3 separate experiments performed under the same experimental conditions. Figure 6 Effects of PMN-conditioned medium alone and supplemented with AAT or C-36 peptide on HCC cell caseinolytic activity. Lane 1-PMN-conditioned medium alone; lanes 2-PMN-conditioned medium supplemented with C-36 peptide and incubated with HCC cells for 24 h, line 3-HCC cells exposed to C-36; lane 4-; PMN-conditioned medium supplemented with AAT and incubated with HCC cells for 24 h, lane 5-HCC cells exposed to AAT; lane 6-PMN-conditioned medium incubated with HCC cells; lane 7-HCC cells alone. This figure represents 1 of 3 separate experiments performed under the same experimental conditions. Molecular profile of AAT in HCC cell culture supernatants Cell culture supernatants collected from HCC cells cultured in a regular medium or in PMN-conditioned medium with and without addition of AAT were analysed for AAT by 7.5% SDS-PAGE followed by immunoblotting using polyclonal antibody against human AAT. As shown in Fig. 7 , line-2, AAT added to HCC cells was unchanged in amount and form compared to native AAT alone (Fig. 7 , line-5), whereas AAT added as a supplement into PMN-conditioned medium is altered in the distribution of its forms: in addition to monomeric AAT, a complexed form of AAT can be detected (Fig. 7 , lane-4). Figure 7 HCC cell culture medium alone and in the presence of AAT studied by Western blot analysis. Cell culture supernatants were applied to 10% SDS-PAGE and immunoblotted with polyclonal antibody against AAT. M, molecular size markers (myosin-205 000, β-galactosidase-123 000, bovine serum albumin-79 000, carbonic anhydrase-45 700), lane 1-HCC cells alone, lane 2-HCC cells stimulated with AAT, lanes 3 and 4-HCC cells cultured in PMN-conditioned medium alone and supplemented with AAT, respectively lane 5-ATT done. The arrow indicates complexed AAT. The AAT profile shown in this figure represents one of three similar experiments. Discussion Local and systemic inflammatory mediators derived by tumor-associated PMN and tumor cells are suggested to play a role in the development of lung tumors [ 36 , 37 ]. Different studies have reported an increase of proteolytic activity in cancer and have suggested a role of proteases in tumor progression and metastasis [ 38 - 40 ]. For example, Guner and co-workers have shown that the plasma concentration of elastase in patients with malignant lung carcinoma is 10-fold higher compared to benign lesions of the lungs [ 41 ]. An increased concentration of neutrophil elastase is found to be closely associated with progression of non-small cell lung cancer (NSCL). In parallel, an increase in levels of proteinase inhibitors, such as AAT, has also been reported in tumor cases, including lung tumors. These elevated levels of proteinase inhibitors are attributed to the inflammatory reactions accompanying the tumorigenesis [ 42 , 43 ]. PMN, the most abundant circulating blood leukocytes, are potent effectors of inflammation and release a wide profile of serine and metalloproteinases [ 44 ]. Infiltration of tumors with PMN is associated with a favourable prognosis in some studies in humans, however for individual patients there is no predictable relationship between PMN infiltration and cancer prognosis [ 45 , 46 ]. Thus, in the present study we aimed to investigate how degranulated neutrophil-conditioned medium alone and supplemented with the serine proteinase inhibitor, AAT, or its C-terminal peptide (C-36), affects functional activities of a non-small cell lung adenocarcinoma cell line (HCC), in vitro . We found that PMN-conditioned medium expresses multiple effects on lung cancer HCC cell functional activities (decreases in cell proliferation and IL-8 release and stimulation of cell invasion through the ECM membrane), but has no significant effects on VEGF release. These dual neutrophil activities toward cancer cells have been described by other investigators. It has been suggested that through the release of cytokines, chlorinated oxidants and defensins, neutrophils may cause direct tumor killing [ 47 , 48 ] while in parallel, through release of serine proteinases and MMPs, breakdown of basement membranes and ECM and increase in tumor invasiveness are promoted [ 49 ]. Our findings that PMN-conditioned medium decreases HCC cancer cell proliferation and IL-8 release, but increases invasiveness, in part support the hypothesis that neutrophils can both, inhibit and promote tumor cell growth and invasiveness. In parallel experiments, we found that PMN-conditioned medium supplemented with AAT loses its effects on HCC cell invasion, proliferation and IL-8 release, but enhances VEGF. In contrast, C-36 peptide added to PMN-conditioned medium abolished the effects of medium on cell invasiveness, but had no significant influence on medium effects on cell proliferation, IL-8 and VEGF release. The major function of AAT is to inhibit neutrophil derived serine proteases, particularly neutrophil elastase [ 50 ]. A local increase of proteinases and PMN elastase-AAT complexes has been demonstrated in bronchoalveolar lavage from patients with lung cancer [ 51 ]. AAT is also known to interact with other components of degranulated-PMN. For example, AAT binds to defensins and neutralizes their effects on cell cytotoxicity and migration [ 52 ]. In our in vitro model, AAT added to degranulated PMN-conditioned medium occurs in a complexed form. Thus, the modulating effects of AAT on PMN-medium activities toward HCC cancer cells can be attributed to AAT alone, but also to its interaction(s) with components of the medium. Verification of the effects of AAT on separate components released by degranulated PMN needs further investigations. It is important to point out that AAT alone as well as in combination with PMN-conditioned medium induced VEGF release from HCC cells. VEGF is one of the most potent angiogenic molecules, regulating both angiogenesis and vascular permeability, and hence promotes tumor progression and development in NSCLC [ 53 ]. Many compounds, including anti-VEGF antibody and anti-VEGF receptor antibody have been developed as VEGF inhibitors and these compounds were reported to inhibit growth of a wide variety of tumor cell lines in vitro and in animal models, in vivo [ 54 ]. Therefore, our in vitro findings that AAT expresses potent effects on VEGF release allow classify AAT as a tumor promoter. Contrary data exist for a relationship between the levels of AAT and cancer advancement. Some investigators propose that AAT plays a protective role in tumorogenesis. For example, Finlay and co-workers showed that AAT might be directly involved in breast cancer progression by acting as a tumor suppressor [ 55 ]. Yavelow and co-workers have shown that AAT inhibits human breast cancer cells MCF-7 growth [ 56 ] and AAT was also shown to block the activation of pro MMP-2 and tumor cell invasion [ 11 ]. However, immunohistochemical studies revealed that patients with AAT-positive colon, gastric and lung adenocarcinomas had a worse prognosis than AAT-negative ones [ 57 - 59 ]. Together these findings suggest that AAT may play multiple roles in cancerogenesis in vivo, in addition to its role as proteinase inhibitor. During tumor progression, inflammatory cells produce large amounts of IL-8, an autocrine growth factor [ 53 , 60 ]. Levels of IL-8 have also been shown to correlate with lung cancer angiogenesis. Current knowledge of the degranulation process of neutrophils in vitro suggests that the chemokine IL-8 promotes rapid release of all neutrophil granular stores after cell exposure to cytoskeleton-disrupting agents [ 61 ]. In the present study, we found that the amount of IL-8 released by HCC can be reduced by AAT or PMN-conditioned medium alone, whereas these inhibitory effects are abolished when AAT and PMN-conditioned media are added together. Based on our findings, we hypothesize that IL-8 and other angiogenic factors released from HCC cells are regulated by PMN components, which may potentially initiate tumor cell activities or induce anti-tumor responses dependent on the tumor cell microenvironment. AAT is the main inhibitor of neutrophil elastase and proteinase 3 [ 62 ] and it has been reported that an imbalance between AAT and neutrophil elastase may predispose to lung cancer development [ 63 ]. Patients who carry the deficiency allele of AAT have a significantly higher risk of developing squamous cell or bronchoalveolar carcinoma of the lungs [ 64 ]. On the other hand it was reported that strong expression of AAT in lung adenocarcinoma correlates with poor prognosis [ 65 ], although it was not shown whether the inhibitory activity of AAT was normal and whether elastase levels were high among this group of patients. Moreover, AAT is a good substrate for MMPs: neutrophil collagenase, gelatinase-B, stromelysin-1 and -3, and matrilysin can effectively cleave AAT [ 66 - 69 ]. The cleavages by these MMPs occurs at peptide bonds within AAT active site loop, resulting in a generation of cleaved forms of AAT. Comparative proteome analysis performed to identify protein alterations in plasma of prostate, lung and breast-cancer patients showed significant elevation of AAT and its N-terminal fragment [ 70 ], which points to a role for different molecular forms of AAT in cancer progression. Recent studies provide evidence that the C-terminal fragment of AAT may enhance tumor growth and invasiveness in vitro and in vivo [ 29 , 30 ]. We found that the C-36 peptide displays striking concentration-dependent pro-inflammatory effects on human neutrophils, including induction of neutrophil chemotaxis, adhesion, degranulation and superoxide generation [ 31 ]. Interestingly, in our HCC cancer cell model, C-36 peptide reduced HCC cell invasiveness and also abolished PMN-conditioned medium-induced cell invasiveness, whereas there were no significant effects on other parameters measured, such as cell proliferation, VEGF and IL-8 release. Conclusions Our data show that AAT in various molecular forms expresses differential effects on lung tumour cell responses and provide an experimental evidence for complexity in the interactions of neutrophil-released molecular species and lung cancer cells. Materials and methods AAT and its C-terminal peptide Native, purified human plasma AAT was purchased from the Clinical Chemistry Department (UMAS, Malmö). The quality of AAT preparations was confirmed by 7.5% SDS-PAGE and determination of anti-elastase activity as described by Gaillard et al. [ 71 ]. Synthetic C-terminal fragment of AAT (C-36, corresponding to residues 359–394) was obtained from Saveen, Biotech AB (Denmark) and was greater than 98% purity. The C-36 peptide was prepared in sterile, endotoxin-free Tris buffered saline (0.015 M Tris, pH 7.4 containing 0.15 M NaCl) just before use. The endotoxin content in AAT was tested by quantitative E-TOXATE Assay (Sigma, USA). According to international standards no more than 0.08 enzyme U/ml of endotoxin is allowed to be present in AAT solutions [ 72 ]. In our experimental model we used AAT preparation with endotoxin levels below 0.05 enzyme U/ml. If necessary, AAT samples were purified using Detoxi-Gel™ endotoxin removing gel (Pierce, Rockford, IL, USA) according to the manufacturers recommendations. Neutrophil isolation Neutrophils were isolated from the peripheral blood of healthy donors using Polymorphprep™ (Axis-Shield PoC AS, Oslo, Norway) according to the manufacturers recommendations. Neutrophils were harvested as the lower cellular band above the red cell pellet. Residual erythrocytes were removed by a hypotonic lysis using ice cold 0.2% NaCl (w/v) for 30 s, followed addition of an equal volume of 1.6% NaCl to restore isotonicity. The neutrophil purity was typically 90% as determined by an AC900EO AutoCounter (Swelab Instruments, Sweden) and cell viability exceeded 95% according to trypan blue staining. Preparation of PMN-conditioned medium PMN-conditioned medium was prepared by degranulating neutrophils (5 × 10 6 cells/ml) according to Videm and Strand [ 72 ]. Briefly, isolated neutrophils were incubated on a shaker for 1 h at 37°C and immediately after on ice for 3 min, and centrifuged at 240 × g for 10 min at 4°C. Supernatants were collected and analyzed for the cytokine release and enzyme activity. Cell culture Human non-small cell lung carcinoma (HCC) cells were established from the lung of 54-year old women with non-small cell lung the adenocarcinoma type (DSMC No. ACC 534). The cells were routinely grown at 37°C in a humid air containing 5% CO 2 , in 125-cm 2 flasks in RPMI-1640 medium, supplemented with 10% FBS, 100 IU/ml penicillin, and 100 μg/ml streptomycin. Medium was changed every 2–3 days. Prior to experiments HCC cells were seeded into 12-wells plates at a density of 3 × 10 5 cells/ml, and cultured until confluent. After, cells were washed and a serum-free medium containing test substances was added for 24 h. Two sets of experiments were designed: the first, in which HCC cells were exposed to native AAT (0.5 mg/ml) or C-36 peptide (0.06 mg/ml) in a RPMI-1640 and the second, in which the cells were cultured in PMN-conditioned medium alone or supplemented with native AAT or C-36 peptide. [ 3 H] Thymidine incorporation assay HCC cells were incubated with test substances for 20 h. [ 3 H] Thymidine was then added (0.2 μCi/ml) for a further 4 h at 37°C. The medium was then aspirated, the cells were washed twice with 0.5 M NaCl and incubated for 5 min with 5% trichloroacetic acid. The cells were then washed with distilled water, dissolved in 1 ml 0.5 M NaOH, neutralised with 200 μl HCl and radioactivity determined in a β-counter (Packard 300CD liquid scintillation spectrometer; Packard Instruments). Protein concentration in cell lysates was determined by the Lowry method using HSA as a standard. Matrigel invasion assay Cell invasion assay was performed in an invasion chamber, a 24-well tissue culture plate with 12 inserts. The inserts contain an 8 μm pore size polycarbonate filters over which is placed a thin layer of ECMatrix™. HCC cells were suspended in a serum free medium containing 1 × 10 6 cells/ml alone or together with test substances. Each suspension was added to the upper chamber of an ECM Invasion system. Medium containing 10% of fetal bovine serum (FBS), serving as a chemoattractant, was added to the lower chamber. The chambers were incubated at 37°C in 5% CO 2 for 24 h. Invasive cells were stained and counted using microscope (Olympus BX41, PC program Olympus MicroImage) at a 100 × magnification. The number of cells per field, from 10 randomly selected fields, are presented. Enzyme activity assay HCC cells were cultured alone or stimulated with test substances for 24 h. Cell free supernatants were then analysed by zymography using 10% and 12% Tris-Glycine gels containing 0.1% gelatin or β-casein, respectively (NOVEX, Invitrogen Life Technologies, UK). Briefly, supernatants were diluted in a sample buffer and separated by electrophoresis at 125 V for 90 min. The gels were then renatured and developed over night in a developing buffer (NOVEX, Invitrogen Life Technologies, UK) at 37°C. After washing, the gels were stained with Coomassie Blue R-250. Proteases that can utilize casein or gelatine as a substrate showed up as clear zones in the gels. Endothelial growth factor (VEGF) and IL-8 analysis HCC cell supernatants were analysed for the VEGF and IL-8 levels by using commercially available quantitative ELISA kits (R&D systems, Minneapolis, USA) according to manufacturers instructions. The lowest detectable concentration for VEGF and IL-8 was 15.6 pg/ml and 10 pg/ml, respectively. Immunoblotting HCC cells supernatants were analysed by 10% SDS-PAGE gels and when transferred to a polivinylidene fluoride (PVDF) membrane (Millipore, Millipore Corporation, Bedford, MA 01730) using a semi-dry immunoblot transfer system. The blot was visualised using polyclonal rabbit antibody to human AAT (1:500) (DAKO, A/S, Denmark), secondary horseradish peroxidase-conjugated swine anti-rabbit antibody (1:800) (DAKO, A/S, Denmark) and peroxidase substrate, DAB (3,3-diaminobenzidine tetrahydrochloride) (Sigma, USA). Statistics The differences in the means of experimental results were analysed for their statistical significance with the one-way ANOVA combined with a multiple-comparisons procedure (Scheffe multiple range test), with an overall significance level of α = 0.05. Statistical Package (SPSS for Windows, release 11.0) was used for the statistical calculations. List of abbreviations AAT – alpha1-antitrypsin; C-36 – C-terminal fragment of AAT; FBS – fetal bovine serum; NSCLC – non small-cell lung cancer; MMPs – matrix metalloproteinases; MT-MMP – membrane type matrix metalloproteinase; PBS – phosphate buffered saline; PMN – polymorphonuclear neutrophil; VEGF – vascular endothelial growth factor. Authors contributions Inga Zelvyte – AB, Tim Stevens – JY, Ulla Westin – JY, Sabina Janciauskiene ES, FG	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC539361.xml
535571	The What and Why of Research on Reinforcement	Reinforcement - a process that helps prevent interbreeding between hybridising populations - is an important and little understood mechanism driving the completion of speciation	Reinforcement, like sympatric speciation (see Box 1 ), has charisma. Evolutionary biologists are still deeply uncertain about how often these processes take place, and hence how important they are in explaining the biological diversity we see today. Empirical and theoretical support for both ideas has waxed and waned over recent decades. Yet both ideas have consistently garnered an unusual amount of attention. Box 1. Glossary Sympatry —Area(s) of overlap in the ranges of populations, enabling potential interbreeding. Allopatry —Area(s) of population ranges that do not overlap with one another, preventing interbreeding. Sympatric speciation —Speciation that occurs within a range of sympatry. Reinforcement —The evolution of mechanisms that prevent interbreeding between newly interacting incipient species, as a result of selection against hybrids (narrow definition) or interspecific matings (broad definition) (See Figure 1 ). Conspecific sperm precedence —Disproportional fertilization of a female by sperm of a conspecific male, when that female has mated with both conspecific and heterospecific males. Much of the appeal of both reinforcement and sympatric speciation lies in the way they unite micro- and macroevolution. Reinforcement, a concept popularized by Dobzhansky (1937) , is a process by which speciation, a macroevolutionary process, can be driven directly by natural selection, one of the primary microevolutionary forces. Sympatric speciation can make the same claim. Because of this close linkage between the concepts, the study of one can tell us a great deal about the other (see Kirkpatrick and Ravigné 2002 ). Such studies can also reveal a lot about the general role of microevolution in species divergence. Reinforcement provides a pathway toward the completion of the speciation process. Imagine that two divergent populations (potentially even classified as separate species) come into contact after a period of allopatry ( Figure 1 ). If the populations have been apart for a long time, evolved differences between them will cause a certain degree of incompatibility when the populations come together. Often, this incompatibility comes in the form of low hybrid fitness (postzygotic isolation) or mismatched mating characteristics (premating isolation). The degree of the development of these isolating mechanisms is roughly proportional to the genetic distance between the populations, reflecting the fact that incompatibilities accumulate over time ( Coyne and Orr 1989 ). Figure 1 Schematic Diagram of Reinforcement (A) Populations diverge by evolving separately for a period of time in allopatry (separated by a geographic barrier). Populations may still be completely compatible in their mating characteristics, or these may also have diverged slightly (represented in the figure by the differences in color). (B) Secondary contact commences. This is shown through the removal of a geographic barrier that allows individuals to migrate between populations (migration is represented by a bidirectional arrow). Secondary contact can also occur by range expansion to produce an area of sympatry, or through other similar mechanisms. Due to the prior divergence between populations, hybrids have low fitness. (C) Selection to avoid producing low fitness hybrids causes the evolution of further divergence in the mating traits (represented by color) of the two populations, reducing interbreeding. If the isolating mechanisms between these populations are only partially complete, extensive hybridization may occur. This can result in fusion back into a single population, or in the swamping of one population's gene pool by the genes of the other (extinction). But there is another possibility, one that can cause the speciation between the two populations to proceed. Remember that if the populations have been separated for long enough, it is likely that hybrids between them will have relatively low fitness. Individuals who mate with members of the opposing population will therefore produce offspring of poor quality, and hence have lower fitness than individuals that mate within their own population. This favors the evolution (or further divergence) of characteristics that cause mating within, rather than between, populations ( Figure 1C ). Speciation between the populations is driven further towards completion through this increase in premating isolation. This process, the evolution of premating isolation after secondary contact due to selection against hybrids, is reinforcement sensu Dobzhansky (1937) . Recent authors have broadened the definition of reinforcement to include as a driving force any form of selection against mating between populations (e.g., Servedio and Noor 2003 ). This could include, for example, lower fertility, or higher mortality of females that mate with members of other populations. In all definitions, however, the microevolutionary process of selection is essential for reinforcement. In fact, in reinforcement, speciation itself can be thought of as an adaptive response to selection. It is little wonder that this causal linking of micro- and macroevolution has appeal for many evolutionary biologists. Reinforcement in the 21st Century Despite the substantial progress in our understanding of reinforcement that has been achieved over the last few decades, many questions remain about the process. These questions lend themselves to exploration by a broad variety of disciplines (evolution, ecology, behavior, phylogenetics, phylogeography, genetics), approaches (experimental, observational, comparative, theoretical) and taxonomic systems. Doubtless, the most important unanswered question about reinforcement is how often it occurs. It is very difficult to prove that reinforcement is occurring, or has occurred, between two species. Reinforcement occasionally leaves a signature, called reproductive character displacement, in which mating characteristics have diverged between populations in areas of sympatry but not areas of allopatry ( Figure 2 ) (the relationship between reinforcement and reproductive character displacement, and controversy over the definition of the latter, is reviewed in Howard 1993 ). In sympatric areas, populations are capable of producing hybrids, which drives reinforcement, while in allopatry hybrid production, and hence the selection for reinforcement, is absent. Reproductive character displacement has been found to be common, suggesting to some that reinforcement may be common as well ( Howard 1993 ). It is universally acknowledged, however, both that reproductive character displacement can be caused by processes other than reinforcement, and that reinforcement can occur without leaving this signature (e.g., when population ranges are completely sympatric). Proving that reinforcement has occurred requires the ruling out of several alternative hypotheses, which are themselves difficult to assess ( Noor 1999 ; Coyne and Orr 2004 ). Figure 2 The Pattern of Reproductive Character Displacement (A) Reproductive character displacement due to the presence of an area of overlap between two populations (differences in mating characteristics are represented by color changes, with the hatched area showing divergence in sympatry). Reproductive character displacement can also occur when the sympatric and allopatric areas are not contiguous. (B) Reproductive character displacement can appear as a cline in mating cues or mating preferences (y-axis), with divergence originating in the area of sympatry and spreading into areas of allopatry. Several isolated examples of reinforcement between specific pairs of species have been demonstrated, fairly conclusively, in a variety of taxa including Drosophila pseudoobscura and D. persimilis ( Noor 1995 ), flycatchers ( Sætre et al. 1997 ), sticklebacks (e.g., Rundle and Schluter 1998 ), spadefoot toads ( Pfennig 2003 ), and walking-stick insects ( Nosil et al. 2003 ) (see also reviews of Noor 1999 ; Coyne and Orr 2004 ). These studies involve a variety of behavioral tests of mate choice, analyses of hybrid fitness and the production of hybrids in the wild, and controls for alternative explanations. While examples such as these provide essential information about reinforcement, their slow rate of compilation and biased reporting do not provide efficient ways to assess how often reinforcement occurs in general. Comparative approaches, which examine patterns across a broader taxonomic group, can also provide support for reinforcement without these detailed mechanistic analyses (review in Coyne and Orr 2004 ). The revival of reinforcement in the late 1980s began with one such study in the genus Drosophila ( Coyne and Orr 1989 ). By comparing patterns across a wide number of species, such studies can give a better assessment of the potential frequency with which reinforcement occurs—without, however, providing conclusive evidence for reinforcement between specific species pairs. Another area where further research is essential is the determination of which biological factors promote reinforcement, as opposed to population fusion. Theoretical studies, using mathematical models and computer simulations, are proving useful in pinpointing the effects of many factors such as migration rates and patterns, the type of selection against interspecific mating, and the genetic basis of premating isolation (reviews in Turelli et al. 2001 ; Servedio and Noor 2003 ). Fortunately, some of the cases of reinforcement in specific species pairs are now being developed to the point where they can address similar questions (e.g., sex linkage of mating genes; Sætre et al. 2003 ). Both theoretical studies and these well developed empirical systems are also starting to address a third important area of research: how reinforcement interacts with other forces, such as ecological selection pressures, that promote speciation (e.g., Servedio 2004 ; Nosil et al. 2003 ). These integrated studies are essential to the correct placement of reinforcement within the bigger context of speciation processes. In recent years, exciting developments have started to take place in the analysis of the genetics of reinforcement (reviewed in Servedio and Noor 2003 ). These developments both parallel and overlap with progress made on the genetics of speciation and species differences in general. For example, significant progress has recently been made in identifying the genetic control of hybrid incompatibilities (e.g., Presgraves et al. 2003 ; Barbash et al. 2003 ). This progress has been accompanied by a new understanding of how chromosomal rearrangements may allow these incompatibilities to be maintained despite hybridization in sympatry ( Rieseberg 2001 ; Navarro and Barton 2003 ; Brown et al. 2004 ). Sympatric maintenance of incompatibilities, of course, has profound implications for reinforcement, which requires these incompatibilities as the force driving divergence ( Noor et al. 2001 ). Genetic analysis is also allowing a new understanding of the mechanisms by which reinforcement might be taking place in specific cases. Work by Ortiz-Barrientos et al. (2004) in this issue of PLoS Biology illustrates the extent of the insights that can be made with this approach. Using high-resolution genetic mapping the authors have identified the locations of genes that cause increased discrimination against Drosophila persimilis males by D. pseudoobscura females, due to reinforcement in sympatry. Surprisingly, these genes map to very different areas of the chromosomes than do genes that cause a basal level of mating discrimination between the species in allopatry. Among other insights, the position of these genes suggests that the reinforced discrimination is based on odor, not on the mechanism used in allopatry, male song. This leads to the novel conclusion that reinforcement is not just increasing the strength of an already existing mechanism of species discrimination, but is occurring through the development of a new discrimination system. These kinds of developments can also motivate more realistic theoretical models of the reinforcement process. Implications and Extensions of Reinforcement What if, when our assessment of the frequency of reinforcement is improved, it turns out to have been a rare occurrence in the generation of current biological diversity? The study of reinforcement is broad and varied enough that many of our findings about the process would still have wide-reaching implications. First, recall the claim, at the start of this article, that studying reinforcement reveals much about the role of microevolution in the macroevolutionary process of speciation. Knowledge gained about this relationship is not only directly applicable to the very similar process of sympatric speciation, but can also tell us a great deal about speciation caused by ecological adaptation and sexual selection, which are critical components of reinforcement in many systems (e.g., Nosil et al. 2003 ; Haavie et al. 2004 ). Studies looking for reinforcement have also led to insights into the formation and maintenance of hybrid zones (e.g., Butlin 1998 ; Britch et al. 2001 ). Situations where reinforcement fails to occur likewise teach a lesson, elucidating possible mechanisms of extinction when secondary contact occurs between species. Second, analysis of reinforcement clarifies the interactions between levels of reproductive isolation that occur at different stages in the life cycle. Reinforcement, broadly defined, can be driven by isolation at the postzygotic level or by incompatibilities that occur between mating and zygote production (postmating-prezygotic incompatibilities; Servedio 2001 ). Postzygotic isolation can likewise cause divergence at the premating stage (reinforcement) or potentially at the postmating-prezygotic stage, through the evolution of conspecific sperm precedence ( Marshall et al. 2002 ). These various stages of isolation have different degrees of importance among plants, free-spawning marine invertebrates, and other internally and externally fertilizing animals ( Bernasconi et al. 2004 ). Analysis of these stages of isolation, their interactions, and the evolutionary pressures they are under therefore has broad implications for comparative reproductive biology across these varied groups. Finally, regardless of whether reinforcement has been a common pathway in speciation, its relevance may be increasing. Reinforcement is a possible outcome anytime species that are capable of hybridization come into contact. Human activity is increasing the incidence of secondary contact by altering habitat and introducing invasive species. This contact often results in hybridization (reviews in Rhymer and Simberloff 1996 ; Mooney and Cleland 2001 ). It is important to identify and understand the properties of species pairs that make extensive introgression, extinction, stable hybrid zones, or reinforcement likely outcomes of such contact. If reinforcement has played a small role in the generation of current diversity, it may be because secondary contact itself has historically been a rare occurrence. It is the frequency of reinforcement among incidences of secondary contact that will determine its importance in the near future.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535571.xml
535559	The in vitro effect of gefitinib ('Iressa') alone and in combination with cytotoxic chemotherapy on human solid tumours	Background Activation of the epidermal growth factor receptor (EGFR) triggers downstream signaling pathways that regulate many cellular processes involved in tumour survival and growth. Gefitinib ('Iressa') is an orally active tyrosine kinase inhibitor (TKI) targeted to the ATP-binding domain of EGFR (HER1; erbB1). Methods In this study we have used a standardised ATP-based tumour chemosensitivity assay (ATP-TCA) to measure the activity of gefitinib alone or in combination with different cytotoxic drugs (cisplatin, gemcitabine, oxaliplatin and treosulfan) against a variety of solid tumours (n = 86), including breast, colorectal, oesophageal and ovarian cancer, carcinoma of unknown primary site, cutaneous and uveal melanoma, non-small cell lung cancer (NSCLC) and sarcoma. The IC50 and IC90 were calculated for each single agent or combination. To allow comparison between samples the Index SUM was calculated based on the percentage tumour growth inhibition (TGI) at each test drug concentration (TDC). Gefitinib was tested at concentrations ranging from 0.0625–2 microM (TDC = 0.446 microg/ml). This study represents the first use of a TKI in the assay. Results There was heterogeneity in the degree of TGI observed when tumours were tested against single agent gefitinib. 7% (6/86) of tumours exhibited considerable inhibition, but most showed a more modest response resulting in a low TGI. The median IC50 value for single agent gefitinib in all tumours tested was 3.98 microM. Interestingly, gefitinib had both positive and negative effects when used in combination with different cytotoxics. In 59% (45/76) of tumours tested, the addition of gefitinib appeared to potentiate the effect of the cytotoxic agent or combination (of these, 11% (5/45) had a >50% decrease in their Index SUM ). In 38% of tumours (29/76), the TGI was decreased when the combination of gefitinib + cytotoxic was used in comparison to the cytotoxic alone. In the remaining 3% (2/76) there was no change observed. Conclusion The in vitro model suggests that gefitinib may have differential effects in response to concomitant cytotoxic chemotherapy with the agents tested during this study. The mechanism involved may relate to the effect of TKIs on growth rate versus their effect on the ability of the cell to survive the stimulus to apoptosis produced by chemotherapy.	Background The epidermal growth factor receptor (EGFR) is involved in many cellular processes including cell proliferation, motility, adhesion and angiogenesis via the activation of three pathways: phosphatidylinositol-3 kinase (PI3)/Akt pathway, the Jak/STAT pathway and the ras/raf pathway. EGFR is expressed or highly expressed in a variety of human tumours including non-small-cell lung cancer (NSCLC), breast, bladder, ovarian and head and neck [ 1 ] and is therefore a promising target for cancer therapy. Gefitinib ('Iressa') is an EGFR-tyrosine kinase inhibitor (EGFR-TKI) that competitively inhibits binding of ATP at the ATP site on EGFR. It also displays remarkable selectivity for EGFR (IC50 = 0.033 microM) compared with other receptor tyrosine kinases (RTKs) that share sequence homology in the ATP binding domain [ 2 ]. In pre-clinical studies, gefitinib has demonstrated in vitro growth inhibition against a variety of human cell lines including NSCLC, ovarian, breast, colon and head and neck and is active in a range of xenograft models, including breast, colon and prostate [ 3 ]. Phase II trials with gefitinib monotherapy have produced encouraging results with clinically significant benefits observed, such as disease control rates at 250 mg/day gefitinib of 54% and 42% in IDEAL 1 and IDEAL 2, respectively [ 4 , 5 ]. Results from Phase III trials investigating gefitinib in combination with cisplatin and gemcitabine (INTACT 1) [ 6 ] and gefitinib in combination with paclitaxel and carboplatin (INTACT 2) [ 7 ] in NSCLC concluded there was no added benefit in patients receiving chemotherapy plus gefitinib; however the tolerability of gefitinib was confirmed. At present, there is conflicting evidence relating the activity of gefitinib directly to the levels of EGFR expression. One group found that the concentration of gefitinib required to inhibit ligand-independent growth by 50% (IC50) in four bladder cancer cell lines ranged from 1.8–9.7 microM and correlated with EGFR protein and transcript level [ 8 ]. However, another study using human tumour xenografts found that gefitinib caused growth inhibition of tumours and enhancement of the activity of a number of cytotoxic drugs, but neither was dependent on high levels of EGFR expression [ 9 ]. Moreover, no consistent association was demonstrated between EGFR expression and clinical outcome in IDEAL 1 and 2 [ 10 ]. Alternative explanations for the activity of gefitinib in systems where EGFR is not over expressed include inhibition of EGFR pathway activation mediated by increased levels of receptor ligands e.g. epidermal growth factor (EGF) and transforming growth factor-alpha (TGF-alpha); heterodimerization with HER2 and cross talk with heterologous receptors; and EGFR mutations yielding a constitutively active receptor that is not down-regulated by endocytosis [ 11 ]. There is evidence that the ras/raf pathway mediates proliferation [ 12 ], whereas the PI3/Akt pathway is essential for cell survival and may be constitutively activated in many tumours by loss of PTEN [ 13 ]. We have previously shown that the ATP-based tumour chemosensitivity assay (ATP-TCA) can be used to measure the effects of cytotoxic agents and antibodies against human tumour-derived cells, and that this matches clinical outcome in a number of tumour types [ 14 , 15 ]. Use of the assay to direct choice of chemotherapy has been shown to improve response rate and progression-free survival in ovarian cancer [ 16 , 17 ] and a fully randomized trial of assay-directed versus physician's choice of chemotherapy for platinum-resistant ovarian cancer is in progress [ 18 ]. The assay system has been used to assist the development of a number of new agents and combinations [ 19 , 20 ], but this represents the first use of a TKI in the assay. EGF and TGF-alpha, ligands of EGFR, act as survival factors for many cells as well as growth factors. As many cytotoxic agents induce apoptosis, gefitinib may be able to potentiate their effects by reducing survival stimuli. The current pilot study was undertaken to assess the effect of gefitinib in combination with existing chemotherapeutic agents (cisplatin, gemcitabine, oxaliplatin, treosulfan) against a wide range of tumour types. Methods Tumours A total of 86 tumours (57 females:29 males) were tested in this study, with a median age of 59 years (range 21–90). The samples tested consisted of the following tumour types; breast adenocarcinoma (n = 8), colorectal carcinoma (n = 18), cutaneous melanoma (n = 7), NSCLC (n = 1), oesophageal adenocarcinoma (n = 4), ovarian carcinoma (n = 26), sarcoma (n = 2), squamous cell carcinoma (n = 2), sweat gland carcinoma (n = 1), uveal melanoma (n = 12) and carcinoma of unknown primary site (n = 5). The 26 ovarian carcinomas were all recurrent stage 3/4 cancers and 25/26 were pre-treated (11 with carboplatin and 14 with carboplatin + paclitaxel). Of the remaining samples, 10/86 had been treated with a variety of chemotherapy regimens and some patients had more than one treatment; epirubicin + cisplatin + 5-Fluorouracil (5-FU) (n = 3), epirubicin + cyclophosphamide (4-HC) (n = 1), 4-HC+methotrexate+5-FU (CMF) (n = 2), cisplatin + vinorelbine (n = 1), mitomycin C + 5-FU (n = 1), mitoxantrone + paclitaxel (n = 1), chlorambucil (n = 1), 4-HC (n = 1) and irinotecan (n = 1). The remaining 51 patients had no previous treatment. In each case only tumour material not required for diagnosis was sent for ATP-TCA, and in all cases consent had been obtained and permission had been granted by the local ethics committee. ATP-TCA The ATP-TCA was performed as previously published [ 14 , 21 ]. Solid tumour or ascites samples were transported to the laboratory in transport medium, consisting of Dulbecco's Eagles Media (DMEM) (Sigma, UK; D6171). Solid samples were dissected under sterile conditions in a BioQ Microfuge Class II Hood and placed into a 0.75 mg/ml collagenase solution (Sigma, UK; C8051) for enzymatic dissociation overnight. Following dissociation, the single celled suspension or ascites sample was washed using DMEM supplemented with 1 M HEPES, (Sigma, UK; H0887), 100 IU/mL penicillin, 10 mg/mL streptomycin (Sigma, UK; P0781) and 10 mg/mL gentamicin (Sigma, UK; G1272). The final cell suspension was then plated in 96-well polypropylene plates (Corning Life Sciences, High Wycombe, UK) at 20,000 (solid sample) or 10,000 (ascites sample) cells/well in a serum-free complete assay medium (CAM, DCS Innovative Diagnostik Systeme, Hamburg, Germany). Drugs were added to triplicate wells at serial dilutions corresponding to 200–6.25% of a test drug concentration (TDC) estimated from pharmacokinetic data, which included the degree of protein binding. Two controls were included in each plate: one with no drug and consisting of media only (MO), and a maximum inhibitor (MI) control which killed all cells present. The plates were incubated for 6 days at 37°C with 5% CO2. At the end of the incubation period, remaining cells were lysed by addition of an ATP extraction reagent (DCS Innovative Diagnostik Systeme). An aliquot of the lysate from each well was added to the corresponding wells of a white 96 well microplate (Thermo Life Sciences, Basingstoke, UK), followed by addition of luciferin-luciferase reagent. The light output corresponding to the level of ATP present was measured in a luminometer (MPLX, Berthold Diagnostic Systems, Hamburg, Germany). These data were transferred automatically to an Excel spreadsheet where the % inhibition achieved at each concentration tested was calculated using the equation; 1-(test-MI)/(MO-MI) × 100. Several parameters of efficacy can be calculated e.g. IC50 and IC90, however previous ATP-TCA studies have found that a natural logarithmic sum index (Index SUM ) calculated by direct addition of the percentage survival at each concentration tested (Index = 600-Σb3;%Inhibition6.25...200) provides a better indication of sensitivity or resistance to different drugs in different tumour types [ 22 ]. The total inhibition of growth resulted in an index of 0, and no inhibition of growth at any concentrations produces an index of 600 [ 23 ]. Area under the concentration-inhibition curve (Index AUC ) was calculated from the data using the trapezoidal rule. Data Analysis The results were entered into an Access 2000 database for further analysis. Statistical tests were performed using non-parametric methods. Drugs The cytotoxic drugs used in the assay were obtained as vials for injection and made up according to manufacturers' instructions. Gemcitabine, oxaliplatin and treosulfan were stored in aliquots at -20°C, while cisplatin was stored at room temperature. Table 1 shows the 100% TDC for each of the drugs used. Drug combinations were tested by combining single agents. The EGFR-TKI, gefitinib (kindly provided by AstraZeneca) was tested at concentrations ranging from 0.06–2 microM (100% TDC = 0.99 microM). Immunohistochemistry Tissue was available for EGFR immunohistochemical staining in 31/86 (36%) cases comprising of 4 breast carcinomas, 12 colon carcinomas, 2 oesophageal carcinomas, 2 ovarian carcinomas, 1 sarcoma, 4 skin melanomas, 5 uveal melanomas and 1 carcinoma of unknown primary site. Paraffin embedded sections of 4 μm thick were dewaxed and rehydrated in preparation for immunohistochemical staining. Endogenous peroxidase was blocked using 3% hydrogen peroxide in methanol. The sections were pretreated with 0.1% Trypsin (CaCl2/Tris buffer pH8.0) for 10 minutes at 37°C for antigen retrieval. Immunohistochemical studies were performed according to manufacturer's instructions of the Vectastain Universal ABC-AP kit (Vector Laboratories, Burlingame, California, U.S.A), which uses an avidin-biotin complex method and Vector red as the chromogen. Monoclonal antibody for EGFR, Clone E30 (Dakocytomation, Cambridgeshire, U.K) was used at a dilution of 1:20 and incubated with sections for 18 hours at 4°C. Positive (squamous cell carcinoma tissue) and negative controls were included in each staining procedure. Samples were assessed by a pathologist using the H-score. Intensity was graded on a scale ranging between 0, 1+, 2+ or 3+, (where 1+ equals weak staining, 2+ equals moderate and 3+ equals intense) and the proportion of cells stained at the highest intensity. The two values were then multiplied together to give the final value. The same tissue available for EGFR staining was also available for pAkt staining. Paraffin embedded sections of 4 μm thick were dewaxed and rehydrated in preparation for immunohistochemical staining. Endogenous peroxidase was blocked using 3% hydrogen peroxide in methanol. The sections were pretreated with 0.1 M citrate buffer in a pressure cooker for 2.5 minutes for antigen retrieval. Immunohistochemical studies were performed according to manufacturer's instructions of the Vectastain Universal ABC-AP kit (Vector Laboratories, Burlingame, California, U.S.A), which uses an avidin-biotin complex method and Fuchsin as the chromogen. Phospho-Akt, Ser473 (#9277 L, Cell Signalling, MA. USA) was used at a dilution of 1:50 and incubated with sections for 18 hours at 4°C. Positive (prostate cancer tissue) and negative controls were included in each staining procedure. Samples were assessed as described previously. Results Gefitinib showed low inhibition (Index SUM >300) across the range of concentrations tested in the ATP-TCA, with little evidence of increasing inhibition with increasing drug concentration. 7% (6/86) of tumours exhibited considerable inhibition (>50% inhibition at 100% TDC), but most showed a more modest response resulting in a low maximum percentage inhibition (Figure 1 ). The estimated median IC50 and IC90 value for single agent gefitinib in all tumours tested was 3.98 microM (<0.1–69.9 microM) and 6.45 microM (2.4–125.9 microM) respectively. The median IC50 for individual tumour types tested is shown in Table 2 . There was heterogeneity in the degree of inhibition observed when tumours were tested against single agent gefitinib (Figure 2 ). To compare between tumours, an Index SUM of <300 corresponding to 50% inhibition across the range of concentrations tested was used to compare results. On this basis, single agent gefitinib was effective against 5% (4/86) of samples, comprising 1 colorectal tumour, 1 ovarian tumour, 1 uveal melanoma and 1 unknown primary carcinoma. In 88% (76/86) of samples there was sufficient material to test gefitinib in combination with different cytotoxics. Table 3 shows the median results for single-agent cytotoxics tested compared to results when tested in combination with gefitinib. In samples tested with gefitinib in combination with cisplatin (n = 6) only 33% (2/6) showed increased sensitivity (i.e. a decrease in their Index SUM ), compared to when cisplatin was used alone. The remaining 67% (4/6) showed increased resistance (i.e. an increase in their Index SUM ). This compares with gefitinib in combination with oxaliplatin (n = 10) where 90% (9/10) of samples showed an increase in sensitivity with the combination, with 1 sample showing a >50% decrease in the Index SUM . When gefitinib was combined with gemcitabine (n = 2), both samples showed an increase in their sensitivity. Of the tumours tested with treosulfan + gefitinib, 38% (13/34) were of ovarian origin. Of these, 62% (8/13) showed potentiation, with 1 sample showing a >50% decrease in Index SUM . 31% (4/13) showed increased resistance with the combination in comparison with treosulfan alone and 1 sample showed no change (Figure 3 ). Of the remaining samples tested with gefitinib + treosulfan, 57% (12/21) showed an increase in sensitivity, 38% (8/21) showed an increase in their resistance and one sample showed no change. Figures 4a and 4b show differential effects of gefitinib in combination with treosulfan in cells derived from a skin melanoma sample (Figure 4a ) and an ovarian carcinoma sample (Figure 4b ). When gefitinib was tested in combination with treosulfan + gemcitabine (n = 24), 54% (13/24) showed an increase in sensitivity, with 3 samples showing a >50% decrease in their Index SUM and 46% (11/24) showed an increase in resistance. In summary, the addition of gefitinib appeared to potentiate the effect of the cytotoxic agent or combination in 59% (45/76) of tumours tested; of these 11% (5/45) had a >50% decrease in their Index SUM . In 38% of tumours (29/76), the combination of gefitinib + cytotoxic caused the Index SUM to increase thereby increasing resistance. In the remaining 3% (2/76) there was no change observed. Immunostaining for EGFR was positive in 32% (10/31) of samples comprising of 1 breast carcinomas, 5 colon carcinomas, 1 ovarian carcinoma, 1 sarcoma, 1 skin melanoma and 1 carcinoma of unknown primary site. Immunostaining for pAkt was positive in 81% (25/31) of samples comprising of 3 breast carcinomas, 10 colon carcinomas, 2 oesophageal carcinomas, 2 ovarian carcinomas, 4 skin melanomas, 3 uveal melanomas and 1 carcinoma of unknown primary site. Of the positive samples, 8 were positive for both antibodies (comprising 1 breast carcinoma, 4 colon carcinomas, 1 ovarian carcinoma, 1 skin melanoma and 1 carcinoma of unknown primary site). In 74% (23/31) of samples that were stained for EGFR and pAkt, there was an IC50, IC90 and Index SUM value available for comparison. In all cases tested there was no relationship with gefitinib activity and EGFR or pAkt staining. Discussion This is the first study in which a TKI has been successfully tested in the ATP-TCA. ATP-TCA has potential to assist drug development for TKIs and possibly to direct therapy for individual patients. It represents one possible answer to the need for predictive oncology testing of these agents, and could be performed alongside clinical trials to obtain correlation data with outcome in patients treated with gefitinib. However, it is difficult to ascertain whether these were specific or non-specific effects of gefitinib and whether similar outcomes would be seen in the clinical setting. This would need to be determined before using this test for routine screening. Gefitinib showed activity in the assay and even though cytotoxic effects were not expected, in some cases the diminution in ATP levels suggests that these may occur. In general, flat concentration – activity curves were observed which are consistent with a cytostatic rather than a cytotoxic effect. Gefitinib alone showed activity in lung, ovarian and colon carcinomas. These results were consistent with previous findings in cell lines [ 24 ]. When gefitinib was tested in combination with a limited number of cytotoxic drugs, increases and decreases in the activity of the cytotoxic agent were observed. For example, gefitinib in combination with cisplatin caused 67% of samples to have a decrease in the activity of the cytotoxic. This compares with gefitinib in combination with a second platinum-containing agent, oxaliplatin, where 91% of samples showed an increase in the activity of the cytotoxic. However, it should be noted that oxaliplatin was virtually ineffective against the cells tested and this is therefore likely to reflect the effect of the gefitinib alone (Figure 5 ). Decreased activity of cytotoxic agents when these were combined with gefitinib was seen in 4 samples with cisplatin, 13 with treosulfan, 1 with oxaliplatin and 11 with treosulfan + gemcitabine. This could be detrimental to patients. It is similar to the effect of tamoxifen treatment on the success of breast cancer chemotherapy [ 25 ]. Although there was heterogeneity in the response of tumours to single agent gefitinib, there was no relationship between immunostaining for EGFR and gefitinib activity, consistent with other published studies [ 26 ]. Sirotnak et al . [ 9 ] showed that gefitinib caused growth inhibition in human tumour xenografts that was not dependent on high levels of EGFR expression. However, EGFR activation leads to activation of at least three separate second messenger cascades. While the ras/raf pathway may mediate the proliferative effects, survival signals are thought be mediated by the PI3/Akt pathway. As cells have to die in the ATP-TCA to register increased inhibition, sensitivity might be related to the degree of activation of the Akt pathway by other mechanisms. Sensitivity to gefitinib and other non-TKI EGFR inhibitors might therefore be related to pathway activation assessed by detection of pAkt, rather than the levels of EGFR expression. However, this study has not found any such relationship and, when EGFR staining was compared to pAkt staining there was no correlation between EGFR levels to pAkt activity. A similar observation was made by Campiglio et al ., [ 27 ] whose data suggested that neither MAPK nor pAkt were reliable markers of gefitinib activity. It should be noted that many receptors lead to Akt activation and that constitutive activation of the PI3/Akt pathway may be the result of PTEN inactivation. Of the 4 samples that had an Index SUM of <300 and the 6 samples that demonstrated >50% inhibition at 100% TDC when tested with single agent gefitinib, 2 samples (a uveal melanoma and an unknown primary) had material available for immunohistochemical staining with EGFR and pAkt. The uveal melanoma was negative for EGFR and positive for pAkt compared to the unknown primary, which was positive for both EGFR and pAkt. However, there were samples with similar IHC results that did not show sensitivity to gefitinib. As the EGFR (HER1) dimerizes with the other HER molecules and mediates greater activity as a heterodimer, it is likely that the expression of these molecules is also important to the activity of gefitinib [ 12 ]. Sensitivity to gefitinib is therefore likely to be the end result of a complex series of interactions within the cell. Conclusion In this study we have found that gefitinib in combination with different cytotoxic agents (cisplatin; gemcitabine; oxaliplatin; treosulfan and treosulfan + gemcitabine) is a double-edged sword: their effect on growth rate may make some tumours more resistant to concomitant cytotoxic chemotherapy, while their effect on cytokine-mediated cell survival (anti-apoptotic) mechanisms may potentiate sensitivity to the same drugs in tumours from other individuals. Competing interests IAC is director of Cantech Ltd. Authors' contributions LAK drafted the manuscript and carried out ATP-TCA assays. FDN, PW, SM, SS and SG also carried out ATP-TCA assays. PJ carried out the immunohistochemical work and IAC conceived the study and participated in its co-ordination. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535559.xml
514554	Estimating the beginning of the waterpipe epidemic in Syria	Background Waterpipe smoking is becoming a global public health problem, especially in the Eastern Mediterranean region (EMR). Methods We try in this study, which is a cross sectional survey among a representative sample of waterpipe smokers in cafes/restaurants in Aleppo-Syria, to assess the time period for the beginning of this new smoking hype. We recruited 268 waterpipe smokers (161 men, 107 women; mean age ± standard deviation (SD) 30.1 ± 10.2, response rate 95.3%). Participants were divided into 4 birth cohorts (≤ 1960, 1961–1970, 1971–1980, >1980) and year of initiation of waterpipe smoking and daily cigarette smoking were plotted according to these birth cohorts. Results Data indicate that unlike initiation of cigarette smoking, which shows a clear age-related pattern, the nineties was the starting point for most of waterpipe smoking implicating this time period for the beginning of the waterpipe epidemic in Syria. Conclusion The introduction of new flavored and aromatic waterpipe tobacco ( Maassel ), and the proliferation of satellite and electronic media during the nineties may have helped spread the new hype all over the Arab World.	Background Waterpipe smoking is becoming increasingly a worldwide phenomenon, with populations in the Eastern Mediterranean region (EMR) being especially affected [ 1 ]. This centuries-old tobacco use method comes under many different names (e.g., shisha, hookah, narghile, arghile), shapes, and sizes, depending on the region, with the term waterpipe implying a unifying feature of all these forms; the passage of smoke through water before inhalation by the smoker [ 2 ], Recent evidence shows that a quarter of some populations in the EMR currently smoke the waterpipe [ 3 ]. This trend is worrisome because of tobacco's known harmful effects to human health, and because prevailing norms the EMR may put certain slices of the society at increased risk of acquiring the habit, particularly women and children [ 6 , 7 ]. Although research on the health effects of waterpipe is still scarce, preliminary evidence links waterpipe use to respiratory, cardiovascular, and cancer diseases [ 8 - 11 ]. Developing effective intervention strategies to curb this emerging public health problem requires clear understanding of factors influencing the take up of this habit, as well its time course [ 1 ]. According to waterpipe smokers, the recent resurgence in waterpipe popularity is due to the introduction of Maassel (a specially prepared tobacco with sweetened fruit flavors and mild aromatic smoke), the media, and social trends [ 6 ]. Understanding the context in which these factors operate as well as being able to follow the secular course of the waterpipe epidemic requires estimation of the time frame for the beginning of the waterpipe hype. In this study we try to identify this time frame as well as provide evidence for the increase in waterpipe smoking. Methods The current analysis is drawn from a survey conducted in 2003 among a representative sample of waterpipe smokers visiting cafes/restaurants in Aleppo, Syria. The survey details can be found elsewhere [ 12 ], but briefly a cross sectional interviewer-administered survey was conducted in 17 randomly selected (out of total 112) café/restaurants in Aleppo, Syria. Overall, 268 waterpipe smokers were recruited (161 men and 107 women; mean age ± SD 30.1 ± 10.2, age range 18–68 years; response rate 95.3%). Participants were asked about their waterpipe use frequency, cigarette smoking status, current age, age of initiation of waterpipe smoking, and age of initiation of daily cigarette smoking. The protocols and informed consent documents for this study were approved by the SCTS' IRB and the University of Memphis' IRB. The questionnaire was anonymous and informed consent was obtained prior to all interviews. Analysis First, year of birth was calculated by subtracting current age of participant from the year of survey (2003), while year of initiation of waterpipe smoking and daily cigarette smoking were calculated by adding the age of initiation of smoking to the year of birth. Year of birth was then divided into four decade-long categories (people born in/before 1960; during 1961–1970; during 1971–1980; in/after 1981), and year of smoking initiation into three decade-long categories (initiation in/before 1990, initiation during 1991–2000, and initiation in/after 2001). The Chi-Square test was used to assess differences between the three smoking initiation time-groups for each birth cohort, with p level <0.05 considered significant. Results and conclusions Figure 1a,1b compares between year of initiation of waterpipe and cigarettes among study participants, respectively, according to their birth cohort. It shows that while cigarette initiation displays an age-related pattern with peak initiation of participants occurring during in their twenties, most of waterpipe initiation and for all birth cohorts is commencing in the 1990s. Figure 2 , depicts number of waterpipe smokers according to their year of initiation, and suggests indirectly the rapid increasing trend of this smoking method. Taking into consideration possible limitations of this study, particularly the use of cross-sectional design to examine time trends and the relative youngness of the study sample, these results suggest that the current hype of waterpipe smoking is a recent one, commencing mostly in the nineties of the 20 th century, and showing an increasing trend. Based on these results and our previous data regarding factors related to the spread of waterpipe [ 6 ], we present the following scenario for the current surge of popularity of waterpipe smoking. In the nineties, Maassel was introduced [ 13 ] simplifying waterpipe preparation and attracting more people to its mild aromatic smoke. Out of curiosity about this new tobacco or of modeling of others, people started experimenting with the waterpipe. The increasing number of waterpipe users together with the spread of satellite channels and electronic media, occurring during the same time period, may have contributed further to the creation and spread of this new smoking trend. Since waterpipe smoke contains considerable amounts of the addictive substance nicotine [ 14 , 15 ], nicotine dependence can sustain the habit among experimenters creating a stable base of waterpipe smokers within the society and contributing further to its spread. It remains to be seen, the possible role of resurgence of local identities in contrast to western culture in the adoption of this "oriental" method of smoking. The dramatic increase of this addictive smoking method within a short period of time and its potential health risks mandate that active surveillance and in depth research into its risk profile should become a priority for health systems in the EMR. Policy makers should also be alerted to this eminent public health threat, which is so far escaping current regulations and restrictions imposed on cigarettes, such as the ban on advertisement and the inclusion of health warnings on waterpipe tobacco products. Figure 1 a: Shows the proportion of current waterpipe smokers of different birth cohorts according to their year of initiation categorized into three decade-long categories. b. The same parameters are shown for cigarette smoking. Figure 2 Shows number of study participants according to their year of initiation of waterpipe smoking. Competing interests None declared. Authors' contributions SR designed the study, conducted the analysis and wrote the first draft of the manuscript. KW and TE contributed to the design of survey and revision of the manuscript. WM contributed to the design of survey and wrote the final draft of the manuscript. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514554.xml
524497	Health, human rights and mobilization of resources for health	Background There has been an increased interest in the role of a human rights framework to mobilize resources for health. Discussion This paper argues that the human rights framework does provide us with an appropriate understanding of what values should guide a nation's health policy, and a potentially powerful means of moving the health agenda forward. It also, however, argues that appeals to human rights may not necessarily be effective at mobilizing resources for specific health problems one might want to do something about. Specifically, it is not possible to argue that a particular allocation of scarce health care resources should be changed to a different allocation, benefiting other groups. Lack of access to health care services by some people only shows that something has to be done, but not what should be done. Summary The somewhat weak claim identified above together with the obligation to realize progressively a right to health can be used to mobilize resources for health.	Background During the past few years there has been an increasing interest in attempts to utilize a human rights framework to argue that we have obligations, in one way or another, to do something about the human suffering among the large number of poor in the world. The argument is that the suffering of the poor is a violation of their human rights, and the international human rights instruments place an obligation on us to do something about it. The exact details of the argument vary among different commentators, but they all have in common this basic argument structure. Rarely, however, are we provided with any details about exactly how one should understand particular violations of human rights or exactly how one arrives at recommending a particular action to rectify the alleged violation of a human right. Nevertheless, many are confident that a linkage to human rights will prove useful when we want to mobilize resources for the world's poor. In the words of Paul Farmer Of course, it is easy to demand more resources; what is hard is to produce them. But if social and economic rights are acknowledged as such, then foundations, governments, businesses, and international financial institutions-many of them awash in resources-may be called on to prioritize human rights endeavors that reflect the paradigm shift advocated here [ 1 ], p. 244. In this paper I want to examine this claim as it relates to health: How can one use a human rights framework to mobilize resources for health? If we want to use this framework presumably we can make demands on the basis of two types of reasons. Those who lack access to resources can argue that the total amount available to promote health should be increased, or they can argue that the allocation of available resources should be made in a different way, giving them access to health care services, but denying other, less deserving groups access. One may, of course, also wish to make both demands. In this paper I shall argue that the human rights framework does provide us with an appropriate understanding of what values should guide a nation's health policy, and a potentially powerful means of moving the health agenda forward. I shall, however, also argue that appeals to human rights may not necessarily be effective at mobilizing resources for specific health problems one might want to do something about. Discussion The international human rights framework One may understand the claim that "health is a human right" in at least three different ways. First, we may simply want to say that health is important, that we should all do what we can to promote health, and we may even expect that referring to health as a human right might produce an emotional response in our audience, motivating them to action. If this is what we want to do, and if claiming that "health is a human right' does in fact produce this type of response, then utilizing a rights framework can indeed be expected to mobilize resources. While it is undoubtedly true that this strategy will sometimes be effective, its effectiveness will more often than not depend on the immediate reaction to the deprivation of a particular group, and that group's ability to elicit sympathy for their cause, rather than a legitimate policy response where all competing claims have been taken into account. Second, we may want to make a moral claim of a particular type. There is an extensive philosophical discussion of how one should understand the concept of "a right" in general, and a "right to health" in particular. While not denying the importance of this discussion, it does not have much relevance to the problem addressed in this paper: does one have a reason to believe that the claim that "health is a human right" will mobilize resources for health? Even if we take the strongest claims to be true, for example that there are moral obligations on each one of us to do something quite specific to improve other people's health, we still have to provide an account of how to translate these moral obligations into effective action. That is why I want to limit my discussion to an understanding of "health as a human right" as those specific legal obligations on states that arise out of international law. International law in this context refers primarily to The Covenant on Social, Economic and Cultural Rights, but also the General Comment by the UN Committee on Economic, Social and Cultural Right on article 12 (on the right to health) of the International Covenant on Economic, Social and Cultural Rights, and the first report of the recently appointed Special Rapporteur with a mandate to focus on the right to health [ 2 , 3 ]. In addition, there are a number of cases that have been decided on the basis of a right to health. These cases are particularly important when we attempt to understand what is meant by a "right to health" in international law. If we take a right to health in this third sense, we have at least in principle identified a way of understanding rights that can lead to an effective mobilization of resources. The issue to be addressed is therefore: Can one on the basis of international law argue that a person's right to health is violated if that person is denied access to health care services or the underlying determinants of health on economic grounds; that is, either because that person does not have sufficient resources herself to pay for the services or the state claims that it does not have sufficient resources to pay for the necessary services. If one could establish that such cases are violations of international law, then one would have an effective way of mobilizing resources for health based on a human rights framework. If one adopts this approach, one challenge is, in the words of the Special Rapporteur, that "although there is a growing national and international jurisprudence on the right to health, the legal content of the right is not yet well established (#39)". Some indication of how one should understand the right to health in international law is nevertheless provided in the key, authoritative documents and in the court cases that have been decided on the basis of right to health challenges. Some of the statements made in the General Comment might indeed lead one to believe that a state has strong legal obligations to provide sufficient resources to ensure adequate health for all. It is said, for example, that "health facilities, goods and services must be affordable for all. Payment for health care services, as well as services related to the underlying determinants of health, have to be based on the principle of equity, ensuring that these services, whether privately or publicly provided, are affordable for all, including socially disadvantaged groups (12 1 iii). One could take this to mean that it prohibits denial of health care services on economic grounds. Other statements seem to support this claim " [F]unctioning public health and health care facilities, goods and services, as well as programmes, have to available in sufficient quantity within the State party (12(1)). Central to the General Comment is the principle of non-discrimination: "...the Covenant [on Economic, Social and Cultural Rights] proscribes any discrimination in access to health care and underlying determinants of health ... on the grounds of race, colour, .... health status ... and civil, political, social or other status" (18). This is reinforced by the special rapporteur: "Accordingly, international human rights law proscribes any discrimination in access to health care, and the underlying determinants of health, on the internationally prohibited grounds, including health status, which has the intention or effect of impairing the equal enjoyment of the right to health" (26). Taken together, these statements seem to give a strong endorsement to the claim that it is a violation of a person's health rights to deny him treatment on the grounds that treating that person is too expensive for the state. If one denied treatment to persons who happen to have diseases that are expensive to treat, one does indeed discriminate on the basis of "health status", one of the prohibited grounds, and one does not ensure equal access to health care services to all. There is, however, one important qualification to this claim. Although the state has an obligation to provide health care in "sufficient quantity", "the precise nature of the facilities, goods and services will vary according to numerous factors, including the State party's developmental level" (12 1). Furthermore, "the Covenant provides for progressive realization and acknowledges the constraints due to the limits of available resources" (30). It is, therefore, in spite of the strong statements that everyone should be assured access to health care, in principle legitimate for a state to claim that it can deny access to health care to patient groups who happen to have diseases that are expensive to treat. A state's claim that it does not have sufficient resources to provide access to health care or its determinants to a particular group because of its costs can, of course, be challenged. We would then need some principled way of adjudicating between the competing claims: on the one hand those of the group denied health care access claiming that its health rights are violated and on the other hand those of state claiming that its resources can be utilized better elsewhere. The official documents on how we should understand a right to health do not provide us with much guidance on how one should adjudicate between such competing claims. As we shall see in the next section, neither apparently do the cases which have been brought forward as violations of a right to health. The lack of a principled way of adjudicating between competing claims The two recent South African cases provide a particular striking example of the challenge of using a reference of a right to health or the courts to mobilize resources for health. The South African constitution, article 27 (1) gives everyone a right to have access to health care services, including reproductive health care and in article 27 (2) it says that the state must take reasonable legislative and other measures, within its available resources, to achieve the progressive realization of these rights. These national provisions, of course, reflect the ones in the UN Covenant. In November 1997, the Constitutional Court of South Africa decided on a case involving the scope of such a "right to health" [ 4 ]. The case involved a diabetic man with ischaemic heart disease and cerebro-vascular disease with chronic kidney failure. He was rejected for the dialysis program on the grounds that there was a shortage of dialysis machines and dialysis had to be reserved for people with acute renal failure or for patients who are candidates for kidney transplantation. He appealed this decision to the Constitutional Court on the grounds that the South African Constitution gives every person a right to life and that nobody may be denied emergency medical treatment. The Court rejected the appeal on the grounds that this was not an emergency treatment, and that a right to life should be interpreted as a right to non-interference, but not necessarily a duty to sustain life. In the words of the Court "It [article 27(3)] provides reassurance to all members of society that accident and emergency departments will be available to deal with unforeseeable catastrophes which could befall any person, anywhere and at any time" (section 51) and The applicant suffers from chronic renal failure. To be kept alive by dialysis he would require such treatment two to three times a week. This is not an emergency which calls for immediate remedial treatment. It is an ongoing state of affairs resulting from a deterioration of the applicant's renal function which is incurable" (section 21) Although the appellant did not appeal the decision on the basis of the South African Constitution's article 27 (1) giving everyone a right to have access to health services, the Court did discuss this matter. The Court pointed out that it is not disputed that the Department of Health "does not have sufficient funds to cover the cost of the services which are being provided to the public ... This is a nation-wide problem and resources are stretched in all renal clinics throughout the land. Guidelines have therefore been established to assist the persons working in these clinics to make the agonising choices which have to be made in deciding who should receive treatment and who not" (section 24). The Court further maintained that the current guidelines are justified by the fact that more patients would benefit from the limited resources available than by any alternative use of resources, and that "it has not been suggested that these guidelines are unreasonable or that they were not applied fairly and rationally when the decision was taken by the Addington Hospital that the appellant did not qualify for dialysis" (section 25). The second case involves use of nevirapine to prevent perinatal HIV transmission[ 5 , 6 ]. As is well know, the South African government has until recently refused or been reluctant to provide antiretroviral treatment for HIV, both for HIV positive people and to pregnant women to prevent perinatal HIV transmission. The South African Treatment Action Campaign brought a suit against the government that the policy regarding the use of nevirapine violated the constitutional right to health. The Supreme Court ruled in 2002 with the Treatment Action Campaign, affirming that the government's policy violated the Constitution's right to health. It is important, however, to note the basis for this ruling. The South African government had claimed that providing nevirapine to pregnant women would be too costly in terms of infrastructure, in particular provision of testing and counseling, and that its safety and efficacy had not been sufficiently demonstrated in a South African context. The Supreme Court disagreed with both of these claims. The issue was "whether it was reasonable to exclude the use of nevirapine for the treatment of mother-to-child transmission at those public hospitals and clinics where testing and counseling are available". Regarding the safety and efficacy issue, the court cited scientific opinions which made these claims completely unreasonable. The Court did not address the issue of an appropriate allocation of resources. Accepting the Court's decision would not require the South African government to allocate additional resources to health care delivery nor would it require it to re-allocate existing resources for health: A potentially lifesaving drug was on offer and where testing and counseling facilities were available it could have been administered within the available resources of the state without any known harm to mother and child. In fact, the Court, as in the dialysis case, rejected any role for the courts in resource allocation decisions, instead adopting a criterion of evaluating government policies on the basis of a criterion of "reasonableness", from a previous right to housing case. In this previous case, the court would not decide "whether other more desirable or favourable measures could have been adopted, or whether public money could have been better spent"[ 7 ] (41). A policy is reasonable if it is comprehensive and well coordinated; is balanced, and does not exclude a significant segment of society; and responds to the urgent needs of those in desperate circumstances. In the nevirapine case the Court affirmed that any right did not impose an obligation on the state "to go beyond available resources or to realise these rights immediately" (para 32). The court did concede that it would be reasonable for the government to carry out a research project to determine the safety of the drug before a wider implementation, but that it would not be reasonable to deny people access to the drug outside of research and training sites. If one therefore follows the views of the South African Court regarding a right to health, it would seem to be difficult to use the courts to challenge a particular allocation of health care resources. The Court felt that the provincial administrations should make decisions as to how funds for health care should be spent and that courts should be "slow to interfere with rational decisions taken in good faith by the political organs and medical authorities whose responsibility it is to deal with these matters" (section 29). And "courts are not the proper place to resolve the agonising personal and medical problems that underlie these choices. Important though our review functions are, there are areas where institutional incapacity and appropriate constitutional modesty require us to be especially cautious. Our country's legal system simply cannot replace the more intimate struggle that must be borne by the patient, those caring for the patient, and those who care about the patient. The provisions of the bill of rights should furthermore not be interpreted in a way which results in courts feeling themselves unduly pressurized by the fear of gambling with the lives of claimants into ordering hospitals to furnish the most expensive and improbable procedures, thereby diverting scarce medical resources and prejudicing the claims of others" (section 58). One might, of course, want to criticize this view of the role of the courts, as has been done by Darrel Moellendorf.[ 8 ] Moellendorf argues that the courts should rule on what is meant by "within available resources" and that the Constitutional Court previously has recognized that rulings on socio-economic rights do have budgetary implications: "the Court's role in upholding socio-economic rights is not foreseen as limited to the framework of existing national or provincial budgetary allocations. Rather the court may pass judgments on these rights, as with other rights, that require a change in fiscal priorities" (p. 331). In this paper, however, I shall suggest a different approach that may be more promising if one wants to use an appeal to a right to health to mobilize resources for health. The problem of resource allocation The central issue with regard to differential access to health care services is, on what basis one can claim that a state does not spend a sufficient amount of resources on health care, relative to its general wealth, and on what basis can one claim that a state does not use the resources it devotes to health care appropriately. If a state does not allocate sufficient resources to health care, or uses its available resources inappropriately, a citizen could claim that the state violates her right to health when a particular health care intervention is denied her, and when that intervention would be available to her if the state increased its allocation to health care to an acceptable level, or re-allocated resources within the health care sector to an acceptable mix of interventions. Focusing attention on this set of questions is important because they are at the heart of debates about health care access, in both resource rich and resource poor settings. We want to know whether denying a person access to a life-saving intervention because it is too expensive for a given health care system is a denial of his right to health. This can be the typical high cost therapies in rich countries, such as bone marrow transplantation, or it can be antiretroviral treatment for HIV in resource poorer countries. If we assume that a state is in principle entitled to make judgments about when a particular treatment is too expensive to provide within its healthcare systems, given available resources, we have to find a way to examine claims about appropriateness, and provide criteria for how these judgments are to be made. At least in the two South African cases the Court has been reluctant to get involved in this type of questions. It should come as no surprise that this is the case. As the ever growing literature on the ethics of health care resource allocation demonstrates, there is no clear consensus about how one should balance the various legitimate concerns and values involved in making these kinds of decisions. Judgments will have to be made, and, unless there are grave violations of due process, or a gross misapplication of the principles of resource allocation, it seems unlikely that one could make a principled criticism of a particular decision that could be used as a basis for a claim against the state. This is probably why the courts have been reluctant to criticize allocation judgments made in good faith by government bodies. In spite of this largely negative conclusion, I shall now nevertheless a strategy that shows that appeals to health as a human right might nevertheless be used to argue that denial of care on economic grounds is a violation of that right. Using human rights to mobilize resources for health The central guiding principle in the international health and human rights documents is that of non-discrimination. It is prohibited to deny a person access to health care or its determinants on the basis of characteristics such as race or religion, but also, as we saw above, on the basis of social and health status: health care services should be accessible to everyone , there should be equal access to health care, and health care should be affordable for all . It is fairly easy, and uncontroversial, to establish that a policy such as the one accepted during apartheid in South Africa, of denying access to health care services on the basis of race, is a violation of a right to health. It is more difficult to agree on how we should understand the principle of non-discrimination on the basis of social and health status. One might be tempted to conclude that the principle of non-discrimination on the basis of social status implies a right for everyone to access the same bundle of health care services. An unqualified acceptance of that claim, however, would obviously be unacceptable: there are going to be health care services which provide very little benefit, but which are hugely expensive, which may be accessible to the very rich, but which should not by any account be included in the bundle of health care services accessible to all, even by the most egalitarian standards of justice. We therefore need some account that will specify which health care services should be accessible to all, even within a rights approach to health care. In order to give a satisfactory account of that problem, we again need an appropriate value framework for prioritization. It seems to me that there is nevertheless something we can say, even before we have a fully worked out framework for prioritization, if we take a right to health approach. If a range of health care services, known to be effective in significantly preventing premature death or significantly increasing quality of life, is available to a substantial portion of a country's population, but is not available to a particular group of people, such as the very poor, this is a violation of the right to health of the members of that group. The qualifiers indicate that a lot of work needs to be done to give precise content to this claim, but for the purposes of this discussion, this somewhat vague claim will be sufficient. It is easy to identify examples: If antiretroviral treatment for HIV is available to most people in country, but is not available to people in the lowest 1% income bracket, we can conclude that there is a prima facie violation of their health rights. Although one might agree that such a principle identified here does indeed follow from the international human rights documents, there are some obvious objections if one's aim is to show that an international human rights framework can be effective at mobilizing resources for health. First, if one accepts this principle of non-discrimination based on social status, why should one not accept a principle of non-discrimination based on health status, and argue analogously that if a small group of persons is denied access to health care services because of the illness that they happen to have, at the same time as the majority of the population has access to health care for their illnesses, this is an illegitimate form of discrimination? Second, can not a state use the same principled objection to mobilization of resources in cases of discrimination based on social status, that its limited resources are better used elsewhere? Third, are any real policy alternatives ruled out by this principle, or is it not so general that all types of health policies is compatible with it. The first criticism says that if we find it objectionable that the poor are treated differently from the rich with regard to health care access, why should we not find it objectionable that those who happen to have late stage breast cancer are treated differently from those who happen to have disease that can be treated more cheaply? Why could not a person with late stage breast cancer argue successfully that a decision not to fund the known effective treatment she suffers from is of the same type as a decision not to fund treatments the poor suffer from? One response to this objection would be that decisions regarding what constitutes a non-discriminatory policy are different in these two cases. When deciding whether who to give access to health care services in general, we do not discriminate on the basis of what illnesses people have: we look at other characteristics, such as gender, age, ethnic background, social status, and income level. When deciding what bundle of health care services to provide within a nation's health care system, other considerations are relevant. We would then look at the effect of the various interventions, their costs, how each individual benefits from the intervention, and what the cumulative benefit for the population is. A discriminatory practice in the first type of decision would be to single out a particular income group for preferential treatment, while a discriminatory practice of the second type of decision would be to single out a particular illness for preferential treatment. It would not be discriminatory to distribute resources on the basis of, for example, health effects of the various interventions. The reason why we have more problems in identifying inappropriate discrimination on the basis of health status is that there is less agreement on how we should identify interventions that have the same effect on health status across illness groups. The second objection is more difficult to answer. Let us assume that there is a country with a government that tries to allocate resources justly. Although the government tries its best, there continues to be lack of access to health care services to the poor. The government has also done its best in raising the resources allocated to health care, but they have still not been able to fund life-saving interventions for a particular group of patients. Because of its generally sound economic policies, the country has experienced more than the expected economic growth over the past few years, and there is a budget surplus. The government now has to decide what to do with the additional money: Should they remove continued barriers of access to the poor, or should they ensure funded for the neglected disease? We assume that the government cannot do both. It seems to me that the principle of non-discrimination within a human rights approach to health care does not provide us with guidance about how we should go about solving this type of problem. In spite of this problem, there is one answer we can give, albeit a limited one, but still with a potential for identifying some inadmissible policies this imaginary country might want to consider. Within the human rights framework there is a recognition that any given country cannot fulfill its obligations immediately, but there is an obligation of progressive realization of a right to health. This places an obligation on the country to do something to increase access to health care services during periods of economic growth. On that basis we could justifiably criticize government inaction , but we would not be justified in proscribe a particular action. The third objection is that the principle of non-discrimination is so general that it would not rule out any particular health policy. This is not the case. One currently fashionable proposal is that governments should identify an essential package of health services based on a criterion of cost-effectiveness. This package should be available to everyone, regardless of ability to pay. Those who are wealthy would be able to, or indeed encouraged to, pay for additional health care services using their own funds. WHO has called this proposal "the new universalism". Classic universalism would obligate governments to provide everything medically useful to everybody. How should we evaluate this proposal in terms of a human rights approach to health care delivery? Let me first examine the arguments in favor of this proposal. First, it is quite clear that the policy is pro-poor, in the sense that it would increase access to health care services for the poor compared with what is currently available to them. Data from many country shows that government subsidies today go disproportionally to the wealthy to pay for interventions of low cost-effectiveness, while the poor suffer from conditions which require interventions of high cost-effectiveness. Requiring that governments fund a package of services of high cost-effectiveness available to everyone would therefore shift government resources from the rich to the poor. Second, according to the principle of non-discrimination within the human rights tradition, governments should not discriminate against a particular group of people in terms of access to health care services. This policy seems to satisfy that requirement. The basic package is accessible to everyone, and the government does not subsidize health interventions for non-poor if they are not also subsidized for the poor, unlike what is the case in many countries today. For these arguments, see [ 9 , 10 ]. The following sentence sums up the rational for this position: Although many countries cite equity as the reason for strong government controls, public sector-controlled policies do not have a good track record on equity. In Indonesia, for example, the rich receive almost three times as much public health care as the poor. In China in the early 1980s, rural households-almost 80 percent of the population-received just 29 percent of public health spending. In Tanzania the richest fifth of the population use more than twice as many government hospital beds and more than four times as many outpatient services as the poorest fifth. In Côte d'Ivoire less than one-quarter of the rural poor who were sick received any form of medical care, as compared with half of the urban rich. In Peru only 20 percent of the poor received care, versus 57 percent of the rich. In general, when government expenditures are concentrated on urban areas and on hospitals rather than on basic services, the results are highly inequitable, governments are essentially subsidizing the rich. [ 11 ] This conclusion only follows, however, if the principle of government non-discrimination only applies to the proportion of public resources which goes to different population groups. If the non-discrimination principle is limited in this way, and if the public system is available to everyone regardless of ability to pay, a two-tiered system where the wealthy have access to a much broader range of health care services because they are able to pay for the additional services from their own income is not in conflict with a right to health. If, on the other hand, one takes the position that a right to health should be understood in terms of equality of access to health services in general, then an explicit acceptance of at least some types of two-tiered systems is not in accordance with a right to health. There are a number of considerations which speak in favor of this expanded notion of government responsibility. A government does not only have an obligation to provide or to finance health care services. In fact, in recent discussions about health sector reform, these types of government obligations have been de-emphasized and instead there has been a focus on a government responsibility to regulate, facilitate and ensure that people have access to health care services, whether they are provided or financed by government, non-profit organizations or for profit companies. If one takes this role seriously, focusing only on how government resources are spent is at best only one component of government obligations to fulfill the right to health of their population. Persistent inequalities of access to health care, whether they are caused by imbalances in public resources, or caused by differences in income between population groups, should therefore be regarded as an essential concern of equity and rights oriented policies. Summary I have in this paper argued that appeals to health as a human right are not going to be helpful if we want advice on how we should allocated resources among different patient groups. Many of the most difficult problems of resource allocation are therefore not addressed by a human rights approach. However, I have also argued that some inequalities of access can be criticized: One can say that it is a violation of human rights if one group systematically has less access to health care services compared with other groups. There is, however, not much more on can say; in particular, one cannot proscribe a particular corrective to that violation of health as a human right. All one can say is that the state has an obligation to do something about the injustice. This result may not be very satisfactory for those who are concerned about doing something about lack of access to health care. However, if one takes the requirement of progressive realization seriously, together with the periodic reporting requirement, one might conceivably have a powerful basis for criticism of government policies. Governments are obligated to report to the UN, and, as I have shown, there is a basis on which one can criticize them for not fulfilling their health rights, also with regard to their economic policies. This creates an obligation on countries to show that something has been done when the next periodic report is done to and they would have to take steps to improve access in some respects. While this may seem to be a modest requirement, it can, together with appropriate political pressure, lead to significant, positive changes in health care systems. But it is undeniable that it is much less than what many who advocate a human rights approach to health care access hope can be achieved by utilizing this framework. It is also true that it is generally recognized that the reporting requirements, at least with regard to health, are not very effective. Strengthening the reporting and monitoring mechanisms of the UN system with regard to health should therefore be the highest priority for those who want to emphasize a human rights approach to health sector reform. Competing interests and acknowledgments The opinions expressed are the author's own. They do not reflect any position or policy of the National Institutes of Health, Public Health Service, or Department of Health and Human Services. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524497.xml
526280	General practitioners believe that hypnotherapy could be a useful treatment for irritable bowel syndrome in primary care	Background Irritable bowel syndrome is a common condition in general practice. It occurs in 10 to 20% of the population, but less than half seek medical assistance with the complaint. Methods A questionnaire was sent to the 406 GPs listed on the West Sussex Health Authority Medical List to investigate their views of this condition and whether they felt hypnotherapy had a place in its management Results 38% of general practitioners responded. The achieved sample shared the characteristics of target sample. Nearly half thought that irritable bowel syndrome (IBS) was a "nervous complaint" and used a combination of "the placebo effect of personal care," therapeutic, and dietary advice. There is considerable divergence in the perceived effectiveness of current approaches. Over 70% thought that hypnotherapy may have a role in the management of patients with IBS; though the majority (68%) felt that this should not be offered by general practitioners. 84% felt that this should be offered by qualified hypnotherapist, with 40% feeling that this should be offered outside the health service. Conclusions General practitioners vary in their perceptions of what constitutes effective therapy in IBS. They are willing to consider referral to a qualified hypnotherapist.	Background This study explores general practitioners' beliefs about irritable bowel syndrome (IBS), and whether they see hypnotherapy as an appropriate complementary therapy for its management; and if so, who should deliver it. IBS is estimated to occur in 10–20% of the population in most countries [ 1 ]. It is known that less than half of subjects seek medical help for their complaints [ 2 ] but it is a common cause for referral to secondary care (referral is advised in patients over 50 years old with changing symptoms) [ 1 ]. Over a period of time of between 2 and 5 years, it is thought that there is a 30% turnover of patients having IBS [ 3 ]. It therefore uses significant primary care resources over many years. IBS is typical of many conditions seen in general practice. There is a risk that some of its symptoms may actually represent serious underlying illness and there are a number of conventional medical interventions, but for many patients the interest of their practitioner is paramount [ 4 ]. The mainstays of medical management are a high fibre diet and pharmacotherapy [ 5 ]. These help to some extent but do not appear to offer cure or permanent remission. There is also little evidence of the effectiveness of dietary advice. Many general practitioners see IBS as a complex bio-psychosocial problem where an appropriate consultation style can be as important as the therapeutic interventions themselves [ 6 ]. UK general practitioners are having a greater say in the management of the local health services through the organisation of local health services into Primary Care Trusts (PCT). The latter are responsible for commissioning the health services for their resident population. Via these, general practitioners have the potential to influence the development of services for the treatment of this condition. Most PCTs are financially constrained and cannot meet all the needs of every patient. Services have to be prioritised according to need. The study investigated general practitioners' opinions as to whether hypnotherapy should be provided by Primary Care Trusts, or provided privately. The possibility of hypnotherapy being used as a primary care level intervention for IBS has developed from the realisation in the 1980's that hypnotherapy was an effective treatment for intractable IBS in a hospital context [ 7 , 8 ]. These trials are good evidence that this therapy is effective in secondary care. The concept of shortening a long-term illness that often requires secondary referral, long-term drug therapy and repeated primary care attendance is an attractive one. No large-scale primary care trial of this sort of treatment has been performed. If a large RCT were to be undertaken it would be important to assess whether the findings would be perceived as relevant and likely to be implemented in a primary care setting. There is a lack of literature about general practitioners' opinions and few validated questionnaires. A literature search identified two recent studies about IBS in primary care [ 9 , 10 ] which comment on the lack of confidence in the diagnosis of IBS in primary care and the over-dependence on treatment using drugs. An interview study explores medical and lay views of IBS [ 11 ]; concluding that patients are affected by medical beliefs about the nature of IBS and suggest that better explanations could be given for the disorder. A questionnaire study of consultants' and GPs' attitudes to functional bowel disorders showed marked differences in the perception of the psychological as opposed to the physical basis for the condition between the two groups – GPs favoured a psychological explanation. However, the study did not investigate alternative therapeutic intervention [ 12 ]. There are few studies describing the experiences and attitudes of GPs towards IBS management in primary care, especially pertaining to the use of alternative or psychological treatments. Although it is very likely that treatments such as hypnotherapy and cognitive behaviour therapy would have a prolonged impact on IBS if they were commonly employed at an early stage in the illness, little is known about the attitudes of primary care teams in encouraging this sort of approach. We therefore conducted this study to discover how general practitioners perceive IBS and if they consider it to be managed effectively by current conventional therapy. We wanted to know if GPs consider hypnotherapy to be an appropriate intervention and whether GPs would refer there patients for it. Finally, we wanted to know whether such a service would be provided within general practice or another setting and if it should be funded by the NHS or be provided from the private sector. Methods Study setting A survey was conducted of all general practitioners in West Sussex. The setting for the study is a mixed area. There are market towns and no large cities. Much of the population lives in sub-urban and rural areas. The area has a lower than average deprivation level than the regional norm, but in common with most of the south east of England, income is higher than the national average. As it is a rural area a larger number of practices dispense their own medicines, rather than send patients with a prescription to a local pharmacy [ 13 ]. Questionnaire design and testing A questionnaire was designed and pre-tested with a pilot group for readability, validity, reliability, acceptability of layout and time taken to complete [ 14 ]. The questionnaire is attached as an appendix [see Additional file 1 ]. The sample was the 406 general practitioners listed as unrestricted principals on the Health Authority's list in 1997. This list included both full and part time general practitioners. The questionnaires were mailed to GPs using the Heath Authority's internal mail system and were returned via the same Health Authority post. At that time, the use of this service was free to the investigator. Once the first group of questionnaires had been returned, the questionnaire was repeated once to those practices that had not responded. The responses were mostly in the form of a Likert scale. This scale enables measurement of degrees of opinion, so increasing the sensitivity of the analysis. A central category was provided for a neutral response. The design purposefully did not force a polarised choice because it was thought that the treatment would be unlikely to be in common use. It was possible that many general practitioners would genuinely not have an opinion about some of the questions. To prevent acquiescence bias, the questions were worded so that expected responses varied unpredictably according to the direction of the scale [ 14 ]. Demographic data were requested, to enable assessment of possibly biased responses due to factors such as age, sex or practice size. Questionnaire themes included GP perception of IBS, its management, and hypnotherapy as a treatment. It also included questions related to the funding of treatment and the acceptability of hypnotherapy as part of the management of IBS in primary care. The questions were grouped together using these themes. Piloting enabled assessment of face validity and content validity [ 15 ]. The constructs being tested were not formally defined, as the purpose of the instrument was to obtain opinion about the subjects of the questions, not to form hypotheses. Data analysis We analysed the data using SPSS (Statistical Package for Social Sciences). The target and achieved samples were compared using the Chi-square test of proportion [ 16 ]. Where there is a statistically significant difference between the samples it is included. The results of the main questionnaire are reported as proportions rather than as a score. The purpose of this approach is to ascertain determine general practitioner opinion. Preliminary analysis found that the proportions for the 'strongly agree' and 'strongly disagree' categories are generally small. For clarity of interpretation and reporting, the 5 point-scale was collapsed into 3 where applicable, amalgamating the 'strongly agree' category with the 'agree' category, and 'strongly disagree' with 'disagree'. Results Response bias 155 (38%) general practitioners of the target sample returned questionnaires after the initial request and one further follow-up reminder. The characteristics of general practitioners in achieved sample were compared with that of the target sample to assess sample bias. No statistically significant differences were found between the achieved and target samples other than fewer GPs from non-training practices and from single-handed practices responded to the survey. With the exception of these factors, the characteristics of the achieved sample are not very different from that of the population of GPs in West Sussex. A comparison of the achieved and target sample are shown in Table 1 . Nine general practitioners (6%) had previously used hypnotherapy. Table 1 Comparison of responders (achieved sample) with all West Sussex GPs (target sample) Characteristic Achieved sample Target sample Test % n % n Aged 35 & under 17.4% 27 16.5% 67 Aged 36–45 43.9% 68 41.6% 169 Aged 46–55 31.0% 48 28.3% 115 Aged 56 & over 6.5% 10 12.3% 50 p = 0.17 X 2 Male 68.4% 106 73.4% 298 Female 31.0% 48 26.6% 108 p = 0.20 X 2 Full time 83.9% 130 88.7% 360 Part time 15.5% 24 11.3% 46 p = 0.09 X 2 Research active % (95%CI) 12.9% (7.1–18.6) 20 5.7% (3.4–7.9) 23 n.s. Test of Proportion Training 53.50% 83 16.05% 65 Non training 43.90% 68 83.95% 341 p < 0.001 X 2 Single handed 5.2% 8 10.2% 41 Group 92.3% 143 89.8% 365 p < 0.05 X 2 Total 155 406 (Source of GP data in West Sussex: NHS Executive Oct 1997) How do general practitioners perceive IBS 45.2% of the respondents agreed with the statement that "IBS is mainly a nervous complaint", and 40% felt that IBS "responds mainly to the placebo effect of personal care and attention". Although many respondents seem to categorise IBS as a 'nervous complaint', conventional medical management with drugs and dietary advice is seen to have a role to play with 45.2% agreeing with the statement that drug therapy works effectively, and 38.7% dietary advice works effectively. A striking finding is that a sizeable minority of GPs were unsure if IBS patients respond to placebo effect (at 37.4%) or to medical therapy (at 41.3%). Combining the 'unsure' and the 'disagree' categories show that the majority of GPs in this survey are uncertain if existing treatment regimes for IBS (drug, dietary advice and placebo effect) are efficacious. These results are set out in Table 2 . Table 2 How general practitioners perceive irritable bowel syndrome. Agree Unsure Disagree n= Irritable Bowel Syndrome is mainly a 'nervous complaint' 45.2 29.7 23.2 155 Irritable Bowel Syndrome responds mainly to the placebo effect of personal care and attention 40.0 37.4 22.6 155 Drug therapy works effectively in my Irritable Bowel Syndrome patients 45.2 32.9 20.6 155 Dietary advice works effectively in my Irritable Bowel Syndrome patients 38.7 38.7 20.6 155 Irritable Bowel Syndrome responds mainly to medical/therapeutic interventions 39.4 41.3 19.4 155 N.B. the rows do not always sum to 100% as missing responses are included in the analysis but not shown. Is care of patients with IBS adequate Just under half of the respondents (45.2%) agreed that IBS requires more attention in primary care and less than a quarter (22.6%) disagreed with the statement that IBS required more attention. The majority (56.8%) of the respondents felt that it would be possible to manage IBS better in their practices. Only 12.9% disagreed with this statement. The vast majority (84.5%) felt that the present management of IBS is variable with less than 10% of practitioners believing that their care is effective. These results are summarised in Table 3 . Table 3 How general practitioners perceive the effectiveness of irritable bowel syndrome management Agree Unsure Disagree n= Irritable Bowel Syndrome requires more attention in Primary care 45.2 29.7 22.6 155 It would be practically possible to manage Irritable Bowel Syndrome better in our practice 56.8 28.4 12.9 155 Is your present management of Irritable Bowel Syndrome effective, ineffective or variable? Effective Variable Ineffective n= 9.7 84.5 4.5 155 N.B. the rows do not always sum to 100% as missing responses are included in the analysis but not shown. General practitioners views about hypnotherapy and its role in IBS Three quarters of general practitioners (75.5%) saw hypnotherapy as an alternative therapy. Notwithstanding this, a large majority of general practitioners in this survey agreed that hypnotherapy could help patients who suffer from both physical and psychological problems, 72.9% and 77.4% respectively. However just over a third of general practitioners (34.8%) saw hypnotherapy as potentially dangerous, with just under half feeling unsure (40.6%). Nearly 84% felt that hypnotherapy was the province of accredited therapists, and only 20% felt that they would be willing to receive training to provide hypnotherapy themselves. The statement that hypnotherapy should be available through an accredited hypnotherapist (83.9% agreed) and not to take this on as a general practitioner (68.4% with 10.3 unsure) were two of the most polarised responses. If a course of 8 × 30 minutes hypnotherapy were shown to be effective, there would be a willingness amongst most respondents (78.1%) to advise hypnotherapy for some, but definitely not all, IBS patients. 83% of those questioned felt that it was not something that should be offered to every patient with IBS. These results are set out in Table 4 . Table 4 What general practitioners think about hypnotherapy and their willingness to refer irritable bowel syndrome patients for hypnotherapy. Agree Unsure Disagree n= Is Hypnotherapy an "alternative" not a mainstream therapy? 75.5 17.4 6.5 155 Hypnotherapy could help a sufferer from a physical illness 72.9 22.6 4.5 155 Hypnotherapy could help a sufferer of a Psychological disorder 77.4 18.7 3.9 155 Hypnotherapy could be dangerous 34.8 40.6 24.5 155 Is Hypnotherapy a treatment that you might advise for your patients? Yes Neutral No n= 38.7 32.9 28.4 155 Hypnotherapy should be available through an accredited Hypnotherapist. 83.9 13.5 1.9 155 Would you be willing to provide Hypnotherapy personally (after training)? 20.6 10.3 68.4 155 If Hypnotherapy took 8 × 30 minutes sessions to ensure long-lasting remission in Irritable Bowel Syndrome: Yes No n= Would this be a cost-effective measure to provide for all Irritable Bowel Syndrome patients? 12.9 83.2 155 Would this be a cost-effective measure to provide for some Irritable Bowel Syndrome patients? 78.1 13.5 155 Would you refer Irritable Bowel Syndrome sufferers to these sessions elsewhere 56.1 28.4 14.8 155 N.B. the rows do not always sum to 100% as missing responses are included in the analysis but not shown. How should hypnotherapy for IBS be resourced There is considerable uncertainty and divergence of opinions amongst the respondents on how hypnotherapy should be funded. Only 36.1% of general practitioners thought that more NHS resources should be used to give IBS sufferers better treatment, yet a slightly larger percentage (44.5%) thought that the money could be better spent elsewhere. The proportion of those answered 'unsure' in response to the above 2 statements are at large at 45.8% and 38.7% respectively. Whilst just under half (49%) would support their primary care organisation investing in hypnotherapy if it were shown to work, there was less support for providing it within the NHS, at 29%. Most general practitioners (56.8%) thought that hypnotherapy should be provided through private hospitals, with a substantial minority (40%) thinking that insurance companies should pay for it. Details of these responses are shown in Table 5 . Table 5 Should hypnotherapy be funded by the National Health Service. Agree Unsure Disagree n= National Health Service resources should be used to give Irritable Bowel Syndrome sufferers better treatment 36.1 45.8 18.1 155 National Health Service resources could be better spent on other illnesses 44.5 38.7 16.8 155 I would support my primary care trust investing in Hypnotherapy (if shown to work) 49.0 29.7 21.3 155 Hypnotherapy should be available through the National Health Service 29.0 41.9 28.4 155 Hypnotherapy should be available through private hospitals 56.8 34.2 9.0 155 Medical insurance companies should pay for Hypnotherapy for their clients 40.0 46.5 12.9 155 N.B. the rows do not always sum to 100% as missing responses are included in the analysis but not shown. Finally despite the divergence and uncertainty on hypnotherapy for IBS sufferers, over three quarters of respondents, 76.3% (table not shown), said that they would be willing to refer patients into a study of hypnotherapy for the treatment of IBS in primary care if it were conducted. Discussion The principal findings of the study are that many general practitioners see IBS as a "nervous condition" to be treated with care and attention, as well as drugs and dietary change. There is considerable divergence in the perception of the effectiveness of current approaches, and a willingness amongst many GPs to refer to qualified hypnotherapists if it can be shown to be effective, even though this treatment is considered "alternative" and potentially dangerous. It is also acknowledged that this could not be made a priority for the NHS, but should be provided privately outside it. Unfortunately there were not sufficient resources available to perform telephone reminders, and this may have in part accounted for the low response rate. Comparison of the achieved sample with the target sample shows that fewer single-handed practices and more training practices took part in the survey, which may reflect a greater willingness of academically oriented practitioners to participate in research. It is beyond the scope of the present study to estimate how representative the sample is of other areas, especially inner cities and other more deprived areas where private hospital and insurance are simply not an option for most patients. Existing literature suggests that around 40% of patients in the UK have access to complementary therapy [ 17 ]. This study indicated that that only 6% of general practitioners had ever provided hypnotherapy as a treatment, although 38% said that they might advise hypnotherapy under some circumstances. Other studies show that doctors hold different 'public' and 'private' attitudes to IBS as an illness, and may classify patients informally into 'good' and 'bad' patients [ 11 ]. This study seems to show general practitioners responding in a 'public' way, saying that IBS needs more resources and that it could be managed more effectively than it is at present. Another study concluded that 'Drug usage in the IBS is more than is justified and should, in our view, be minimised [ 10 ]. The present study seems to show that just under half of the general practitioners felt that drugs were effective with about a third unsure (whether through pharmacological or placebo effects was not shown). More research is needed into the spectrum of views of general practitioners about IBS. A qualitative study might have provided more information about general practitioner beliefs about IBS and what they feel should be done to improve services. Additional quantitative work might identify whether there are age-sex differences in practitioners attitudes. The general practitioners were clearly open to hypnotherapy to help manage IBS. It is unclear whether this was a specific attitude towards hypnotherapy, or a general willingness to promote any psychological or supportive therapy for this illness. It would also be interesting to investigate the opinions of gastroenterologists or psychosomatic therapists. Conclusions The vast majority of general practitioners think that current management options for IBS are variable in their effectiveness. Many agreed that this condition needs improved treatment options in primary care; and general practitioners seem willing to consider hypnotherapy as one such treatment option if it can be shown to be effective. Competing interests The authors declare that they have no competing interests. Authors' contributions SC conceived the study and developed the questionnaire, TC and S de L analysed the data, S de L with contributions from the other authors wrote the paper. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 Appendix 1: The questionnaire Questionnaire used in the study Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526280.xml
520829	Analysis of cardiac signals using spatial filling index and time-frequency domain	Background Analysis of heart rate variation (HRV) has become a popular noninvasive tool for assessing the activities of the autonomic nervous system (ANS). HRV analysis is based on the concept that fast fluctuations may specifically reflect changes of sympathetic and vagal activity. It shows that the structure generating the signal is not simply linear, but also involves nonlinear contributions. These signals are essentially non-stationary; may contain indicators of current disease, or even warnings about impending diseases. The indicators may be present at all times or may occur at random in the time scale. However, to study and pinpoint abnormalities in voluminous data collected over several hours is strenuous and time consuming. Methods This paper presents the spatial filling index and time-frequency analysis of heart rate variability signal for disease identification. Renyi's entropy is evaluated for the signal in the Wigner-Ville and Continuous Wavelet Transformation (CWT) domain. Results This Renyi's entropy gives lower 'p' value for scalogram than Wigner-Ville distribution and also, the contours of scalogram visually show the features of the diseases. And in the time-frequency analysis, the Renyi's entropy gives better result for scalogram than the Wigner-Ville distribution. Conclusion Spatial filling index and Renyi's entropy has distinct regions for various diseases with an accuracy of more than 95%.	Background Bio-signals are essentially non-stationary signals; they display a fractal like self-similarity. They may contain indicators of current disease, or even warnings about impending diseases. The indicators may be present at all times or may occur at random – in the time scale. However, to (study and) pinpoint anomalies in voluminous data collected over several hours is strenuous and time consuming. Therefore, computer based analytical tools for in-depth study and classification of data over day long intervals can be very useful in diagnostics. Electrocardiography deals with the electrical activity of the heart. Monitored by placing sensors at defined positions on chest and limb extremities of the subject, electrocardiogram (ECG) is a record of the origin and propagation of the electric action potential through cardiac muscle. It is considered a representative signal of cardiac physiology, useful in diagnosing cardiac disorders. The state of cardiac health is generally reflected in the shape of ECG waveform and heart rate. It may contain important pointers to the nature of diseases afflicting the heart. However, bio-signals being non-stationary signals, this reflection may occur at random in the time scale. (That is, the disease symptoms may not show up all the time, but would manifest at certain irregular intervals during the day.) Therefore, for effective diagnostics, the study of ECG pattern and heart rate variability signal (instantaneous heart rate against time axis) may have to be carried out over several hours. HRV is a useful signal for understanding the status of the autonomic nervous system (ANS). The interest in the analysis of heart rate variability (HRV), that is, the fluctuations of the heart beating in time, is not new. And much progress was achieved in this field with the advent of cheap and massive computational power, which provoked many recent advances. HRV is a non-invasive measurement of cardiovascular autonomic regulation. Specifically, HRV is a measurement of the interaction between sympathetic and parasympathetic activity in autonomic functioning. There are two main approaches for analysis: time domain analysis of HRV [for standard deviation of normal to normal intervals (SDNN)]; and frequency domain analysis [for power spectrum density (PSD)]. The latter provides high frequency (parasympathetic activity) and low frequency (sympathetic and parasympathetic activity) and total power (sympathetic/parasympathetic balance) values [ 1 - 3 ]. Recent results on HRV signal analysis show that its dynamic behavior involves non-linear components that also contribute in the signal generation and control [ 4 ]. The autonomic nervous system (ANS) modulates the cardiac pacemaker and provides beat-to-beat regulation of the cardiovascular rhythm. Application of wavelet transformation techniques to beat-to-beat heart rate variations (HRV) provides an important non-invasive tool for monitoring the autonomic nervous system functioning. The cardiovascular system is a complex system that includes heart and vessels. ECG and HRV are two methods for study it. Hence, many attempts have been made to analyze these signals and extract information about the cardiovascular system. Most of the methods used are linear and it has been recognized that nonlinear methods may be more suitable for analyzing signals that originate from complex nonlinear living systems [ 5 ]. Recent developments in non-linear analysis have provided various methods for the study of the complex cardiovascular system [ 6 ]. It is now generally recognized that many processes generated by the biological system can be described in an effective way by using the methods of nonlinear dynamics. The nonlinear dynamical techniques are based on the concept of chaos, which was first introduced with applications to complicated dynamical systems in meteorology [ 7 ]. Since then, it has been applied to medicine and biology [ 8 , 9 ]. A particularly active area for the application of chaos theory has been cardiology [ 10 , 11 ], where many aspects have been addressed including whether chaos can be used to represent healthy or diseased state [ 12 ]. A complex system like cardiovascular system can not be linear in nature and by considering it as a nonlinear system can lead to better understanding of the system dynamics. Recent studies have also stressed the importance of nonlinear techniques to study HRV in both health and disease. The progress made in the field using measures of chaos has attracted scientific community applying these tools in studying physiological systems, and HRV is no exception. There have been several methods of estimating invariants from nonlinear dynamical systems reported in the literature. Recently, Fell et al and Radhakrishna et al have tried the nonlinear analysis of ECG and HRV signals respectively [ 13 , 14 ]. Also, Addison at al showed that coordinated mechanical activity in the heart during ventricular fibrillation may be made visible in the surface ECG using wavelet transform [ 15 ]. Rajendra et al, [ 16 ] have classified the HRV signals using Artificial Neural Networks (ANN) and Fuzzy equivalence relation. Recently, Renyi's entropy is used for texture analysis by Grigorescu et al [ 17 ]. Gokcay et al have applied Renyi's entropy to clustering and analyze the resulting staircase nature of the performance function that can be expected during learning [ 18 ]. In this work, different heart rate signals are analyzed using spatial filling index and time frequency techniques. Renyi's entropy is evaluated for the different cardiac abnormalities. Methods ECG data for the analysis was obtained from MIT-BIH arrhythmia database [ 19 ]. Prior to recording, the ECG signals were processed to remove noise due to power line interference, respiration, muscle tremors, spikes etc. The R peaks of ECG were detected using Tompkins's algorithm [ 20 ]. The ECG data contains eight different classes representing eight different diseases. The number of datasets chosen for each of the eight classes is given in Table 1 . The Normal class contains datasets from people where no cardiac abnormality was diagnosed. The remaining classes are named according to the diagnosed cardiac abnormality, premature ventricular contraction (PVC), Complete Heart Block (CHB), Sick Sinus Syndrome (SSS), Congestive heart failure (CHF), Ishemic/Dilated cardiomyoapathy (ISCDIL), Atrial Fibrillation (AF), and ventricular fibrillation (VF). Table 1 Number of subjects in various groups Type Normal PVC CHB SSS CHF ISCDIL AF VF Number of datasets 60 60 20 20 40 20 35 45 Each dataset is taken consists of more than10,000 samples and the sampling frequency of the data is 360 Hz. The interval between two successive QRS complexes is defined as the RR interval (t r-r seconds) and the heart rate (beats per minute) is given as: HR = 60/t r-r (1) Spatial Filling Index Let the signal be represented by the coordinates of a point X ( k ) in phase space. Then the dynamical behavior of the signal is reconstructed by succession of these points X ( k ) in the phase space. Phase space reconstructions are based on the analysis of dynamic systems by delay maps. The vectors X ( k ) in the multidimensional phase space are constructed by time delayed values of the time series, which determine the coordinates of the phase space plot. X ( k ) = { x ( k ), x (( k + τ ), ..., x ( k + ( E -1) τ )} for k = 1,2,..., N - ( E - 1) τ (1) where X ( k ) is one point of the trajectory in the phase space at time k , x ( k + τ ) are the coordinates in the phase space corresponding to the time delayed values of the time series, τ is the time delay between the points of the time series considered and E is the embedding dimension, which is the number of coordinates of the phase space plot. The attributes of the reconstructed phase space plot depend on the choice of value of τ . One way to choose τ is to take it as the time it takes the autocorrelation function of the data to decay to 1/ e [ 21 ]. Another method is to take the first minimum in the graph of average mutual information [ 22 ], which appears to be better since it considers the nonlinear structure in the signal. It has been established using this method that the value of 7 for τ is the best choice for ECG signals and 5 for HRV signals [ 23 ]. From the given signal x (1), x (2), ..., x ( N ), a matrix A E is obtained as where E is the number of dimensions and M is related to N by the equation: M = N - ( E -1) τ (3) By plotting column 2 of matrix A against column 1 (for the case E = 2), the phase space plot for two dimensions is obtained. Similarly, the first three columns of matrix A 3 represent a phase space plot in three dimensions. Now, a normalized matrix B E is obtained by dividing each element of A E by x max where x max = max \| x ( k )\| 1 ≤ k ≤ N (5) The matrix B 2 (in two dimensions) is hence represented as In two dimensions, the phase space plot corresponding to the normalized matrix spans from -1 to +1 on either axis. The phase space area is now divided into small square areas of size { R × R \| R ∈ Real , 2/R ∈ Integer }. Then the number of grids in the normalized phase space is n = 2/ R . A matrix C is now obtained with its elements c ( i,j ) equal to the number of phase space points falling in a grid g ( i , j ). The matrix C is called the phase space matrix and its elements are divided by m , where This division yields P ( i , j ), the probability of a phase space point falling in a grid g ( i , j ). A matrix Q is now formed by squaring each element of P to get q ( i , j ) as the elements of Q . The sum of elements of matrix Q is calculated as The spatial filling index η is defined as: η = s / n 2 (9) Now the value of η is used to quantify the degree of variability in the test signals. Time-Frequency analysis There are three common approaches to generating the time-frequency (TF) plots. These are the short Time Fourier Transform; the Wigner-Ville based bilinear distributions and the Continuous Wavelet Transform. In this investigation the latter two were used. Wigner-Ville analysis The Wigner-Ville distribution (WVD) is defined as: where z ( t ) is the analytic signal and h ( τ ) is a window function. The results where obtained using a Hamming window. This window attenuates the interferences oscillating perpendicularly to the frequency axis. The WVD satisfies a large number of desirable mathematical properties. In particular, the WVD is always real-valued; it preserves time and frequency shifts and satisfies the marginal properties. Moreover, the WVD conserves the Energy of the signal. We obtain the Energy ( E x ) by integrating the WVD of z all over the time frequency plane: With the Energy conservation property the WVD can be interpreted in terms of probability density: expression (10) is the Fourier transform of an acceptable form of characteristic function for the distribution of energy. Therefore, the WVD can be used to obtain the information content of a signal; this thought is further extended in Section 4.3. Continuous Time Wavelet Transform (CWT) analysis A 'wavelet' implies a small wave of finite duration and finite energy, which is correlated with the signal to obtain the wavelet coefficients [ 24 ]. The reference wavelet is known as the mother wavelet , and the coefficients are evaluated for the entire range of dilation and translation factors [ 25 ]. Initially the mother wavelet is shifted (translated) continually along the time scale for evaluating the set of coefficients at all instants of time. In the next phase, the wavelet is dilated for a different width – also normalized to contain the same amount of energy as the mother wavelet – and the process is repeated for the entire signal. The wavelet coefficients are real numbers usually shown by the intensity of a chosen color, against a two dimensional plane with y-axis representing the dilation (scaling factor) of the wavelet, and the x-axis, its translation (shift along the time axis). Thus the wavelet transform plot ( scalogram ) can be seen as a color pattern against a two dimensional plane. In the CWT the wavelet coefficients are evaluated for infinitesimally small shifts of translation as well as scale factors. That is, the color intensity of each pixel in the scalogram is separately evaluated, and the resulting pattern contains information about the size and location of the 'event' occurring in the time domain [ 26 , 27 ]. Since the dilated wavelet is normalized to contain the same amount of energy as the mother wavelet; the scalogram representation of even high frequency, low energy 'events' occurring in the time scale are more conspicuous than in the Fourier Transform. Thus the color patterns in the scalogram can be useful in highlighting the abnormalities specific to different types of disease. MATLAB version 6.1 is used to plot the various scalogram plots. For a given wavelet Ψ a , b ( t ), the coefficients are evaluated using Eq. (12): The wavelet, defined as ,...small wave of finite duration and finite energy...' has also zero mean value, is energy normalizing coefficient, and Ψ a , b ( t ) is the mother wavelet; a → scale factor ; b → translation factor. Just like the WVD, the CWT representation preserves also the energy of the signal. The total energy ( E x ) is obtained by integrating over all scale and translation factors: The scalogram patterns thus obtained also depend on the wavelet chosen for analysis. Bio-signals usually exhibit self similarity patterns in their distribution, and a wavelet which is akin to its fractal shape would yield the best results in terms of clarity and distinction of patterns. In the present work, the analysis is based on the Morlet wavelet. This wavelet gives good result compared to all the other wavelets. The Morlet wavelet function is given by: Renyi's Entropy (RE) The previous sections detailed WVD and CWT as two methods to represent a signal in the time-frequency domain. This section is concerned with the interpretation of the time-frequency representation. The signals represent measurements taken from patients being either normal or suffering from different vascular diseases. The goal is to find a measure which allows classifying the different signals according to the medical conditions. One interesting information that one may obtain from the time-frequency representation is the number of elementary signals present in the current observation. This leads to the following question: How much separation between two elementary signals must one achieve in order to be able to conclude that there are two signals present rather then one? A solution to this problem is given by applying an information measure to a time-frequency distribution of a signal. This can be done, because CWT and WVD preserve the energy of the signal. Unfortunately, the well known Shannon information can not be applied to the time-frequency representation of a signal, because it contains negative values. One information measure, which allows negative values in the distribution, is Renyi's entropy . This information measure was used to analyze the time-frequency representation of the measurement data. Renyi's entropy definition is derived from his proposed theory of means [ 28 ] where φ (.)- is a continuous and strictly monotonic function subclass of Kolmogorov-Nagumo functions. To satisfy the constraints of an information measure I ( p k )- any information measure Simplifying the above relation, we have The third order Renyi's entropy ( α = 3) is calculated from the WVD time-frequency representations as follows: similarly, the third order Renyi's entropy is calculated form the CWT as follows: The result produced by this measure ( and ) is expressed in bits : If one elementary signal yields zero bit of information (2 0 ), then two well separated elementary singles will yield one bit of information (2 1 ), four well speared elementary singles will yield two bits of information (2 2 ), and so on. It shows that for different cardiac signals the Renyi's entropy in the time-frequency domain is different. One-Way Analysis of Variance (ANOVA) The purpose of one-way ANOVA is to find out whether data from several groups have a common mean. That is, to determine whether the groups are actually different in the measured characteristic. One-way ANOVA is a simple special case of the linear model. The one-way ANOVA form of the model is where: y ij = α . j + ε ij • y ij is a matrix of observations in which each column represents a different group. • α . j is a matrix whose columns are the group means. (The "dot j" notation means that applies to all rows of the jth column. That is, the value α ij is the same for all i.) • ε ij is a matrix of random disturbances. The model posits that the columns of y are a constant plus a random disturbance. You want to know if the constants are all the same. Results The result section compares the three different analyzing methods. Each of these methods results in a single parameter for each of the datasets. For Phase Space the parameter used for comparison is the spatial filling index ( η ) defined in Equation (9). For WVD the parameter is the third order Renyi's entropy ( ) defined in Equation (14). For CWT the parameter is also the third order Renyi's entropy ( ) defined in Equation (15). For each dataset these three parameters ( , and η ) where calculated. Table 2 shows the mean and the variance (normalized by N -1 where N is the sequence length) of these parameters for each of the data classes. The p-value, also shown in Table 2 , results form the ANOVA test for each of the parameters. Table 2 Results for various cardiac abnormalities Type SSS PVC CHB NORMAL CHF AF ISCDIL VF p-value η Phase Space 1.56 ± 0.08 3.77 ± 11.82 7.07 ± 0.57 6.76 ± 1.61 7.71 ± 0.20 2.26 ± 0.05 7.44 ± 1.24 3.77 ± 8.78 0.00005 Wigner-Wille 3.38 ± 2.74 4.79 ± 1.46 5.65 ± 0.61 4.00 ± 1.52 2.10 ± 0.61 3.31 ± 2.45 4.07 ± 2.78 4.44 ± 0.96 0.021 Scalogram 2.84 ± 1.89 2.25 ± 1.06 3.01 ± 0.29 1.67 ± 0.84 1.78 ± 0.82 2.04 ± 1.37 2.15 ± 0.57 3.29 ± 0.33 0.001 The proposed technique was applied to a number of different signals, both normal and abnormal. Some of the normal and abnormal signals used in the analysis, along with their two dimensional plots are shown in the Figures 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 . Figure 1 Heart rate in representative subject with SSS; (a) Phase space plot (b) Scalogram (c) Wigner-Ville distribution Figure 2 Heart rate in representative subject with PVC; (a) Phase space plot (b) Scalogram (c) Wigner-Ville distribution Figure 3 Heart rate in representative subject with CHB; (a) Phase space plot (b) Scalogram (c) Wigner-Ville distribution Figure 4 Heart rate in representative subject with Normal; (a) Phase space plot (b) Scalogram (c) Wigner-Ville distribution Figure 5 Heart rate in representative subject with CHF; (a) Phase space plot (b) Scalogram (c) Wigner-Ville distribution Figure 6 Heart rate in representative subject with AF; (a) Phase space plot (b) Scalogram (c) Wigner-Ville distribution Figure 7 Heart rate in representative subject with Ishemic/Dilated Cardiomyopathy; (a) Phase space plot (b) Scalogram (c) Wigner-Ville distribution Figure 8 Heart rate in representative subject with VF; (a) Phase space plot (b) Scalogram (c) Wigner-Ville distribution For the time frequency plots the normalized frequency is shown over the hart rate values. It is not useful to state an absolute frequency, because such a value is not relevant for the cardiac system under observation. Moreover, the relative frequency representation allows comparing the time frequency analysis results form varying observation intervals. As example, the observation interval for the VF data is significantly shorter as for the rest of the data, but still the results can be compared. Discussion The resulting phase space plots for various types of disease are shown in Figure 1(a) , 2(a) , 3(a) , 4(a) , 5(a) , 6(a) , 7(a) , 8(a) . In SSS – III (Sick Sinus Syndrome – III, Bradycardia-Tachycardia) there is a continuous variation of heart rate between Bradycardia and Tachycardia. The phase space plot spreads over a larger area (Figure 1(a) ). In the Ectopic beat abnormality; there would be a sudden impulsive jump in the heart rate. This may be due to a Premature-Ventricular beat in the ECG signal. This is indicated as a sudden spike in the phase space plot (Figure 2(a) ). In Complete Heart Block (CHB) cases, as the atrio-ventricular node fails to send electrical signals rhythmically to the ventricles, the heart rate remains low. The phase space plot reduces almost to a point, indicating very little change with time (Figure 3(a) ). For Normal cases, the phase space plot looks like a cluster of points (Figure 4(a) ). In the Congestive heart failure (CHF), the heart rate variation is lower and hence the phase space plot spread in a very small area (Figure 5(a) ). In the Atrial Fibrillation (AF), heart rate signal records highly erratic variability; this is depicted as scattering of points in the phase space plot (Figure 6(a) ). In the case of Ischemic/Dilated cardiomyopathy, the ventricles are unable to pump out blood to the normal degree. Here the heart rate variation is very small. And hence the phase space plot will be almost a point (Figure 7(a) ). And its phase space plot resembles that of Normal class. Finally, in VF, the heart rate variation is high and hence the phase space plot is randomly distributed (Figure 8(a) ). The contour plots of scalogram and Wigner-Ville distribution plot for the different abnormalities are shown in figures 1(b),1(c) , 2(b),2(c) , 3(b),3(c) , 4(b),4(c) , 5(b),5(c) , 6(b),6(c) , 7(b),7(c) , 8(b),8(c) respectively. In the contour plot of scalogram (Figure 1(b) ), for SSS, there is clear indication of variation of high frequency and low frequency in the form of irregular circles at these frequencies. In PVC (Figure 2(b) ), a irregular circle is shown at high frequency indicating the spike of the signal. These irregular circles or contours are at low frequencies for CHB (Figure 3(b) ), indicating smaller R-R variation. In normal case (Figure 4(b) ) these contours are in the middle frequency due to variation in the R-R interval. In CHF (Figure 5(b) ) and Ischemic/Dialted cardiomyopathy (Figure 7(b) ), the R-R variation is extremely low. Hence the contours are aligned at the low frequency. In AF (Figure 6(b) ), due to very high R-R variation are shown as irregular contours at high frequency. For VF, this R-R variation is slightly low and as result the contours are aligned at the middle of the contour plot (Figure 8(b) ). The contour plots of the Wigner-Ville distribution does not indicate as clearly as contour plot of scalogram for various cardiac diseases. The spatial filling index decreases or increases from the normal class for the abnormal subjects in different ranges (Table 2 ) depending on the R-R variation. This value decreases for the abnormalities of high R-R variation and increases for CHB, CHF and Ishemic/Dilated cardiomyoapathy, which has low R-R variation. This parameter has excellent 'p' value for various classes (0.00005). The Renyi's entropy has high value for cardiac abnormalities like Ischemic/Dilated cardiomyopathy, CHB, VF and it decreases for Normal, PVC, AF, SSS and CHF. This RE gives lower 'p' value for scalogram than Wigner-Ville distribution and also, the contours of scalogram visually show the features of the diseases. Hence, in the time-frequency analysis, the Renyi's entropy gives better result for scalogram than the Wigner-Ville distribution. Conclusion Considering heart as a nonlinear complex system and processing various cardiovascular signals (HRV) seems to provide very useful information for detection of abnormalities in the condition of the heart that is not available by conventional means. In this paper, a phase space and the time-frequency analysis of these cardiac signals using spatial filling index and Renyi's entropy has been proposed for detecting cardiac dysfunction. The ANOVA test was used to compare the different analyzing methods. The Renyi's entropy gives better result for the scalogram than the Wigner-Ville distribution. The evaluation of the proposed technique on a larger data set will improve the efficacy of the technique. It is left as future work to compare the different methods with more sophisticated statistical methods, such as post hoc comparisons. It is hoped that the graphical representation along with its corresponding analytical index and Renyi's entropy proposed here will find potential applications in computer analysis of cardiac patients' status in intensive care units.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC520829.xml
546206	Effectiveness and cost-effectiveness of a multidisciplinary intervention programme to prevent new falls and functional decline among elderly persons at risk: design of a replicated randomised controlled trial [ISRCTN64716113]	Background Falls are common among community-dwelling elderly people and can have a considerable impact on quality of life and healthcare costs. People who have sustained a fall are at greater risk of falling again. We replicated a British randomised controlled trial which demonstrated the effectiveness of a multidisciplinary intervention programme to prevent falls. The objective is to describe the design of a replication study evaluating a multidisciplinary intervention programme on recurrent falls and functional decline among elderly persons at risk. The study consists of an effect evaluation, an economic evaluation and a process evaluation. Methods/design The programme is aimed at community-dwelling elderly people aged 65 years or over who have visited an accident and emergency department (A&E department) or a general practitioners' cooperative (GP cooperative) because of a fall. The design involves a two-group randomised controlled trial. Participants are followed for twelve months after baseline. The intervention programme consists of a detailed medical and occupational therapy assessment with referral to relevant services if indicated. People in the control group receive usual care. The main outcome measures of the effect evaluation are number of falls and daily functioning. The economic evaluation will be performed from a societal perspective. A process evaluation will be carried out to evaluate the feasibility of the intervention programme.	Background Publishing the design of a study This article describes the design of a replication of a randomised controlled trial (RCT) evaluating the effectiveness and cost-effectiveness of a multidisciplinary intervention programme to prevent further falls among elderly people at risk. Publishing the design and protocol of a study before results are available is important for several reasons. A published protocol allows easier comparison between what was originally intended and hypothesised and what was actually done [ 1 ], and it gives readers greater insight into the methodological quality of a study. Furthermore, it has often been recognised that negative or adverse outcomes are less likely to be published [ 1 - 3 ]. Publishing the design of a study before its start announces that a study will be undertaken, which encourages publication of the results and in any case informs researchers where they can find the data for inclusion in systematic reviews [ 1 , 2 ]. Thus, publishing a design article can prevent publication bias. Prevention of falls About one-third of people over the age of 65 fall at least once a year [ 4 ]. People who have fallen show an increase in morbidity, mortality and healthcare utilisation [ 5 ], which implies increased healthcare costs. In addition, people who have sustained a fall are at greater risk of falling again [ 5 ]. Since preventing falls has been a matter of interest for years, many programmes aimed at preventing falls have been developed and evaluated. Unfortunately, many of these have turned out to be ineffective [ 4 ]. However, there is now considerable evidence of the effectiveness of multifaceted interventions. Programmes likely to be effective in preventing falls among elderly people are multidisciplinary, multifactorial programmes screening for health and environmental risk factors and offering interventions, both for the general population of community-dwelling elderly people and for elderly people with a history of falling selected because of known risk factors [ 4 ]. An example of such an intervention programme is the successful programme developed by Close et al. [ 5 ]. This programme is aimed at people aged 65 years or older who live in the community and have visited an accident and emergency department because of a fall. The intervention programme consists of a detailed medical and occupational therapy assessment with referral to relevant services if indicated. The intervention has been evaluated in a randomised controlled trial, which demonstrated that this multidisciplinary intervention implemented among people at risk was highly effective in reducing the number of recurrent falls and associated injuries in London (United Kingdom) [ 5 ]. Because details of the status of the participants, the context of the intervention and the content and presentation appear to be critical, it has been recommended to re-evaluate effective intervention programmes in different healthcare systems [ 4 ]. We therefore decided to evaluate the effectiveness of the intervention developed by Close [ 5 ] in Dutch healthcare, by replicating this study in the Netherlands. Objective and research questions The main objective of our current study is to evaluate the effects of a multidisciplinary intervention programme on recurrent falls and functional decline among elderly persons who have visit a general practitioners' cooperative (GP cooperative) and/or an accident and emergency department (A&E department) because of a fall. This objective has resulted in the following research questions. • Is a multidisciplinary intervention programme more effective than usual care in preventing new falls and functional decline among community-dwelling elderly people who visit a GP cooperative and/or A&E department at a hospital because of a fall? • Is the multidisciplinary intervention programme cost-effective compared to usual care when assessed from a societal perspective? Besides the effect evaluation and economic evaluation, a process evaluation is being carried out to assess the feasibility and applicability of the intervention programme for those receiving and implementing the intervention. Design and methods Design Figure 1 shows the design of the study presented, which is a two-group randomised controlled trial. At this stage, the randomisation process has already been completed. Randomisation was achieved by means of computerised alternative allocation and performed by an external agency. Randomisation took place after completion of a self-administered baseline questionnaire. People allocated to the control group received usual healthcare, while people in the intervention group underwent a medical and occupational therapy assessment. The intervention period is scheduled to last for a maximum of 3.5 months after the baseline measurement. After baseline measurement, all subjects are followed for a twelve-month period. During this follow-up period, falls and healthcare utilisation are measured continuously. Subjects are contacted monthly by telephone for an interview about their falls and healthcare utilisation. In addition, self-administered questionnaires are sent to the subjects after four and twelve months. Figure 1 Study Design We have taken various measures to ensure blinding in the data collection process. Questionnaires are collected anonymously and sorted by number. Follow-up measurements by phone are contracted out to an independent call centre, whose operators are unaware whether the subjects have been allocated to the intervention or the control group. The study design and protocols were approved by the Medical Ethics Committee of the University Hospital and University of Maastricht. Target population Various studies have been conducted to assess the effectiveness of programmes to prevent falls. Although most studies were aimed at the general population of elderly people, details of the status of the participants appear to be critical [ 4 , 6 ]. Several authors have suggested that interventions are likely to have greater effect when targeting people at risk [ 7 , 8 ]. People who attend an A&E department with an injurious fall form a high-risk group, and are expected to be more receptive to an intervention programme aimed at reducing falls than the general population of community-dwelling elderly people. In a study by Close et al., about half of the patients who attended an A&E department with a fall had experienced an earlier fall in the previous year, compared to about one third of the elderly people in the general population [ 5 ]. Like Close et al. [ 5 ], we chose community-dwelling elderly people aged 65 years or over who had sustained an injurious fall as the target population of our intervention programme. The following definition of a fall was used: 'A fall is an event which results in a person coming to rest inadvertently on the ground or other lower level'. This definition is derived from that used by the Kellogg International Work Group [ 9 ]. Recruitment of the study population Recruitment of subjects took place at the local GP cooperative and the A&E department of the University Hospital in Maastricht. The Maastricht GP cooperative is a group of GPs from practices in the town of Maastricht and the surrounding area who have founded a non-profit organisation to provide care for their own patients after hours [ 10 ]. The Maastricht GP cooperative has been set up at the hospital's A&E department and covers the out-of-hours service for all local GPs [ 11 ]. The following inclusion criteria were used: age 65 years or older, community-dwelling, having visited the A&E department or GP cooperative at the University Hospital Maastricht for the consequences of a fall, and living in Maastricht or its surrounding area. People were only allowed to enter the programme after completing and returning an informed consent form. Exclusion criteria were: not able to speak or understand Dutch, not able to complete questionnaires or interviews by telephone, cognitive impairment (a score of less than 4 on the Abbreviated Mental Test 4 (AMT 4) [ 12 , 13 ], long-term admission to a hospital or other institution (more than four weeks from the date of inclusion), permanently bedridden, or fully dependent on a wheelchair. Sample size calculation Sample size calculations were based on the expected effects of the intervention on the main outcome measure, the percentage of people sustaining a fall during one year of follow-up. The study by Close et al. [ 5 ] found that the percentage of persons who sustained a recurrent fall was 52% in the usual care group and 32% in the intervention group. If we want to detect the same reduction in the percentage of persons sustaining a recurrent fall in our study, with a power (1-beta) of 90% and alpha of 0.05, we need 123 patients in each group (a total of 246). Based on the experiences of Close et al. and our own experiences in trials among elderly people in the Netherlands [ 14 ], we expect a dropout rate of about 25% during the one-year follow-up period. This means that about 164 persons per group (a total of 328) have to be included in the study. The inclusion period was 14 months. Intervention programme To adapt the programme developed by Close et al. [ 5 ] to the Dutch situation, and to make improvements based on recent insights, we performed a review of the literature, convened a consensus meeting and tested the adapted version in a pilot study (n = 36). Based on this process, we made some improvements to Close et al.'s programme. The final programme includes a medical and occupational therapy assessment resulting in recommendations. The medical assessment consists of an examination performed by a geriatrician, a geriatric nurse and a rehabilitation physician to identify and address risk factors for falling. The examination includes a comprehensive general examination, but in addition focussed on a more detailed assessment of visual acuity, stereoscopic vision, mobility, balance, cognition, affect, use of medication and examination of feet and footwear. Recommendations or indications for referral resulting from this examination are sent to the patient's GP. After the medical assessments, an occupational therapist visits the patients to identify possible risk factors for falling in the home environment. The therapist makes recommendations regarding adaptations to the home environment, assistive devices, home care and behavioural change. Recommendations by the occupational therapist are sent directly to the subjects themselves and to their GPs. As stated before, the intervention period is scheduled to last for a maximum of 3.5 months after the baseline measurement. An important addition to Close et al.'s protocol [ 5 ] is the collaboration with a rehabilitation physician (physiatrist) in the medical part of the intervention. In addition to the screening by a geriatrician, our programme also involves screening by a rehabilitation physician who examines the subjects' feet and the shoes which the subjects wore at the time of the fall. Details of the process of adaptation and the contents of the intervention programme will be published elsewhere. Usual care in the Netherlands People in the control group receive usual care. At present, no guidelines exist for the systematic assessment of the underlying causes of an injurious fall presented at an A&E department or GP cooperative in the Netherlands. In usual care, medical risks and other risk factors such as environmental hazards in the home and patients' risk behaviour are not systematically registered and addressed by hospital physicians, specialists or general practitioners. Moreover, no systematic attention is currently being paid to the specific consequences of an injurious fall for the daily functioning of individual patients in their unique situation. We placed no restrictions on co-interventions. Effect evaluation The primary outcome measures of the effect evaluation are number of falls and daily functioning. Number of falls is subdivided into three separate measures: the percentage of elderly people sustaining a fall during the one-year follow-up period, recurrent falls during follow-up (i.e., the percentage of elderly people sustaining two or more falls), and injurious falls during follow-up (the percentage of elderly people receiving medical care after a fall). Falls are recorded continuously by means of a fall calendar during the twelve-month follow-up period. Subjects are called monthly to report their falls as recorded on the fall calendar relating to the previous month. We decided to measure daily functioning by means of the Frenchai Activity Index (FAI) [ 15 ], in contrast to Close et al. [ 5 ], who used the Barthel Index. Our reason for choosing this instrument was that the FAI has proved to be suitable for the general population of elderly people [ 14 ] and has at least two advantages over the Barthel Index. One is that the Barthel Index shows a ceiling effect when applied to elderly people who have sustained a fall [ 5 ]. The other is that most activities of daily living (ADL) scales, like the Barthel Index, do not refer to complex activities like housekeeping, recreation, hobbies and social interaction. These so-called instrumental abilities (IADL) may affect the quality of life considerably, and the FAI focuses primarily on these IADL abilities [ 15 ]. The FAI is measured by means of self-administered questionnaires at baseline and after four and twelve months. Our secondary outcome measures are: recuperation from the fall, health complaints, perceived health measured by means of the first two items of the RAND-36 [ 16 ], ADL and IADL disability measured by means of the GARS (Groningen Activity Restriction Scale) [ 17 ], mental health measured by means of the HADS (Hospital Anxiety and Depression Scale) [ 18 , 19 ] and quality of life measured by means of the European Quality of Life instrument (EuroQol) [ 20 ]. The secondary outcome measures are assessed by means of self-administered questionnaires at four and twelve months. Besides the primary and secondary outcome measures, we assess some background variables which are considered to be predictors, confounders or effect modifiers. The following personal characteristics are assessed: age, sex, marital status, living alone and socio-economic status. In addition, we assess the circumstances and causes of the falls reported at the GP cooperative and/or A&E department, the consequences of the falls (using the Falls Handicap Inventory [ 21 ]), the type of injury, falls in the previous year (retrospective), the patient's height, weight, use of medication and social contacts (using an adjusted version of items 4 and 5 of the Rand Social Health Battery)[ 22 , 23 ], and the occurrence of life events. All background variables are measured at baseline. Economic evaluation The economic evaluation in our study is being performed from a societal perspective, which implies that all costs and outcomes are taken into account if possible. The economic evaluation will be a combination of a cost-effectiveness and a cost-utility analysis. The primary outcome measure for the cost-effectiveness analysis will be the percentage of people sustaining a fall during one year of follow-up. As mentioned above, falls are recorded by means of a calendar. Within the cost-utility analysis, the effects are measured in terms of generic health-related quality of life descriptions, measured according to the standard Dutch version of the EuroQol (EQ 5-D) [ 20 ] in self-administered questionnaires at baseline and after four and twelve months. A direct value for every state of health is generated using the social tariff [ 24 ], which involves an algorithm for interpolating EuroQol results to population utilities. We will assess programme costs, healthcare costs and patient and family costs. All costs are measured by means of a cost diary [ 25 ], in which patients continuously record volumes of health care utilisation during the twelve-month follow-up period. Subjects are asked to report their cost diary relating to the previous month during the same monthly telephone interview in which they report falls from the calendar. The volume of each category we measure will be multiplied by the cost price of each category. Cost prices are presented in Euros. Health care costs are estimated according to the Dutch guideline for cost analysis in health care research [ 26 ]. Where such guidelines are not available for a specific category, real costs or tariffs are used to estimate costs. Process evaluation The process evaluation involves assessing the extent to which the intervention programme is performed according to protocol, the nature of the recommendations made to the participants, participants' compliance with these recommendations and the opinions of participants, physicians and therapists about the intervention programme and recommendations. Data on these topics are collected using the following methods: structured registration forms for the medical and occupational parts of the intervention programme; self-administered evaluation forms filled in by the participants after the medical intervention; interviews by telephone with the participants six weeks or longer after the recommendations are sent and interviews with all participating physicians and therapists at the end of the intervention period. Analysis Data will be primarily analysed according to the intention-to-treat principle, i.e., including all participants with valid data, regardless of whether they received or did not receive the intervention. Subsequently, the results of the intention-to-treat analysis will be compared with the results of an on-treatment analysis, to assess whether protocol deviations have caused bias. Participants with documented deviations from the study protocol (i.e., participants in the intervention group who did not receive the entire intervention or participants in either the intervention or the control group with incomplete follow-up data) will be excluded from this on-treatment analysis. Comparability between the intervention and control groups will be assessed at baseline to check for differences between the two groups. Outcome at four and twelve months will be compared between the intervention and control groups by both univariate and multivariate techniques. We will use multivariate analysis to adjust for possible differences in baseline scores and background variables between the intervention and control groups. Dropouts and losses-to-follow up will be described. The economic evaluation will involve calculating cost-effectiveness and cost-utility ratios. The additional costs and additional benefits of the intervention programme compared with usual care will be examined by calculating incremental cost-effectiveness and cost-utility ratios. These incremental ratios represent the difference in mean costs between the intervention and usual care groups in the numerator and the difference in mean effects in the denominator[ 27 ]. Since the recruitment period is only 14 months, and the follow-up period is also relatively short (12 months), it is unlikely that there will be substantial differences between costs made by and for patients who started in the first part of the recruitment period and those who started in the last part. Therefore, discounting of costs is not required. Finally, a sensitivity analysis will be done to assess the generalisability of the assumptions made in the costing process. This sensitivity analysis, which involves calculating the upper and lower limits of the confidence interval of cost and effect variables, will allow us to explore and quantify the uncertainty not related to sampling variations. The process evaluation will mainly be analysed by means of descriptive techniques. Progress of the study Recruitment of eligible subjects commenced in December 2002 and ended in February 2004, resulting in a total of 333 eligible subjects being included in the trial. Randomisation resulted in the allocation of 166 participants to the intervention group and 167 to the control group. Of the 333 persons recruited, 105 (32 %) are male and 228 (68 %) are female. The follow-up period is to end in May 2005. Results on the effects of the programme will be available in 2006 and will be published in the relevant journals. Discussion Although the intervention has been subject of earlier research, this study will provide new information about the effectiveness in the Dutch situation. Furthermore, the results of the economic evaluation can provide information about the cost-effectiveness of the intervention and the effects on quality of life. In case of shown effectiveness and cost-effectiveness, we aim to implement this intervention into usual healthcare. Competing interests The author(s) declare that they have no competing interests. Authors' contributions All author's read and approved the final version of the manuscript. MH is the investigator and performed most of the writing of this manuscript. JH supervises the planning of the project and wrote the study design. JD supervises the planning and methodological aspects of the project. SA supervises the economic evaluation. HC supervises the clinical part of the project. JE is the principal investigator of the study. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546206.xml
514568	Tissue Doppler echocardiography – A case of right tool, wrong use	Background The developments in echocardiography or ultrasound cardiography (UCG) have improved our clinical capabilities. However, advanced hardware and software capabilities have resulted in UCG facilities of dubious clinical benefits. Is tissue Doppler echocardiography (TDE) is one such example? Presentation of the hypothesis TDE has been touted as advancement in the field of echocardiography. The striking play of colors, impressive waveforms and the seemingly accurate velocity values could be deceptive. TDE is a clear case of inappropriate use of technology. Testing the hypothesis To understand this, a comparison between flow Doppler and tissue Doppler is made. To make clinically meaningful velocity measurements with Doppler, we need prior knowledge of the line of motion. This is possible in blood flow but impossible in the complex myocardial motion. The qualitative comparison makes it evident that Doppler is best suited for flow studies. Implications of the hypothesis As of now TDE is going backwards using an indirect method when direct methods are better. The work on TDE at present is only debatable 'research and publication' material and do not translate into tangible clinical benefits. There are several advances like curved M-mode, strain rate imaging and tissue tracking in TDE. However these have been disappointing. This is due to the basic flaw in the application of the principles of Doppler. Doppler is best suited for flow studies and applying it to tissue motion is illogical. All data obtained by TDE is scientifically incorrect. This makes all the published papers on the subject flawed. Making diagnostic decisions based on this faulty application of technology would be unacceptable to the scientific cardiologist.	Background Echocardiography or Ultrasound Cardiography (UCG) is a key investigation in cardiology. Its non-invasive nature makes it a widely acceptable and safe form of investigation. In many cases it has surpassed cardiac catheterisation in diagnostic yield and has become the investigation of choice. The rapid development of UCG has made available to the clinician newer insights into the anatomy and physiology of the heart. With the availability of digital technology it is possible to manipulate raw data in different ways. This has also spawned UCG facilities of dubious clinical benefits. Tissue doppler echocardiography (TDE) is one such example. When we consider UCG we are interested in 2 distinct aspects of the heart: 1. Structure 2. Blood flow. Both these ultimately indicate cardiac function. B-mode/M-mode imaging or simply Ultrasound Imaging (USI) studies structure. Blood flow is studied by Doppler studies (DS). In each stream every development should incrementally improve our understanding of the heart, its structure and function. Or the development should improve the ease of operation or permit better documentation. In USI we had the following developments and each development added to the diagnostic yield. M-mode – temporal resolution; 2D – spatial resolution; Harmonic imaging – improvement in image quality; B-color – enhanced tissue perception; Ultrasonic tissue characterization and acoustic quantification – quantitative analysis of tissue; Omni-plane M-mode – temporal resolution at different segments and 3D imaging – another dimension of imaging. In DS too we had many developments, each improving our perception. Continuous wave (CW) – detected and measured high velocity flows; Pulse wave (PW) – located the abnormal flow; Color flow mapping (CFM) – made it possible to 'image' blood flows and power amplitude Doppler – to study vascular flow. Contrast, trans-oesophageal and intravascular UCG are the clinical extensions and exploitation of these developments What we see in TDE is a retrograde development. In the context of UCG, DS is an indirect method to study blood flow since USI in its present state cannot image fast moving blood cells. Spontaneous echo contrasts could mark the beginning of USI of blood cells. By using TDE we are using an indirect method where direct USI methods are better. Besides, TDE is an inappropriate application of the Doppler methodology. Presentation of the hypothesis TDE is a by-product of color flow mapping (CFM) technology. In CFM, tissue signals are suppressed and flow signals are analyzed. The reverse is true in TDE. Doppler velocities associated with tissue motion are slower than blood flow. Flow signals are eliminated on the basis of signal amplitude. The amplitude of tissue motion is about 40 db greater than the corresponding flow amplitude. Blood flow imaging applies a high pass filter to exclude the strong but low frequency tissue signals and other 'noise' before the signal is input into an autocorrelator that estimates velocity.[ 1 ] Erroneous filter settings could cause the autocorrelator to include components of tissue signals so that tissue velocities become encoded. This principle has been legitimized to color code tissue motion and we get an 'image' which is entirely Doppler information. In this discussion TDE includes both pulse tissue Doppler and color tissue Doppler (also known as Doppler tissue imaging or tissue velocity imaging). The first report on the use of TDE appeared in 1989.[ 2 ] However the real development and the widespread use of this technique started after the publication of the validation work of TDE setting the 'scientific' basis for the quantitative analysis of myocardial velocities in real time.[ 3 ] There are several papers on the use of TDE.[ 4 , 5 ] Various values and indices are already in place.[ 6 , 7 ] Over the past few years technological advances in TDE like curved M-mode, strain rate imaging and tissue tracking have been developed. These are in addition to modalities like the phase imaging, amplitude imaging and acceleration imaging.[ 8 ] Text books on TDE has also been added to the number of books already available on UCG.[ 9 ] Even now there are a number of papers on tissue Doppler echocardiography (TDE) appearing in leading journals.[ 10 ] Is TDE methodology correct? Is it in agreement with the Doppler principles? Has TDE made any tangible improvement in the already available UCG techniques? Has TDE improved our diagnostic yield? This article presents the hypothesis that TDE is a flawed application of Doppler and hence data collection with TDE is erroneous. Testing the hypothesis The article explains the flaws in tissue Doppler echocardiography. As the concept itself is flawed, all data using this modality is flawed . You cannot even think of designing a study to prove the point because the mensuration technique itself is wrong. So we have to use the scientific methods of comparison and reductio ad absurdum to verify it. Any new modality of diagnosis or treatment should be compared with the existing systems. In clinical trials we usually employ quantitative comparison with statistical methods. In this case, since the data acquisition is flawed we have to use a qualitative comparison. This is accepted practice in clinical medicine – It is like comparing a weak limb with the strong one in neurology. One also needs to understand the Doppler methodology to test the hypothesis. The Doppler methodology To estimate clinically significant peak Doppler velocity the following steps are required: 1. You must have prior knowledge of the line of motion of the object. Only if you know this line of motion, you can apply the interrogating signal along the path of motion. This is because of the directional bias of Doppler. 2. Next you will know the direction of motion – whether the object is moving towards the interrogating signal or away. 3. Only after the first two steps you can measure the clinically significant peak velocity. Based on the above discussion it is clear that the following information can be derived from Doppler: 1. Is there motion? Does the object move? This is a random application of Doppler. This is one aspect of tissue Doppler. But this information is redundant as we can already see the movement better by ultrasound imaging. 2. What is the direction of motion? Is it moving towards or away from the interrogating signal? This is also redundant information in tissue Doppler for the reason given above. 3. To measure the peak significant velocity we must know the clear-cut line of motion. Application of Doppler along this line enables us to measure the meaningful peak velocity. Thus the primary step is to know the line of motion. This is possible in flow Doppler but impossible in tissue Doppler. For example we know for sure that the blood moves from left atrium to left ventricle through the mitral valve. At the mitral valve this is a unique free linear motion towards the apex. This produces a clinically significant velocity. That is why we place the sample volume at the mitral valve from apical views so as to align the beam in the line of anticipated flow. That is also why we do not measure the mitral flow velocity from the parasternal views. Next we come to learn whether the blood is flowing towards or away from the transducer. After all these steps it is possible to measure velocity. In tissue Doppler we can measure several velocities in several directions. But we can never know which is the clinically significant peak velocity. Flaws in TDE TDE is a clear case of misuse of technology. To understand this, a comparison between flow Doppler and tissue Doppler is in order. CFM allows us to 'see' what we cannot see with ultrasonic 'eyes' hence its value is great. In color TDE we see more or less what we already see by USI hence its value is marginal. In CFM the anatomic landmarks are intact as the color is superimposed on the B-mode image. In color TDE the B-mode is eliminated and the entire 'picture' is Doppler information. It is difficult to determine the different anatomic regions on TDE. For example it would be very difficult to delineate the blood-endocardial boundary. In fact, the word 'tissue Doppler' is a misnomer and this is one of the reasons for the prevailing confusion. It gives the impression that only myocardial tissues are studied. The appropriate term would have been low velocity Doppler. Any movement in the low velocity range will be detected by 'tissue Doppler'. Myocardial tissue movement is just one among them. The discriminatory power of this modality is unsatisfactory. Myocardial motion is very complex and not amenable to Doppler studies. In the cardiac motion there are translational, rotational and deformational movements. Besides many tissues near the heart move – due to transmitted cardiac motion, vessel pulsation, respiratory motion and involuntary muscle movements and these interact with cardiac motion further and cause false Doppler shifts.[ 1 ] Velocity is a vector quantity and so Doppler interrogation at one point will determine the velocity of the resultant of all these movements projected in the line of the Doppler beam with angle correction. Similarly at a particular point there are movements in several axes and we can never predict the sum resultant vector. Even if known, the resultant is accurately recorded only if it is in the line of the Doppler beam. This is due to the inherent problem of directional bias. It is like measuring the length of a twisted rod directly with a straight rigid ruler. In the case of strain rate imaging, the protagonists have substituted a vector quantity (velocity) for a scalar quantity (length) in the original formula of strain calculation.[ 11 ] This would be mathematically unacceptable. Cardiac motion becomes more complicated in the presence of wall motion abnormalities. Blood flow is simple and suitable for Doppler study. In flow Doppler at a particular point the linear projectile motion of free moving blood cells are studied. At the current interrogation points there is a unique unidirectional flow in one part of the cardiac cycle. For example in mitral valve Doppler interrogation, the unique directional signal is obtained only in diastole. If there is a signal in the in systole, it becomes abnormal and this information has great value. In TDE the to- and- fro motion (systole and diastole) of a tethered interconnected syncytium of myocytes is imaged. Such information is useless. This can be even otherwise seen and analyzed by USI. In flow Doppler there are definite 'points of interrogation', which are the normal and abnormal orifices. In TDE there are no such definite points. While in flow Doppler higher velocities are studied, TDE studies lower velocities i.e. the study of hypo functioning myocardial segments. Lower velocities are difficult to appreciate. The higher velocities are easier to appreciate due to aliasing and variance. There are no such indicators for low velocity. Thus hypo function is difficult to analyze. The derivations from flow Doppler are based on accepted hydrodynamic formulae (like Bernoulli's equation) and allows us to get orifice size, amount of flow and pressure gradients, which are clinically of great importance. The derivations from TDE are useless. Thus Doppler in its present form, is best suited to study flow. Implications of the hypothesis TDE has shown some promise in the Wolff-Parkinson-White syndrome.[ 12 ] But even here by having a high frame rate or by using the omni plane M-mode, we could be able to locate the pre-excited tissue by USI. Diastolic dysfunction in pseudo-normalized mitral Doppler spectrum and in atrial fibrillation are other areas where TDE is found useful.[ 13 , 14 ] Here the benefits are marginal as there are other parameters to study diastolic dysfunction.[ 15 ] Besides, the determination of diastolic dysfunction by Doppler in clinical practice may not be sacrosanct.[ 16 ] It has been used as a method to differentiate between constrictive pericarditis and restrictive cardiomyopathy.[ 17 ] Here its role is supplementary and does not provide critical distinction. The main role of TDE as a method of detecting regional wall motion abnormalities has been stressed.[ 18 ] Here again it is complementary to routine USI. TDE is being used in the study of cardiac transplant rejection.[ 19 ] In this situation the results are not clear and there are other better UCG methods to study rejection.[ 20 , 21 ] TDE has been advocated in planning and follow-up of cardiac re-synchronization therapy.[ 22 , 23 ] Here again the conventional methods are adequate[ 24 , 25 ] and the false-positive data (see below) may be confounding. Besides several issues regarding the therapy need to be settled.[ 26 ] Mitral annular tissue velocities have been used in the determination of left ventricular filling pressures.[ 27 ] Here again an 'hydrodynamic' information (fluid pressure) is derived from non-hydrodynamic (tissue velocity) measurement. Mitral annulus is a circular area. As mentioned earlier here also we cannot define the precise point of interrogation. An infinite number of values can be obtained all around the mitral annulus. It is important to note that in all the above papers the data acquired is flawed due to the basic problem in technology application . Hence the correlations obtained in these papers are spurious or non-sense correlations . One paper presents a misuse of statistics when tissue Doppler velocities have been correlated with myocardial interstitial fibrosis and myocyte interstitial beta-adrenergic receptor density.[ 28 ] By analogy, using a little statistical jugglery, we could find correlation between blood cholesterol levels and flow Doppler indices. All these papers are mentioned just for completeness. Accepting the data in these papers would be tantamount to rejecting the Doppler principle and methodology. As of now TDE is going backwards using an indirect method where direct methods are available. The work on TDE at present is only debatable 'research and publication' material and does not translate into tangible clinical benefits. The prime use of TDE is to study regional wall motion abnormalities. It is claimed that it is possible to quantify wall motion at rest and during stress. However with the above-mentioned limitations it is unlikely to prove more effective than USI. The measurements are ultra sensitive and represent gross movements rather than myocyte contractions. With the advent of omni-plane M-mode it is possible to get temporal resolution in inaccessible planes and study wall motion abnormalities in a better way. The Doppler methods use mathematical formulae to indirectly study motion that cannot be directly measured. Examples are movements of stellar bodies and in the present context movements of blood cells. We use the indirect method based on Doppler principle because we cannot 'image' blood cells by USI. When the imaging of blood cells become possible, Doppler studies would probably be relegated to the background. Color TDE would have been valuable if direct USI was not available. Tissue Doppler has been a disappointing modality in clinical ultrasound cardiology. This is due to the basic flaw in the application of the principles of Doppler. Doppler is best suited for flow studies and applying it to tissue motion is not acceptable. In flow Doppler we suppress the tissue 'noise' and display flow.[ 1 ] In TDE it is the other way round. Besides TDE is ultra sensitive and so the information gathered is almost useless (too many false positive information). In fact excellent cardiac waveforms can be obtained by placing the sample volume just outside the cardiac region! This is also the reason why it is useless for even studying the temporal aspects of the cardiac cycle and waveform analysis. Once the foundation of a modality is wrong, all derivations tend to be wrong. However the author is of the view that TDE could probably have some role in diagnostic UCG. What is required is that Doppler should be used for the purpose it was intended i.e. to study blood flow. Thus we will have 2 types of Doppler studies: High velocity flow Doppler (HVFD) studies and Low velocity flow Doppler (LVFD) studies. Conventional Doppler would be HVFD. TDE would be LVFD and we would be imaging and analyzing very low velocity flows. The possible uses of color LVFD studies would be the study of low flow thrombogenic states, flow in the atria, the blood-endocardial/endothelial interface, low velocity turbulences and blood flow in organs. Pulse wave spectral LVFD could be useful for the detection of low velocity turbulences as a cause for murmurs (not detectable with HVFD). These are potential areas for investigation and research. It is time to look at TDE on more realistic terms. As a new modality of imaging it appears exciting. But its real clinical utility is doubtful. TDE does not give any additional information over the conventional modalities. In fact due to the above-mentioned deficiencies it could give misleading information. Making diagnostic and therapeutic decisions based on this faulty application of technology would be unacceptable to the scientific cardiologist. List of abbreviations used CFM Color Flow Mapping CW Continuous Wave DS Doppler Studies HVFD High Velocity Flow Doppler LVFD Low Velocity Flow Doppler PW Pulse Wave TDE Tissue Doppler Echocardiography UCG Ultrasound Cardiography USI Ultrasound Imaging Competing interests None declared.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514568.xml
509288	Specific TATAA and bZIP requirements suggest that HTLV-I Tax has transcriptional activity subsequent to the assembly of an initiation complex	Background Human T-cell leukemia virus type I (HTLV-I) Tax protein is a transcriptional regulator of viral and cellular genes. In this study we have examined in detail the determinants for Tax-mediated transcriptional activation. Results Whereas previously the LTR enhancer elements were thought to be the sole Tax-targets, herein, we find that the core HTLV-I TATAA motif also provides specific responsiveness not seen with either the SV40 or the E1b TATAA boxes. When enhancer elements which can mediate Tax-responsiveness were compared, the authentic HTLV-I 21-bp repeats were found to be the most effective. Related bZIP factors such as CREB, ATF4, c-Jun and LZIP are often thought to recognize the 21-bp repeats equivalently. However, amongst bZIP factors, we found that CREB, by far, is preferred by Tax for activation. When LTR transcription was reconstituted by substituting either κB or serum response elements in place of the 21-bp repeats, Tax activated these surrogate motifs using surfaces which are different from that utilized for CREB interaction. Finally, we employed artificial recruitment of TATA-binding protein to the HTLV-I promoter in "bypass" experiments to show for the first time that Tax has transcriptional activity subsequent to the assembly of an initiation complex at the promoter. Conclusions Optimal activation of the HTLV-I LTR by Tax specifically requires the core HTLV-I TATAA promoter, CREB and the 21-bp repeats. In addition, we also provide the first evidence for transcriptional activity of Tax after the recruitment of TATA-binding protein to the promoter.	Background In eukaryotes, transcription by RNA polymerase II requires the orderly recruitment of basal transcription factors and activators to the core promoter and enhancers, respectively [ 1 , 2 ]. The core promoter contains the transcription initiation site, and it provides the docking sites for the basal transcription factors that nucleate the assembly of a functional preinitiation complex (PIC). The TATA box is one of four major core promoter elements, and it is specifically recognized by the TATA-binding protein (TBP), a subunit of the basal transcription factor TFIID which also contains at least 14 TBP-associated factors (TAFs). On the other hand, enhancers are bound by sequence-specific transcriptional activators that are thought to promote PIC assembly through interactions with components of the basal transcription machinery. Human T-cell leukemia virus type I (HTLV-I) Tax protein is a unique transcriptional regulator [ 3 ]. Tax can modulate the HTLV-I long terminal repeats (LTR), heterologous viral promoters, and a variety of cellular genes. In most context, Tax acts as a potent transcriptional activator through Tax-responsive DNA elements that are recognized by cellular transcription factors CREB, NFκB and serum response factor (SRF) [ 4 - 6 ]. For activation of the HTLV-I LTR, Tax targets three imperfectly conserved 21-bp direct repeats flanked by GC-rich sequences. In this scenario, Tax forms a ternary complex with CREB and the 21-bp repeat through physical interaction with CREB and direct contact with the flanking GC-rich sequences [ 7 - 9 ]. Tax-induced activation of other promoters is thought to be mediated through protein-protein interactions. Thus, Tax is a pleiotropic transcriptional activator that targets multiple enhancer elements through multiple cellular transcription factors. To date, the molecular mechanisms for Tax trans-activation have been well studied. Due to its pleiotropic activities, there are likely nuances to Tax's activity which remain unrevealed. Currently, we understand Tax to harbor a minimal activation domain [ 10 ], to interact with basal transcription factors such as TBP [ 11 ], to form a homo-dimer [ 12 - 14 ], and to stimulate the dimerization of cellular regulatory factors such as CREB [ 15 , 16 ] and IKK-γ [ 17 ]. Moreover, we also know that Tax can directly engage transcriptional coactivators such as CREB-binding protein, p300 and P/CAF [ 18 - 20 ]. However, it remains unclear what is Tax's optimal preference for an enhancer – TATAA configuration. It has also been unaddressed whether Tax has a transcriptional activity after the formation of an initiation complex at the TATAA-box. In mammalian cells, the artificial recruitment of TBP sufficiently activates transcription from some promoters [ 21 - 24 ]. It is understood that the structure of core promoter is one important determinant for this activation [ 23 ]. On the other hand, DNA-tethered TBP can also work synergistically with selective natural activators such as human immunodeficiency virus type 1 (HIV-1) Tat protein [ 21 - 23 ] and cytomegalovirus IE2 protein [ 25 ]. In this regard, it is not known whether TBP recruitment suffices for activation of HTLV-I minimal promoter. Nor is it clear whether Tax can cooperate with promoter-tethered TBP. Here, we have constructed a series of chimeric enhancer-TATAA reporters to analyze the functional roles of these transcription elements in Tax-mediated activation. We observed that Tax activates the HTLV-I 21-bp repeats more potently than other enhancer elements. Analysis of ten mutants of Tax revealed that Tax utilizes different domains to target different cellular factors. We also found that multiple bZIP transcription factors including the newly-identified LZIP are involved in Tax activation of HTLV-I LTR. Finally, two other salient findings are that optimal Tax-responsiveness is specified by the HTLV-I-specific TATAA element, and that Tax synergizes with artificially recruited, DNA-tethered, TBP in a phase of transcription after the assembly of an initiation complex at the promoter. Results Specific preference by Tax for only one enhancer element Tax can activate transcription through 21-bp repeats, CRE, κB site or SRE [ 4 - 9 ]. However, a direct head-to-head comparison between the relative preferences of Tax for each of these elements is complicated by the context of additional DNA elements in the various promoters tested to date (i.e. the HTLV-I LTR versus the HIV-1 LTR versus the interleukin-2 promoter). To directly compare enhancer motifs, they should be placed in identical TATAA-context and tested in identical experimental settings. Towards this end, we constructed a series of six reporters to dissect the ordered preference of Tax for various enhancers. Each reporter contains two copies of enhancer motifs (21-bp repeats, CRE, AP1, Sp1, κB or SRE) and a minimal HTLV-I TATAA promoter (Fig. 1A ). Because all reporters have the same HTLV-1 minimal promoter and are otherwise devoid of any known enhancer elements, side-by-side comparisons would reflect directly the contribution of the variously added cis-enhancer. We observed that the κB- and CRE- motifs had the highest basal activities in HeLa cells in the absence of Tax (Fig. 1B , lanes 3, 4, 9 and 10; and Fig. 1C , columns 3 and 6 compared to column 1). Of significant interest, in stark contrast to the cellular CRE elements, the reiterated HTLV-I 21-bp repeats (normally considered as viral CRE elements) and the SRE exerted little or no basal activity (Fig. 1B , lanes 1, 2, 11 and 12; and Fig. 1C , lanes 2 and 7 compared to lane 1). The AP1 and Sp1 sites were moderately active (Fig. 1B , lanes 5–8 and Fig. 1C , lanes 4 and 5). Hence for basal expression in the context of the HTLV-I TATAA promoter, κB, CRE > AP1, Sp1 >> 21 bp, SRE. Figure 1 Relative responsiveness of enhancers to Tax in HeLa cells. ( A ) CAT reporter plasmid. Each plasmid contains two copies of enhancer elements (21-bp repeats, CRE, AP1, Sp1, κB and SRE) and one copy of HTLV-I minimal promoter (HTLV TATAA). The enhancer (Enh.) sequences are shown in green. ( B ) A representative example of CAT assay. Increasing amounts (5 to 10 μg) of p21-HTLV-CAT (lanes 1 and 2), pCRE-HTLV-CAT (lanes 3 and 4), pAP1-HTLV-CAT (lanes 5 and 6), pSP1-HTLV-CAT (lanes 7 and 8), pKB-HTLV-CAT (lanes 9 and 10) and pSRE-HTLV-CAT (lanes 11 and 12) were transfected into HeLa cells. CAT assays were performed 48 h after transfection. AcCM: acetyl chloramphenicol. CM: chloramphenicol. ( C ) Basal transcriptional activities of enhancer elements. Five microgram of plasmids containing the HTLV TATAA alone (pHTLV-CAT; column 1) or the indicated enhancer elements (columns 2 to 7) were transfected into HeLa cells and the relative CAT activities were compared. CAT activity from pKB-HTLV-CAT-transfected HeLa cells was taken as 100% (lane 6). ( D ) Tax-dependent transcriptional activities of enhancer elements. The same plasmids as in C plus 1 μg of Tax-expressing plasmid pIEX were co-transfected into HeLa cells and the CAT assays were performed. Fold activation in the presence of Tax versus in the absence of Tax was calculated and compared. All CAT results are representative of three independent experiments. When the reporters were tested in the presence of Tax, a different pattern emerged. Transcription from the 21-bp repeats was stimulated approximately 70-fold (Fig. 1D , lane 2 compared to lane 1) while that from the Sp1 site, not prototypically known to be responsive to Tax, was not activated significantly over the activity of the HTLV-I minimal promoter (Fig. 1D , lane 5 compared to lane 1). All other responses to Tax were markedly weaker than that seen from the 21-bp repeats. Hence, for all practical purposes, only a duplicated 21-bp repeat in the context of isolated placement upstream of an authentic HTLV-I minimal TATAA box could be regarded as significantly Tax-responsive in HeLa cells. We repeated the experiments in Jurkat T lymphocytes and obtained similar results (Fig. 2 ). Thus, while the κB and CRE enhancers displayed the highest activities in the absence of Tax (Fig. 2B , lanes 3 and 6 compared to lanes 5, 4, 2, 1, and 7), only the 21-bp repeats were highly responsive to Tax (Fig. 2A , lanes 1 and 2; Fig. 2C , lane 2). Our results from HeLa and Jurkat cells consistently support the preferential activation of the 21-bp repeats by Tax. Figure 2 Relative responsiveness of enhancers to Tax in JPX9 cells. ( A ) A representative example of CAT assay. Tax-expressing plasmid pIEX (1 μg) and increasing amounts (0.5 to 1 μg) of p21-HTLV-CAT (lanes 1 and 2), pCRE-HTLV-CAT (lanes 3 and 4), pAP1-HTLV-CAT (lanes 5 and 6), pSP1-HTLV-CAT (lanes 7 and 8), pKB-HTLV-CAT (lanes 9 and 10) and pSRE-HTLV-CAT (lanes 11 and 12) were transfected into Jurkat cells. CAT assays were performed 48 h after transfection. AcCM: acetyl chloramphenicol. CM: chloramphenicol. ( B ) Basal transcriptional activities of enhancer elements. One microgram of plasmids containing the HTLV TATAA alone (pHTLV-CAT; column 1) or the indicated enhancer elements (columns 2 to 7) were transfected into Jurkat cells and the relative CAT activities were compared. CAT activity from pKB-HTLV-CAT-transfected Jurkat cells was taken as 100% (lane 6). ( D ) Tax-dependent transcriptional activities of enhancer elements. The same plasmids as in C plus 1 μg of Tax-expressing plasmid pIEX were co-transfected into Jurkat cells and the CAT assays were performed. Fold activation in the presence of Tax versus in the absence of Tax was calculated and compared. All CAT results are representative of three independent experiments. Multiple activation surfaces are configured in Tax In Fig. 1D , the 21-bp repeats were activated by Tax >75 fold, while κB and SRE motifs were activated five and three fold, respectively. The low activation of the latter motifs, although comparatively less significant than that from the 21 bp elements, was real and reproducible. To further understand how Tax works, we wondered whether the different magnitudes of activation were due to quantitative or qualitative differences in protein-protein interaction. To address this question, we examined the separate responses of the three motifs to a battery of Tax mutants. Previously we had characterized 47 mutations in Tax that affect transcriptional activity [ 26 ]. Here we selected 10 of these Tax mutants to shed light on the discrete surfaces used by Tax to mediate effects on 21-bp repeats, κB and SRE. All mutants were expressed to comparable levels in HeLa cells (data not shown). Their relative activities on 21-bp repeats, κB and SRE were assessed (Fig. 3 ). Figure 3 Differential activities of Tax mutants on 21-bp repeats ( A ), κB ( B ), and SRE ( C ) motifs. One microgram of plasmid expressing the indicated Tax mutants plus 5 μg of p21-HTLV-CAT, pKB-HTLV-CAT or pSRE-HTLV-CAT was individually transfected into HeLa cells. CAT activity from wild type Tax-transfected cells (lane 1) was taken as 100%. Based on percentage of activation relative to wild type Tax, we saw three patterns of mutant activity for 21 bp, κB and SRE (Fig. 3 ). Hence, the activation domain mutant Tax L320G [ 10 ] and the zinc finger mutant Tax H52Q [ 26 ] were defective in activating either 21-bp repeats or SRE, but were fully competent for κB (Fig. 3 , lanes 4 and 10). By contrast, the N-terminal mutant Tax Δ3–6 and the point mutant Tax S258A activated 21-bp repeats and SRE well, but did not activate κB (Fig. 3 , lanes 2 and 7). Additionally, mutants Tax Δ94–114, Tax S150A and Tax Δ337–353 were active on all three motifs (Fig. 3 , lanes 5, 6 and 11), while neither Tax Δ2–58, Tax Δ 284–353 nor Tax L296G (Fig. 3 , lanes 3, 8 and 9) activated any of the motifs. These non-identical patterns suggest that Tax may use different contact surfaces to target factors docked at the 21-bp repeats, κB or SRE. We note some similarity in the Tax mutant activity profiles for the 21-bp repeats and SRE suggesting that overlapping surfaces may be utilized. Amongst bZIP factors, CREB is specifically preferred by Tax Tax activates the HTLV-I LTR through the viral 21-bp repeats [ 7 - 9 ]. When compared to κB and SRE, the activation of 21-bp repeats by Tax is particularly effective (Fig. 1 and Fig. 2 ) and, based on mutant profiles (Fig. 3A ), relies upon unique structural surfaces. Previously, it has been proposed that bZIP cellular transcription factors including CREB [ 9 , 27 , 28 ], ATF4 [ 29 , 30 ] and c-Jun [ 31 ] play roles in Tax activation of 21-bp repeats. However, the relative contribution of these bZIP factors has not been compared directly in the same experimental setting. Furthermore, it remains undetermined whether additional newly identified bZIP proteins may also participate in Tax activation of 21-bp repeats. We next used dominant-negative proteins to assess the contributory roles of different bZIP transcription factors on Tax-dependent activation. We employed several well-documented dominant-negative inhibitors of CREB and Jun proteins including KCREB [ 32 ], A-CREB [ 33 ], A-Fos [ 34 ] and TAM67 [ 35 ]. In addition, we constructed dominant-negative versions of ATF4 and LZIP [ 36 ] using the strategies suggested by Vinson et al. [ 37 ]. The dominant-inhibitory activities of the latter two proteins A-ATF4 and A-LZIP were verified using electrophoretic mobility shift assay and CAT reporter assay (data not shown). We interrogated these dominant negative bZIP proteins for inhibition of Tax activation of HTLV-I LTR (Fig. 4A ). All, KCREB, A-CREB, A-ATF4 and TAM67, suppressed Tax activation in a dose-dependent manner (Fig. 4A , lanes 3–10 compared to lane 2). However, different dominant negative inhibitors constructed to the same protein using different strategies might have different potencies. For example, KCREB contains a mutation of a single amino acid in the CREB DNA-binding domain [ 32 ], whereas A-CREB was constructed by fusing a designed acidic amphipathic extension onto the N terminus of the CREB leucine zipper region [ 33 ]. Differential inhibitory effects of KCREB and A-CREB were observed (Fig. 4A , lanes 3–6). In light of this, we quantitated and compared the inhibitory activities of dominant negative proteins all constructed using the same strategy (Fig. 4B ). Since NFκB is not involved in Tax activation of HTLV-I LTR, we included a dominant negative form of IKKβ, IKKβ DN, as a neutral control (Fig. 4B , group 7). When we compared four dominant negative bZIP proteins, A-CREB, A-LZIP, A-Fos and A-ATF4, constructed using the identical molecular strategy, we observed the most dramatic suppression of Tax activation of HTLV-I LTR with A-CREB (Fig. 4B , group 3, red column). The second most significant reduction in activity was seen with A-LZIP [ 36 ] (Fig. 4B , group 6, red column). Thus, although several bZIP proteins can redundantly serve to mediate Tax-activation of the LTR, a clear preference for CREB is revealed by our assay. Figure 4 Specific preference for CREB by Tax. ( A ) An example of CAT assay. HeLa cells were transfected with pU3RCAT alone (lane 1), pU3RCAT plus Tax expression plasmid pIEX (lane 2) or pU3RCAT plus pIEX plus increasing amounts (5 to 10 μg) of plasmids expressing the indicated dominant-negative proteins (lanes 3–10). D-Threo-[dichloroacetyl-1- 14 C]-chloramphenicol was as used as substrate in the CAT assay. ( B ) Influence of dominant-negative proteins on Tax activation. The cells received pU3RCAT (red) or pKB-SV40-CAT (blue) only (group 1), pU3RCAT/pKB-SV40-CAT plus Tax-expressing plasmid pIEX (group 2) or pU3RCAT/pKB-SV40-CAT plus pIEX plus plasmids expressing the indicated dominant-negative proteins. The empty vector was used to normalize the amount of plasmids given to each group of cells. DN: dominant-negative. To verify the specificity of dominant negative effects, we also tested the activities of dominant negative proteins on an NFκB-dependent reporter (Fig. 4B , blue columns). Noticeably, none of the dominant negative bZIP proteins had an effect on Tax activation of NFκB (Fig. 4B , groups 3–6 compared to group 2, blue columns). In contrast, the expression of IKKβ DN led to more than 50% suppression of NFκB activity (Fig. 4B , group 7, blue column). These results ruled out the possibility that A-CREB, A-ATF4, A-Fos and A-LZIP might non-specifically inhibit transcription. Functional significance of the HTLV-I TATAA element to transcriptional activation by Tax In the course of our analyses, we noted that Tax can activate the HTLV-I minimal TATAA-promoter without any known enhancer element by approximately 4-fold (Fig. 1D , lane 1). This responsiveness of the HTLV-I minimal promoter is compatible with the concept that the core promoter can also be an important determinant of transcriptional specificity [ 2 ]. We next asked whether all TATAA-elements are recognized by Tax in the same way for purposes of activated transcription. Hence, we constructed reporter plasmids that contain two 21-bp repeats and a minimal TATAA promoter from HTLV-I, HIV-1 or SV40 (Fig. 5A ). Since the TATAA promoters were all placed within the same context, we consider this a valid comparison of their relative responsiveness to Tax activation. Figure 5 Tax preferentially activates the HTLV-I minimal TATAA promoter. ( A ) CAT reporter plasmid. Each plasmid contains two 21-bp repeats and one copy of minimal promoter (TATAA) from HTLV-I, HIV-1 and SV40. The minimal promoter sequences are shown in blue. ( B ) A representative example of CAT assay. The cells received 0, 0.5 and 1 μg of Tax-expressing plasmid pIEX and 5 μg of the indicated CAT reporter constructs (p21-HTLV-CAT, p21-HIV-CAT and p21-SV40-CAT). ( C, D ) Basal and Tax-induced transcriptional activities. HeLa cells were co-transfected with 5 μg of the indicated CAT reporter plasmids (p21-HTLV-CAT, p21-HIV-CAT and p21-SV40-CAT) plus 0.5 μg of pCMV empty vector (w/o Tax) or pIEX (w/ Tax). Basal CAT activity from p21-SV40-CAT-transfected cells was taken as 100% ( C , column 3). While the basal activities of HIV-1 and SV40 minimal promoters were measurably greater than that from HTLV-I (Fig. 5C ), replacement of the HTLV-I TATAA with the counterpart element from either HIV-1 or SV40 led to a significant reduction in Tax responsiveness (Fig. 5B , lanes 4–9; and Fig. 5D ). To further verify the importance of the TATAA-promoter, we asked the same question using a different approach. Above, Tax was recruited presumably to the downstream TATAA-box via factors bound to the HTLV-1 21bp repeats (see Fig. 5A ). We next investigated whether the same conclusion could be established if a Gal4 DNA-binding domain-Tax fusion protein (Gal4-Tax) was delivered to downstream TATAA element by tethering to upstream Gal4-binding sites (see Fig. 6A for reporter schematic). For this assay, we tested the HTLV-I, the HIV-1, and the E1b TATAA-elements. Consistent with the results from the 21 bp-TATAA experiments (Fig. 5 ), Gal4-Tax activated most strongly the HTLV-I TATAA element (Fig. 6B , lane 9 and Fig. 6D , group 3) and was minimally potent for the adenoviral E1b promoter (Fig. 6B , lane 7 and Fig. 6D , group 1). As a control for Gal4-Tax, we checked in parallel the activity of the artificial Gal4-VP16 activator. In contrast with Gal4-Tax, Gal4-VP16 showed no preference for the various TATAA elements (Fig. 6B , lanes 4–6 and Fig. 6D ). Thus, two lines of evidence here support that the HTLV-I TATAA promoter is an additional Tax-specific responsive element. Figure 6 DNA-tethered Tax is specifically active on the HTLV-I minimal promoter. ( A ) CAT reporter plasmid. Each plasmid contains five tandem copies of Gal4-binding sites and one copy of minimal promoter (TATAA) from adenovirus E1b, HIV-1 and HTLV-I. The minimal promoter sequences are shown in blue. ( B ) A representative example of CAT assay. The cells were co-transfected with 2 μg of a Gal4DB plasmid (pM vector alone for lanes 1–3, pGal4-VP16 for lanes 4–6, and pGal4-Tax for lanes 7–9) and 5 μg of a CAT reporter construct (pG5-E1B-CAT for lanes 1, 4 and 7; pG5-HIV-CAT for lanes 2, 5 and 8; and pG5-HTLV-CAT for lanes 3, 6 and 9). ( C, D ) Basal and activated transcriptional activities. HeLa cells were co-transfected with 5 μg of the indicated CAT reporter plasmids (pG5-E1B-CAT, pG5-HIV-CAT and pG5-HTLV-CAT) plus 2 μg of pM empty vector ( C ), pGal4-VP16 ( D , blue) or pGal4-Tax ( D , yellow). Basal CAT activity from pG5-HIV-CAT-transfected cells was taken as 100% ( C , column 2). Evidence for Tax activity after assembly of an initiation complex Artificial recruitment of TBP to some higher eukaryotic promoters bypasses transcriptional activation by a DNA-tethered activator [ 21 - 24 ]. When observed at such promoters, this finding is evident that those activators act mechanistically to enhance TBP recruitment to the TATAA box. For general transcriptional activation, additional events subsequent to TBP recruitment are also known to be functionally critical [ 21 - 23 , 25 ]. To date, it remains unclear whether Tax works transcriptionally through a mechanism solely to recruit TBP or whether there are additional mechanistic implications after TBP is recruited to the TATAA-element. To investigate the mechanism(s) of Tax function with respect to TBP recruitment, we constructed a series of reporter plasmids (Fig. 7A ) with two copies of 21-bp repeat, five copies of Gal4-binding sites and a minimal TATAA sequence from one of four viral promoters (HTLV-I, HIV-1, SV40 and E1b). We artificially delivered TBP to each promoter by provision of Gal4-TBP, and we asked whether Tax has an additional transcriptional effect which is independent of TBP-recruitment to the TATAA-element. If Tax were to serve only for TBP-recruitment, then when TBP is tethered to the TATAA via Gal4-TBP one should expect to see no transcriptional enhancement from Tax. Provocatively, for both the HTLV-I and HIV-1 TATAA elements, Tax stimulated reporter expression greatly over that already achieved with Gal4-TBP (Fig. 7 , groups 1 and 2). Consistent with above findings, the SV40 and E1b TATAA elements appear to be transcriptionally rate-limited by TBP recruitment, and Tax has minimal activity on these promoters. However, the findings from the HTLV-I and the HIV-1 reporters provide evidence that more than simply accelerating TBP recruitment Tax can serve transcriptional function(s) subsequent to TBP (TFIID) assembly at the core promoter. This is the first time that Tax has been shown to have a role subsequent to transcriptional initiation complex formation at the promoter. Figure 7 Tax further activates a promoter with DNA-tethered TBP. ( A ) CAT reporter plasmid. Each plasmid contains two copies of 21-bp repeat, five copies of Gal4-binding sites and one copy of minimal promoter (TATAA) from adenovirus HTLV-I, HIV-1, SV40 and adenovirus E1b. ( B ) CAT assay. HeLa cells were co-transfected with 5 μg of the indicated CAT reporter plasmids (p21-G5-HTLV-CAT, p21-G5-HIV-CAT, p21-G5-SV40-CAT and p21-G5-E1B-CAT) and 2 μg of pGal4-TBP (yellow) or 2 μg of pIEX (Tax; pink) or 2 μg of pGal4-TBP plus 2 μg of pIEX (Gal4-TBP + Tax; blue). Basal CAT activity from cells transfected with pGal4-TBP plus p21-G5-E1B-CAT was taken as 100% (group 4, yellow). Discussion Here, we have delineated functional requirements for both the TATAA promoter and the 21-bp enhancer elements in HTLV-I Tax mediated activation of the viral LTR. To date Tax has been considered solely to initiate transcription. Our study shows for the first time that Tax has a transcriptional role after assembly of an initiation complex at the promoter. Preferential requirements for 21-bp repeats, CREB, and the HTLV-I TATAA box HTLV-I is etiologically associated with adult T-cell leukemia [ 38 , 39 ]. Expression of Tax leads to immortalization of T lymphocytes [ 40 - 42 ] and transformation of rat fibroblasts [ 43 , 44 ]. Tax is a transcriptional activator that can interact pleiotropically with several different enhancers. In addition to the HTLV-I 21-bp repeats, κB and SRE elements can also mediate Tax activation [ 4 - 6 ]. Amongst these three enhancers, it is clear that the viral 21-bp repeats are the most highly responsive to Tax-activation (Fig. 1D ). However, data elsewhere have raised questions as to the identity of the 21-bp binding bZIP factor which is best used to mediate Tax activation [ 30 ]. In direct comparisons, we have used matched A-CREB, A-Jun, A-ATF4 and A-LZIP dominant negative mutants to ask which bZIP factor is most contributory to Tax activation. In our cell system, a novel bZIP factor called LZIP [ 36 ] can apparently participate in LTR transcription; however, for Tax activation CREB is preferred over ATF4 or c-Jun (Fig. 4 ). Beyond the requirement for the 21-bp enhancer, our experiments revealed that the HTLV-I TATAA is also specifically preferred by Tax (Fig. 5 and Fig. 6 ). This finding is consistent with the general notion that core promoters can contribute specificity to transcriptional regulation [ 2 ]. Indeed, core promoter preference by other cellular and viral activators such as Sp1, VP16 and Tat have been documented previously [ 45 - 47 ]. However, the reasons underlying core promoter preferences are poorly understood. TAFs have been suggested to be responsible for the core promoter selectivity of some activators [ 48 - 50 ]. In this vein, the interaction of Tax with TBP [ 11 ] and TBP-associated factors such as TAF II 28 [ 51 ] might provide mechanistic explanations. Roles of Tax subsequent to TBP recruitment A provocative notion which emerges from our study is that Tax can further activate a promoter at which TBP has already been artificially tethered (Fig. 7 ). Experiments in yeast and mammalian cells indicate that many genes can be activated through artificial recruitment of TBP and other components of the basal transcription machinery to their promoters [ 52 , 53 ]. In yeast, artificial recruitment of TBP bypasses the effect of DNA-tethered activators whereas the activators fail to activate transcription when physically fused to components of the basal transcription machinery [ 54 ]. This and other lines of evidence support the notion that activator-dependent recruitment of TBP and basal transcription machinery is a major mechanism for transcriptional activation in yeast cells [ 54 , 55 ]. In contrast, artificial recruitment of TBP to mammalian promoters has not yet been extensively studied. Among the few promoters examined, some such as the ones from E1b and thymidine kinase genes can be fully activated by artificially recruited TBP, while others such as HIV-1 and c-fos promoters are stimulated weakly [ 21 - 25 ]. On the other hand, some activators such as VP16, E1A, Tat, E2F1 and IE2 work synergistically with artificially recruited TBP, while others such as Sp1 cannot further enhance the activity of DNA-tethered TBP [ 21 , 22 ]. Thus, artificial recruitment of TBP might insufficiently activate transcription in mammalian cells and different activators might function at different steps with respect to TBP recruitment. Our results indicate that DNA-bound TBP can activate HTLV-I LTR only weakly, but its activity is further enhanced by Tax (Fig. 6 ). While such experimental results do not exclude that under physiological circumstances the primary function of Tax may be to enhance initiation complex formation (i.e. TBP-recruitment), they do indicate that Tax has an additional transcriptional activity that extends to phases after transcriptional initiation. Currently, we do not know whether this is at the step of promoter clearance, transcriptional elongation, or some other processes. However, we do believe that Tax should be added to the list of mammalian activators that can function at steps subsequent to TBP recruitment [ 21 - 25 ]. All the transcriptional assays in the present study were based on transiently transfected reporters. We noted that transiently transfected and stably integrated promoters might behave differently [ 24 , 56 ]. Obviously, chromatin structure and copy numbers can account for significant differences [ 56 , 57 ]. Future experiments are required to verify whether the observations established here also hold for stably integrated HTLV-I LTRs. Methods Plasmids Chloramphenicol acetyltransferase (CAT) reporter plasmid pG5CAT was from Clontech. CAT plasmid pU3RCAT containing the HTLV-I LTR has been previously described [ 13 ]. Other CAT plasmids were derived from pCAT-basic (Promega). For each construct, one copy of a minimal promoter and two copies of an enhancer were chemically synthesized and cloned into pCAT-basic. For example, pCRE-HTLV-CAT contains two copies of canonical CRE motif plus one copy of HTLV-I minimal promoter (Fig. 1A ). Five copies of Gal4-binding sites as in pG5CAT were also inserted in some reporters. All constructs have the same spacing between the TATAA box and the CAT open reading frame (44 bp) or between the enhancer and the TATAA box (23 bp). Sequences of canonical CRE, Sp1, AP1 and κB motifs in the reporter plasmids have been described [ 36 , 58 , 59 ]. HTLV-I 21-bp repeats and serum response element (SRE) in the plasmids were derived from the following synthetic oligonucleotides: 21-bp repeats, 5'-AGCTTAGGCC CTGACGTGTCCCCCTGGATCCTAGGCCCTGACGTGTCCCCCTA-3' and 5'-AGCTTAG GGGGACACGTCAGGGCCTAGGATCCAGGGGGACACGTCAGGGCCTA-3'; SRE, 5'-AGCTACCATATTAGGATCCATATTAGGT-3' and 5'-AGCTACCTAATATGGATCCTAATATGGT-3'. Sequences of the minimal promoter elements from HTLV-I, HIV-1, SV40 and adenoviral E1b have been described [ 60 ]. The SV40 early promoter naturally used for expression of the viral T/t antigens was used. Expression plasmids for wild type and mutant Tax have been described elsewhere [ 26 , 61 ]. pIEX is a Tax expression vector driven by cytomegalovirus IE promoter. Tax mutants are indicated by the amino acid to be changed, the position of the residue, and the replacement amino acid (e.g. Tax S150A). Amino acids that were removed in mutants are indicated as in Tax Δ3–6. Expression vector pM for Gal4 DNA binding domain (Gal4DB; amino acids 1–147) was from Clontech. Tax, human TBP and the activation domain of VP16 fused to Gal4DB were designated Gal4-Tax, Gal4-TBP and Gal4-VP16, respectively. Expression plasmids for Gal4-Tax and Gal4-TBP have been described [ 10 , 21 ]. Expression plasmid for Gal4-VP16 was from Clontech. Expression plasmid pRSV-KCREB for the dominant-negative CREB protein KCREB [ 32 ] was kindly provided by Dr. Richard Goodman. Expression plasmids pCMV-ACREB and pCMV-AFOS for dominant-negative CREB and AP1 proteins A-CREB [ 33 ] and A-Fos [ 34 ] were gifts from Dr. Charles Vinson. Expression plasmid pCMV-TAM67 for dominant-negative c-Jun protein TAM67 [ 35 ] was from Dr. Michael Birrer. Expression plasmids pCMV-AATF4 and pCMV-ALZIP for dominant-negative ATF4 and LZIP proteins A-ATF4 and A-LZIP were derived from pCMV500 provided by Dr. Charles Vinson [ 33 , 37 ]. A-ATF4 contains 304–352 amino acids of human ATF4 and A-LZIP contains 175–223 amino acids of human LZIP. A-ATF4 and A-LZIP can specifically and dominantly inhibit the CRE-binding and CRE-activating activities of ATF4 and LZIP, respectively, in electrophoretic mobility shift assay and CAT reporter assay (data not shown). Expression plasmid for dominant-negative IKKβ (IKKβ DN) was a gift from Dr. Michael Karin [ 62 ]. Reporter assay HeLa cells were grown in Dulbecco's modified Eagle's medium supplemented with fetal calf serum and antibiotics, seeded at 5 × 10 5 cells/well into six-well culture plates and transfected using calcium phosphate method as described [ 13 ]. Jurkat cells were cultured in RPMI 1640 medium and transfected by FUGENE 6 reagents (Roche). CAT activity was assayed as previously described [ 63 ]. Briefly, transfected cells were harvested and lysed by freezing and thawing. Protein concentration of clarified lysates was determined by Bradford reagent (Bio-Rad). Equal amounts of lysates were mixed with 14 C-labeled chloramphenicol (Amersham) and acetyl coenzyme A (Calbiochem) for CAT reaction. CAT activities were detected using thin-layer chromatography and quantified by phosphorimager (Molecular Dynamics). For transfection of cells, each well received the same dose of plasmids. The empty vector or pUC19 was added to compensate for the different amounts of plasmids when necessary. Competing interests None declared.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC509288.xml
545069	Computer-aided DSM-IV-diagnostics – acceptance, use and perceived usefulness in relation to users' learning styles	Background CDSS (computerized decision support system) for medical diagnostics have been studied for long. This study was undertaken to investigate how different preferences of Learning Styles (LS) of psychiatrists might affect acceptance , use and perceived usefulness of a CDSS for diagnostics in psychiatry. Methods 49 psychiatrists (specialists and non-specialists) from 3 different clinics volunteered to participate in this study and to use the CDSS to diagnose a paper-based case (based on a real patient). LS, attitudes to CDSS and complementary data were obtained via questionnaires and interviews. To facilitate the study, a special version of the CDSS was created, which automatically could log interaction details. Results The LS preferences (according to Kolb) of the 49 physicians turned out as follows: 37% were Assimilating, 31% Converging, 27% Accommodating and 6% Diverging. The CDSS under study seemed to favor psychiatrists with abstract conceptualization information perceiving mode (Assimilating and Converging learning styles). A correlation between learning styles preferences and computer skill was found. Positive attitude to computer-aided diagnostics and learning styles preferences was also found to correlate. Using the CDSS, the specialists produced only 1 correct diagnosis and the non-specialists 2 correct diagnoses (median values) as compared to the three predetermined correct diagnoses of the actual case. Only 10% had all three diagnoses correct, 41 % two correct, 47 % one correct and 2 % had no correct diagnose at all. Conclusion Our results indicate that the use of CDSS does not guarantee correct diagnosis and that LS might influence the results. Future research should focus on the possibility to create systems open to individuals with different LS preferences and possibility to create CDSS adapted to the level of expertise of the user.	Background Different types of decision support (DS) methods have been used in medicine for long. Computerized decision support systems (CDSS) including so-called "expert systems" can be used in for example interpretation of medical images, medical diagnostics or other areas. Examples are MYCIN for antibiotic treatments [ 1 ] and deDombal's system for acute stomach pain [ 2 ]. In Psychiatry, the DIAGNO system by Spitzer and Endicott was described 1968 as a computer program that simulated a DSM-1 diagnosis based on data from the psychiatric status schedule [ 3 ]. The CATEGO and DETOX systems are other examples [ 4 , 5 ]. For a more general reading on CDSS adoption in medical practice, see Fieschi et al [ 6 ] and for different approaches used to implement computers as diagnostic aids in medical decision making see for example Engle Jr. [ 7 ] and Miller [ 8 ]. Another study has been published by Berner and colleagues [ 9 ] who compared performance scores between four different diagnostic decision support systems. Various models, or mode of actions, of DS exists including textual guidelines based on if-then-else strategies that forces the decision maker to make decisions in a logical and sequential manner; more advanced systems using fuzzy logic; neural networks; and systems using so called Artificial Intelligence. Many DS systems have been developed with aims to enable more accurate, consistent diagnosis and faster diagnostic procedures. CDSS have been applied in many medical disciplines, but have also been discussed in terms of e.g. reliability, usefulness, and user-acceptance. For example, Lu et al [ 10 ] found that the willingness to use CDSS rely heavily on preferences and perceived usefulness. Another example is Ridderikhoff and van Herk [ 11 ] who stated that although physicians indicate a need for diagnostic support, medical diagnostic support systems are not in widespread use. Miller [ 8 ] pointed out that it is misleading regarding the state of the art of these systems to just focus on the lack of widespread use. Miller's bibliography of systems from 1954 to 1993 convinced him that diagnostic systems nowadays can be seen as ubiquitous and "The prospects for adoption of large-scale diagnostic systems are better now than ever before, due to enthusiasm for implementation of the electronic medical record in academic, commercial, and primary care settings." Friedman et al [ 12 ] indicated that CDSS should be able to improve healthcare quality by providing accurate, useful and timely diagnostic information to clinicians and that most studies have emphasized the accuracy of the computer system alone without placing clinicians in the role as direct users. In exploring the extent to which CDSS might improve the diagnostic capability of clinicians, the success rate varied between different groups with different training levels. Larger improvements were observed for students than for residents and faculty. They concluded that "hands on" use of CDSS might influence diagnostic reasoning of clinicians. Regarding the decision procedure of a real expert, there are a number of theories. Many of these points out that there are differences between the decision process of an expert and a beginner (novice) [ 13 , 14 ]. It is unclear if the perceived usefulness of CDSS only is due to the DS model itself, or if the design and management of the computerized system is as important. Furthermore, there have been discussions about the design of computer systems, and how they might suit different users with different user-characteristics. Allen [ 15 ] argues that individual differences between users of information systems might influence search performance. Different types of cognitive resources such as topic knowledge, search skills, cognitive abilities, cognitive styles and learning styles have been shown to be related to a variety of search tactics and to tendencies to use certain information system features [ 16 ]. Learning styles The concept of learning style (LS) might be regarded as an important characteristic and independent variable when individual differences in perceiving and processing information are investigated. A deep understanding of the user, his tasks and his environment is required to design a well-accepted (useful) computer program. Learning styles and its importance for users of computer systems has been demonstrated in various areas, for example in Internet use [ 17 ], web-based teaching [ 18 ], interactive multimedia environment [ 19 ], efficacy of computer training methods [ 20 ], hypertext environments [ 21 ], and in interactive learning systems [ 22 ]. There are a number of learning style models and inventories [ 23 - 27 ]. Claxton and Murrell [ 28 ] systematized the various learning style models based on Curry's [ 29 ] previous work on learning style constructs. The Kolb LS model [ 25 ], classified by Claxton and Murrell [ 28 ] as an information-processing model, has recently been further developed. The Kolb LS model has been widely applied during the years [ 30 ] and the latest version of the Kolb Learning Style Inventory instrument, is version 3 (LSI 3) [ 31 , 32 ]. This is a model of experience-based learning where all processes of the model are vital for the learning result. According to this model, the user/learner moves around the four modes in a circular direction (Figure 1 ). First there is an actual concrete learning experience. Second, the learner reflects on this experience. Third, the learner conceptualizes his/her observations and/or reflections into abstract theories or ideas. Fourth, the learner tests the theories or ideas by active experimentation. In this model, there are two processes for perceiving information: concrete experience mode and abstract conceptualization mode and two processes for processing experience into learning: active experimentation mode and reflective observation mode. Figure 1 Structural Dimensions Underlying Learning Styles (After Kolb 1984) These four processes combine into four learning styles: Converging (abstract conceptualization mode and active experimentation mode), Accommodating (concrete experience mode and active experimentation mode), Diverging (concrete experience mode and reflective observation mode) and Assimilating (abstract conceptualization mode and reflective observation mode). Diverging learning style is associated with value generating skills: building relationships, helping others and sense making (reasoning). Assimilating learning style is associated with thinking skills: information gathering, information analysis and theory building. Converging learning style is associated with decision skills: quantitative analysis, technical device use and formulation of goals. Accommodating learning style is associated with action skills: leadership, initiative and action [ 25 , 33 ]. Learning styles of the Kolb model are not only associated with skill, but also with adaptivity and flexibility concerning management of different situations. Curry [ 34 ] points out that a learning style is different from ability, strategy and tactic. Styles might be observed across content domains, abilities, personalities and interpersonal behaviors and they are measured in terms of typical performance. According to Curry, learning style is spontaneously demonstrated without conscious awareness or choice across a wide variety of situations with similar requirements. Strategies, in contrast are the result of conscious decisions and tactics are specific observable activities in specific performance situations [ 34 ]. Diagnostics in psychiatry In Psychiatry, the diagnostic process is mainly based on medical history, which is not often to be confirmed by lab tests or physical examinations. This leaves it more open for inter-personal differences in diagnostics, which might be a serious problem. To facilitate the diagnostic procedures, and increase certainty and consistency in Psychiatry, a system called DSM has been created [ 35 ]. The DSM (Diagnostic Statistical Manual) system is a kind of information management system, an instrument that sort symptoms and handles them according to what the user judges as important symptoms. DSM is designed to assist the user by making the system criteria-based and by multi-axial descriptions, create a conflict-free base and thereby increase the reliability of the diagnosis. Currently, DSM version 4 (DSM-IV) is most used. The multi-axial DSM-system is based on the following reasoning: 1. Which symptoms have currently forced the patient to seek help? 2. How does the patient's overall pattern of experiences and behavior, compared with what is generally expected in the patients socio-cultural milieu, look like? 3. Are there any somatic diseases, which have to be attended to? 4. Have there occurred any stressful events in the patients' life along with the initial symptoms? 5. How serious are the problems just now, how is the patient functioning? However, even DSM is not easily implemented in all situations, and therefore SCID, Structured Clinical Interview for DSM-diagnoses was created to further increase the reliability in psychiatric diagnostics [ 36 ]. SCID is a semi-structured interview instrument for DSM-IV-diagnoses, and is widely used in psychiatry internationally. DSM training for physicians has been going on since 20 years during the Psychiatric course (9th semester) at Karolinska Institutet. The DSM-training is integrated in the course and the amount of time spent on DSM is approximately 4 hours. To our knowledge, there is no SCID-training in the Psychiatric course. However, in the training to become a specialist in Psychiatry at Karolinska Institutet, the DSM-framework is always used in teaching diagnostics and the amount of formal SCID-training is about 8 hours. Complementary to this, physicians becoming specialists in psychiatry are further trained in how to use SCID1 (axis 1 in DSM-IV) during their clinical training. During a SCID interview, "jigsaw puzzle bits" are gathered, where DSM functions as a method to sort and put together these puzzle bits to known clinical syndromes. DSM is criteria based. For each criteria the users have to consider if the patient's symptom reach clinical significance so that the criteria can be regarded as fulfilled or not. There is some help for the interviewer in the DSM system in the form of a "decision tree" for axis 1 diagnosis. Computer support for SCID CB-SCID1 (Computer-Based SCID for axis 1 diagnostics) is a software program that is reported to have advantages as compared to the ordinary SCID-interview (according to the CB-SCID1-manual) [ 37 ]. The program handles most administration tasks for e.g. summing up of fulfilled criteria and also presents an overview of noted diagnoses. The order of questions and some control of possible conflicting diagnoses are also taken care of by the program. The "decision tree" mentioned above is integrated in the software. Objectives This study was undertaken to investigate and describe how different learning style preferences among psychiatrists might affect acceptance , use and perceived usefulness of the CDSS CB-SCID1 for DSM-IV-diagnostics. Methods Subjects A number of practicing Psychiatrists, working at three different clinics, with different degree of expertise were asked to participate in this study. A fourth clinic was invited to participate in the study but could not do this due to lack of time. Out of 67 invited physicians, 49 volunteered to participate in the study. Out of these, 31 were experts at a senior level (being registered as specialists in psychiatry), and 18 were non-specialists (physicians with a position in psychiatry but without a specialist exam in psychiatry). In this study the groups are called "specialist group" (experts) and "non-specialist group" (not experts), respectively. They were all asked to complete a questionnaire regarding learning style preferences and to use the CB-SCID1 computer program for diagnostics of a real patient case (described in text) collected from the DSM-IV Case Book [ 38 ]. To be able to relate the Learning Styles of the 49 physicians in the study to the general situation in Sweden, a random sample of 250 (out of 1900 practicing psychiatrists in Sweden) were asked to fill in the same Learning Style inventories as mentioned above. This part of the study was done by sending out a letter including details of the study, which was followed up by a second letter as a reminder some weeks after the first one. All data were kept unidentified. Survey instrument for LS The learning style preferences for all participating physicians were measured according to the Kolb model [ 25 ] using the Kolb Learning Style Inventory instrument, version 3 (LSI 3) [ 31 , 32 ]. This instrument (questionnaire) presents specific questions and statements, which the test person enters his personal views on. The responses entered are then used to categorize the LS preferences for the person under study [ 31 , 32 ]. CDSS under study The standardized terms and concepts of the DSM-system are the fundamentals of the CB-SCID1 system. CB-SCID1 uses logical inference of logical data (true, false), symbolically representing connections and dependency between components in the psychiatric knowledge base and presents questions according to the "paper" SCID-manual. CB-SCID1 takes care of the administration (for instance ordering of questions), correction possibility in criteria judgment, suggestion of answers according to previous in-data, summing up of fulfilled criteria, and also presentation of noted diagnoses. The system also handles some conflict control upon diagnoses. The user is asked to determine if various criteria are fulfilled or not and the system chooses how to go on, based on the user's input. If the number of fulfilled criteria reaches a certain level (according to DSM-IV) the program is automatically suggesting the corresponding DSM-IV diagnosis. The program is designed in a way that little training should be needed. CB-SCID1 also has a built-in, context sensitive help function in the consultation form, which put forward reminders and appropriate text information in tune with the decision problem at hand. The "assistance" from the system is based on the users input and also combined with the data driven rules derived from DSM-IV. The physicians were instructed to use the CB-SCID-1 program as a tool to find the correct DSM-IV diagnoses of the paper case, and use the system as if the case had been a real one. Data collection A special version of CB-SCID1, CB-SCID1_Log, was created, with a logging function that automatically stored a number of data in a separate log file, while the physician was using the program trying to make DSM-IV diagnosis on the patient case. The data logged (outcome log-file variables) were: • Total session time (total time spent in the CB-SCID1-program) • Total decision time (total time used to decide about the different criteria used in CB-SCID1) • Average decision time (mean time to decide about a criterion) • Total "non decision" time (time spent in the program not making decisions) • Total number of criteria judged (total number of criteria decided about) • Total number of diagnoses (total number (correct and incorrect), of proposed diagnoses by CB-SCID1) • Total number of correct diagnoses (total number of correct diagnosis according to the DSM-IV Case Book) • Total number of incorrect diagnoses (all other diagnoses proposed by CB-SCID1 and not correct according to the DSM-IV Case Book) • Ratio between correct diagnoses and proposed number of diagnoses • Total number of regretted criteria-judgments (total number of times the user clicks on the Regret-button in the CB-SCID1 while deciding about a criterion) • Total number of criteria judged unclear (total number of times using the Unclear-button in the CB-SCID1 while deciding about a criterion) • Sum of numbers of regretted criteria-judgments and unclear criteria judgments The CB-SCID1_Log system (identical to CB-SCID1 for the user) was installed at the office computers of the clinicians and a short oral introduction of the system functionality was given, explaining both the use of the system, its online help system, and the aim of the study. Paper case used The case used was picked from the DSM-IV Case Book [ 38 ]. The cases in this book are real, but unidentified, patients. These have been collected from a large number of clinicians (experts in particular areas of diagnosis and treatment). According to the Case Book, the recommended use of the cases is for example for researchers to assess the level of diagnostic expertise and the reliability with which members of their staff can make diagnostic assessments. A senior psychiatrist who is very experienced in DSM-IV and SCID training picked the actual case to be used. This specific case was chosen because it reflects multi-axial assessment (especially Axis 1 diagnoses), which was considered to be well suited for the CD-SCID1 program (which is aimed for Axis 1 diagnostics). The chosen case (called "Sickly" in the DSM-IV Case Book) is rather complex including three different diagnoses forcing the user to use the CB-SCID1 in full, taking decisions in problem areas like somatic problems, psychiatric problems and addiction problems. The correct Axis 1 diagnoses (that is the assessment of Clinical Disorders and Other conditions that may be a focus of clinical attention) were in the actual case: • Major Depressive Disorder, recurrent, mild • Somatization Disorder • Alcohol Dependence, in sustained full remission To create a more realistic situation, the correct diagnoses were not revealed to the participating 49 psychiatrists, who all were given the same case with the aim to study the variation amongst the physicians using the CB-SCID1. They were not told what kind of case it was, neither the name of it ("Sickly"), nor where it came from. Investigation procedure The investigation procedure was performed in four steps as follows: 1. General information : The project leader gave oral information at each clinic on a regular meeting for psychiatrists about the study 2. Individual information and questionnaire : Each participating physician was given further oral information about the study and was asked to fill in a form about gender, age, professional training, DSM/SCID training, computer skill and attitude towards computer-aided diagnostics. This was followed by filling in the Kolb Learning Style Inventory questionnaire LSI 3. 3. CB-SCID1 test : The physicians received oral instructions on how to use the computer program and that it was more or less self-instructive compared to normal "paper" SCID-training, before using CB-SCID1. They were then instructed to diagnose the patient case with the help of the CB-SCID1-system. Ten minutes were offered reading the patient case before starting the CDSS-system 4. Follow-up interview and questionnaire : A follow-up interview within a week from the first interview was done. There were open-ended questions about the pros and cons about the CB-SCID1 (pro and con categories were built on the basis of the answers content). Also structured questions using a four-graded scale were given about clinical interviewing skill (without computer aid), perceived usefulness of CB-SCID1 and computer anxiety, which all were graded using a four-graded scale. Statistical methods and analysis Structured questions, constructed by the authors, and graded on a four-grade scale were given in the pre-assessment survey about computer skill and attitude to computer-aided diagnostics. In the post-assessment survey, structured questions were given about computer anxiety, clinical interviewing skill (without computer aid) and perceived usefulness of CB-SCID1. The questions were put in a clear statement which they could agree to/not agree to in an ordered categorical scale (Strongly disagree = 1, Disagree = 2, Agree = 3, Strongly agree = 4). The subject areas asked about were well defined and familiar to the users, why standardized attitude scales were not used. An open-ended question about the pros and cons about the CB-SCID1 were also given in the post-assessment survey. The answers to this question were grouped into several categories. An analysis of correlations between the dimensions of LSI and the outcome log-file variables, as well as a comparison between specialists and non-specialists in this respect was also performed. Results were calculated as mean, standard deviation, median and lower – and upper quartile, where appropriate. Comparison between the two independent groups (specialists and non-specialists) was performed by the Mann-Whitney U Test and comparison between more than two independent groups (LS groups) was performed by the Kruskal-Wallis ANOVA by Ranks Test. Association between variables was calculated by Spearman Rank Order Correlations. Ethical approval All parts of this study have been approved by the ethical committee of Karolinska Institutet. Results General results All of the 49 psychiatrists volunteering to participate in the original study group fulfilled all phases of the study. The randomly sampled 250 psychiatrists had a response rate of 42% (95). Only 226 questionnaires could actually be sent out because 24 of the randomly chosen 250 psychiatrists could not be reached due to for example that they had moved abroad or retired. One of the 95 actual responses was incomplete. Demographic data of the study groups can be seen in Table 1 . Table 1 Demographic variables in study groups Variables Original study group (n = 49) Original study group – specialists (n = 31) Original study group – non-specialists (n = 18) Random sample group – specialists (Valid n = 93) Age (mean ± SD) Male 48 ± 9 (n = 27) Male 53 ± 7 (n = 14) Male 43 ± 8 (n = 13) Male 52 ± 8 (n = 46) Female 47 ± 11 (n = 22) Female 52 ± 7 (n = 17) Female 32 ± 6 (n = 5) Female 50 ± 8 (n = 47) DSM/SCID-training (hours, mean) Male 7 Male 10 Male 3 Male 13 Female 9 Female 11 Female 3 Female 10 The 49 physicians in the original study group were reporting a value of 3 for general computer skill and a value of 1 for computer anxiety, both median values on a four-graded scale (where 1 is very negative or very low and 4 is very positive or very high). There were no differences in computer skill between specialists and non-specialists. Computer skill median was 3 in the random sample group. The physicians were reporting to have a good clinical interviewing skill (without computer-aid), the median was 4 for specialists and 3 for non-specialists, which would predict high expected values on correct diagnosis and low values on incorrect diagnoses. In the original study group, the attitude to computer-aided diagnostics presented as medians and lower – and upper quartile were for, male specialists 3 (2–4), female specialists 3 (2–3) and male non-specialists 3 (3-3), female non-specialists 3 (3-3). In the random sample group, where all were specialists, the results were for males 3 (2–4) and females 3 (2–3). Other general variables like Gender, Age, Level of professional training, Computer skill, Attitude to Computer-aided diagnostics, DSM-IV/SCID-training were not found to be statistically correlated to the dependent log file variables (Total session time, Total decision time, Average Decision Time, Total "non decision" time, Total Number of Criteria judged, Total Number of Diagnoses, Total Number of Correct Diagnoses, Total Number of Incorrect Diagnoses, Correct Diagnoses ratio, Total Number of regretted criteria-judgments, Total Number of Criteria judged Unclear and Sum of numbers of regretted criteria-judgments and unclear criteria judgments). Learning styles The Learning Styles of the 49 physicians (tested by the LSI for learning style preferences) are shown in Table 2 , where it is seen that the most common LS was Assimilating, followed by the Converging style. No major differences in Learning Style preferences were found between males (27) and females (22). Table 2 Learning style preferences in the original study group and the random sample group Group Assimilating Accommodating Converging Diverging Row Totals Original study group 18 (37 %) 13 (27 %) 15 (31 %) 3 (6 %) 49 Specialists 10 (32 %) 6 (19 %) 12 (39 %) 3 (10 %) 31 Non-specialists 8 (44 %) 7 (39 %) 3 (17 %) 0 (0 %) 18 Random sample group 34 (36 %) 23 (24 %) 17 (18 %) 21 (22 %) 95 All groups 52 36 32 24 144 In comparison, the random sample of the psychiatrists in Sweden (also tested by the LSI for learning style preferences) turned out as indicated in Table 2 . Here was also the most common LS Assimilating, followed by Accommodating style. No significant differences were found among the genders. Among the 49 psychiatrists, it was found a correlation (p < .01) between learning styles and reported computer skill. The persons with highest score on computer skill were Converging, followed by Accommodating and Assimilating. Diverging styles were found to have the lowest computer skill. A positive attitude to computer-aided diagnostics and learning styles were also found to correlate (p = .04). Most positive were Assimilating, followed by Converging and Accommodating. Diverging had the lowest figures in terms of attitude to computer-aided diagnostics. Finally, it was also found that the distribution of learning styles and the Number of Criteria judged in the system significantly correlated (p < .01). The Accommodating group used the highest number of criteria, followed by the Converging, Diverging and Assimilating groups in that order. Diagnostic results Interestingly, the 49 physicians gave a rather low percentage of correct diagnoses. The correct diagnoses Major Depressive Disorder was found by only 27%, Somatization Disorder by 55% and Alcohol Dependence by 78%. Only 10% had all three diagnoses correct. 41% two correct diagnoses, 47% one correct diagnose and 2% no correct diagnose at all. Median values of proposed, correct, incorrect and ratio correct/proposed diagnoses for specialist and non-specialists are shown in Table 3 . Table 3 Median values of proposed, correct, incorrect and ratio correct/proposed diagnoses Group Proposed diagnoses Correct diagnoses Incorrect diagnoses Ratio correct/proposed Original study group (n = 49) 4 2 3 40 % Specialists (n = 31) 4 1 3 33% Non-specialists (n = 18) 4 2 3 50 % As seen in Table 3 , the non-specialist group seems to produce slightly better results (although not significant) than the specialist group. Also visible in the table, the median values for the number of proposed diagnoses (as a result from CB-SCID1 use) was as high as 4 for both specialists and non-specialists compared to the three correct diagnoses, which makes the above findings even more interesting and indicates a risk of over-diagnosing. Acceptance The indicator of acceptance, user's attitude to computer-aided diagnostics, was found to be 3 on a four-graded scale for both specialists and non-specialists in the original study group as well as for the random sample. No difference was found among genders. When correlating this to computer skill and computer anxiety it was found an overall negative significant correlation (-.48, p < .001) between positive attitude to computer-aided diagnostics and computer anxiety. Within the Accommodating group there was an even higher negative significant correlation (-.80, p < .001) between positive attitude to computer-aided diagnostics and computer anxiety. There was only a tendency to a correlation between computer skill and a positive attitude toward computer-aided diagnostics (.27, p = .06) in the original study group. However, within the Accommodating group, there was a positive significant correlation (.62, p = .02). In the random sample group, a significant correlation between computer skill and positive attitude to computer-aided diagnostics (.43, p < .001) was found. Finally, there was an overall negative significant correlation between computer skill and computer anxiety (-.51, p < .001). Use The use of CB-SCID1 was found to be varying among the 49 psychiatrists. For example, a positive significant correlation (.31, p = .03) was found between the active experimentation information processing mode (Converging and Accommodating) and the Total Number of Criteria judged in the program. No other significant correlations between the active experimentation-reflective observation dimension and the other 11 outcome log-file variables were found. The total number of criteria judged for the patient case used varied from 44 to 119. There were no significant correlations between the abstract conceptualization-concrete experience dimension and the 12 outcome log-file variables. The comparison between the specialist group and non-specialist group concerning the 12 log-file variables revealed no significant differences of medians (See Table 4 ). Table 4 Comparison of medians, lower – and upper quartile between specialists and non-specialists in outcome log-file variables Variable Original study group (n = 49) Specialist group (n = 31) Non-specialist group (n = 18) p-value Tot time (seconds) 1895 (1476–2405) 1710 (1476–2132) 2064 (1457–2532) 0.32 Decision time (seconds) 1193 (870–1559) 1035 (844–1486) 1269 (1107–1758) 0.20 Average decision time (seconds) 17 (13–22) 15 (12–20) 19.5 (14–22) 0.18 "Non decision time" (seconds) 661 (495–800) 650 (495–772) 677 (454–900) 0.83 Criteria judged (number) 68 (61–78) 69 (61–78) 67.5 (61–78) 0.74 Proposed diagnoses (number) 4 (3–5) 4 (3–5) 4 (4–5) 0.88 Correct diagnoses (number) 2 (1–2) 1 (1–2) 2 (1–2) 0.50 Incorrect diagnoses (number) 3 (2–4) 3 (2–4) 2.5 (2–3) 0.57 Ratio correct/prop diagnoses 40 (25–50) 33.3 (25–50) 50 (25–50) 0.24 Regretted judgments (number) 3 (1–8) 2 (1–8) 3.5 (2–10) 0.32 Unclear judgments (number) 7 (3–15) 7 (3–15) 8.5 (3–15) 0.80 Sum regretted unclear judgments (number) 14 (8–23) 14 (6–22) 13 (8–29) 0.56 Perceived usefulness When analyzing the follow-up interview, (after trying CB-SCID1) a significant correlation (-.32) between perceived future usefulness and the Abstract Conceptualization – Concrete Experience dimension (p = .02) was found. The Abstract Conceptualizations orientated group had more Pros while the Concrete Experience orientated had more Cons on perceived future usefulness. This indicates that the Assimilating and Converging learning styles, which perceive information by abstract conceptualizations, are favored by the CB-SCID1-system. According to the interviews, the perceived usefulness of the CB-SCID1-system was more negative than positive. 27 (55%) of the psychiatrists noted more Cons than Pros, 22 noted more Pros than Cons (45%). The responses were also categorized according to their general content. 6 positive and 7 more negative general categories were found, see Table 5 . Table 5 Pro and Con categories of Perceived usefulness of CB-SCID1 Pros (6 different categories) Cons (7 different categories) • Structure "there is a structure to hold on to in the program" • Appropriateness "not suited for the diagnostic interview situation" • Accurate and reliable diagnoses "contradicts diagnoses by feeling", "more exact diagnose" • Empathy and intuition "risk of missing emotional and non-verbal information" • Feedback "what works in treatment or not" • Conflict "managing the patient contact and the CB-SCID1-system at the same time" • Help "help in diagnostic thinking while working with the program" • Underestimation "risk of underestimating your own skill, risk of getting dependent of the program" • Correction "you will be noticed if you are on the wrong track" • Routine questioning "promotes exhaustion effects and lack of initiative" • Timesaving "the program runs all administration and presents the diagnostic results" • Dialogue "breaking up of dialogue, missing emotional states and risk of irrelevant questions" • General picture "risk of losing the overall picture" Discussion General results Given the reported high clinical interviewing skill, high computer skill and relatively positive attitude to computer-aided diagnostics for the group, the low number of correct diagnoses and high number of incorrect diagnoses is very interesting. The DSM-IV-diagnosis is a symptom-diagnose derived from a deliberately limited amount of relevant diagnostic information, pattern of symptoms and development within predefined limits. This is not intended to be compared with the clinical diagnosis, which is building on patterns of symptoms, complete development, actual circumstances, anamnestic data, etiological discussion, laboratory tests, psychological tests etc. The symptom-diagnose and clinical diagnose perspective might be complementary to each other and it is usually recommended that a clinical interview always should be done before the SCID or CB-SCID1 interview. One possible explanation of our findings is that diagnosis in psychiatry is so complex that neither the DSM-system nor the "paper-SCID" or the CB-SCID1 systems might help. Another possible explanation is that CDSS have their limitations. This is in accordance with for example Dreyfus and Dreyfus who argued that computers could be good competent manipulators of symbols according to prepackaged algorithms, but they lack the type of intuition that real experts have [ 39 ]. There are alternatives to the SCID and CB-SCID1 to make DSM-IV-diagnosis. For example, the SCAN (Schedules for Clinical Assessment in Neuropsychiatry) and the computerized version, the SCAN 2.1 system, are systems developed by the WHO (World Health Organization) [ 40 , 41 ]. The computerized SCAN system is more based on "facts" from the patients' answers transformed by algorithms to DSM-IV diagnosis compared to the CB-SCID1 where the interviewer's judgment of each criterion is of importance. If a system like SCAN would have given other results, remains to be investigated. The fact that this study did not diagnose living patients and that it was done with the help of a computer program may affect the results in that tacit knowing could not come into play as much as in a real situation. Polanyi [ 42 ] mentions that we can notice and do things without being able to tell how we recognize something or tell exactly what we do. Maybe the results of the experts, number of correct diagnoses for instance, are most affected by this in their intuitive way of functioning due to extensive personal experience, exercise and experience of former master-trainee relationship that is, "tacit knowing knowledge". We also have to keep in mind that the 49 physicians tried the CB-SCID1 for the first time and with limited training in the program (even if the system is said to be possible to be used with little training). When trying to evaluate the results, we have to consider the different needs, habits and working style of experts and non-experts of different levels. For example, any expert, with a possible intuitive way of thinking, might be confused when they use a program addressed to non-experts that emphasize rule-following and logical step-by-step working procedures. But our interpretation is still that the explanation is that CDSS do not suit all clinicians. This is in accordance with for example Ridderikhoff and van Herk [ 11 ] who found that despite need for diagnostic help, computer-aided support systems was ranked lower than other computer-aids and the use of a diagnostic computer-aided support system. Limitations of the study A limitation of this study is that only one case was used. The main reason for only using one single case was that the 49 physicians participating were very busy and reported that they had little time to spend for research projects like this. Furthermore, most of the 18 physicians not accepting to participate in the study reported lack of time as the main reason; asking the participants to use more cases, would most probably have resulted in more dropouts. The fact that physicians decided not to participate due to time constraints, could also raise a concern about selection bias since they might actually be less prone to use computer technology, differ in learning style or else. However, as the real case used is judged as being rather standardized and has been used successfully in training, we consider the risk of incorrect results being rather low. Another limitation is that the computer skill was self-reported, and that no standardized skill test was used. Finally, one limitation is that some of the psychiatrists participating in this study might have seen the used case before. However, since the DSM Case Book covers more than 200 cases, the chance that those psychiatrists remembered both the actual case and its three specific diagnoses is considered to be very low. LS and psychiatry It has been reported that domain specialists might have different LS preferences, for example Baker III, et al [ 43 ] using the Kolb model found that there is an identifiable surgical learning style: Converging (46 %). The other styles were Accommodating (26%), Assimilating (20%) and Diverging (8 %). This is in line with the Plovnick [ 44 , 45 ] results, which suggests surgeons as Converging in Medicine. Our results concerning learning styles did not confirm the Plovnick results that psychiatrists should be Diverging. This could depend on various reasons. The role for the psychiatrist today is not the same as it was 30 years ago. The "Diverging" aspect of relational skill is less vital compared to diagnostic methods of today and various modern treatment methods. Another possible explanation is, according to the Kolb theory, Diverging-preference persons with their Concrete perceiving and Reflective observation are predicted to be the least interested in computer work, and thus that there is a possibility that the result is due to selection bias and low response rate. Theoretically the nearly 60 % of the 226 Swedish psychiatrists who did not answer the questionnaire, might have a Diverging preference. If so psychiatrists would have a Diverging preference, and the CB-SCID1 perceived usefulness values would probably be even more negative. Divergers also had the lowest figures in terms of attitude to computer-aided diagnostics and computer skill. Could it be that the low number of Diverging preferences, 6% in the original study group is the result of "selection bias"? After all, they were volunteers to participate in the study. They were asked to try the CB-SCID1 and may have hesitated to participate due to lack of computer skill? The random sample group was not asked to try the CB-SCID1 and Diverging preferences in that group were 22%. The higher numbers with Converging preference (high computer skill) in the Original study group compared to the Random sample group might have been explained by the same reason. The fact that there were no significant correlations between the general variables and the dependent log-file variables indicate that further analyses of for example attitudes, use, perceived usefulness and learning style must be made in the future. This is supported by our findings that individual values regarding computer skill, computer anxiety and attitude to computer-aided diagnostics are related to learning styles and acceptance, use and perceived usefulness. Acceptance Computer skill, computer anxiety and the attitude to computer-aided diagnostics are interpreted to be important variables in the acceptance of the CB-SCID1-system. Accommodating psychiatrists with computer anxiety have a very low value on the indicator of acceptance. This means that this group has difficulties accepting the CB-SCID1. Again within the Accommodating group, those with computer skill are positive to the indicator of acceptance. This highlights the importance of training as a means to increase computer skill and ease the acceptance of systems like CB-SCID1. Maybe the acceptance would increase if only targeting beginners or persons on "medium level" of training. As shown, a tendency for a more positive regard to computer-aided diagnostics exists in the non-expert group. The specialists, as proficient and on expert level, might not have a need of such computer-aid, or at least a different need in accordance with their way of working as experts with an intuitive frame of reference compared to the step by step, rule-following work of the novice. Use A possible explanation for the low number of significant results between LS and the outcome log-file variables might be that learning style is used with more flexibility in a real situation (using CB-SCID1) compared to more "attitude-like" variables like acceptance and perceived usefulness. It might be so that LS must be linked to learning skill and adaptivity/flexibility in different specific situations to give significant results against the log-file variables. The LSI is measuring learning style preferences not associated to specific situations. Perceived usefulness Our findings that the program seems to be more attractive to psychiatrists with learning styles which prefer Abstract Conceptualizations (Assimilating and Converging learning styles) and in the majority negative to the Concrete Experience (Accommodating and Diverging learning styles) is in accordance with Lu et al [ 10 ] who found that the willingness to use CDSS rely heavily on preferences and perceived usefulness. The Cons in future perceived usefulness against the CB-SCID1 about the appropriateness of such a program, the conflict perspective about using it and the breaking up of dialogue contain an all or none view. However, the Cons arguments have much in similar with the Dreyfus and Dreyfus standpoint regarding the limitations of computer use and seems to be reflecting the expert view, for instance lack of appropriateness, empathy and intuition and underestimation of your own skill. The Pros arguments seem to meet the non-expert view of learning to diagnose, for instance the program as a structure, help and correction facility. How to use, by whom and when, in which situations the CB-SCID1 is to be used must be elaborated upon. Non-experts with Diverging preferences might increase their flexibility using learning styles as a result of training and maybe a more positive regard to computer-aided diagnostics. That is, going around the "Kolb circle" in a more flexible and balanced way, using all learning styles, and give up their strong Diverging preferences. Overall results from acceptance, use and perceived usefulness Our results indicate that computer skill is of importance, and computer anxiety of negative impact on the attitude to computer-aided diagnostics. The highest computer skills were found within the Converging and Accommodating groups, which use Active Experimentation as information processing mode. Most positive to computer-aided diagnostics were the Assimilating and Converging groups, which in perceiving information are using Abstract Conceptualization. The results also indicate that the Active Experimentation information-processing mode (Accommodating and Converging learning styles) is significantly correlated to number of criteria judged in the program. Furthermore, the results also indicates, that although the 49 psychiatrists reported a positive attitude to computer-aided diagnostics, physicians with computer anxiety are less positive. Moreover, the CB-SCID1 CDSS seemingly invites the Accommodating and Converging learning styles to significantly adopt an Active Experimentation information-processing mode with a high number of criteria judged. However, this experimental mode may increase the number of incorrect diagnoses. The Assimilating and Converging learning styles also seems to be favored by the computer program concerning the Abstract Conceptualization mode in perceiving information. Conclusions The results of this study suggests that a CDSS is no guarantee of improved diagnostic procedures in Psychiatry and that even a clinically experienced user might end up with several incorrect diagnoses using such a system. The results also indicate that the use of CDSS-tools seems to favor users with learning style preferences using abstract conceptualization information perceiving mode. Furthermore, our results indicate that future research on CDSS should focus on the possibility to create systems open to individuals with different LS preferences. Future research should also focus on the importance of computer training and different professional levels to optimize the usefulness of CDSS. The relationship between learning style preferences and working style habits at different professional levels might also be elaborated upon as well as the importance of learning style flexibility and decision modes in various diagnostic situations. List of abbreviations used CDSS Computerized decision support system DS Decision support DSM Diagnostic Statistical Manual, "DSM-system" DSM-IV Diagnostic Statistical Manual, version 4 LS Learning Style SCID Structured Clinical Interview for DSM-diagnoses CB-SCID1 Computer-Based Structured Clinical Interview for DSM-diagnoses (axis 1) Competing interests Competing interests exists between the commercial CB-SCID1 system from Pilgrim Press and the SCAN-system developed by WHO, World Health Organization. However, the authors have no interests in any of these competitors. Authors' contributions UF has supervised LB in the planning, carrying out and analysis of this research work. LB has carried out all experiments. The manuscript has been written on close collaboration between LB and UF. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545069.xml
533869	Isoform-specific expression of the Coxsackie and adenovirus receptor (CAR) in neuromuscular junction and cardiac intercalated discs	Background The Coxsackie and adenovirus receptor (CAR) has a restricted expression pattern in the adult. In skeletal muscle, although CAR is expressed in immature fibers, its transcript levels are barely detectable in mature muscle. This is in contrast to the robust expression observed in the heart. However, both heart and skeletal muscle are susceptible to infection with the Coxsackie B virus which utilizes primarily CAR for cellular internalization. The specific point of viral entry in skeletal and heart muscle remains unknown. Results Using antibodies directed against the extracellular and the cytoplasmic domains of CAR, we show CAR in normal human and mouse skeletal muscle to be a novel component of the neuromuscular junction. In cardiac muscle, CAR immunoreactivity is observed at the level of intercalated discs. We demonstrate a single isoform of CAR to be expressed exclusively at the human neuromuscular junction whereas both predominant CAR isoforms are expressed at the intercalated discs of non-diseased human heart. Conclusion The localization of CAR to these important junctional complexes suggests that CAR may play both a structural and a regulatory role in skeletal and cardiac muscle, and that these complexes may serve as a point of entry for Coxsackie B virus.	Background The Coxsackie and adenovirus receptor (CAR) [ 1 , 2 ], a transmembrane protein of the immunoglobulin super-family, serves as a receptor for adenovirus (Ad) subgroups A, C, D, E and F [ 3 ] as well as Coxsackie B viruses (CVB) [ 4 ]. CAR is a highly conserved protein with two predominant isoforms, produced through differential splicing, and having cytoplasmic domains of either 107 residues (ending in SIV) or 94 residues (ending in TVV) [ 2 , 5 ]. The extracellular domain mediates homophilic cell adhesion [ 6 - 8 ] and ectopically-expressed CAR localizes to homotypic intercellular contacts [ 8 ]. The expression of CAR is regulated developmentally [ 6 , 9 - 12 ] as well as in a tissue-specific manner [ 2 , 5 ]. To date, most studies on CAR expression in the adult have resorted to analysis of transcript levels. These have revealed that the pattern of tissue-specific expression differs between humans and mice. In humans, a predominant transcript of ~6 – 6.5 kb is observed in heart, testis, prostate and pancreas while much less expression is detected in liver, brain, colon and small intestine. In the mouse on the other hand, the most abundant expression is in liver, kidney, lung and heart. Interest in CAR stems from its function as the primary high affinity receptor for Ad serotype 5, the most commonly used adenoviral vector in gene therapy protocols. CAR expression is the main determinant in gene transfer to normal tissue as ectopic expression of CAR in transgenic mice leads to several magnitudes of increase in adenovirus transducibility of tissues that are otherwise refractory to Ad-mediated gene expression [ 13 - 17 ]. As well, although decay accelerating factor (DAF, CD55) was the first described CVB receptor [ 18 , 19 ], CAR is necessary and sufficient for CVB infection in vitro [ 20 ]. Thus, the expression levels of CAR may also govern the susceptibility to CVB diseases and the pathological consequences of CVB viral infection. In this context, acute viral myocarditis and myositis are inflammatory diseases affecting cardiac and skeletal muscle that can result from infection by the Coxsackie B virus. In both humans and rodents, heart is among the tissues showing the greatest abundance of CAR transcript while its transcripts are barely detectable in skeletal muscle even with the more sensitive reverse-transcriptase (RT)-PCR-based assay [ 21 ]. In contrast to heart, DAF expression is absent in mature skeletal muscle [ 22 ]. Despite the absence of DAF and low CAR transcript levels, skeletal muscle is nevertheless susceptible to Coxsackie virus-induced myositis. Indeed, human patients suffering from inflammatory muscle diseases have tested positive for CBV RNA [ 23 ]. This suggests that the low CAR transcript level in skeletal muscle may produce functional receptor. Therefore, to examine CAR localization in skeletal and cardiac muscle, we used antibodies directed against the extracellular domain of CAR [ 21 ] as well as antibodies that can differentiate between the two major CAR isoforms [ 24 ] with alternate 3' splicing [ending in the amino acids SIV or TVV] [ 2 , 5 ] [Fig. 1 ]. Figure 1 Depiction of the CAR sequences recognized by the anti-C-terminal antibodies. The chicken antibody ChCT was raised against a fusion protein containing C-terminal sequence common to both CAR isoforms (from aa 259 – 339). The rabbit antibodies (RP291 and RP194) were raised against peptides that were specific to each isoform. The diagram does not show the immunogen of the rabbit antibody 2240 which is the extracellular N-terminal domain of murine CAR. Note that each antibody has a distinct recognition sequence. TM – transmembrane sequence. Results and Discussion To immunolocalize CAR, frozen sections of normal human muscle biopsies were probed with polyclonal antibodies raised against the extracellular domain of CAR [ 21 ] [ab 2240], or against the cytoplasmic tail, antibodies which specifically recognize the two predominant isoforms of human CAR (referred to as SIV [ab RP291] and TVV [ab RP194] respectively) [ 24 ]. We localized CAR exclusively to the neuromuscular junction in human skeletal muscle (Fig. 2 ) where CAR immunoreactivity coincided with acetylcholine receptors as revealed by α-bungarotoxin binding. Similarly, in mature murine skeletal muscle, CAR expression was restricted to the neuromuscular junction (Fig. 3 ). Interestingly, we find CAR expression at the neuromuscular junction to be isoform specific, with the CAR SIV isoform accounting for all CAR immunoreactivity (Fig. 2G ). In contrast, the TVV isoform is present in the vasculature (Fig. 2I ). CAR is therefore identified as a novel component of the adult neuromuscular junction, joining other homotypic cell adhesion molecules such as N-cadherin and neural cell adhesion molecule (N-CAM) which have been localized previously to this specialized site. As with NCAM [ 25 ], although CAR is uniformly expressed in immature mouse muscle (data not shown), its expression becomes downregulated within a few weeks of birth [ 21 ] and its localization is confined to the neuromuscular junction. As well, in skeletal muscle undergoing regeneration, as is the case in the dystrophic mdx mouse, CAR is re-expressed [ 21 ] as is NCAM [ 26 ]. Figure 2 Immunolocalization of CAR to the neuromuscular junction in human skeletal muscle . Frozen sections of human muscle were incubated with Alexa Fluor-conjugated α-bungarotoxin and antibodies to CAR as described in Materials and Methods. Panels A and D show α-bungarotoxin staining of acetylcholine receptors at neuromuscular junctions (green). Panel B shows immunofluorescent staining of the section in panel A with a polyclonal antibody (ab 2240) to the extracellular domain of CAR (red). Panel E shows immunofluorescent staining of the section in panel D with a mixture of the isoform-specific C-terminal antibodies RP194 and RP 291 (red). Panel C is a merge of panels A and B . Panel F is a merge of panels D and E . These merges [ C, F ] show that CAR colocalizes with α-bungarotoxin at neuromuscular junctions (in yellow). Of the two C-terminal antibodies, only RP291 ( G ) demonstrates the typical neuromuscular junction staining while signal is absent when sections are incubated under similar conditions with RP194 ( H ) although RP194 does label blood vessels ( I ). Sections incubated with secondary antibody alone did not reveal any signals ( J ). Bar = 25 μm (A, D); 50 μm (G, I). Figure 3 Immunolocalization of CAR to the neuromuscular junction in mouse skeletal muscle . Frozen sections of mouse muscle were incubated with Alexa Fluor-conjugated α-bungarotoxin and antibodies to CAR as described in Materials and Methods. Panels A and D show α-bungarotoxin staining of acetylcholine receptors at neuromuscular junctions (green). Panel B shows immunofluorescent staining of the section in panel A with a polyclonal antibody (ab 2240) to the extracellular domain of CAR (red). Panel E shows immunofluorescent staining of the section in panel D with a chicken polyclonal antibody (ChCT) directed against a C-terminal portion of CAR conserved in both CAR isoforms (red). Panel C is a merge of panels A and B . Panel F is a merge of panels D and E . These merges [ C, F ] show that CAR colocalizes with α-bungarotoxin at murine neuromuscular junctions (in yellow). Sections incubated with the secondary antibodies alone did not give any signal – anti-rabbit IgG [ G ], anti-chicken IgY [ H ]. Bar = 35 μm CAR has been suggested to serve as a structural component at the intercalated discs in murine cardiac muscle [ 8 ] and recently confirmed to reside primarily at the intercalated discs in both rat heart and neonatal cultured cardiomyocytes of the rodent [ 12 ]. Using antibodies directed to the specific SIV and TVV isoforms of the receptor as described above, we localized CAR to the intercalated discs (Fig. 4 ), demonstrating for the first time expression of both receptor isoforms in human and murine heart. To confirm the presence of both isoforms, murine cardiac muscle homogenates were analyzed by SDS-PAGE followed by electrotransfer to nitrocellulose membranes. This immunoblot analysis revealed the presence of both the SIV and TVV isoforms (Fig. 5A ). Furthermore, when murine cardiac muscle homogenates were immunoprecipitated with either RP194 or RP 291 rabbit polyclonal antibodies, Western blot analysis of the resultant fractions with a chicken antibody against CAR C-terminal domain showed a single 46 kDa band (Fig. 5B ), confirming the specificity of the antibodies for CAR. Figure 4 Immunolocalization of CAR to intercalated discs in human and murine cardiac muscle. Human [ A, B ] and murine [ C, D ] cardiac muscle sections were reacted with the polyclonal anti-C-terminal antibodies RP194 [ A, C ] and RP291 [ B, D ]. Note the intense staining of both isoforms at the human intercalated discs (arrow) [ A, B ] and murine intercalated discs [ C, D ]. Neither the human cardiac tissue [ E ] nor the murine cardiac muscle [ F ] gave any signal when incubated with the anti-rabbit IgG. Bar = 10 μm (A, E); 30 μm (C, F). Figure 5 Murine cardiac muscle homogenates express both isoforms of CAR. A. Immunoblot analysis of cardiac muscle homogenates (10 μg) reveals that a single polypeptide of 46 kDa is detected by the anti-C-terminal antibodies recognizing the two different CAR isoforms (RP194 and RP291). B. Murine cardiac homogenates were immunoprecipitated with either RP194 (lanes 1,2) or RP291 (lanes 3,4) followed by immunoblotting with the chicken ChCT antibody that recognizes the C-terminal portion of CAR that is common to both isoforms. The ChCT antibody detected as a single band the increasing amounts of CAR that was immunoprecipitated with increasing amounts of the rabbit polyclonal antibodies (2 μl, lanes 1,3; 4 μl, lanes 2,4). Both of the cytoplasmic variants contain a PDZ recognition motif at their distal end, implicating CAR as a putative member of multiprotein complexes. This is indeed the case in polarized epithelial cells in which CAR is expressed at the tight junction where it associates with the tight junction scaffolding protein ZO-1 [ 8 ] and contributes to maintenance of transepithelial resistance. The localization of CAR to cardiac intercalated discs is in agreement with CAR having a structural or regulatory role as a transmembrane member of junctional complexes. The intercalated discs are composed of at least three structurally distinct cellular junctions – desmosomes, the adherens junctions, and gap junctions [ 27 ]. The SIV and TVV isoforms may localize to separate components of the intercalated disc, or given that CAR has been implicated as a transmembrane component in tight junctions [ 8 ], both isoforms may be localized to the adherens junction, a structure analogous to tight junctions. It is interesting that a correlation can be drawn between disease incidence and expression levels of the CAR receptor. Cases of viral myocarditis in the human population outnumber those of myositis. This may be attributable to the difference of endogenous CAR expression between the two tissues. Considering that induction of CAR accompanies myocarditis [ 28 ] and its dramatic upregulation has recently been demonstrated in patients suffering from multiple diseases of the heart [ 29 ] including dilated cardiomyopathy, a pathological phenotype linked to persistent acute myocarditis, upregulation of the receptor may render the heart even more susceptible to further viral infection. Conversely, the low level of endogenous CAR expression in skeletal muscle may safeguard against the wide-spread viral infection seen in myocarditis and may be responsible for the less severe clinical features of myositis. Conclusions CAR is a novel member of the neuromuscular junction. In cardiac muscle, both CAR isoforms are found at the intercalated discs. The localization of CAR to these important junctional complexes suggests that CAR may play both a structural and a regulatory role in skeletal and cardiac muscle, and that these complexes may serve as a point of entry for Coxsackie B virus. Methods Antibodies Figure 1 depicts the sequences used as immunogen to generate the various antibodies. There was no overlap (i.e. common epitopes) between any of the antibodies. The rabbit polyclonal antibodies used in this study were described previously [ 21 , 24 ]. Briefly, the N-terminal polyclonal antibody (ab 2240) was prepared against a His-tagged fusion protein which encoded amino acid residues 22–208 of the extracellular domain of mouse CAR [ 21 ]. This antiserum cross-reacts with human CAR on Western blots and in indirect immunofluorescence. The two C-terminal polyclonal antibodies [ 24 ] were generated using peptides encompassing the last 13 amino acids of the two predominant human CAR isoforms [ 2 , 5 ]. Antiserum RP194 was raised against the sequence FKYAYKTDGITVV while RP291 was raised against the sequence VMIPAQSKDGSIV. Both these antisera cross-react with the mouse CAR homologs (the peptides are conserved 100% between the two species [ 30 ]). All antisera were affinity purified prior to use. To raise the chicken anti-CAR antibody (ChCT), purified His-tagged fusion protein encoding the C-terminal portion of CAR that is common to both isoforms (amino acids 259–339) was emulsified in an equal volume of TiterMax Gold adjuvant (CytRx Corp., Norcross, GA) and injected intramuscularly into chickens. One month post-injection, IgY antibodies to CAR were obtained from the eggs of injected chickens and subjected to affinity purification. Immunocytochemistry Immunolabelling was performed using standard techniques. Briefly, frozen sections (5 μm) of normal human skeletal and cardiac muscle biopsies, murine skeletal and cardiac muscle were fixed in 2% paraformaldehyde (pH 6.8) for 1–2 minutes, followed by overnight incubation at 4°C with the primary antibodies (a 1:30 dilution was used for ab 2240, and 1:200 dilution for the abs RP291 and RP194, in blocking solution made of 3% bovine serum albumin and 0.05% Tween-20 in phosphate-buffered saline). Incubation with a mouse anti-rabbit biotin-conjugated secondary ab (1:120; Jackson Immunoresearch Laboratories, West Grove, PA) was followed by Cy-3-conjugated streptavidin (1:1000; Jackson Immunoresearch Laboratories). Controls consisted of sections treated in the absence of primary antibody. Neuromuscular junctions were revealed with Alexa-488-conjugated-α-bungarotoxin [α BTX] (1:40) (Molecular Probes, Eugene, OR). Slides were viewed on a Leica microscope-based imaging system using OpenLab imaging software (Quorum Technologies, St Catharines, ON). Western blot analysis and immunoprecipitation Cardiac muscle tissue was homogenized in extraction buffer [1% Triton X-100; 0.1 mM EDTA; 0.1 mM EGTA; 50 mM Tris-HCl; pH 8.0; with protease inhibitors (Roche)] at 4°C. After a 30 second sonication, samples were centrifuged at 3000 × g for 30 seconds at 4°C. Protein samples (10 μg) were anayzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) using 10% (w/v) polyacrylamide gels, followed by electrotransfer to nitrocellulose. The blots were blocked in 10% BLOTTO (skim milk powder) in Tris buffered saline – Tween 20 (TBS-T) for 45 minutes at room temperature. Anti-CAR antibody was added in 10% BLOTTO at a dilution of 1:2500. Following incubation with peroxidase-labeled goat-anti-rabbit secondary antibody (Jackson Immunoresearch Laboratories), the signal was visualized by enhanced chemiluminescence (Amersham Pharmacia Biotech, Baie d'Urfe, QC). Immunoprecipitation was carried out on cardiac muscle homogenates that had been pre-cleared with Protein-A Agarose slurry (Sigma), followed by overnight incubation at 4°C with RP291 or RP194. The samples were further incubated with Protein-A Agarose for 2 hours, washed twice with extraction buffer and then eluted with 2 X Laemmli SDS sample buffer with 5% mercaptoethanol. Following SDS-PAGE and electrotransfer, nitrocellulose membranes were probed with a primary polyclonal chicken anti-CAR (ChCT) in blocking solution (diluted 1:500) overnight at 4°C. Signal was revealed following incubation with a peroxidase-conjugated donkey anti-chicken IgY (Jackson Immunoresearch Laboratories) at a dilution of 1:2500 for 40 minutes, and enhanced chemiluminescence. Authors' contributions CAS performed the immunoprecipitation, the immunoblotting, participated in the immunofluorescent staining experiments and drafted the manuscript. PCH and JN conceived the study, participated in its design and interpretation as well as drafting the manuscript. CA participated in the immunofluorescent staining experiments. KS prepared and characterized the isoform-specific antibodies. MS and GK participated in the design and interpretation of the studies. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC533869.xml
514540	Similarities in transcription factor IIIC subunits that bind to the posterior regions of internal promoters for RNA polymerase III	Background In eukaryotes, RNA polymerase III (RNAP III) transcribes the genes for small RNAs like tRNAs, 5S rRNA, and several viral RNAs, and short interspersed repetitive elements (SINEs). The genes for these RNAs and SINEs have internal promoters that consist of two regions. These two regions are called the A and B blocks. The multisubunit transcription factor TFIIIC is required for transcription initiation of RNAP III; in transcription of tRNAs, the B-block binding subunit of TFIIIC recognizes a promoter. Although internal promoter sequences are conserved in eukaryotes, no evidence of homology between the B-block binding subunits of vertebrates and yeasts has been reported previously. Results Here, I reported the results of PSI-BLAST searches using the B-block binding subunits of human and Shizosacchromyces pombe as queries, showing that the same Arabidopsis proteins were hit with low E -values in both searches. Comparison of the convergent iterative alignments obtained by these PSI-BLAST searches revealed that the vertebrate, yeast, and Arabidopsis proteins have similarities in their N-terminal one-third regions. In these regions, there were three domains with conserved sequence similarities, one located in the N-terminal end region. The N-terminal end region of the B-block binding subunit of Saccharomyces cerevisiae is tentatively identified as a HMG box, which is the DNA binding motif. Although I compared the alignment of the N-terminal end regions of the B-block binding subunits, and their homologs, with that of the HMG boxes, it is not clear whether they are related. Conclusion Molecular phylogenetic analyses using the small subunit rRNA and ubiquitous proteins like actin and α-tubulin, show that fungi are more closely related to animals than either is to plants. Interestingly, the results obtained in this study show that, with respect to the B-block binding subunits of TFIIICs, animals appear to be evolutionarily closer to plants than to fungi.	Background Phylogenetic relationships among animals, fungi, and plants have been a controversial issue. Although fungi traditionally had been considered more closely related to plants than to animals, Whittaker and Margulis [ 1 ] classified the fungi as a separate kingdom in their five-kingdom classification: the three major multicellular groups of animals, fungi, and green plants were each given the status of kingdoms derived from different protistan lineages of uncertain affinities. With the determination of the primary structures of homologous macromolecules in various organisms, spanning several kingdoms, molecular phylogenetic techniques resulted in new hypotheses about the relationships among eukaryotes. Small subunit rRNA, and the proteins like actin and α-tubulin, exist ubiquitously and their primary structures are highly conserved. Thus, these sequences have been used to make molecular trees [for examples, [ 2 , 3 ]]. Most of these studies place fungi as more closely related to animals than either is to plants [ 3 - 6 ]. Eukaryotic RNA polymerase III (RNAP III) transcribes a variety of small RNAs like tRNAs, 5S rRNA, and several viral RNAs [ 7 ]. Short interspersed repetitive elements (SINEs) are also transcribed by RNAP III [ 8 ]. Genes for these small RNAs have internal promoters that consist of two regions called the A and B blocks [ 9 ]. These promoter sequences are well-conserved in diverse eukaryotes [ 9 ]. Transcription by RNAP III requires the multisubunit transcription factor TFIIIC, which plays an important role in transcription initiation [ 10 ]. TFIIIC contains a B-block binding subunit, which recognizes the RNAP III promoter in the transcription of tRNAs and several viral RNAs, orienting its associated subunits along the DNA [ 11 ]. TFIIIC that is oriented toward the start site, promotes TFIIIB binding and assists in directing accurate initiation by RNAP III [ 12 , 13 ]. In TFIIIC of Saccharomyces cerevisiae , a subunit of 138 kDa binds to a B-block, and its gene which is called TFC3 , has been cloned [ 14 ]. The open reading frame for the B-block binding subunit is interrupted by one intron [ 14 ]. TFC3 is a single-copy gene and essential for cell viability [ 14 ]. In human and rat, the B-block binding subunits of TFIIICs are 243 kDa and 220 kDa respectively [ 14 , 15 ], and there is great similarity between them at the amino acid sequence level [ 15 ]. However, B-block binding subunit from mammals has been thought to show no homology to the S. cerevisiae 138 kDa subunit, although all of them bind to similar DNA regions, suggesting a significant degree of evolutionary divergence for RNAP III factors [ 15 , 16 ]. Huang et al. [ 17 ] identified several subunits of S. pombe TFIIIC from the S. pombe sequence database by homology searches using the S. cerevisiae TFIIIC subunits as queries; one of these subunits named Sfc3p, is similar to the S. cerevisiae B-block binding subunit. It has been thought that, like the B-block binding subunit from S. cerevisiae , Sfc3p does not share homology with the human B-block binding subunit. On the other hand, Sfc1p, Sfc4p, and Sfc6p, which are other subunits of S. pombe TFIIIC, show homologies not only to S. cerevisiae TFIIIC subunits but also to human TFIIIC subunits [ 17 ]. It has been shown that Sfc1p, Sfc3p, Sfc4p, and Sfc6p are associated in vivo, and the isolated Sfc3p complex is active in the in vitro RNAP III-mediated transcription of S. pombe tRNA genes [ 17 ]. The lack of clear homology between the yeast and human B-block binding subunits is strange, because transcription of tRNA genes by RNAP III is initiated by the binding of these subunits to the B-block regions, and the internal promoter sequences are highly conserved between vertebrates and yeasts (see above). Interestingly, bacterial tRNA genes also have conserved RNAP III promoters (Fig. 1 ) [ 18 ]. Although these genes are transcribed from the upstream promoters by the bacterial RNA polymerase, RNAP III can transcribe some of them in vitro [ 18 ]. Thus, understanding the relationship of the B-block binding subunits from human and yeast TFIIICs is important for understanding the evolution of the RNAP III transcription machinery. Here, I demonstrate that, at the amino acid sequence level, the B-block binding subunits of the vertebrate and yeast TFIIICs do have important homologies. These homologies are found by comparing the B-block binding subunits with Arabidopsis proteins, which appear to be the homologs of both the human and yeast subunits. Figure 1 Sequences of the internal promoters for RNA polymerase III. In the sequences of the Xenopus tRNA Met gene [37], the Xenopus 5S rRNA gene [38], adenovirus VAI [39], human Alu [40], and the E. coli tRNA Asp gene [18], the nucleotides identical to those in the consensus sequence [9] are shown by capital letters. Results PSI-BLAST search using the human B-block binding subunit as a query When the B-block binding subunits of human and S. cerevisiae are aligned using the early Clustal program, they show little sequence similarity [ 16 ]. To search for potentially homologous B-block binding regions, first I carried out a PSI-BLAST search where the human subunit was used as a query. PSI-BLAST is known to be useful in finding distantly related proteins [ 19 ]. After three iterations, twelve sequences were found with strong similarities to the sequence of human B-block binding subunit. Fig. 2A shows a summary of the result of a PSI-BLAST search, including proteins with alignment scores better than 200. Proteins that showed only local alignment similarities (shorter than about 300 amino acids) were omitted. The human B-block binding subunit was very similar to the rat subunit (75% identity and 84 % positive) as reported by Lagna et al. [ 15 ] (Fig. 2A ). The nuclear protein of 238 kDa in the mosquito Chironomus tentans (GenBank identification number: 18073910) is structurally similar to the human B-block binding subunit [ 20 ]; in BLAST and FASTA searches of non-redundant databases using C. tentans protein as a query, the best hits were the human and rat B-block binding subunits with 28 % identities. A comparable level of similarity has been found also in the Drosophila hypothetical protein, but no other related proteins were identified in the database [ 20 ]. Immunoelectron microscopy shows that this mosquito protein is located at sites of transcription, suggesting the role of the protein in transcription initiation [ 20 ]. In this study, when a PSI-BLAST search was performed using the human sequence as a query, the Chironomus and Drosophila proteins were hit (23 % and 20 % identities respectively) (Fig. 2A ). In addition to these, hypothetical proteins from the mosquito Anopheles gambiae (GI 31212899), Mus musculus (GIs 38087408, and 21595152), Caenorhabditis briggsae (GI 39594187), and Arabidopsis thaliana (GIs 25402830, 9665127, 15218016, 25404859, and 30685327), were also hit with low E -values: E -values were 0, 0, 0, 0, 0, e -169 , e -167 , e -148 , and 5e -87 , respectively (Fig. 2A ). Interestingly, in Arabidopsis proteins of GIs 25402830, 9665127, and 15218016, both the N- and C-terminal end regions corresponded to those of the human subunit, suggesting that these Arabidopsis proteins are orthologs of the B-block binding subunits (Fig. 2A ). Figure 2 Proteins which show homologies to the B-block binding subunits of human and S. pombe , and the Arabidopsis protein GI 25402830. A. The result of a PSI-BLAST search using the human subunit (GI 4753161) as a query. B. The result of a PSI-BLAST search using the S. pombe subunit (GI 19112919) as a query. C. The result of a PSI-BLAST search using the Arabidopsis protein (GI 25402830) as a query. In A-C, only the proteins that had bit scores more than 200 are shown. Bold horizontal lines (black, dark gray, and pale gray lines) are the regions that appeared as convergent iterative alignments. The numbers above the lines are the amino acid positions in the query sequence, and those beneath the lines are the amino acid positions in the hit sequences. When more than two regions were hit in the same sequence, the highest bit score and E -value are shown. Black lines are the alignments that had the best bit scores and E -values. Gray-colored alignments had the worse scores. Dark gray lines represent the alignments with the bit scores more than 200, and the scores of the pale grey alignments are lower than 200. The proteins that had similarities only in short regions (smaller than about 300 amino acids) were omitted, even if the bit scores were more than 200. The case of the Ustilago sequence is exceptional. This result was from the PSI-BLAST which was limited a search of the fungi database. The bold horizontal lines are the regions that appeared as convergent iterative alignments, and below them E -values are shown. PSI-BLAST using the yeast B-block binding subunits as queries The S. cerevisiae B-block binding subunit exhibits 21 % identity and 39 % similarity to the S. pombe Sfc3p protein, and these similarities extend to the overall sequences (Background; [ 17 ]). When I performed a PSI-BLAST search using the S. cerevisiae B-block binding subunit as a query, five sequences were hit with E -values better than threshold after three iterations: the Neurospora crassa hypothetical protein (GI 32412546) with 0.0, S. pombe Sfc3p with e -159 , the Magnaporthe grisea hypothetical protein (GI 38108450) with e -121 , the Saccharomyces bayanus hypothetical protein fragment (GI 10863079) with e -100 , and the Arabidopsis hypothetical protein (GI 15218016) with 0.002 (data not shown). It is noteworthy that the Arabidopsis protein of GI 15218016 was hit although the E -value was not so good. This protein was hit also in the PSI-BLAST search using the human B-block binding subunit as a query (Fig. 2A ). Next, I performed a PSI-BLAST search using the S. pombe Sfc3p protein as a query. Fig. 2B is a summary of the result. The Magnaporthe grisea and Neurospora crassa hypothetical proteins (GI 38108450 and GI 32412546), were hit with very good E -values (0 and e -168 respectively) (Fig. 2B ). The Aspergillus nidulans protein GI 49107000 also was hit with a robust E -value (data not shown). Four hypothetical proteins of Arabidopsis thaliana were hit with E -values worse than those of the two fungi proteins but well above the threshold: GI 15218016 with e -113 , GI 25402830 with e -88 , GI 9665127 with 2e -88 , and GI 25404859 with 7e -70 (Fig. 2B ). The Arabidopsis protein GI 15218016 was found also in the result of the PSI-BLAST search using the S. cerevisiae subunit as a query, but in the search with S. pombe Sfc3p it had a much better E -value. Surprisingly, these four Arabidopsis proteins were identical to the proteins that were hit with low E -values in the PSI-BLAST search using the human B-block binding subunit as a query (Fig. 2A ). While both the N-terminal half regions and C- terminal end regions were similar between the Arabidopsis proteins and the human subunit, their similarities to S. pombe Sfc3p were only in the N-terminal halves (Figs. 2A and 2B ). The B-block binding subunits of rat and human also were hit in this search, but with E -values of 5e -5 and 0.20, respectively (data not shown); the N-terminal 350 amino acid sequences of the rat and human subunits showed similarities to the N-terminal region of S. pombe Sfc3p protein. PSI-BLAST using the B-block binding subunit homolog found in Arabidopsis as a query The four Arabidopsis proteins (GIs 25402830, 9665127, 15218016, and 25404859), were hit with low E -values in both of the PSI-BLAST searches using the human and S. pombe subunits as queries. Thus, I performed a PSI-BLAST search using one of these Arabidopsis proteins (GI 25402830) as a query. Fig. 2C shows a summary of the result. Six hypothetical Arabidopsis proteins were hit with E -values of 0, and three of them were identical to the proteins which were hit in the PSI-BLAST searches using the human and S. pombe subunits as queries. In addition to these, the human and rat subunits were hit with low E -values (e -132 and e -108 respectively). The N-terminal 700 amino acids of the human and rat subunits were most similar to the Arabidopsis proteins, but the short regions of the C-terminal ends also were similar (Fig. 2C ). The hypothetical mouse protein (GI 38087408) which was hit in the PSI-BLAST search using the human subunit as a query, also was hit in the PSI-BLAST search with Arabidopsis GI 25402830 (Fig. 2C ). The B-block binding subunits of S. pombe and S. cerevisiae were hit with E -values worse than threshold (0.024 and 1.1 respectively) (data not shown). However, it should be noted that these Arabidopsis proteins were hit with E -values better than threshold when a PSI-BLAST search was performed using the S. pombe B-block binding protein as a query. It is interesting that the four Arabidopsis proteins were similar to the B-block binding subunits of vertebrates and yeasts, despite the fact that vertebrate and yeast subunits share no recognizable homology [ 15 - 17 ]. These results seemed to imply that the Arabidopsis proteins of GIs 25402830, 9665127, 15218016, and 25404859 represent a 'missing link' between the vertebrate and yeast B-block binding subunits. Alignment of the B-block binding subunits and their homologs When the primary structures of human and rat B-block binding subunits and their homolog in Drosophila are compared, the most conserved sequences are located within the N-terminal two-thirds of the proteins, while the C-terminal one-third is much less conserved [ 20 ]. The Chironomus tentans protein, which probably binds to the B-block in the RNAP III promoter, also has similarities to the human, rat, and Drosophila proteins in the N-terminal region [ 20 ]. The results of PSI-BLAST searches here also showed that conserved sequences in the human subunit, and the three Arabidopsis homologs (GIs 25402830, 9665127, and 15218016), mainly are located in the N-terminal halves of the proteins (Fig. 2A ). Similarly, conserved sequences in the S. pombe subunit and the three Arabidopsis homologs (GIs 25402830, 9665127, and 15218016), are located in the N-terminal one-third regions of the proteins (Fig. 2C ). Thus, I used the Clustal W program [ 21 ] to align the N-terminal one-third regions of the proteins of human, rat, mouse, Drosophila , Chironomus , Arabidopsis , and yeasts. The Clustal W alignment of the N-terminal ends of about 60 amino acids corresponded to each of the alignments obtained by the PSI-BLAST searches between the query and hit sequences (Fig. 3A ). However, the rest of the Clustal W alignment did not correspond to the alignments obtained by PSI-BLAST (data not shown). Therefore, I compared all of the PSI-BLAST alignments further by eye. Three regions including the N-terminal ends were found to be most conserved between the PSI-BLAST alignments; these three regions were aligned separately by Clustal W (Fig. 3 ). Previously, Rozenfeld and Thuriaux [ 22 ] performed PSI-BLAST using the S. cerevisiae B-block binding subunit as a query. They identified two domains of about 70 amino acids, which were conserved in S. cerevisiae subunit and Arabidopsis thaliana protein of 1808 amino acids [ 22 ]: amino acid (aa) positions 333–361 of the S. cerevisiae subunit correspond to positions 334–362 of the A. thaliana protein, and positions 1079–1111 in S. cerevisiae correspond to positions 1716–1748 in A. thaliana . These local homologies are detected also in the human B-block binding subunit by visual inspection (aa positions 367–397 and 1987–2019) (Fig. 2 in [ 22 ]). The Arabidopsis protein reported in Rozenfeld and Thuriaux [ 22 ] appears to be the GI 9665127 protein hit in the PSI-BLAST searches here, because the lengths of their amino acid sequences are the same and the partial sequences shown in their paper are identical to those in the GI 9665127 protein. The alignment shown in Fig. 3C contains the domain reported by Rozenfeld and Thuriaux [ 22 ], and the sequences in the alignment shown in their paper are identical to those in this study. Figure 3 Clustal W alignments of the sequences conserved in the B-block binding subunits and their homologs. The GI numbers of the proteins are shown in parentheses. The amino acid positions of the sequences are shown to the right. The cases of the Clamydomonas sequences are exceptional. These sequences were hit by tblastn of the C. reinhardtii genome sequence using the Arabidopsis protein GI 25402830 as a query (see also Fig. 5). A. Alignment of the N-terminal end sequences. B and C. Alignments of the internal sequences of the proteins. D. Alignment of the C-terminal end sequences. When a PSI-BLAST search was performed using the human B-block binding subunit as a query, it was shown that the C-terminal region also is conserved (Fig. 2A ); for examples, Chironomus , Anopheres , Drosophila , and Arabidopsis (GIs, 25402830, 9665127, and 15218016) proteins are hit with E -values of e -29 , 3e -7 , 6e -54 , 2e -20 , 2e -39 , and 2e -21 respectively (data not shown). These regions contain the domains shown to have sequence similarities [ 22 ] (see above). When a PSI-BLAST search was performed using the S. pombe B-block binding subunit as a query, alignments consisting of its C-terminal region and each of the Magnaporthe , Neurospora , and S. cerevisiae sequences were generated, but no homology to the C-terminal regions of the Arabidopsis proteins was suggested (Fig. 2B ). However, in agreement with the result of Rozenfeld and Thuriaux [ 22 ], when the S. cerevisiae B-block binding subunit was used as a query, the C-terminal region of the Arabidopsis protein (GI 9665127) was hit after four iterations; aa positions 1032–1146 of the S. cerevisiae subunit aligned to positions 1673–1774 of the Arabidopsis protein (GI 9665127) with an E -value of 5e -21 (data not shown). Consequently, the C-terminal sequences of the human, rat, mosquitoes, Drosophila , Arabidopsis and fungi proteins, can be aligned by Clustal W (Fig. 3D ). Are the HMG boxes in the B-block binding proteins? The aa positions 1–68 and 1037–1110 of the S. cerevisiae B-block binding subunit had been tentatively identified as HMG boxes [ 14 ]. The HMG box is a small eukaryotic DNA binding motif (70–80 amino acids in size) found in many proteins including transcription factors [ 23 ]. Interestingly, the regions of HMG boxes predicted for the S. cerevisiae subunit overlap the N- and C-terminal regions conserved in many B-block binding proteins. Therefore, I investigated whether these conserved regions could be homologs of HMG boxes. HMG boxes are diverse, and it is sometimes difficult to determine whether a given protein belongs to the HMG box superfamily [ 24 ]. However, in alignments of known HMG boxes a loose consensus sequence can be defined, in which many basic and aromatic residues are conserved [ 24 - 26 ] (see also Fig. 4B ). Structures of several HMG boxes have been determined by the NMR spectroscopy and X-ray diffraction [for examples, [ 25 , 27 , 28 ]] (Fig. 4B ); most of them have three α-helices arranged in L-shapes, and tertiary structures stabilized by conserved aromatic residues. It should be noted that in alignments of the N- and C-terminal end regions of B-block binding proteins, several basic and aromatic residues also are conserved (Figs. 3A and 3D ). Since HMG boxes contain three helices (see above), I predicted the secondary structures of N- and C-terminal regions of B-block binding proteins using the PSIPRED method [ 29 ]. The results are shown in Figs. 4A and 4C . In the N-terminal end regions, all of the sequences (except for Drosophila ) were predicted to contain three α-helices with corresponding locations (Fig. 4A ). Although the Drosophila N-terminal sequence was predicted to have only two helices, their locations corresponded to the middle and posterior helices in the other sequences (Fig. 4A ). In the C-terminal end regions, all of the sequences were predicted to contain one helix in the same corresponding location (Fig. 4C ). Fig. 4 shows alignments of B-block binding protein sequences and HMG boxes, arranged to show relationships between them. Numerous gaps were inserted into the sequences by visual inspection in order to relate the positions of basic and aromatic residues and locations of the α-helices (Fig. 4 ). It appears possible that HMG boxes are present in B-block binding proteins, particularly in their N-terminal regions, however, strong evidence for this relationship is not clear from comparing their amino acid sequences. Figure 4 Alignments of the N- and C- terminal end sequences of the B-block binding proteins, which are compared with the HMG box sequences. Gaps were inserted into the sequences by visual inspection to relate the positions of the conserved basic and aromatic residues and the locations of the α-helices between the alignments. The amino acid residues common among the three alignments are shown in boldface. A. Alignment of the N-terminal end sequences. Above the amino acid sequences, are the secondary structures predicted from them, where H, E, and C represent helix, strand, and coil respectively. B. Alignment of HMG boxes. The identification codes for PDB entries are shown to the left. The regions shown by black characters form α-helices (see ). C. Alignment of the C-terminal ends, with the predicted secondary structures shown above the amino acid sequences. Evolutionary relationships of the B-block binding proteins PSI-BLAST searches presented here provide the evidence that the B-block binding subunits of vertebrates and yeasts are homologous, and that the Arabidopsis proteins can be used to link these subunits. These results suggest that, with respect to the B-block binding subunits of TFIIICs, animals are evolutionarily closer to Arabidopsis than to yeasts. These results are intriguing because phylogenetic analyses using sequences of small subunit rRNA, elongation factor 1, actin, α-tublin, β-tubulin, and heat shock protein 70, show that animals and fungi are most closely related, to the exclusion of the broad diversity of eukaryotic phyla including plants [ 3 - 6 ]. To confirm that the B-block binding subunits of animals and plants are closely related, I decided to carry out an extensive comparison of the sequences from additional plant taxa. However, the PSI-BLAST searches using the B-block binding subunits of human and S. pombe , and Arabidopsis homolog (GI 25402830) as queries, did not significantly hit any of the plant sequences except the Arabidopsis proteins shown in this study. Therefore, I performed a tblastn search of the Viridiplantae database using the Arabidopsis protein GI 25402830 as a query. The Oryza sativa sequence GI 37990182 was hit with the best E -value (8e -85 ) among Viridiplantae sequences, except the Arabidopsis genes already described (Fig. 5A ). In this Oryza sequence, there are four regions that show high similarities to the query Arabidopsis protein (Fig. 5A ). This result suggests that the Oryza sequence GI 37990182 encodes a B-block binding protein. Subsequently, I performed the PSI-BLAST searches using three conserved domains from the Oryza sequence of GI 37990182 (Fig. 5B ). The query sequence from the 5' end of the Oryza gene had greater similarities to the B-block binding proteins in animals than those in fungi: for examples, the human and rat subunits were hit with E -values of e -26 and 7e -26 respectively, while the S. pombe subunit was hit with an E -value of 0.005, and other fungal sequences were not hit with E -values better than 10 (Fig. 5B ). The results for the query sequence from 3' region of the Oryza gene also indicate that the animal and plant proteins are more closely related: for examples, the human and rat subunits were hit with E -values of 0.009 and 0.12 respectively, while the yeast subunits were not hit with E -values better than 10 (Fig. 5B ). The result for the query sequence from the middle region of the Oryza gene, however, did not retrieve animal subunits, although the S. cerevisiae subunit was hit with an E -value of 0.21 (Fig. 5B ). Figure 5 Protein coding regions in the nucleotide sequences of Oryza sativa and Chlamydomonas reinhardtii with predicted amino acid sequences similar to B-block binding proteins. A. Protein coding regions in the Oryza sativa nucleotide sequence GI 37990182 with amino acid sequences similar to the Arabidopsis protein GI 25402830. Coding regions are shown as boxes and the regions hit by a tblastn search are the filled boxes. E -values are shown below the filled boxes. B. The B-block binding proteins hit by PSI-BLAST searches using the Oryza three amino acid sequences as queries. The corresponding bp positions are indicated above the queries, and hit sequences are below the Oryza queries. Only proteins with E -values better than 10 are shown. C. Protein coding regions in the Chlamydomonas reinhardtii nucleotide sequence with amino acid sequences similar to the Arabidopsis protein GI 25402830. Coding regions are shown as boxes, and the regions hit by a tblastn search are shown as filled boxes. E -values are shown below the filled boxes. D. The B-block binding proteins which were hit by PSI-BLAST searches using the Chlamydomonas amino acid sequences as queries. The corresponding bp positions are indicated above the queries, and hit sequences are below the Chlamydomonas queries. Only proteins with E -values better than 10, are shown. I also searched for a B-block binding protein homolog in the genome of green alga Chlamydomonas reinhardtii . I performed a tblastn search using the Arabidopsis protein GI 25402830 as a query and the C. reinhardtii genome sequence ver2 in the Joint Genome Institute website (see Methods). The Arabidopsis sequences at aa positions of 83–144, 234–262, 698–713, and 1812–1847 showed similarities to sequences corresponding to bp positions of 539027-538842, 527862-537776, 534044-533997, and 526942-526835 in the C. reinhardtii scaffold 16, with E -values of 9.7e -5 , 80.2, 80.2, and 9.7e -5 respectively (Fig. 5C ). The amino acid sequences deduced from bp positions 539027-538842 and 526942-526835 corresponded to the domains conserved among the B-block binding proteins that were aligned by Clustal W, as shown in Figs. 3B and 3D . These results suggest that the Chlamydomonas B-block binding protein is encoded in these DNA regions. Subsequently, I performed PSI-BLAST searches using the two amino acid sequences of C. reinhardtii with the highest similarities to the Arabidopsis protein (Fig. 5D ). The query sequence from bp positions 539231-538197 in C. reinhardtii , showed similarities to the rat B-block binding subunit and its homolog in mouse ( E -values of 7.1 and 5.8 respectively) (Fig. 5D ). Although these E -values are not robust, no fungal B-block binding proteins was hit with E -values better than 10. The other query sequence encoded in bp positions 527524-526466 of C. reinhardtii also had greater similarities to the B-block binding proteins in animals than to those in fungi: for examples, the human and rat subunits were hit with E -values of 4e -68 and 2e -56 respectively, while no fungal proteins were hit with E -values better than 10 (Fig. 5D ). The results of these PSI-BLAST searches with Oryza and Chlamydomonas query sequences indicate that the greater similarity in TFIIIC B-block binding proteins between animals and plants, with yeast as more distant, across the broad diversity of the animal and plant kingdoms. Because yeasts may not be representative of all fungi, it is important to demonstrate that the greater similarity between the animal and plant B-block binding proteins extends beyond the yeast taxa. To this end, I searched for homologs of the B-block binding protein in the basidiomycete genomes. I performed a PSI-BLAST search using the S. pombe subunit as a query, limiting the search to the fungi database. The sequence hit with the best E -value among the basidiomycete sequences, was the Ustilago maydis protein GI 461005911 (Fig. 2B ). The Cryptococcus and Coccidiodes sequences were not hit with E -values better than 10. In the Ustilago sequence GI 461005911, three regions show similarities to the S. pombe subunit, particularly the N-terminal one-third and the C-terminal regions as is true of other homologs in fungi (Fig. 2B ). Subsequently, a PSI-BLAST search was performed using the human B-block binding subunit as a query of the fungi database. Although the S. cerevisiae B-block binding subunit was hit with an E -value of 5e -7 , the Ustilago sequence of GI 46100591 was not hit with an E -value better than 10 (data not shown). Moreover, a PSI-BLAST search performed using the Arabidopsis homolog (GI 25402830) as a query of the fungi database also did not hit the Ustilago sequence of GI 46100591 with an E -value better than 10, although the S. cerevisiae and S. pombe B-block binding subunits were hit with E -values of 0.93 and 9.7 respectively (data not shown). These results indicate that animal and plant B-block binding subunits are more similar to the yeast subunits than to the Ustilago protein GI 46100591. The overall results in this section demonstrate that the greater similarity between the plant and animal B-block binding proteins extends to the green alga protein, and the greater differences in fungi go beyond the yeast taxa. Discussion In this study, I have demonstrated that the B-block binding subunits of TFIIICs in vertebrates are apparently homologous to those of yeasts, by identifying the homologs of each in Arabidopsis . The Arabidopsis proteins (GIs 25402830, 9665127, 15218016, and 25404859), which show strong similarity to B-block binding subunits, are the hypothetical proteins translated conceptually from the nucleotide sequences of the chromosome I [[ 30 ]; see also ]. The lengths of the inferred amino acid sequences of three of these Arabidopsis proteins (GIs 25402830, 9665127, and 15218016), are close to those of the amino acid sequences of the vertebrate subunits (Fig. 2 ). These Arabidopsis proteins probably function as the B-block binding subunits in vivo . B-block binding subunits act on the RNAP III promoters, the sequences of which are conserved in diverse eukaryotes (Background; Fig. 1 ). Thus, the domains of the subunits that bind to these promoters also should be conserved in vertebrates and yeasts. The N-terminal one-third regions of the human, yeast, and the Arabidopsis homologs found in this study probably associate with the B-block sequences. It is interesting that the HMG boxes predicted in the S. cerevisiae B-block binding subunit [ 14 ] overlap with the regions conserved in many of these putative B-block binding subunits. There is a striking degree of similarity in most of the RNAP III transcription machinery in human, S. pombe , and S. cerevisiae ; RNAP III, TFIIIA, TFIIIB, and the TFIIIC subunits that interact with the transcription initiation site, are highly conserved in these three organisms [ 31 ]. On the other hand, the TFIIIC subunits, which interact with downstream promoter regions including the B-block binding subunits, are more divergent [ 31 ]. There is the possibility that substitution rates of the amino acid residues in the B-block binding subunits vary among animals, fungi, and plants, resulting in the high divergence between the human and fungi proteins, and the similarity between the human and plant proteins. Alternatively, evolutionary inferences based on the RNAP III transcription machinery may be different from those of the genes that generally have been used to examine phylogenetic relationships in animals, fungi, and plants. RNAP III transcribes genes encoding tRNA, 5S rRNA, and several viral RNAs, and SINEs (Background). It was reported that molecular phylogenies based on tRNA sequences place plants as the sister group to the animals, although the tRNA data set available at the time was small [ 32 ]. Generally, it is thought that 5S rRNA is convenient for intrakingdom phylogenies, but cannot resolve the question of the animal-plant-fungal divergence because of its short length and high divergence [ 32 , 33 ]. It should be noted that more recent investigations of the proteins involved in RNA metabolism, the mRNA capping apparatus, and several key components that regulate the cell cycle, also suggest a close relationship between animals and plants, with fungi as more distant [ 34 - 36 ]. Conclusions Previously, no evidence of homology between the B-block binding subunits of TFIIICs of vertebrates and yeasts has been reported. PSI-BLAST searches presented here provided the evidence that these subunits are homologous, and that the Arabidopsis proteins can be used to link them. These results imply that, with respect to the B-block binding subunits, animals are evolutionarily closer to Arabidopsis than to yeasts. Comparisons of the B-block binding proteins from additional plant taxa showed that the greater similarity between plants and animals extends to the green algae Chlamydomonas . It was also demonstrated that the differences in fungi go beyond the yeast texa, and occur in basidiomycetes. These are interesting because molecular phylogenetic analyses using the small subunit rRNA and ubiquitous proteins, show that fungi are more closely related to animals than either is to plants. Methods To search for similarities between the B-block binding subunits of vertebrates and yeasts, I used the PSI-BLAST program in the NCBI website [ 19 ]. PSI-BLAST searches were performed by default: matrix was BLOSUM62; gap costs were Existence 11 and Extension 1; and the E -value of threshold was 0.005. Peptide sequence databases used for the PSI-BLAST searches were all non-redundant GenBank CDS translations, RefSeq proteins, PDB, SwissProt, PIR, and PRF (total 1605642 sequences). The PSI-BLAST was limited searches of the eukaryota databases, when the amino acid sequences from the Oryza and Chlamydomonas coding regions were used as queries. The fungi database was used in a search for homologs of the S. pombe B-block binding subunit in basidiomycetes. PSI-BLAST was run three times for each of the queries. To search for homologs of the Arabidopsis protein GI25402830 in the plant sequences, the tblastn program was used [ 19 ]. To search for homologs of the Arabidopsis protein GI25402830 in the Chlamydomonas reinhardtii sequences, the tblastn program at the Joint Genome Institute website , was used. Clustal W in the EMBL-EBI website was used to align the multiple amino acid sequences [ 21 ]. Clustal W was performed by default: matrix was Gonnet 250; the penalty for opening a gap was 10; the penalty for extending a gap was 0.05; and gap separation penalty was 8. Secondary structures of the proteins were predicted by using the PSIPRED protein structure prediction server (PSIPRED v2.4 in ) [ 29 ].	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514540.xml
524483	Evolution and distribution of RNA polymerase II regulatory sites from RNA polymerase III dependant mobile Alu elements	Background The primate-specific Alu elements, which originated 65 million years ago, exist in over a million copies in the human genome. These elements have been involved in genome shuffling and various diseases not only through retrotransposition but also through large scale Alu-Alu mediated recombination. Only a few subfamilies of Alus are currently retropositionally active and show insertion/deletion polymorphisms with associated phenotypes. Retroposition occurs by means of RNA intermediates synthesised by a RNA polymerase III promoter residing in the A-Box and B-Box in these elements. Alus have also been shown to harbour a number of transcription factor binding sites, as well as hormone responsive elements. The distribution of Alus has been shown to be non-random in the human genome and these elements are increasingly being implicated in diverse functions such as transcription, translation, response to stress, nucleosome positioning and imprinting. Results We conducted a retrospective analysis of putative functional sites, such as the RNA pol III promoter elements, pol II regulatory elements like hormone responsive elements and ligand-activated receptor binding sites, in Alus of various evolutionary ages. We observe a progressive loss of the RNA pol III transcriptional potential with concomitant accumulation of RNA pol II regulatory sites. We also observe a significant over-representation of Alus harboring these sites in promoter regions of signaling and metabolism genes of chromosome 22, when compared to genes of information pathway components, structural and transport proteins. This difference is not so significant between functional categories in the intronic regions of the same genes. Conclusions Our study clearly suggests that Alu elements, through retrotransposition, could distribute functional and regulatable promoter elements, which in the course of subsequent selection might be stabilized in the genome. Exaptation of regulatory elements in the preexisting genes through Alus could thus have contributed to evolution of novel regulatory networks in the primate genomes. With such a wide spectrum of regulatory sites present in Alus, it also becomes imperative to screen for variations in these sites in candidate genes, which are otherwise repeat-masked in studies pertaining to identification of predisposition markers.	Background In the post genome sequence era, repetitive sequences, erstwhile considered junk and devoid of function, are increasingly being implicated in many cellular functions, genome organization and diseases [ 1 - 8 ]. Alu repeats, which belong to SINE (short interspersed nucleotide elements) family of repetitive sequences, are present exclusively in the primate genomes. These elements which are ~300 bps in length have originated from the 7SL RNA gene and comprise of two similar, but not identical subunits [ 9 - 12 ]. Each element contains a bipartite promoter for RNA polymerase III, a poly (A) tract located between the monomers, a 3'-terminal poly(A) tract, a number of CpG dinucleotides, and is flanked by short direct repeats [ 13 , 14 ]. Based on certain diagnostic site mutations, they have been broadly classified into three subfamilies: Old (Alu Js), Middle (Alu S) and the Youngest (Alu Ys) [ 15 , 16 ]. Further, some of the Alu Y sequences are very new and exhibit polymorphisms, indicating that they have recently undergone retropositioning process [ 17 ]. Alus have been shown to harbor a number of regulatory sites like hormone response element (HRE), and a couple of ligand activated transcription factor binding sites [ 18 - 24 ]. These sites regulate the expression of downstream genes, in some cases in a temporal or tissue specific manner. Most of the regulatory sites in Alus have been reported during the course of characterization of specific genes [ 25 - 32 ]. Besides, the intrinsic A-Box and B-Box RNA polymerase III (RNA pol III) sequences and the recombinogenic sites present in these elements are involved in retrotranspositional and recombination process [ 10 ]. Alus originally demonstrated to have non uniform distribution on the chromosomes through banding studies [ 33 , 34 ] have been recently substantiated by genome sequence analysis [ 35 ]. It has been observed that that Alus not only show a non- random pattern of distribution in the human chromosomes but also varying densities within genes. Additionally, in a genome wide expression analysis, co-variation of expression of gene pairs has been attributed to sequence similarity metric in the upstream region of promoter predominantly contributed by Alu repeats present in these regions [ 36 ]. These effects of Alu have been shown to be completely independent of the effects of isochoric (GC) composition on Alu density as well as gene expression [ 34 - 36 ]. Identification and analysis of various permutations and combinations of these regulatory elements in otherwise conserved repetitive Alus are mostly excluded from genetic analysis. Since, Alus occupy a tenth of the human genome, it is imperative to identify those, which might assume function in the proper context. Our primary aim in this analysis is to find out if any bias exists in the distribution of transcriptional regulatory sites in Alus of various evolutionary ages and their distribution with respect to the functional classes of genes. Results and Discussion Distribution of functional sites in Alus is position specific As a first step toward examining the role of these regulatory sites, we mapped their most probable positions on Alus, using in house developed algorithms (Figure 1 ). This was carried out on 500 Alus, each of Alu Jo, Alu Jb, Alu Sx, Alu Sc, Alu Yb8 and Alu Y subfamilies. The classification of these evolutionarily distinct subfamilies are based on diagnostic sites [ 15 , 16 , 37 , 38 ]. Besides, members of the most recent and retropositionally active and polymorphic Alus were also included in the analysis [ 39 , 40 ]. Though the polymorphic Alus belong to Alu Y subfamily, these were treated as a separate category since insertion/deletion of these Alus have been associated with many phenotypes/diseases [ 2 ]. The regulatory sites show positional conservation across all subfamilies in which they are represented (Table 1 ). However, these sites are distinct from the diagnostic sites, which are used for classifying Alus, which suggests that they have not arisen randomly in different subfamilies. Table 1 Position of sites analysed in Alu repeats in various subfamilies. Family A-box B-box AML MPO CETP Rec AP1 ERE RARE TRE nCaRE LXR Jb 5 76 48 48 47 22 13/221 80 57–76 -67 289 Jo 5 76 48 48 47 22 13/221 80 66 -67 289 224–240 Sx 5 76 48 48 47 22 13 80 60 -67 289 237–250 Sc 5 76 48 48 47 22 13/267 80 68 -67 289 Y 5 76 48 48 47 22 80 -67 289 Yb8 5 76 48 48 47 22 13/270 80 60–66 -67 289 230–240 POLY 5 76 48 48 47 22 13/267 80 60 -67 289 Figure 1 Representation of regulatory sites on Alu elements. 500 representative Alu sequences each of distinct evolutionary ages were selected for identification of most probable regulatory sites. 126 polymorphic Alus (POLY) from younger subfamilies which show insertion – deletion polymorphisms were also analysed. Sites were identified using local alignment based program as well as by probabilistic modelling approach. These sites are positionally conserved in all subfamilies. Evolution of regulatory sites is biased and clustered in Alus Nearly all the analyzed regulatory sites for RNA polymerase II (RNA pol II) are distributed in the region between A- Box and B-Box with more clustering near the B-Box region (Figure 1 ). There is an evolutionary age specific loss / gain of these sites in various subfamilies leading to a bias in their distribution (Figure 2 ). Newly transposing Alus have methylated CpG sites, which are prone to transition. Many sites seem to have evolved as a consequence of these transitions. The regulatory elements are most abundant in the middle subfamilies and least represented in the younger Alus. Some sites like AP1, ERE, nCARE are present in older and middle Alus but rarely so in the younger as well as polymorphic Alus. An opposite trend is observed for CETP, wherein the highest density is observed in the younger active and polymorphic Alus. RARE and TRE sites are retained in all subfamilies whereas LXR is specific to only middle Alu subfamilies (Figure 2 ). It is curious, nCARE which is also present in the 7sl RNA, the progenitor of Alus, is not equally represented in all Alus and has higher density in the older Alus and middle and is very poorly represented in the younger subfamilies. Figure 2 Distribution of regulatory sites in various Alu subfamilies as well as polymorphic Alus. On the X-axis Alus of different evolutionary ages as well as polymorphic Alus (POLY) are represented. On the Y-axis the percentage of elements carrying these sites in various subfamilies is indicated. Evolution from retropositionally active to transcriptionally active Alu elements Majority of Alu retroposition has ceased at least 30 million years ago and only a few Alu subfamilies are still active [ 15 , 17 , 41 ]. Transcription of Alus is a prerequisite for retrotransposition and there is regulation not only during transcription initiation but also at the level of stability of transcripts [ 42 ]. Alu elements are transcribed by RNA pol III which are composed of two properly spaced conserved sequence motifs, an upstream element named the A-Box and a downstream element called the B Box which are essential for efficient transcription. Deletion of the Box B sequences within the Alu repeat completely abolishes the transcriptional activity. In the absence of box A sequences even though there is a reduction in efficiency of transcription by 10 to 20 fold, B-Box sequence is still capable of initiating transcription 70 bps upstream [ 43 , 44 ]. An intact A Box is therefore a critical determinant for RNA pol III retropositional activity. Besides, it has been shown by in vitro as well as in vivo studies in the 'B' Box that 'G' and 'T' residues at the 1 st and 3 rd positions respectively are very critical for it's functioning [ 45 ]. Our analysis on the distribution of these promoter elements show that the polymorphic Alu sequences have the highest density of A Box (70%) and is almost absent in older subfamilies (Figure 2 ). Since the younger Alus are considered to be transcriptionally more active, this fits in well with the loss of this site in the course of evolution due to accumulation of mutations. The B Box motif with the sequence G(A/T)T(C/T)RANNC shows a similar trend as the A Box. Interestingly, a fraction of older Alu subfamily still retains the B-Box sequence. However, 'A' residue at the second position which has not been shown to be critical for transcription is a diagnostic nucleotide [ 39 ] for the younger subfamilies. This could result in the increased proportion for B-Box in the younger families. We observe a very curious distribution of the B Box motif if we consider the sequence GTT(C/T)GAGAC (B'Box in Figure 2 ) wherein we restrict the pattern to the experimentally validated sequence. Alu Sx and Alu Sc have the highest density match with this pattern, followed by the older subfamilies and it is present in only < 2% frequency in AluY and polymorphic Alus. The "C" at the 4 th position in this case is mutated to "T" in the older families. The Yb8 family that has been reported to be transcriptionally and retropositionally active amongst the younger subfamilies, retains the B'-Box element in a significant fraction. This suggests that even though retropositionally competent younger Alus are hypothesized to be transcriptionally active, only a minority retains consensus B'-Box. It is possible that the enhancing activity of the A Box is sufficient to drive transcription from the weaker B'- Box in the younger subfamilies. Our findings corroborates well with an earlier study in which presence of all subfamilies in the RNA polymerase III driven Alu transcript pool was reported [ 46 ]. Additionally, it was also observed that though there was a quantitative bias towards younger subfamilies and younger members of these subfamilies (based on their relative presence in the transcript compared to their abundance in the genome), there was a preferential expression of the middle subfamilies relative to the most active subfamilies. Our observations, therefore, further rules out the hypothesis that transcription may be biased only towards retropositionally active subfamilies of Alu elements. This could be the reason why only a fraction of younger Alus is currently retrotranspositionally active. The presence and retention of B-Box coupled with near absence of A Box in the Alu Sx and AluSc families suggests basal level of transcription from these elements which could be enhanced through binding of other regulatory proteins under certain conditions such as stress [ 47 ]. Additionally, with evidence of presence of naturally occurring Alu antisense as well as edited Alu transcripts [ 48 , 49 ], transcribing Alus could play a major role in yet unknown biological processes. Exaptation of Alus in the transcriptional regulatory repertoire Alus have been demonstrated to exert effects at transcription, post-transcription as well as at the translation level. In an earlier study on complete chromosomes 21 and 22, we have demonstrated that the Alu elements are clustered in genes of signaling, metabolic and transport proteins and rarely present in the structural and information proteins [ 50 ]. This clustering bias was found to be irrespective of genomic location, GC content, length of genes or intronic content. To further address whether the Alus harboring transcriptional regulatory sites also show a selective distribution and thereby exert effects on transcription, we analyzed their distribution in the genes of various functional categories of chromosome 22. Two different datasets 1) Promoter region Alus and 2) Intronic region Alus, harboring regulatory sites were analyzed. The promoter region Alus of genes involved in metabolism, signaling were significantly rich in regulatory sites compared to those of information, structure and transport (F value = 4.86, df = 4, 40, p-value < 0.0027). In the intronic regions, distinction in their distribution with respect to functional categories was not so significant though the intronic regions also harboured Alus containing regulatory sites (F value = 2.92, df = 4,40, p-value = 0.032). Since the genes of the signaling and metabolic pathway are more subject to regulation by cellular cues like hormonal triggers, this observation is significant. Most of the Alus in the promoters belong to the middle Alu S families and rarely Younger Alus are present. Since younger Alus also harbour few regulatory sites and actively retropose, it is possible that there is a negative selection against their insertion in the promoter sites of genes of information pathways and structural proteins [see the supplementary data ]. Alu movements and aberrant gene expression Gene inversions, duplications and formation of pseudogenes have been extensively reported to be mediated both through retrotransposition as well as recombination of Alus. This, in many cases, has also been associated with aberrant gene expression. For instance, presence of AML sites in an Alu upstream of MPO gene, has been first demonstrated to be associated with Acute Myelocytic Leukemia [ 20 ]. This is due to the presence of a strong SP1 site within AML which leads to over expression of MPO gene. AML sites are most abundant in younger and polymorphic Alus and a single base pair transition results in MPO site, present predominantly in the members of older subfamilies. In the case of polymorphic Alus, many sequences that do not show 100% conservation of AML site still retain the SP1 site. Interestingly, the core recombinogenic site is also most predominant in younger and polymorphic Alus. The presence of recombinogenic sites in polymorphic Alus, could therefore not only contribute to genome shuffling but also serve to distribute ectopic sites such as AML through retrotransposition and recombination (Figure 2 ). Regulatory region distribution through Alu expansion Analysis of regulatory sites within Alus suggests that a polymorphic Alu has the potential to transpose and recombine which allows it to integrate at random sites in the genome. They also harbour potential regulatory sites, which could evolve to become accessory sites for RNA pol II transcription as revealed by their clustering in older subfamilies. Further, the Alu sequence due to acquisition of novel functions could form a part of the transcription repertoire involved in the regulation of the downstream /associated genes and create novel regulatory networks (Figure 3 ). These results also corroborate with the hypothesis of evolution of transposable elements of Kidwell [ 51 ] wherein they had proposed a 3 stage life cycle of class II Transposable elements:- invasion and amplification followed by mutations and maturity and finally senescence and fading. In the case of Alu, instead of fading, they could also evolve to become members of host regulatory machinery. Figure 3 Alu expansion and evolution of regulatory sites. With the help of LINEs, Alu may keep on retro-transposing or may get inactive/negatively selected. Alternatively, it may integrate upstream of a gene, accumulate mutations, evolve RNA pol II regulatory sites, get stabilized and control gene expression. This is supported by the presence of sparse regulatory sites, unhindered A box, recombinogenic sites initially in the younger and active Alus and its accumulation in older Alu subfamilies as well as significant presence of Alus harbouring regulatory sites in the promoter encompassing regions of the genes of signaling and metabolic pathways. Conclusions Comparison of sequences in the regulatory regions of many homologous genes in human have shown accumulation of Alus, not only post divergence from non-human primates but also during primate evolution [ 52 ]. Perhaps, recruitment of cis regulatory elements responsive to cellular cues through Alu elements could result in altered spatial and temporal transcription of genes as well as create novel metabolic and signaling networks. These might contribute to the observable physiological complexity in human and primates [ 53 ]. Additionally, the underlying events which would be defining event of speciation of human from chimpanzee (with which it shares nearly 99% homology at coding level) still eludes identification and might to some extent reside in such genomic elements. These issues can now be addressed through comparison of these sites in human and chimpanzee. Currently, Alus are repeat-masked in all studies pertaining to identification of predisposition markers in complex disorders. With such wide spectrum of nuclear receptors, which play a major role in maintaining normal physiological state and affect as diverse processes as development, reproduction, general metabolism, residing in Alus, it therefore becomes imperative to screen for variations in these sites. This might have important consequences in the candidate genes for those complex diseases that are triggered in response to hormonal imbalances as well as other environmental cues. Methods 126 polymorphic Alu sequences cited in literature [ 39 , 40 ] were retrieved using NCBI BLAST and Repeat Masker software[ 54 , 55 ]. The analysis was carried out on Alu repeats of human chromosome 22. A randomly selected representative set of approximately 500 Alu sequences, each of distinct evolutionary ages, Alu Jb, Alu Jo, Alu Sx, Alu Sc, Alu Yb8 and Alu Y were used for the analysis. Sequences were retrieved from Sanger Institute Home Page, June 2001 release [ 56 ]. Besides, Alus were also analyzed within 5000 base pairs upstream of genes of chromosome 22 in the regulatory regions encompassing promoter sequences as well as inside their intronic regions. Collection of biologically active sites Information about the regulatory sites and their sequences was collected from various literature sources (Table 2 ). Characteristic features of the sites are given below. We selected those regulatory sites, which have been shown to have function in the Alu elements. The A Box and B Box sequences define the bipartite internal promoters, which bind RNA polymerase III. MPO and AML sites, which are 14 nucleotides differ by an A / G at 5th position of the sequence and transition from G to A at this site converts the MPO allele to AML, resulting in the formation of a strong SP-1 binding site and over expression of the following gene. AP1 sites bind AP-1 transcription factor, which is a dimeric complex that contains members of the JUN, FOS, ATF and MAF protein families. Hormone responsive elements (HRE) are super family of binding sites for ligand activated nuclear hormone receptors for thyroid hormone (TRE), retinoic acid (RARE) and vitamin D, which regulate gene transcription. Estrogen response elements (EREs) are sites for binding of estrogen receptor (ER), a ligand-activated enhancer protein that is a member of the steroid/nuclear receptor super family and transactivates gene expression in response to estradiol. The negative calcium response element type 2 (nCARE) is a regulatory DNA sequence, which inhibits transcription in response to raised extra cellular calcium levels. The nuclear receptors liver X (LXR) is involved in different cell-signaling pathways. CETP site is an orphan receptor site in the Alu in promoter of cholesteryl ester transfer protein (CETP) which plays a key role in reverse cholesterol transport in mediating the transfer of cholesteryl ester from HDL to atherogenic apolipoprotein B-containing lipoproteins. Table 2 Sequences of regulatory elements analysed in Alu repeats. Site Sequence Retinoic acid response element (RARE) 5'(AG)G(GT)TCA 3' Estrogen Response Element (ERE) 5'(GA)(GA)TCA(CG)(AC)(CG)TGACC 3' Negative calcium response element (nCARE) 5' TGAGACNNNGTCTCAAAAA 3' Liver X receptor 5' GACCTNNNNTGATCC 3' Cholestryl esterase transferase response element (CETP) 5'CCGNGGCGGGC 3' AP1 site 5' T(GTA)A(GC)TCA 3' Acute Myelocytic Leukemia (AML) site 5' AGGCGGGTGGATCA 3' Myelo Peroxidase (MPO) site 5' AGGCAGGTGGATCA 3' Recombinogenic site 5'CCCTGTAATCCTAGCACTTTGGAGGC 3' A-Box 5' GGGCGCGGTGGC 3' B-Box 5' G(A/T)T(C/T)RANNC 3' B'Box 5' G TT(C/T)GAGAC 3' Nucleotide sequences in parenthesis indicate alternate nucleotides and have been written in increasing order of their preference. Computational methods Two different programs were written in order to locate the most probable biologically significant regions. A local alignment based program, Xalign, was implemented in C++, Red Hat 7.3 based Linux. This program finds the probable sites by aligning the consensus of regulatory site with the query sequence. Multiple queries with a size upto 600 nucleotides can be taken at a time. Another program, Promotif, was implemented in C++, Red Hat 7.3 based Linux, using the probabilistic modeling approach. It uses the position weight matrix, normalization of the positions with conservation index (Ci Value), and inter-nucleotide dependence in terms of transition matrix to find out the sites. Position weight matrices were generated using Gibbs Motif Sampler, for every site included in the program. The sequences for position weight matrix generation were carefully selected based on the sequence and length reported for each binding site. The final length for search was fixed at the lowest length observed. This provides element specific matrix with lesser chance for the selection on non-RE regions. For the sites analyzed, it had an in built transition matrix, position weight matrix and conservation index. Batch analysis of over a thousand Alu sequences can be performed with this program. Using the annotated sequences from literature as well as from NCBI web page, training set for the probabilistic model was created. Training was done for approximately 70% sequences and rest of the sequences were taken as test set. Details of the program along with the equations used are available on request. Mapping of recently integrated and younger Alus About 126 recently integrated Alus from younger subfamilies were searched in the human genome using BLASTn at NCBI server and regulatory sites were mapped in these regions using the programs discussed above. Association analysis Alus in the promoter regions and intronic regions of functionally classified genes [ 50 ] of chromosome 22 were mapped and pattern of distribution of biologically significant sites were analyzed by ANOVA. Authors' contributions RS developed the algorithms and programs for identifying regulatory and significant regions, carried out the analysis of distribution of these sites in Alu subfamilies, association analysis and drafted the manuscript. DG was involved in chromosome 22 analyses. SKB participated in the design of the study. MM conceived of the study, participated in its design, analysis, coordination and manuscript preparation. All authors read and approved the final manuscript. Supplementary Material Supplementary data The analysis over the promoter and intronic regions has been performed through the data given in the supplementary table file, supplementary table 3_ravishankar et al. Format: .xls. For human chromosome 22, the data contains the accession number, associated Alu family, the respective positions, functional class of the region and further details, for each associated regulatory element found within the Alu repeats in the 5' flanking promoter and intronic regions. The zipped file name is supplementary 1.zip. Details about programs used are on request for academic users. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524483.xml
368175	mRNA Targets of RNA-Binding Proteins Suggest an Extensive System for Post-Transcriptional Regulation	null	The single-celled Saccharomyces cerevisiae , commonly known as baker's yeast, measures just 2 microns—it takes about 4 billion to fill a teaspoon. But as a eukaryote (its cells have nuclei), its genes function in much the same way a human's do. For a gene to function, its DNA sequence must first be transcribed into RNA (called messenger RNA, or mRNA), whose sequence can then be translated into a specific string of amino acids to form the unique protein that the gene encodes. The population of mRNA transcripts in each cell (its “transcriptome”) is dynamic—the genome uses its vocabulary of genes to write an ever-evolving script for the cell as its life unfolds and its environment changes. By binding to specific sequences of DNA, proteins called transcription factors process signals from the cell's sensory and information-processing systems to control which genes are transcribed in each cell, under what conditions, and at what rate. While the steps and regulatory programs that govern gene expression at this level are reasonably well known, much less is known about the orchestration of the later steps in the gene expression program—where in the cell each mRNA molecule goes when it leaves the nucleus, at what rate and under what conditions it is translated into protein, and how long it survives. Cluster of RNA targets for Puf proteins RNA-binding proteins (RBPs) have been implicated in diverse aspects of post-transcriptional gene regulation. Hundreds of RBPs are encoded in the eukaryotic genome, but because few have been studied in detail and few of their mRNA targets are known, the nature and extent of an RBP-mediated post-transcriptional program has been obscure. Now a systemic analysis of a specific family of RBPs and their mRNA targets in yeast by André Gerber, Daniel Herschlag, and Patrick Brown, of Stanford University, suggests that such a program may exert detailed control over the life history of every mRNA. By selectively binding and regulating specific classes of mRNAs, RBPs may provide a mechanism to coordinate the collective fate of these transcripts and serve as an integral part of the global transcriptome. Gerber, Herschlag, and Brown focused on the binding targets of a family of RBPs called Pumilio-Fbf (Puf) proteins, which are defined by the presence and configuration of an amino acid domain that mediates RNA-binding. Little is known about the physiological function of the five yeast Puf proteins the researchers studied here (called Puf1p-Puf5p). After using “affinity tags” to snag each of the five Puf proteins from yeast cells, together with their bound mRNA targets, the researchers identified the associated mRNAs with microarray analysis. They found more than 700 mRNAs bound by at least one Puf protein, with each Puf RBP targeting a distinct group of mRNAs. The group of mRNAs associated with each Puf protein turned out to encode proteins with strikingly similar functions and locations in the cell. Many of the mRNA sets encode proteins that reside in the same cellular location, are part of the same protein complexes, or act in the same signaling pathway. Some Puf proteins target mRNAs that encode membrane proteins while others preferentially bind to mRNAs that encode proteins involved in cell division. The most pronounced bias occurs with Puf3p, which overwhelmingly binds mRNAs that encode proteins destined for the mitochondria, the cell's power generators. This selective tagging of functionally related mRNAs by specific RBPs suggests a mechanism for coordinated global control of gene expression at the post-transcriptional level. Just as transcription factors regulate transcription by binding to specific DNA sequences, RBPs may mediate regulation of the subcellular localization, translation, and degradation of the set of specific mRNAs they target. Noting the striking themes in the subcellular localization of the proteins encoded by the mRNAs bound by each Puf protein, Gerber, Herschlag, and Brown propose that RBPs may play important roles in the subcellular localization and efficient assembly of protein complexes and functional systems by ensuring that the location in the cell at which mRNAs are translated “is not left to chance.” Since the number of RBPs encoded in eukaryotic genomes approaches that of transcription factors, the regulatory program that controls the post-transcriptional fate of mRNAs—their localization, translation, and survival—may prove to be nearly as diverse and complex as the regulation of transcription itself.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC368175.xml
517713	An economic analysis of premarriage prevention of hepatitis B transmission in Iran	Background To assess the economic aspects of HBV (hepatitis B virus) transmission prevention for premarriage individuals in a country with cultural backgrounds like Iran and intermediate endemicity of HBV infection. Methods A cost-effectiveness analysis model was used from the health care system and society perspectives. The effectiveness was defined as the number of chronic HBV infections averted owing to one of the following strategies: 1) HBsAg screening to find those would-be couples one of whom is HBsAg positive and putting seronegative subjects on a protection protocol comprising HBV vaccination, single dose HBIG and condom protection. 2) HBsAg screening as above, in addition to performing HBcAb screening in the HBsAg negative spouses of the HBsAg positive persons and giving the protocol only to HBcAb negative ones. Sensitivity and threshold analyses were conducted. Results The cost of each chronic infection averted was 202$ and 197$ for the strategies 1 and 2, respectively. Sensitivity analysis showed that strategy 2 was always slightly cheaper than strategy 1. The discounted threshold value for the lifetime costs of chronic liver disease, above which the model was cost saving was 2818$ in strategy 1 and 2747$ in strategy 2. Conclusions Though premarriage prevention of HBV transmission in the countries with cultural backgrounds similar to Iran seems cost saving, further studies determining precise costs of HBV infection in Iran can lead to a better analysis.	Background Hepatitis B is an important health problem and a major cause of acute and chronic hepatitis, cirrhosis and hepatocellular carcinoma. Approximately 30% of the world population (1.8 billion people) have the serologic evidence of HBV infection of whom 350 million are estimated to suffer from chronic HBV (hepatitis B virus) infection; at least 500,000 chronically infected people die of liver malignancy and cirrhosis each year [ 1 ]. According to Iranian studies, about 22% to 37% of general population in Iran are HBcAb positive [ 2 , 3 ] (e.g. previous exposure to HBV) and about 1.3% to 8.69% of the population are chronic HBV carriers [ 2 - 6 ]. Compared to the United States where HBV is the cause of 25% of chronic hepatitis cases, HBV accounts for up to 70% to 80% of chronic hepatitis cases in Iran [ 7 ]. Therefore, HBV alone is the leading cause of chronic liver disease (CLD) in Iran and it is evident that HBV transmission prevention can be one of the health priorities in the country. We believe that major routes of HBV transmission, local epidemiological factors and the already performed prevention programs of each region are among important factors in identifying the populations at risk of the infection and planning the region-specific prevention strategies. In Iran, universal neonatal vaccination against HBV started in 1993 according to WHO recommendations. It means that Iranians 10 years of age or older at the time of this study received no prevention services against HBV and so most of them could contract the infection if attacked by the virus. For this at-risk population, several preventive strategies can be suggested which of course should have economic justification. Premarriage transmission prevention can be considered as one of the possible solutions to protect this population though it should not be regarded as the only or the best solution available. Sexual contact is one of the common routes of HBV transmission. In Iran, due to a particular cultural and religious background, homosexuality is not known as a common phenomenon compared to the western countries. For the same reasons, it is very unlikely for an individual to have sexual contact (especially in the form of intercourse) with his/her would-be spouse. On the other hand, almost all premarriage individuals (those considering legal marriage) are obliged by Iranian law to undergo a predefined battery of screening tests in government-designated laboratories; this can make premarriage individuals an accessible group for a preventive intervention. Finally, since most premarriage individuals are in young age groups, they have a rather long life expectancy, which allows enough time for them to suffer from chronic complications of HBV infection in productive years of life. We decided to perform this study to provide health policy makers in Iran and those countries with similar demographic conditions (especially in the Middle East) with an economic analysis of premarriage prevention of hepatitis B transmission. Methods Model The economics of performing two rather similar strategies in addition to no intervention strategy were compared using a decision tree (Fig. 1 ). The software used for analysis was Decision Analysis by TreeAge (DATA™, Williamstown, MA, USA). As shown in Fig. 1 , the overall options available for premarriage individuals can be one of the major strategies mentioned below: 1. Screening all premarriage individuals for HBsAg and then performing the following prevention protocol (marked as P.P. in Fig. 1 ) for HBsAg negative individuals whose would-be spouse is HBsAg positive: a. Three-dose HB vaccine (0, 1, 6 mo) b. Single dose HBIG injection c. Using condoms (2 boxes/mo) during all intercourses for 7 months d. Measurement of HBsAb (hepatitis B surface antibody) 1 month after the 3rd dose of the vaccine e. An extra dose vaccine and additional condom protection for another month for the persons whose HBsAb is not in protective ranges (lower than 10 IU/l) No protection protocol is considered for those couples who are both HBsAg positive or negative. 2. Screening all premarriage individuals for HBsAg followed by rescreening of the HBsAg negative spouses of HBsAg positive persons for HBcAb and finally performing the prevention protocol (e.g. the a to e steps above) only for HBcAb (and HBsAg) negative individuals whose would-be spouse is HBsAg positive. No protection protocol is considered for those couples who are both HBsAg positive or negative. 3. No screening and no prevention. The main analysis considered the perspective of health care system. However, in the final analysis (including the sensitivity analysis), a threshold analysis was performed from societal perspective. The results were expressed by the cost per chronic infection (e.g. more than 6 months HBsAg positive) averted (e.g. average cost effectiveness). Assumptions The assumptions in the model included: 1) HBV vaccination is not harmful for the individual or community; 2) the efficacy of preventive methods such as 3-dose HB vaccine, HBIG injection and condoms in preventing sexual transmission of HBV do not significantly vary by geographic region; 3) the compliance of the population for receiving the preventive methods is 100%; 4) the preventive methods are available throughout the country and can cover 100% of the population; 5) HBcAb positive persons are not at risk of HBV infection and this group will not be considered for receiving preventive measures; 6) all HBsAg positive individuals are HBcAb positive too; 7) the sensitivity and specificity of the screening tests (for HBsAg and HBcAb) are 100%; 8) the average age at marriage is 25 years of age for both sexes; 9) the costs associated with CLD are incurred during a 10-year period starting at the age of 50 [ 1 , 24 ]; 10) the transmission rate of HBV from men to women and vice versa are equal; 11) premarriage individuals do not have intercourse before marriage; and 12) Iran is located in a region with intermediate prevalence of hepatitis B. Probabilities and costs The probabilities included in the decision analysis model (Table 1 ) were assumed to be of 2 types: 1) the probabilities that do not significantly vary by geographic location (e.g. the efficacy of 3-dose HB vaccine, HBIG injection and condom in preventing sexual transmission of HBV) and 2) the ones that seem to be significantly different in various geographic locations (e.g. HBsAg or HBcAb prevalence rates in the society). To have preliminary estimates of the latter category probabilities, all available and accurate Iranian medical literature (published from 1993 to 2003) were reviewed. For the former category probabilities, international resources (PubMed) were included in addition to the Iranian sources mentioned above. If a probability was not found in medical literature, the consensus of an expert team including 5 gastroenterologists collaborating with our research center was considered as the base case value. A review of Iranian studies showed that the prevalence of HBsAg in general population varied from 1.2% to 8.69% in different parts of Iran [ 2 - 6 ]. In this study, we considered an average rate of 2% as the baseline in a range of 1% to 9% (Table 1 ). The Iranian studies also showed that the prevalence of HBcAb varied from 15% to 37% in different parts of Iran. Therefore, we assumed an average rate of 20% as the baseline in a range of 15% to 40% (Table 1 ) [ 2 , 3 ]. The review of international and Iranian medical literature revealed that the probability of being HBsAg positive for an HBsAg positive person's spouse (P3) is about 4% to 15% [ 8 - 11 ]. We assumed a baseline probability of 5%. (Table 1 ) The probability of becoming HBsAg positive for an HBsAg positive person's spouse after receiving the prevention protocol (P4) was indirectly calculated by the formula below: P 4 = P 3 (1 - efficacy of preventive protocol) The baseline value for the efficacy of the prevention protocol used in our model to protect against spouse-spouse HBV transmission (3-dose HB vaccine, HBIG injection and condom protection up to complete immunity) was assumed to be 90%. The figure was reached by considering the values found in the literature for the efficacy of 3-dose HB vaccine [ 12 , 13 ], HBIG injection [ 14 - 19 ] and condom protection [ 20 - 23 ] which was finally modified by consensus from the expert team described above. The most pessimistic and optimistic estimations for the efficacy of the prevention protocol were assumed to be 75% and 100%, respectively. Therefore, the baseline value for P 4 was assumed to be 0.05% in a range of 0% to 1.25%. The direct medical costs of interventions (Table 2 ) were extracted from the resources and tariffs of Iranian Health Ministry, Iran Pasteur Institute and Iranian Transfusion Organization in 2003 (unpublished data) and were used as baseline costs in the model. The indirect medical costs of the intervention such as transportation and time costs for the recipients of the preventive methods (to receive the services) were assumed to be zero and were not included in the model. The costs of the averted morbidity (HBV infection especially chronic liver disease) were not directly put in the model because of unavailability of relative Iranian studies. However, as a solution to better analysis, the latter costs were calculated as a final variable in a threshold analysis, considering the fact that of the adults in chronic carrier state, 15% will eventually develop chronic liver disease (CLD) [ 1 , 24 , 25 ]. In most economic analyses, the costs are modified for the outcomes occurring in the future, a process called discounting. Considering the average age at marriage and the age at which CLD starts (see assumptions) the cost of CLD calculated through the threshold analysis was discounted by a discount rate of 3% to the beginning of the 10-year period of CLD development. The costs are expressed in the text and tables in US $ and Iranian Rials (1US $ = 8300 Iranian Rials). The currency conversion rate reported here is the one for mid-2003 when the study was performed. Sensitivity analysis Uncertainty management, which is one of the central processes in decision-making, usually involves working with probabilities that usually vary in different circumstances. Therefore, the outcome values and final decision is prone to change when the value of probabilities change. Sensitivity analysis is a method in which the final decision and the value of outcomes are estimated while each probability (univariate) or combinations of probabilities (multivariate) are varied in a reasonable range. This will reveal the variables whose change the model is sensitive to. In this study, univariate sensitivity analysis was performed for prevalence of HBsAg positivity in general population (P 1 ), prevalence of HBcAb positivity in general population (P 2 ), probability of becoming HBsAg positive for an HBsAg positive person's spouse (P 3 ) and probability of becoming HBsAg positive for an HBsAg positive person's spouse after receiving the prevention protocol (P 4 ). Multivariate sensitivity analysis was performed for P 1 and P 2 , considering the fact that the two probabilities were dependent. All of the direct medical costs of the intervention (Table 2 ) were assumed to be constant in the sensitivity analyses. The results of the above sensitivity analyses were evaluated in respect of their impact on the value of the cost per chronic HBV infection averted and the preference of strategies 1 and 2. Results Having run the model for baseline values (Tables 1 and 2 ), the average cost effectiveness of strategies 1 (without additional screening for HBcAb) and 2 (including additional screening for HBcAb) were 1,675,500 Rials (202 $) and 1,633,200 Rials (197 $) for each chronic HBV infection prevented, respectively. The worst-case analysis (e.g. setting all input probability values so that they would act to decrease effectiveness and increase the costs) was performed setting P 1 , P 2 and P 3 at their minimum values and P 4 at its maximum value. It showed that the average cost-effectiveness ratio of strategy 1 would be 2,460,204 Rials (296 $) and that of strategy 2 would be 2,440,015 Rials (293 $) in the worst case. It is noteworthy that the costs were not varied and were kept at their baseline values in this sensitivity analysis. A threshold analysis was performed to find the threshold value for the lifetime cost of CLD for one individual (from a societal perspective) below which the interventions in the model were not cost saving (e.g. the net benefit was negative). The analysis was performed keeping all other variables at their baseline values. The preliminary (non-discounted) threshold value for the lifetime cost of CLD was found to be 11,170,000 Rials (1346 $) and 10,888,000 Rials (1312 $) using strategies 1 and 2, respectively. After discounting, the threshold figures for CLD costs were 23387500 Rials (2818 $) for strategy 1 and 22797054 Rials (2747 $) for strategy 2. This showed that for the cost of CLD higher than the thresholds above, the respective strategies used for HBV transmission prevention would be cost saving. Sensitivity analysis Sensitivity analysis showed that when the prevalence of HBsAg positivity in general population (P 1 ) varied from the minimum to maximum, the cost per chronic HBV infection averted varied from 1,503,440 Rials (183 $) to 2,740,560 Rials (330 $) in strategy 1 and from 1,482,620 Rials (179 $) to 2,568,310 Rials (309 $) in strategy 2 (Fig. 2 ). When the prevalence of HBcAb positivity in general population (P 2 ) increased from the lowest to highest, the cost per chronic HBV infection prevented did not vary in strategy 1; but in strategy 2, it decreased from 1,645,720 Rials (198 $) to 1,561,530 Rials (188 $). Therefore, the higher rates of HBcAb positivity made the cost of strategy 2 become remarkably lower than that of strategy 1 (Fig. 3 ). Changing the probability of becoming HBsAg positive for an HBsAg positive person's spouse after marriage (P 3 ) from minimum to maximum varied the cost-effectiveness ratio from 2094370 Rials (252 $) to 558,500 Rials (67 $) in strategy 1 and from 2,041,500 Rials (246 $) to 544,400 Rials (66 $) in strategy 2. It shows that higher spouse-to-spouse transmission rates significantly increase the cost of both strategies (Fig 4 ). When the probability of becoming HBsAg positive for an HBsAg positive person's spouse after receiving prevention protocol (P 4 ) changed from the highest to lowest , the cost-effectiveness ratio decreased from 2,010,600 Rials (242 $) to 1,507,950 Rials (182 $) in strategy 1 and from 1,959,840 Rials (236 $) to 1,469,880 Rials (177 $) in strategy 2. It shows that higher efficacy of the preventive protocol results in lower costs-effectiveness ratios (Fig. 5 ). Strategy 2 was always cheaper than strategy 1 for all values of P 1 , P 2 , P 3 and P 4 in the univariate sensitivity analyses (explained above). The results of multivariate sensitivity analysis of the two variables P 1 (the prevalence of HBsAg positivity in population) and P 2 (the prevalence of HBcAb positivity in population) revealed that the strategy 2 was always cheaper than strategy 1 while the two probabilities varied. Discussion Preventing sexual transmission of HBV is not a new issue; however, the authors did not encounter any studies directly addressing the economic aspects of premarriage prevention of hepatitis B in their literature review. The reason can be the particular cultural backgrounds of Iranian community in which extramarital sexual relationships with the would-be spouse is expected to be rare due to strong traditional and religious bans against it. Consequently, the model used in this study may not be an appropriate one for countries with remarkable cultural difference in terms of extramarital sexual relationship such as Western countries. On the other hand, the cultural similarities between Iran and some other Eastern countries can increase the external validity of our model for policy makers in such countries. One of the important challenges in our study was the lack of precise data regarding some probabilities. Most of the studies in our literature review contained the prevalence of HBsAg positivity (chronic HBsAg carrier state) in the spouses of HBsAg positive people and the extent to which it was different from the rate in general population [ 8 - 11 ]. Because such a rate existed in literature, we preferred to use this prevalence rate representing the initial outcome of sexual contact with an HBsAg positive spouse after marriage. Thus, what is seen in the model (Fig. 1 ) as HBsAg+ final outcome does not mean getting infected with HBV; it means getting into a chronic HBsAg carrier state. In our model, we assumed all such cases to be chronic HBsAg carriers that entered the chronic carrier state asymptomatically or following an acute infection; 15% of such carriers could finally develop CLD during their lifespan [ 1 , 24 , 25 ]. Therefore, it is obvious that our model have primarily focused on more chronic outcomes of HBV infection (e.g. we ignored the costs of acute infections). However, this will not endanger the data robustness in our study; instead, it will always guarantee that all of the cost-effectiveness ratios calculated here are actually higher than that would be resulted with including the costs for acute cases which usually comprise a considerable portion of the symptomatic cases in adults [ 1 , 24 , 25 ]. The costs of acute and chronic liver disease (CLD) due to HBV infection in Iran were other variables for which data was lacking. One of the important reasons we did not use a Markov model to calculate CLD costs for Iran was the lack of important necessary data for running a Markov model (e.g. lack of data on age specific mortality rates, etc.). We first calculated cost-effectiveness ratios ignoring the costs of acute and chronic liver disease (the costs of the morbidity averted); this surely exaggerated the calculated cost-effectiveness ratios in the study. Thus, one should judge them from a more optimistic point of view. In the second step, to compensate for the lack of lifetime CLD costs in Iranian literature, we performed a threshold analysis using the baseline values for input probabilities and determined a threshold value for the lifetime costs of CLD in Iran, above which the preventive interventions were cost saving. The threshold levels for CLD lifetime costs estimated above do not seem high costs compared with the costs that CLD can impose on the society in terms of direct and indirect medical costs and productivity losses due to time spent sick or years of life lost because of premature death. The relevant medical costs or the costs associated with the productivity losses due to CLD was not accessible at the time of this research, so we discussed the point through some indirect comparisons considering the threshold we calculated for lifetime CLD costs. Though comparing the costs of CLD in Iran with that in the United States does not seem a standard approach, the large differences between the threshold figure we calculated for CLD costs in Iran and the CLD costs estimations (including productivity losses) mentioned in the studies for the United States in 2001 (64,382 $) [ 25 ] may partially reveal some facts [ 1 , 24 , 25 ]. To perform a more realistic comparison, we used Purchasing Power Parity (PPP) rates instead of exchange rates to convert the threshold cost in Rials into PPP dollars. PPP is defined as the numbers of units of a country's currency needed to buy in the country the same amounts of goods and services as, say, one US dollar would buy in the United States. The PPP rate for Iranian Rial was extracted from the National Health Account 2002 by Planning and Management Organization of Iran (unpublished data) and from figures reported by the World Bank Group. The threshold cost using PPP rates would be equal to about 9615 to 9863 PPP $ (for 2 strategies). Therefore, even if the total CLD costs in a country like Iran were 7-fold smaller than that in the United States, the strategies 1 and 2 mentioned in the model could still be cost saving. To give a sketchy view of some of CLD costs in Iran, the costs of a liver biopsy and those of a pharmacotherapy regimen related to CLD was retrieved by contacting discharge and accounting departments of a state-run hospital and a major drug distributor in Tehran. The discounted cost for an uncomplicated liver biopsy needing 1 to 2 days of hospitalization, blood coagulation serial tests and a special liver biopsy needle turned out to be about 2,093,778 Rials (252 $) in all state-run hospitals in Iran. Pharmacotherapy with new antiviral and immunomodulatory drugs is employed for treatment of active chronic liver disease in patients with hepatitis B [ 26 ]. The discounted cost of lamivudine, a typical example of these drugs, can be another instance of costs CLD imposes on many patients. A 12-month course of lamivudine (Iranian brand) in Iran can cost a patient 2,521,956 Rials (304 $). If the same calculation is performed for the foreign brand of lamivudine available in Iran (Zeffix), the discounted cost will be 12,991,892 Rials (1565 $). In addition, these costs may be much higher when some more expensive drug regimens are used or more prolonged regimens are repeated due to chronicity or intractability of disease. The costs mentioned here can not directly give a clue to total CLD costs and mentioning them was to help the reader get a view of the scale of some familiar CLD costs in Iran. In a different approach, we converted the threshold value into the number of productive months for a national of a country having a GDP (Gross National Product) per capita like Iran. GDP per capita shows the amount that an individual contributes to domestic income every year. According to the World Bank Group, Iran had a GDP per capita of 1641 $ in 2002, and the average annual growth of GDP per capita in Iran for 2002 to 2006 was 4.5%. From here, the GDP per capita in 2003 comes to be 1790 $. Assuming that the average growth of GDP per capita for Iran will remain at 5% in the coming 25 years (the period used for discounting the CLD threshold costs), the GDP per capita will be 6062 $ at the end of the period. Considering such figures, the threshold calculated above would be equal to about 6 months of productivity based on a GDP view. The average number of Quality Adjusted Life Years (QALY) that CLD can deduct of an Iranian's life is not yet determined. Nevertheless, a 6-month period does not seem a long time compared with the life years lost due to premature mortality and the QALYs lost due to sickness in the proportion of CLD patients with cirrhosis and hepatocellular carcinoma only. Finally, the fact that the strategies would be cost effective can be further emphasized when taking into account the costs of acute cases of HBV infection that comprise the majority of symptomatic cases in adults and were ignored in the model due to lack of Iranian data and for sake of simplicity. Another similar topic that merits discussion here is the topic of mother-to-child vertical transmission of HBV. If a female gets into a chronic HBsAg carrier state and remains positive for HBeAg during pregnancy, it is very likely that her child is infected with HBV. Since a considerable proportion of infected infants will get chronic HBV carriers, this can lead to newer CLD cases further increasing the costs of CLD. Considering this fact, the cost of averted morbidity owing to the preventive strategies will increase even more and the model would seem more cost-effective. On the other hand, the prevention protocol in our model might seem a bit extravagantly designed when looking at the final HBsAb test and the extradose of HB vaccine administered when HBsAb serum levels are insufficient. In addition, one may argue that condoms may be more available while being nearly as protective against HBV transmission as HBIG injection. This can be useful in modifications poor countries can make to the model to get similar results in lower price or with more flexible/available choices. The average cost-effectiveness ratios associated with the two preventive strategies shown in the model did not differ much. The strategy 2 was always slightly cheaper than strategy 1. When the prevalence of HBcAb positive people in the general population (P2) rose, strategy 2 would get remarkably cheaper. Finally, it is noteworthy to mention the issue of compliance. As explained in methods, we assumed all recipients of the preventive interventions (strategies 1 and 2) would be 100% compliant and there was a full coverage of such services in the country. The reason for such assumption was the strict regulations set by Iranian government for all premarriage individuals to undergo a battery of screening tests in which the strategies in our model could be integrated. Nevertheless, we can assume that compliance variations can affect the efficacy of our prevention protocol (e.g. with lower rates of compliance for accepting preventive strategies, we will have lower efficacy of the prevention protocol). As stated in results, we performed sensitivity analysis for P4 (probability of becoming HBsAg positive for an HBsAg positive person's spouse after receiving the prevention protocol), which is a variable dependent on the efficacy of the prevention protocol (see the formula in methods). Therefore, we indirectly incorporated compliance into our model's sensitivity analysis. Considering compliance issue, we may prefer strategy 1 because it includes fewer steps (e.g. it does not include screening for HBcAb) and may be easier to administer in a low compliance population especially that it is negligibly more expensive than strategy 2. Conclusions Finally, we conclude that applying the preventive strategies in our model for HBVsexual transmission prevention before marriage in the countries with cultural backgrounds similar to Iran seems cost saving. Further investigations in the country for precise calculation of costs of HBV infection especially the costs associated with CLD is necessary for more accurate economic evaluations. Competing Interests Nond declared. Author's Contributions PA : Proposing the main idea, supervising the project and counseling the methodology development MR : Developing methodology, critical appraisal, rewriting the final article from draft, literature review, and responding to reviewers DR : Literature review, methodology development, decision tree development, calculations and results, and writing the first draft of the article NB : Contributing to decision tree development, searching literature for costs SA : Review literature for probabilities, contributing to decision tree MHS : Counseling the scenarios of the decision tree model SS : Counseling the methodology, model development and discounting problems MRZ : Senior supervisor of the research project All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517713.xml
517707	Bayesian model accounting for within-class biological variability in Serial Analysis of Gene Expression (SAGE)	Background An important challenge for transcript counting methods such as Serial Analysis of Gene Expression (SAGE), "Digital Northern" or Massively Parallel Signature Sequencing (MPSS), is to carry out statistical analyses that account for the within-class variability, i.e., variability due to the intrinsic biological differences among sampled individuals of the same class, and not only variability due to technical sampling error. Results We introduce a Bayesian model that accounts for the within-class variability by means of mixture distribution. We show that the previously available approaches of aggregation in pools ("pseudo-libraries") and the Beta-Binomial model, are particular cases of the mixture model. We illustrate our method with a brain tumor vs. normal comparison using SAGE data from public databases. We show examples of tags regarded as differentially expressed with high significance if the within-class variability is ignored, but clearly not so significant if one accounts for it. Conclusion Using available information about biological replicates, one can transform a list of candidate transcripts showing differential expression to a more reliable one. Our method is freely available, under GPL/GNU copyleft, through a user friendly web-based on-line tool or as R language scripts at supplemental web-site.	Background An important challenge in Serial Analysis of Gene Expression (SAGE) [ 1 ] analysis is the decision whether a gene is differentially expressed between two classes, for example tumoral vs. normal classes. In statistical terms, this essential step is to test the null hypothesis H 0 : "gene has no differential expression between the two probed classes". A much more usual approach is to assign an index ( P -value or Bayes factor, for example) that measures the confidence/significance of the hypothesis and let the biologists themselves to establish a cutoff of what they call significant. This necessity arises because counting sequenced SAGE tags is a process prone to random and systematic errors that affect gene expression abundance estimates. Systematic errors may come from various sources such as GC content bias [ 2 ], sequencing errors [ 3 , 4 ] as well as the possibility of non unique tags. This kind of error can be detected/corrected using some bioinformatics procedures such as quality control of automatic sequencing pipe-line [ 5 ], or statistical estimation procedures such as "denoising" [ 6 , 7 ]. Random errors are due to the inherent stochastic characteristic of SAGE data acquisition: sampling from automatic sequencing. Like colored balls in an urn, sampling and counting SAGE tags from a library is commonly modeled by a Bernoulli Process relying on an infinite population sampling approximation. If an Expressed Sequence Tag (EST) library is non-normalized, its counting data, also known as "Digital-Northern", reflects the abundance of genes. Likewise, the Massively Parallel Signature Sequencing (MPSS) [ 8 ] technique counts tags to infer the transcriptome, but using a completely different strategy from traditional DNA sequencing methods, that allows augmented high-throughput capability. Therefore, all the results discussed here are readily applicable to "Digital-Northern" or MPSS context since, from a mathematical viewpoint, all represent the same bioinformatics problem: counting transcripts (as balls in urns). Nowadays, the variability in SAGE abundance data is modeled only as due to sampling from sequencing, since almost all statistical procedures are performed after aggregation of observations from various libraries of the same class, creating a "pseudo-library". See [ 9 - 11 ] for good reviews on statistical techniques used in SAGE analysis. This extensively used trick tacitly ignores the within-class variability, i.e., the biological variability among individuals within a class (different patients having the same cancer diagnosis, for example), and could lead to overconfident conclusions. Results Here we propose a Bayesian model of mixtures to account for within-class variability as a generalization of the Beta-Binomial model [ 12 ]. We also show that the usual "pseudo-library" construction is a particular case of our mixture model. Finally, we propose the use of the Bayes Error Rate to intuitively rank the differential expression hypothesis under a Bayesian framework, avoiding several technicalities and difficulties such as: typeI and typeII error analysis, Bonferroni-like multiple testing correction, asymptotic results evocation, imposition of a test statistic and null probability density function (pdf), and so on. Statistical model The counting process from automatic sequencing of one single i -th library is often modeled as a Bernoulli Process and a fixed unknown tag abundance π i is implicitly assumed. The pdf of the random variable of interest, "expression abundance" π ∈ [0;1] among all n libraries is unknown, thus each library could be regarded as being created by a realization of π . These features lead naturally to mixture models [ 13 , 14 ]: where: f (·) is the unknown pdf of the abundance among same-class libraries parameterized by a vector θ , X = ( x 1 ,..., x n ) is the vector of counts in all n libraries of same class, M = ( m 1 ,..., m n ) is the vector of library sizes and L is the likelihood of each i -th observation. The common procedure of merging all observations from libraries of the same class, constructing a "pseudo-library" before statistical inference, is recognized as a particular case of this mixture model: just assume that all libraries have strictly the same abundance, with no biological variability. Mathematically, this is a function with infinite probability density over one single abundance value π = θ and zero over every other π ≠ θ , or a Dirac's Delta function. Using f (·) as a Dirac's Delta function constrained to [0;1], turn Eq.1 into the familiar and commonly used binomial distribution (see derivation in the Methods section). We believe that Dirac's Delta is a naive description of real-life SAGE libraries. The Beta distribution is an alternative with non-zero within-class variance to account for intuitively expected biological differences among them. Using f (·) as a Beta in Eq.1, yields the so-called Beta-Binomial model (see derivation in the Methods section). Given the parameter vector θ that describes the random variable π of some fixed gene G , we must decide if there is a difference between A and B classes (e.g, tumor vs. normal classes). We propose to consider genes as being differentially expressed based on non-superposition of the predictive Beta pdfs of both A and B classes. By "predictive" we mean that we use the a posteriori mode in the Beta pdfs. The "non-superposition" intuitive feature is mathematically written as the Bayes Error Rate E [ 15 ]: where f (·) is the Beta pdf and "hat" over the parameters indicate the values that lead to an a posteriori pdf maximum. The a posteriori distribution is obtained as usual from Bayesian Statistical Theory ( a priori pdf choice and detailed derivation are in the Methods section). Intuitively, if the pdfs are "far apart", the gene probably has reproducible differential expression between classes. In this case, rarely could one misclassify class A as B and vice-versa. Figure 1 gives some insight about this fact. Using our proposed approach, the "far apart" notion means a small Bayes Error Rate E . For adepts of the Frequentist Statistics, this evidence measure could resemble a typeI and typeII errors sum, however it is just an illustrative analogy. As in any significance test method, the experimenter must define what is a high significance E value. This cutoff should be guided by external and independent confirmatory assays. To avoid crude decision boundaries, one could rank their significance results but there is no way to avoid some arbitrariness in any kind of statistical test. In the classical Frequentist Statistics framework, it is common to call a result as significant if it presents a P -value ≤ 0.01 in a t -like-test, hoping that this could control the error at this level. However, due to technical difficulties such as lack of sensitivity of posterior confirmatory methods or high absolute expression (not differential expression) necessity, this apparent statistically sound results could be not useful in a pragmatic sense. That is why we prefer to rank the differential expression results and allow researchers to establish a cutoff compatible with their subsequent application for the selected genes, rather than split them based in assumption-derived error-rate cutoffs. People familiar with the Frequentist Statistics framework could miss multiple testing considerations, typeI/typeII error studies, and so on. However, in the Bayesian framework, several of these concerns are meaningless since we work with parameters space and not with sample space. The bayesians avoid statements about "data that could be observed but was not" and work only with available information (prior and experimental), extracting all possible information from data effectively observed. For those genes classified as differentially expressed, one should aggregate intuitive information adding "error-bars" to expression ratios. Recently we have developed a method to add credibility intervals to gene expression ratio [ 16 ], which could improve posterior analyses such as clustering [ 17 ] or comparison with microarray data. Comparison with available methods using publically available data To show the model is usefulness, we applied it to a tumor vs. normal two-classes comparison problem. We chose a subset of brain tumor SAGE data from The Cancer Genome Anatomy Project's SAGE Genie public database web-site [ 18 ]. The SAGE Genie performs several bioinformatics protocols to assure the quality of its data with systematic errors cleaning/correction [ 19 ]. We used all 4 available libraries in SAGE Genie until Jan/2004 from astrocytoma grade III tumors and almost all (except the fetal library) normal brain libraries (see Methods section for details about libraries). We want to stress 3 typical and important cases: (i) when our measure agreed with other evidence measures accepting null hypothesis H 0 , i.e., there is no evidence of differential behavior between tumor and normal classes; (ii) when our method agreed with others rejecting H 0 , i.e., there is evidence of differential expression; and (iii) when our method showed evidence in favor but other evidence measures showed evidence against the H 0 . Case (iii) is the main motivation of our method since it reveals situations that researchers may call a gene differentially expressed and, in fact, it could be not so significant if biological replicates are taken into account. The other evidence measures used were: the Audic-Claverie bayesian evidence [ 20 ], the classical Fisher Exact Test P -value, and the classical χ 2 P -value, all obtained using the IDEG6 web-interface [ 21 , 22 ] (see Methods section). A case (i) prototype is the TTTCAATAGA tag with X T = (0, 2, 5, 8) and X N = (1, 1, 0, 0, 0, 7, 2). The Audic-Claverie, Fisher and χ 2 methods yield P -values of 0.06, 0.44, 0.41, respectively, indicating low evidence against H 0 for all mystical significance level cutoffs ≤ 0.01, ≤ 0.05 or ≤ 0.1. The Bayes Error Rate evidence is E = 0.61, an intuitively unacceptable superposition level between the normal and tumoral predictive Beta pdfs, showing that there is no separable behavior between classes. Figure 2a shows an obvious superposition between pdf and observations of this two classes. A case (ii) prototype is the AAAAGAAACT tag with X T = (7, 11, 18, 10) and X N = (7, 1, 2, 1, 2, 0, 3). All P -values are 0.00 (zero), significant at any cutoff level. Our evidence is E = 0.03, showing safely that this gene behaves differentially between normal brain and astrocytoma grade III patients. Figure 2b shows that two Betas are apart from each other and, even observing clear within-class variability, the expression is different. A case (iii) prototype is the TTGGAGATCT tag with X T = (7, 239, 244, 123) and X N = (54, 27, 33, 21, 40, 196, 28). All P -values are 0.00 (zero), indicating significant difference between classes. On the other hand, our evidence E = 0.73 indicates high superposition between tumor and normal classes. Figure 2c shows that within-class variability for tumor class is not negligible. It is obvious that individual libraries confound their results with normal brain libraries, and the Betas have a relatively high intersection. Using a common "pseudo-library" approach, one is lead to call this gene as a strong discriminator between classes. We believe that this is a suspect conclusion. There are several other obvious case (iii) examples, such as tag TACAGTATGT in Figure 2d , that received P -values < 0.01 from all other methods, and they are the main concern of our method since they may lead to waste of resources in clinical validation efforts of genes that, by SAGE itself, could be left behind in favor of other promising genes. All tag results are available as additional file and graphics for all tags are at the supplemental web-site [ 23 ]. One could think about a case (iv) when considering within-class variability leads one to H 0 rejection, but considering "pseudo-libraries" leads to H 0 acceptance. This seems to be inconsistent since one expects that, once H 0 is accepted in a simplified model, it should also be accepted in the complete model. In fact, we do not observe such a situation, except by tags with P -values or Bayes Error Rate very close to arbitrarily defined cutoff values. We believe that these occurrences are just "edge effect" manifestations. Discussion In order to assure that we are dealing with a fundamental question in SAGE analysis, we show more insights analyzing the method's robustness using the same data but excluding "small" libraries. Also, we draw some parallels between our proposed method and the only available published solution for dealing with within-class variability, a t -test approximation [ 12 ]. We used our method with all available libraries but some of them are smaller than 50,000 tags (see Table 1 ). In the SAGE community, libraries smaller than this arbitrary limit are considered "small". Several researchers claim that these are non-representative and should be excluded from analysis. We observed several case (iii) tag examples which remain as case (iii) if we use libraries with size > 40,000 and > 50,000 (shown at the supplemental web-site only). Figure 3 shows a tag example analyzed in these tree setups and it is clear that inclusion of "small" libraries gave pretty much the same result, indicating robustness of our method against small class size variations and against "small" sized libraries. Moreover, these libraries are not always outliers from biological sampling but seem to be samples like any other. These results suggest that one can use the "small" libraries, jointly with non-"small" ones, because biological variability seems to be greater than binomial sampling variability. Obviously, we are not recommending to use only "small" libraries in SAGE analysis, but suggesting that our method is relatively robust. For low expression genes, the binomial sampling variability should become more relevant as the library size decreases. Also, the results obtained using two/three libraries could be very different from using just one. These proprieties could be tag dependent since some tags could be much noisier than others for biological reasons. Some "denoising" procedure could be used before application of our method [ 7 ]. Therefore, our findings should be carefully interpreted. To prove that the incoherence of using "pseudo-libraries" methods is not a prerogative of tags showing small fold-changes, we analyzed another three very illustrative examples: ATGGCAACAG, GGATGTGAAA, and GTATGGGCCC; which are case (iii) tags. These tags present high fold changes: 7.59, 8.15 and 25.80 fold-change respectively, augmented in pooled tumor libraries. Using the well-known Fisher Exact test, χ 2 classical test and the Audic-Claverie's method, we get 0.00 (zero!) for all P -values of the no differential expression null hypothesis. Using the conceptually different Bayesian P -value implemented at SAGE Genie [ 24 , 25 ] we obtain 0.01, 0.00 and 0.00 respectively for posterior probabilities of fold-changes greater than 4-fold. Finally, using our own proposed measure, applied to the pool, we get E = 0.00 meaning no superposition between the two classes pdfs. All these results indicate strong significance in differential expression of these tags. However, if we consider within-class variability, the test proposed by Baggerly et al. [ 12 ] yields 0.08, 0.07 and 0.15 respectively for t -test P -values, and our method yields Bayes Error Rates of 0.38, 0.37, 0.43 respectively; indicating not so significant evidence in favor of the differentially expressed hypothesis. A closer look at the graphics of these tags induces one to believe that there is no reproducible differential expression because several observations of tumor and normal are superimposed (all graphics available at supplemental web-site [ 23 ]). Since we show clearly that methods that use "pseudo-library" aggregation could be incoherent in some cases, a natural question is how our proposed method performs compared to the only published solution that accounts for within-class variability, the Baggerly et al. [ 12 ] t -test approximation. Without knowing the true state of all tags, it is impossible to carry out a serious benchmark. Since the interpretation of evidence measures is very different, the performance could be subjected to an arbitrary cutoff selection for each method. Figure 4 shows a scatter-plot of evidence measures obtained for each of the two methods. It is clear from this graphic that there are many more tags considered as differentially expressed using our method than the t -test approximation, considering E ≤ 0.1 and P -value ≤ 0.01. There are also some tags selected by t -test and ignored by ours. It is impossible to know which method perform better without the true unknown status of those tags. Looking at individual libraries results, constructed as depicted in Figure 2 for example, could help in this analysis but this is a subjective procedure. It is important to bear in mind that a difficulty is hidden in the Beta modeling imposed in the very first beginning. If Beta is not a good model for an unknown biological behavior, then some apparent inconsistency could appear in both Baggerly et al. [ 12 ] and our approaches. However, our general mixture model allows another propositions. Other simplex constrained pdfs, different from Beta, exist but the tractability is much more difficult [ 26 ]. We believe that to build a fully non-parametric approach to this problem is a very hard issue, but should be considered as a future challenge. Conclusion Until now, almost all statistical methods for SAGE data analysis tacitly ignore the within-class variability. To our knowledge, the firsts to formally address this issue was Baggerly et al. [ 12 ] who introduced the Beta-Binomial model as the correct way to model the probability of counting tags instead of Binomial models. They also proposed a t -like statistics, outlined a possible hypothesis test using the classical Frequentist Statistics framework and evoked some asymptotic results for t pdf justification. In this work we presented the Bayesian alternative for this problem and defined a theoretical model that views Baggerly's Beta-Binomial approach or even the common Binomial approach as particular cases of mixture models. Other models are possible modifying the mixing distribution, such as Beta-Poisson [ 14 ], or using other simplex constrained pdf [ 26 ] to model expression abundance. At last, but not at least, we proposed a method for ranking differentially expressed genes between two classes using the Bayes Error Rate as an intuitive measure of separation between the classes pdfs, avoiding statistical test formalism and its conceptual/practical difficulties. We show that there are cases in which approaches that ignore within-class variability will lead to high significance in differences between tumor and normal classes, but looking carefully at individual observations jointly, one should not attribute such high significance to them since abundance probability density functions have considerable superposition. In conclusion, we recommend that within-class variability must be taken into account in any statistical analysis of SAGE data if replicates are available. We suggest that biological replication should be considered in planning new SAGE experiments. Methods General bayesian model To start generically, suppose that the probability density function (pdf) for the random variable of interest "expression abundance" π ∈ [0;1] of some gene G is indexed in a model family by means of a parameter vector θ . Therefore, following the usual Bayesian framework, the a posteriori pdf that describes the class is: where: X = ( x 1 ,..., x n ) is the vector of counts in all n libraries of same class, M = ( m 1 ,..., m n ) is the vector of total observations in all n libraries of same class, g (·) is the a priori pdf, and L is the likelihood of each i -th observation. Note that the product of all likelihood functions over all observations is the so-called Likelihood Function. The counting process from automatic sequencing is often modeled as a Binomial. Since the sample size and the stopping rule are not known in advance the model is not strictly Binomial. We do not need the combinatorial constant in the model, but we write it just because it is commonly used and will vanish in posteriori expression anyway. "Pseudo-Library" method as particular case Merging all observations from the same class libraries and constructing "pseudo-libraries", with the sum of their components, is the standard procedure to use replicates. Our general model is reduced to this (unrealistic) one if one uses f (·) as a Dirac's Delta in Eq.1: where: 1 {·} is the indicator function. Using Eq.4 in Eq.3 yield: where: g( θ ) = 1, the non-informative uniform a priori distribution. The expert recognizes that θ ~ Beta(1 + Σ x i , 1 + Σ m i - Σ x i ), and the sum of observations is the mathematical translation of "pseudo-libraries" construction. Beta-Binomial method as particular case The only published solution that allows non-zero within-class variance in SAGE analysis is the Beta-Binomial model [ 12 ]. Using f (·) as a Beta in Eq.1 we get the Beta-Binomial model as a particular case of general model: where: B (·) is the beta special function, and: Again, an expert recognizes θ = ( θ 1 , θ 2 ) as the mean and standard deviation (stdv) of a Beta random variable. We prefer this parameterization of Beta distributions instead of the common ( α , β ) one because: (i) it is much more intuitive to biologists to deal with mean and stdv than with abstract α and β , and (ii) as α , β > 0, the domain Θ = {( θ 1 , θ 2 ): 0 ≤ θ 1 ≤ 1, 0 ≤ θ 2 2 < θ 1 (1- θ 1 ) ≤ 1/4} is bounded and much more amenable to perform the necessary numerical computations. Using Eq.6 in Eq.3 yield: where: g ( θ 1 , θ 2 ) is the priori pdf. A Priori distribution definition To complete a Bayesian model, it is necessary to choose the a priori pdf. We use an uniform distribution over Θ. On the other hand, we know in advance that variance of this model cannot be smaller than the variance eventually obtained if we do not consider within-class variability. Even if the within-class variability is very small, it cannot be estimated as being smaller than the simple sampling error because they are inseparable, and sampling error is the lower limit [ 12 ]. As an illustration, the same situation could occur if one takes several diameter measurements of a folded paper ball and a perfect sphere using a common ruler. In the first case, the intrinsic nature of the measured object dominates the measurement variability but, in the second case, we cannot know the diameter of the perfect sphere with better precision than our ruler can measure. This kind of knowledge is naturally incorporated in Bayesian statistics by means of a priori distributions. To match our desired features, it is sufficient to define an uniform over the Θ parameter space but constrained at a minimum stdv σ, obtained from the result of no within-class variance model: over domain Θ = {( θ 1 , θ 2 ): 0 ≤ θ 1 ≤ 1, 0 ≤ θ 2 2 < θ 1 (1 - θ 1 ) ≤ 1/4}. Since we showed (Eq.5) that the no within-class variance model is θ ~ Beta(1 + Σ x i , 1 + Σ m i - Σ x i ), it is easy to obtain our lower stdv boundary from Beta variance: Therefore, using Eq.9 and Eq.10 in Eq.8, our posteriori is completely defined. Differential expression detection We detect a tag as differentially expressed using the Bayes Error Rate E [ 15 ] in both predictive Beta pdfs: where: Note that f (·) is the Beta pdf, as in Eq.6 development. The "hat" over θ = ( θ 1 , θ 2 ) indicates values that leads Eq.8 to maximum. As usual, the maximization, subject to constrain Θ defined previously, is made upon logarithm of posteriori's core since it gives the same estimates as the posteriori itself: Figure 5 shows an example of this process. See Results section to get an intuitive notion of this evidence measure. Implementation – numerical analysis The method was implemented as R language [ 27 , 28 ] scripts which are freely available under GPL/GNU copyleft license at supplemental web site [ 23 ]. At this web page there are details on how to run it locally. Our method is computer-intensive mainly because some numeric maximization and integration are needed. We used efficient R built-in routines to perform such numerical tasks. Remember that maximization needed in Eq.12 is constrained, thus we used simply auxiliary re-parameterization to obtain linear constrains and used the constrOptim R routine. For numerical integration we used the 1-dimensional gaussian quadrature integrate R built-in function. Although numerical integration of Eq.9 should be performed in all [0;1] support, the relevant contribution for this integral is concentrated in a much smaller region. Integrating over the formal limits will cause serious numerical errors, and to avoid this problem we approximate our integration region to an interval delimited by 0.005 and 0.995 quantile of each Beta pdf since the relevant density lie in there. The credibility intervals ("error-bars") for the expression ratio of interesting tags were obtained as described in our recent work [ 16 ]. We chose arbitrarily 68% credibility intervals. Implementation – Web based interface We have also developed an easy-to-use web-based service that performs all calculations at our server and provides password-protected results. Although desirable, for the sake of automatic web hyperlink with SAGE Genie database, it is not necessary to explicitly identify the tags analyzed but rather any (custom) i.d. string. This could increase privacy or make our web-interface useful for "Digital-Northern", MPSS or any mathematically related problem of mixtures from binomial sampling. Figure 6 shows snapshots of the interface. Publically available data The Table 1 list the SAGE Genie's library name, Gene Expression Omnibus (GEO) [ 29 ] accession code and size of all used libraries. For our aims, it is sufficient to focus the analysis at the tag level. Thus, we process the tag counts and let the identification of tag's best gene match as a posterior question that could be carefully done only to really interesting tags. We choose not to process tags whose counts appear only in libraries of one class. It is important to note that all libraries are from bulk material, without cell-lines, and came from patients with similar disease description. The normal libraries came from different normal regions of the brain. We think that this data set is very illustrative since there are biological replicates in the tumor class allowing clear verification of within-class biological variability. On the other hand, taking only one kind of disease, astrocytoma grade III, instead of all brain tumors in the database, leads one to believe that the within-class variability is in fact due to biological diversity of the patients and not due to very distinct molecular profile of distinct brain tumors stored in SAGE Genie's database. Therefore, we believe that this in silico comparison is well-suited to demonstrate the necessity of dealing with within-class effect, although it is not our aim here to make a detailed or biological analysis of brain tumor data. Comparison with other methods To bring some intuition about our differential expression evidence measure, we tabulated evidence measure obtained from the famous Fisher Exact Test, the classical Pearson's χ 2 proportion test and the bayesian Audic-Claverie's method. All these tests were performed using the easy-to-use web-interface IDEG6 [ 21 , 22 ]. The " P -values" are conceptually very different from our evidence measure but are the most used evidence measures. Although numbers cannot be compared, the conclusions obtained from these methods should be since graphical representation of each library observation gives clear indication of incoherence of "pseudo-library" methods. The results of the significance measures for all tags are available as additional Excel © or OpenOffice © interactive files in which the user can set cutoffs for the significance measures, and explore the differences in conclusions. We carry out a qualitative comparison of our method with Baggerly et al. [ 12 ] t -test approximation in a graphical way since it is impossible to judge them without the unknown true status of analyzed tags, given the too different interpretation of numeric values returned. In their Frequentist framework, the estimator p i = x i / m i is used for π i and a linear combination of these abundances is proposed as the correct way to combine results from different libraries: where w i are the weights that yield an unbiased minimum variance estimator V u for weighted proportion's variance and θ = ( α , β ) are the Beta pdf parameters. However, this unbiased variance could be unrealistically small when it becomes smaller than the sampling variability. We know that the variance of this model cannot be smaller than the variance eventually obtained if we do not consider within-class variability. Therefore, they propose the final ad hoc estimator: V = max [ V u ; V pseudo-lib ] (14) where: The max(·) function assure that V is not unrealistic small when V u is unrealistic small. To fit all these parameters, they used the computationally practical Method of Moments. Once p A , p B , V A and V B are found for classes A and B, these authors test if the proportions are significantly different proposing the use of a t w statistics as following a Student's t df pdf: List of Abbreviations SAGE: Serial Analysis of Gene Expression MPSS: Massively Parallel Signature Sequencing EST: Expressed Sequence Tag pdf: probability density function GEO: Gene Expression Omnibus Authors' Contributions RV conceived and executed this work. HB helped with all biological issues. DFCP helped in differential expression detection methods and implemented the on-line web-based tool. CABP helped with Bayesian statistics and proposed the mixture ideas. Supplementary Material Additional File 1 Results for all evidence measures . This file allows the user to interactively define significance cutoffs for ranked tags. The ranks are based on evidence measures against "no differential expression" hypothesis, i.e., evidences closer to 0 (zero) denote higher confidence in differential expression and closer to 1 (one) denote no evidence of differential expression. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517707.xml
548954	Progress on the Origin of Species	Two new books on speciation update the classic texts by Mayr and Grant and help set the stage for a renaissance of research into one of the most important processes in evolution	Despite the biases that a single author inevitably brings to a subject, only one or a few closely interacting authors can bring coherence, synthesis, and vision to a broad and complex topic. A symposium volume just doesn't do the job. Few topics in biology are as simultaneously encompassing, complex, and controversial as the origin of species, i.e., speciation. Speciation is, after all, the process responsible for biological diversity, at least of sexual organisms, so it is hardly a minor topic. But even though recent years have seen the publication of symposia on speciation and books on the ever-contentious issue of species concepts, it has been 23 years since Verne Grant's authoritative Plant Speciation [ 1 ] and 41 years since Ernst Mayr's magisterial and highly influential Animal Species and Evolution [ 2 ]—the last syntheses of research on speciation. Now, two outstanding new books not only treat speciation as a conceptually unified topic in both plants and animals for the first time, but also provide rich review and analysis of a vast subject that has progressed at least as much since Mayr and Grant wrote as in the century that preceded their work. These books are very different, but wonderfully complementary. Gavrilets reviews and adds to mathematical theories and simulation studies of speciation and related issues, such as fitness landscapes and selection in heterogeneous environments. A deep reading of his book will require considerably more mathematical competence than most evolutionary biologists (including this reviewer) have, but Gavrilets provides excellent verbal explanations of the models' assumptions and conclusions, as well as comparisons and critiques of related models. Gavrilets cites empirical studies (with which he has very broad familiarity) plentifully, but as a theoretician, he does not evaluate them or describe them in depth. That task is undertaken by Coyne and Orr, who introduce most topics with a verbal overview of theory, review empirical evidence and its bearing on hypotheses, and conclude with incisive assessments of what they think we know and what remains uncertain or unexplored. Like Gavrilets, they offer a number of novel ideas or suggestions about how to proceed. Coyne and Orr have both worked mostly on speciation genetics in Drosophila , so it is hardly surprising that their treatment of speciation bears a strong genetic emphasis and draws heavily on Drosophila work (perforce, since this is almost the only source of evidence on some topics, such as the genetics of hybrid sterility and inviability). Even the most drosophilophobic readers, however, will be pleased by the extent to which Coyne and Orr have conscientiously scoured the literature on nonmodel animals and plants. To appreciate the landmark status of these books, consider what has happened in speciation studies since Mayr and Grant published theirs. Mayr and Grant articulated positions on species and speciation that had developed during and soon after the Modern Synthesis of the 1930s and 1940s, when modern evolutionary theory developed from a reconciliation of genetics, systematics, and paleontology. Mayr and Grant drew on abundant systematic data on patterns of divergence and experimental data on genetic differences between related species. They rightly identified reproductive isolation (RI) as a critical, even defining, property of species, and allopatric divergence (i.e., in disjunct geographic areas) as the major geographic mode of speciation. They recognized that trait differences between species, including RI, usually have a polygenic basis, and that different coadapted (epistatically interacting) sets of genes underlie incompatibility (e.g., hybrid sterility). They emphasized the role of ecological selection as a driving force in speciation, largely by extrapolation from the primacy of selectionist thinking that developed during the Synthesis. They accepted that natural selection can reinforce prezygotic isolation (i.e., lack of mating or zygote formation) between species and thereby reduce production of unfit hybrids, even if Mayr did not share Dobzhansky's belief that this was the norm. Mayr combined selection with genetic drift in his theory of founder-effect speciation (divergence in populations founded by just a few individuals), which became widely accepted. (It became Eldredge and Gould's [ 3 ] theoretical foundation for punctuated equilibrium nine years after Mayr's book appeared.) Mayr and Grant wrote against a background that almost entirely lacked any mathematical theory of speciation (which I suspect neither of them would have drawn on even if it had been developed), any relevant molecular data (other than early allozyme studies by the time Grant published), and any rigorous phylogenetic methodology. Kimura's neutral theory of molecular evolution [ 4 ] had not yet been published when Mayr wrote, and had not been vindicated when Grant wrote, so genetic drift and a neutral (nonselectionist) interpretation of molecular data were still suspect. Detailed analysis of genetic architecture was decades away, and of course insights into selection and historical demography from DNA sequence data were a dim dream at best. Coyne and Orr and Gavrilets analyze a new world of speciation studies. Theoretical studies of speciation, for example, now include more than 100 papers on one topic alone, the evolution of prezygotic isolation. (Gavrilets laments that the theoretical work suffers from domination by simulation rather than analysis, so that it is often hard to draw general conclusions from models that use different assumptions, but he nevertheless draws some fairly strong conclusions, as I note below.) Molecular studies have provided important data on such issues as the absolute dates of speciation events, the duration of speciation, and the time course of the evolution of RI. The field of molecular phylogeography, which documents the history of spatial isolation and geographic expansion of populations, has developed. The relation between range overlap of related species and their molecularly dated time of divergence provides some evidence on the role of geographic versus sympatric speciation (i.e., speciation without geographic segregation). In all these areas and others, our knowledge has increased steadily. For instance, molecular markers enable more detailed dissection of the genetic architecture of species differences, and support the conclusion that they are usually rather highly polygenic, but that much of the variance can be explained by a few major gene substitutions. We now have good evidence, as Coyne and Orr emphasize, that at least in animals, hybrid infertility is caused by differences in gene action, not by structural chromosome differences or failure of meiosis. Regarding the mechanisms of speciation, evidence for the role of divergent ecological selection in allopatric speciation is sparse, because this crucial topic has been unaccountably neglected until recently. Very different kinds of data, ranging from DNA sequences to correspondence between RI and ecological divergence, support natural selection, but there is hardly enough evidence, in my opinion, to support Coyne and Orr's strong conclusion that “at least one important debate has been settled: selection plays a much larger role in speciation than does drift” (p. 410). Even more astonishing than the paucity of studies of the role of ecological selection in speciation is the fact that the likely role of sexual selection was not even recognized until almost 20 years after Mayr's book. I agree with Coyne and Orr that the theory and evidence for speciation by sexual selection is one of the most important advances in speciation studies, but it is important to recognize that the evidence consists mostly of correlations between diversification rates and indices of the likely strength of sexual selection; as Coyne and Orr note, there are no cases in which we understand just how sexual selection has caused speciation. This is a rich, largely unexplored area. Gavrilets feels that divergent evolution by sexual conflict (in which females evolve resistance to males' advances) is a potentially important process, whereas Coyne and Orr are skeptical that this will prove widespread. Coyne and Orr remark that populations may diverge in male signals because of intrasexual selection (competition among males), and that female mate preference may follow. Quite so, but even though Berglund et al. [ 5 ] summarized many examples in which male signals appear to serve both inter- and intrasexual functions, this topic has been almost ignored in the literature of both speciation and sexual selection. On a related theme, an important speciation process appears to be the extraordinarily rapid evolution of male reproductive proteins (e.g., sperm surface proteins), which may contribute to the “faster male evolution” that is a cause of “Haldane's rule” (that hybrid sterility and inviability first appear in the sex that has two unlike sex chromosomes). Despite their very different approaches, Coyne and Orr and Gavrilets arrive at rather similar conclusions on some of the most controversial issues in speciation. One such is the role of genetic drift in speciation. Gavrilets analyzes founder-effect speciation (which combines drift and selection), agrees with most other theoreticians (e.g., Barton and Charlesworth) [ 6 ] that it is very improbable, and argues instead for his model of evolution on “holey landscapes,” whereby allopatric populations can evolve by genetic drift along ridges of roughly equal fitness to different, incompatible gene constitutions. He admits that the time to speciation under this process will ordinarily be very long unless selection is involved. Coyne and Orr fully accept both Barton and Charlesworth's critique and Gavrilets's alternative model. But while admitting the plausibility of Gavrilets's models of speciation by genetic drift, they nevertheless maintain that “the models seem unnecessary when compared to adaptive ones” (p. 398). Coyne and Orr appear to adopt selection as the null hypothesis for speciation, whereas drift is generally taken as the null hypothesis in much of evolutionary genetics, for the simple reason that drift operates at all loci in all finite (i.e., real) populations, whereas selection need not. The burden of demonstrating that selection is not responsible for an evolutionary event (i.e., demonstrating a negative) is, of course, far heavier than the burden of demonstrating selection; indeed, Coyne and Orr do not address the difficult question of what would constitute evidence for drift. Having, perhaps, stacked the deck, Coyne and Orr find almost no evidence that drift has contributed to speciation in nature, but conclude that there is “considerable evidence” that selection has done so. However, the amount of evidence is hardly on a par with, say, the evidence for allopatric speciation. It consists of only about eight studies of ecological selection, indications that diversification rates are associated with greater scope for sexual selection, selection signatures in a few genes that underlie genetic incompatibility, and a paucity of molecular evidence for bottlenecks (i.e., opportunities for founder events) in the history of recently formed species. But the evidence on the role of sexual selection is very indirect, and the high levels of genetic variation revealed in molecular studies argue against past bottlenecks only if this is ancestral variation, rather than variation generated anew since a possible bottleneck—a question that has been addressed in only a few cases. Assuming that experiments with laboratory populations can be validly extrapolated to natural speciation processes, founder-effect speciation may indeed be a moribund hypothesis, but I do not believe long-term genetic drift can yet be ruled out, and cannot agree that this “important debate has been settled” (p. 410). The geography of speciation continues to be one of the most difficult and contentious topics, and undoubtedly will remain so despite the careful analyses by these authors. They agree that parapatric speciation (evolution of RI between neighboring populations that exchange genes) is theoretically plausible, but Gavrilets notes that although it has become clear that its likelihood is sensitive to several model parameters, parapatric speciation is difficult to model and has been shamefully neglected. Coyne and Orr do not doubt that it is a fairly common mode of speciation, yet “it is almost impossible to demonstrate parapatric speciation in nature” (p. 118), and no cases have been well documented. Gavrilets provides an exhaustive analysis of the many models of sympatric speciation, and identifies some key issues that have been underemphasized. For example, the sympatric evolution of behavioral isolation by “matching traits” (e.g., genetically independent male signal and female preference) is generally much more difficult than “similaritybased” mating (in which females choose males that have the same phenotypic trait as themselves). Just how common the latter is in animals is an open question that Coyne and Orr unfortunately do not address. Gavrilets also identifies the cost of female choosiness as a critical issue: many models of sympatric speciation depend on the assumption that females always succeed in mating even if the male type they prefer is rare, so their choosiness has no cost. Gavrilets criticizes some popular models of sympatric speciation on these and other grounds, and while granting that sympatric speciation by divergent habitat or host preference is plausible, he concludes that it need not be faster than allopatric speciation and that “contrary to common claims in recent theoretical papers, conditions for sympatric speciation are not wide and sympatric speciation does not occur easily” (p. 404). For their part, Coyne and Orr feel that the prevalence of sympatric speciation is an empirical issue (but a very difficult one), and undertake a broad, detailed review. They identify three examples of completed speciation in which a sympatric scenario “seems plausible.” I see no reason to accept one of these cases, a pair of sister species of “parasites” (fig wasps) on the same host species, since allopatric speciation of a widespread parasite need not be accompanied by speciation of its host. Moreover, Coyne and Orr note weaknesses in all three cases, as well as in examples of “host races” that have been advanced as species in statu nascendi . Coyne and Orr's conclusion echoes Gavrilets's: “It is hard to see how the data at hand can justify the current wave of enthusiasm for sympatric speciation” (p. 178). Bravi! I have indicated some disagreements with Coyne and Orr, and could certainly cite others. But whatever weaknesses their book may have (more ecology and phylogeny, anyone?) are much less important than its strengths. The strengths of Speciation are not only Coyne and Orr's comprehensive, scholarly coverage of an exceedingly broad subject, but also, and especially, their rigorous, incisive analysis, coupled with strongly stated conclusions and suggestions for how to resolve controversies. Many readers will have a visceral reaction against their position on sympatric speciation, reinforcement, founder-effect speciation, or other issues—but can these readers counter Coyne and Orr's arguments with equally cogent analysis? Or are these subjects that simply require more, and perhaps more imaginative, research? Together, these books provide a comprehensive, thoughtful synthesis of our current understanding of one of the most important processes in evolution. They are required reading for anyone who studies species and speciation. I recommend Speciation and the nonmathematical final chapter (“General Conclusions”) of Fitness Landscapes and the Origin of Species to all evolutionary biologists, students, and professionals alike. It may not take another two decades for the next foundational books on speciation to appear, but these books will fill that role for a long time to come.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC548954.xml
555534	Low frequency of defective mismatch repair in a population-based series of upper urothelial carcinoma	Background Upper urothelial cancer (UUC), i.e. transitional cell carcinomas of the renal pelvis and the ureter, occur at an increased frequency in patients with hereditary nonpolyposis colorectal cancer (HNPCC). Defective mismatch repair (MMR) specifically characterizes HNPCC-associated tumors, but also occurs in subsets of some sporadic tumors, e.g. in gastrointestinal cancer and endometrial cancer. Methods We assessed the contribution of defective MMR to the development of UUC in a population-based series from the southern Swedish Cancer Registry, through microsatellite instability (MSI) analysis and immunohistochemical evaluation of expression of the MMR proteins MLH1, PMS2, MSH2, and MSH6. Results A MSI-high phenotype was identified in 9/216 (4%) successfully analyzed patients and a MSI-low phenotype in 5/216 (2%). Loss of MMR protein immunostaining was found in 11/216 (5%) tumors, and affected most commonly MSH2 and MSH6. Conclusion This population-based series indicates that somatic MMR inactivation is a minor pathway in the development of UUC, but tumors that display defective MMR are, based on the immunohistochemical expression pattern, likely to be associated with HNPCC.	Background Upper urothelial carcinomas (UUC) represent about 5% of the urinary tract tumors, with transitional cell carcinomas of the renal pelvis and the ureter being the most common [ 1 ]. Exogenous agents such as smoking and occupational exposures to e.g. acrylamines constitute risk factors that are estimated to cause up to half of the tumors [ 2 ]. Hereditary factors also contribute to the development of UUC with a 2-fold increased risk among first-degree relatives [ 3 ]. The familial cases develop due to site-specific inheritance as well as within the hereditary nonpolyposis colorectal cancer (HNPCC) syndrome [ 1 , 3 - 5 ]. Individuals with HNPCC are at increased risk for several types of cancer, with the highest life-time risks for colorectal cancer (80%), endometrial cancer (40–60%), ovarian cancer (10–15%), cancer of the small intestine and upper urothelial cancer [ 4 ], and the revised Amsterdam criteria for the diagnosis of HNPCC consider these tumor types to be associated with the syndrome [ 6 ]. Although HNPCC-patients have a 14 to 75-fold increased risk of UUC, with the highest risks reported for carriers of mutations in MSH2 , the absolute lifetime risk for mutation carriers to develop UUC is <10% [ 7 - 9 ]. HNPCC is caused by a germline mutation in a DNA mismatch-repair (MMR) gene, most commonly affecting either of the genes MLH1 (40%), MSH2 (50%) or MSH6 (10%) [ 10 , 11 ]. Over 95% of the HNPCC-tumors are characterized by wide-spread microsatellite instability (MSI) and 90% by loss of immunohistochemical expression of the MMR protein affected [ 12 ]. Hence, these analyses are used in the clinical diagnosis of suspected HNPCC cases. However, somatic MMR defects occur in a subset of certain sporadic tumor types, e.g. in 15–20% of gastrointestinal and endometrial cancer, and are in most of these tumors caused by somatic hypermethylation of the MLH1 promoter [ 13 , 14 ]. Studies of the contribution of defective MMR to the development of urothelial carcinomas, assessed using MSI analysis, loss of MMR protein expression, and MMR gene mutations, have found a low frequency (<10%) of MMR defects in urothelial carcinomas of the urinary bladder [ 15 ], but have indicated a high frequency (15–45%) of MMR defects in UUC [ 16 - 19 ]. Since data on the frequency of MMR defects in UUC are scarce and in order to characterize the contribution of the different MMR proteins to development of UUC, we assessed MSI and immunohistochemical expression of MLH1, MSH2, and MSH6 in a population-based series of UUC. Methods Patient Material In Sweden a population-based national Cancer Registry was started in 1958 and applies mandatory registration by both clinicians and pathologists in order to achieve maximal coverage (estimated to be 98%). We applied the southern Swedish part of the registry, which currently contains about 300.000 entries, to identify all carcinomas of the upper urothelial tract that had developed between 1992 and 1999. We identified 262 patients with a median age of 70 (range 34–90) years and a male:female ratio of 1.8:1. For further analyses, 27 patients were excluded because of lack of tumor blocks, and 19 because of autopsy-based diagnosis with autolysis that prevented good quality immunostaining. Hence, 216 patients with a median age of 69 (range 34–89) years were analyzed. Tumor location was as follows for the whole material (cases analyzed within parenthesis): renal pelvis 173 (154), ureter 75 (60) and an unspecified tumor location in 14 (2) patients. Data on family history of cancer or blood samples for mutation analysis were not available since the study was retrospective and register-based. Ethical approval for the study was obtained from the ethics committee at Lund University. Analysis of microsatellite instability (MSI) Representative tumor blocks containing at least 20% tumor tissue were selected and DNA was extracted from 3 × 10-μm sections of formalin-fixed, paraffin-embedded tissue through incubation of the samples in EDTA-Tris-buffer with Proteinase K at 65°C for at least 2 hours, followed by boiling, centrifugation, and removal of the aqueous phase, which was stored at 4°C. MSI was assessed with the National Cancer Institute (NCI) panel; BAT25, BAT26, BAT34, BAT40, D2S123 and D5S346 [ 20 ]. These markers identify MSI with high accuracy in colorectal cancer, but the sensitivity of individual markers may vary between different tumor types [ 21 ]. The primer sequences used have been reported previously [ 22 ]. The markers were fluorescencely-labeled as follows: NED™ (yellow) for BAT25, 6-FAM™ (blue) for BAT26, BAT34C4 and D2S123, and HEX™ (green) for BAT40 and D5S346. The DNA microsatellite sequences were amplified by PCR according to the following programme; 94°C for 7 minutes, 10 × (94°C for 15 seconds, 45°C (BAT 25) / 50°C (other markers) for 15 seconds, and 72°C for 15 seconds), 23 × (89°C for 15 seconds, 45°C / 50°C for 15 seconds, and 72°C for 15 seconds), 72°C for 7 minutes, followed by a final cooling step at 4°C. 0.5–2 μl PCR product was mixed with 12 μl deionized formamide (Hi-Di Formamide, Applied Biosystems) and 0.5 μl ROX™ 500 Size Standard (Applied Biosystems), denatured at 95°C for 2 minutes, chilled on ice, and separated in Performance Optimized Polymer-4 (POP-4™) on the ABI PRISM™ 3100 Genetic Analyzer (Applied Biosystems) for fragment analysis. MSI was defined by the presence of extra peaks demonstrating altered length of the repetitive sequence. Data from at least three markers were required for the classification of tumors as microsatellite stable (MSS). The tumors were regarded as MSI-high if at least two microsatellites showed instability and as MSI-low if only one marker showed instability. All cases with suspected MSI were verified through repeated analysis. Immunohistochemistry Immunohistochemical staining was performed using 4-μm sections of formaline fixed, paraffin-embedded tissue, which were mounted on DAKO ChemMate Capillary Gap Microscope Slides (DAKO A/S BioTek Solutions, USA) and dried at room temperature overnight followed by incubation at 60°C for 1–2 hours. The tissue sections were deparaffinized in xylol and rehydrated through descending concentrations of alcohol. Antigen retrieval was achieved by microwave-treatment in 1 mM EDTA, pH 9.0, at 900 W for 8 minutes followed by 15 minutes at 350 W. The slides were then allowed to cool for at least 20 minutes in the EDTA-solution. Immunohistochemical staining was performed in an automated immunostainer (TechMate 500 Plus, DAKO), according to the manufacturers instructions. The main steps were as follows: Mouse monoclonal IgG antibodies to MLH1 (clone G168-15, dilution 1:100, PharMingen, San Diego, CA, USA) MSH2 (clone FE-11, dilution 1:100, Oncogene research products, Boston, MA, USA), MSH6 (clone 44, dilution 1:1000, BD Transduction Laboratories) and PMS2 (clone:A16-4, dilution 1:500, BD Pharmingen) were applied and the sections were incubated at room temperature for 25 minutes. Thereafter, the slides were incubated with biotinylated anti-mouse antibody (DAKO) for 25 minutes (for MLH1 and MSH2) or with rabbit anti-mouse immunoglobulins (DAKO, dilution 1:400) for 20 min (for MSH6 and PMS2). Endogenous peroxidase activity was blocked in Peroxidase-blocking solution (DAKO) for 3 × 2,5 minutes. This was followed by incubation with streptavidin-horseradish peroxidase for 25 minutes for MLH1 and MSH2, whereas EnVision™/HRP rabbit/mouse (DAKO) incubation for 25 min was used for MSH6 and PMS2. Finally, the tissue sections were treated with diaminobenzidine (DAB) for 3 times 5 min, counterstained with hematoxylin for 1 min, rinsed in running tap water for 10 min, dehydrated in ascending concentrations of alcohol and mounted. After each step, the sections were rinsed in Tris buffered saline, pH 7.4, and Tween-20. In order to block nonspecific protein binding, bovine serum albumin was added to the buffer before the antibody incubation steps in the MLH1 and MSH2 stainings. A detailed protocol is available from the authors upon request. Two of the authors (K.E. and M.N.), who were blinded regarding the MSI status, independently evaluated all stained sections. Sections without nuclear staining in the tumor cells, in the presence of normal nuclear staining in lymphocytes and normal epithelial or stromal cells in the same section, were considered to have a lost expression (Fig. 1 ). The expression was classified as present, absent or non-evaluable without grading of the staining intensity. Results Microsatellite analysis For the MSI analysis of the 216 cases, 16 tumors were excluded because of small tumor size or less than 20% tumor tissue in the samples, and 6 tumors were excluded because of lack of information from at least 3 MSI markers, which left 194 tumors successfully analyzed. A MSS phenotype was identified in 180 tumors, MSI-low in 5, and MSI-high in 9 tumors (table 1 , figure 1 ). Immunohistochemistry Immunohistochemical staining for the MMR proteins gave evaluable results for MLH1 in 211 tumors, MSH2 in 216, MSH6 in 200 tumors and PMS2 in 215 tumors. Of the 180 MSS tumors, 177 showed retained expression for all evaluable proteins, as did also the 5 MSI-low tumors. One MSS tumor that was not assessed for MSI due to a small amount of tumor material, showed loss of MSH2 and MSH6 expression, one MSS tumor showed loss of MLH1 and PMS2, one MSS tumor showed loss of MSH6 expression, and one MSS tumor showed loss of MSH2 expression. Among the 9 MSI-high tumors, 5 showed a concomitant loss of expression of MSH2 and MSH6, 1 tumor showed loss of expression of MSH6 and 1 tumor showed loss of expression of MLH1 and PMS2. Retained expression of all four proteins was found in 2 MSI-high tumors of the renal pelvis (table 1 , figure 2 ). Synchronous and metachronous tumors Eleven patients had developed synchronous tumors of the upper urinary tract, and these cases were all analyzed. In one patient with synchronous urethral tumors the tumor tissue showed MSI and loss of expression of MLH1 in both tumors (U2-229), and the other patients with synchronous tumors had MSS tumors all of which showed retained expression of all three MMR proteins (table 1 ). Metachronous UUC occurred in 3/262 patients, 2 of whom were included in the series analyzed, and these tumors were MSS and MSI-low, respectively, but both tumors showed retained MMR protein expression. In the whole series, 122 (97 among the analyzed cases) patients had been diagnosed with another malignancy, which was bladder cancer in 67 cases (54 among the cases analyzed). Among the cases with MSI and/or immunohistochemical MMR protein loss, 8 metachronous tumors developed and 5/5 analyzed (a leiomyosarcoma, a colon tumor, an endometrial cancer and 2 bladder tumors) displayed MSI and immunohistochemical loss of the concordant MMR protein (table 1 ). Discussion Urothelial carcinomas of the upper and the lower urothelial tract share many clinical and epidemiological traits. However, the UUC have specifically been associated with HNPCC, and in line with this observation the contribution of defective MMR has been reported to differ between these tumor types. In bladder cancer, a MSI-high phenotype has been found in 3–10% of the tumors [ 15 , 23 ], whereas elevated microsatellite alterations at selected tetranucleotides (EMAST) has been described at a higher rate in bladder cancer and the latter phenomenon is being perused as a tumor marker in urine [ 24 ]. Higher MSI rates have been reported in UUC with 13–31% of the tumors showing MSI [ 16 - 19 ]. A similar anatomical specificity has been described in the ventricle with a higher number of MSI tumors in the antrum, and in the colorectum with 20% MSI tumors in the cecum and <5% in the rectum [ 25 - 27 ]. We applied the population-based southern Swedish Cancer Registry to assess the contribution of defective MMR to the development of UUC. The results are based on 216/262 (82%) of the tumors that occurred in the southern Sweden health care region between 1992 and 1999. A MSI-high phenotype was found in 9/216 (4%) patients and a MSI-low phenotype in 5/216 (2%). In 11/216 cases synchronous tumors occurred within the urothelial tract and 1 patient (U2-229) had synchronous MSI-high tumors, all of which displayed a concordant immunohistochemical loss of MLH1. Thus, the vast majority of synchronous UUC does not display MMR defects and does not develop within the HNPCC syndrome. The overall frequency of MSI tumors detected in our study, 4%, is lower than the 13–31% previously reported (table 2 ) [ 16 - 19 ]. Possible reasons for the discrepancy include that our study was unselected and population-based. Furthermore, Müller et al . [ 28 ] have suggested that microsatellite instability analysis should optimally be performed by using microdissection, where analysis is made on DNA extracted from tumor cells without dilution of DNA from normal cells. However, this was not available at our institution at the time the study. The marker selection is probably not the cause of discrepancy since the NCI marker panel for MSI analysis has proven effective in several extracolonic tumor types such as endometrial, ovarian and gastric cancer [ 20 , 29 ]. Hartmann et al. [ 18 ] identified BAT40 (93% detection rate) and BAT25 (53%) as the best markers for the detection of MSI also in UUC and indeed reported that using a combination of the markers BAT40, BAT25, and BAT26 allowed identification of all MSI tumors. Whereas our finding of 5% MMR defects in renal pelvis tumors is in accordance with the 5–8% previously reported, we identified MMR defects in a lower (4%) fraction of the urethral tumors than the 25–41% previously reported [ 16 - 19 , 23 ]. Loss of immunostaining was in our series detected in 7/9 MSI-high tumors, in one tumor biopsy that was to small to allow MSI analysis, and in 2 MSS tumors (figure 2 ). The immunohistochemical expression loss affected MSH2/MSH6 in 6 cases, MSH2 in one, MSH6 in two, and MLH1/PMS2 in two cases. Concordant loss of the same MMR protein was observed in the patient who had developed multiple synchronous tumors (table 1 ). This frequency of immunohistochemical loss of expression in MSI-high tumors is similar to that previously reported (table 2 ) and thus demonstrates that loss of immunostaining for at least one of the MMR proteins investigated is found in about 85% of MSI-high UUC [ 18 , 23 ]. Regarding histological grade and stage among the tumors with MSI and/or MMR protein expression loss, the majority of the tumors were moderate differentiated (WHO-grade 2–3) and of early stages (table 1 ). Synchronous/metachronous tumor development is common in urothelial cancer, mainly through intraepithelial migration or intraluminal dispersion of tumor cells [ 30 ]. In our series, 54/216 (25%) patients had developed metachronous bladder tumors. An increased incidence of metachronous tumors has been observed in patients with MMR defective UUC [ 19 ], and synchronous UUC is found in 1–2% of UUC patients [ 31 ]. Of the 8 patients with MMR defective tumors in our study, 5 had developed other malignant tumors, including two cancers of the urinary bladder, one colon cancer, one rectal cancer, one endometrial cancer, one cervical cancer, one soft tissue sarcoma, and one patient who had developed myelofibrosis (table 1 ). Among these tumors, 5 could be retrieved and were immunohistochemically stained. The leiomyosarcoma, the colon tumor, the endometrial cancer and 2 bladder tumors showed loss of expression for MSH2/MSH6, which suggests an association with HNPCC. About 1/3 of HNPCC patients develop metachronous primary tumors, and the concordant MSI and loss of MMR protein expression in these cases strongly suggests HNPCC, although mutation analysis was not performed. The lifetime risk of developing UUC in HNPCC mutation carriers is estimated to be 4–10%, and UUC is in the revised Amsterdam criteria considered to be a HNPCC-associated tumor type, and screening for UUC is generally recommended in HNPCC-families ( ), with sonogrophy and urinary analysis. None of the patients in this series with MSI and/or IHC loss of MMR protein expression are previously known HNPCC patients in our health care region. Mutation analysis is not planned. A higher frequency of extraintestinal tumors has been reported in families with germline mutations in MSH2 , and from the data available, MSH2 seems to play a predominant role also in UUC; loss of MSH2 expression has been reported in 33–60% of MSI-high UUC tumors [ 18 , 19 , 32 , 33 ]. Although our data suggest that MMR defects represent a minor tumorigenic pathway in the development of UUC, the high frequency of MSH2/MSH6 loss in MMR-defective tumors should caution clinicians to obtain an individual and a family history of cancer in patients with carcinomas of the renal pelvis and the ureter. Competing interests The author(s) declare that they have no competing interests. Authors' contributions KE conceived of the study, carried out the MSI analysis, performed immunohistochemical validation and drafted the manuscript. AI also carried out the MSI analysis. BI performed immunohistochemical validation. MN conceived of the study, and participated in its design and coordination and helped to draft the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC555534.xml
423139	A Calculus of Purpose	Biological systems are so complex that we must ask: "What purpose does all this complexity serve?" Lander argues that computational biology may help provide answers	Why is the sky blue? Any scientist will answer this question with a statement of mechanism: Atmospheric gas scatters some wavelengths of light more than others. To answer with a statement of purpose—e.g., to say the sky is blue in order to make people happy—would not cross the scientific mind. Yet in biology we often pose “why” questions in which it is purpose, not mechanism, that interests us. The question “Why does the eye have a lens?” most often calls for the answer that the lens is there to focus light rays, and only rarely for the answer that the lens is there because lens cells are induced by the retina from overlying ectoderm. It is a legacy of evolution that teleology—the tendency to explain natural phenomena in terms of purposes—is deeply ingrained in biology, and not in other fields ( Ayala 1999 ). Natural selection has so molded biological entities that nearly everything one looks at, from molecules to cells, from organ systems to ecosystems, has (at one time at least) been retained because it carries out a function that enhances fitness. It is natural to equate such functions with purposes. Even if we can't actually know why something evolved, we care about the useful things it does that could account for its evolution. As a group, molecular biologists shy away from teleological matters, perhaps because early attitudes in molecular biology were shaped by physicists and chemists. Even geneticists rigorously define function not in terms of the useful things a gene does, but by what happens when the gene is altered. Molecular biology and molecular genetics might continue to dodge teleological issues were it not for their fields' remarkable recent successes. Mechanistic information about how a multitude of genes and gene products act and interact is now being gathered so rapidly that our inability to synthesize such information into a coherent whole is becoming more and more frustrating. Gene regulation, intracellular signaling pathways, metabolic networks, developmental programs—the current information deluge is revealing these systems to be so complex that molecular biologists are forced to wrestle with an overtly teleological question: What purpose does all this complexity serve? In response to this situation, two strains have emerged in molecular biology, both of which are sometimes lumped under the heading “systems biology.” One strain, bioinformatics, champions the gathering of even larger amounts of new data, both descriptive and mechanistic, followed by computerbased data “mining” to identify correlations from which insightful hypotheses are likely to emerge. The other strain, computational biology, begins with the complex interactions we already know about, and uses computer-aided mathematics to explore the consequences of those interactions. Of course, bioinformatics and computational biology are not entirely separable entities; they represent ends of a spectrum, differing in the degree of emphasis placed on large versus small data sets, and statistical versus deterministic analyses. Computational biology, in the sense used above, arouses some skepticism among scientists. To some, it recalls the “mathematical biology” that, starting from its heyday in the 1960s, provided some interesting insights, but also succeeded in elevating the term “modeling” to near-pejorative status among many biologists. For the most part, mathematical biologists sought to fit biological data to relatively simple mathematical models, with the hope that fundamental laws might be recognized ( Fox Keller 2002 ). This strategy works well in physics and chemistry, but in biology it is stymied by two problems. First, biological data are usually incomplete and extremely imprecise. As new measurements are made, today's models rapidly join tomorrow's trash heaps. Second, because biological phenomena are generated by large, complex networks of elements, there is little reason to expect to discern fundamental laws in them. To do so would be like expecting to discern the fundamental laws of electromagnetism in the output of a personal computer. Nowadays, many computational biologists avoid modeling-as-data-fitting, opting instead to create models in which networks are specified in terms of elements and interactions (the network “topology”), but the numerical values that quantify those interactions (the parameters) are deliberately varied over wide ranges. As a result, the study of such networks focuses not on the exact values of outputs, but rather on qualitative behavior, e.g., whether the network acts as a “switch,” “filter,” “oscillator,” “dynamic range adjuster,” “producer of stripes,” etc. By investigating how such behaviors change for different parameter sets— an exercise referred to as “exploring the parameter space”—one starts to assemble a comprehensive picture of all the kinds of behaviors a network can produce. If one such behavior seems useful (to the organism), it becomes a candidate for explaining why the network itself was selected, i.e., it is seen as a potential purpose for the network. If experiments subsequently support assignments of actual parameter values to the range of parameter space that produces such behavior, then the potential purpose becomes a likely one. For very simple networks (e.g., linear pathways with no delays or feedback and with constant inputs), possible global behaviors are usually limited, and computation rarely reveals more than one could have gleaned through intuition alone. In contrast, when networks become even slightly complex, intuition often fails, sometimes spectacularly so, and computation becomes essential. For example, intuitive thinking about MAP kinase pathways led to the long-held view that the obligatory cascade of three sequential kinases serves to provide signal amplification. In contrast, computational studies have suggested that the purpose of such a network is to achieve extreme positive cooperativity, so that the pathway behaves in a switch-like, rather than a graded, fashion ( Huang and Ferrell 1996 ). Another example comes from the study of morphogen gradient formation in animal development. Whereas intuitive interpretations of experiments led to the conclusion that simple diffusion is not adequate to transport most morphogens, computational analysis of the same experimental data yields the opposite conclusion ( Lander et al. 2002 ). As the power of computation to identify possible functions of complex biological networks is increasingly recognized, purely (or largely) computational studies are becoming more common in biological journals. This raises an interesting question for the biology community: In a field in which scientific contributions have long been judged in terms of the amount of new experimental data they contain, how does one judge work that is primarily focused on interpreting (albeit with great effort and sophistication) the experimental data of others? At the simplest level, this question poses a conundrum for journal editors. At a deeper level, it calls attention to the biology community's difficulty in defining what, exactly, constitutes “insight” ( Fox Keller 2002 ). In yesterday's mathematical biology, a model's utility could always be equated with its ability to generate testable predictions about new experimental outcomes. This approach works fine when one's ambition is to build models that faithfully mimic particular biological phenomena. But when the goal is to identify all possible classes of biological phenomena that could arise from a given network topology, the connection to experimental verification becomes blurred. This does not mean that computational studies of biological networks are disconnected from experimental reality, but rather that they tend, nowadays, to address questions of a higher level than simply whether a particular model fits particular data. The problem this creates for those of us who read computational biology papers is knowing how to judge when a study has made a contribution that is deep, comprehensive, or enduring enough to be worth our attention. We can observe the field trying to sort out this issue in the recent literature. A good example can be found in an article by Nicholas Ingolia in this issue of PLoS Biology ( Ingolia 2004 ), and an earlier study from Garrett Odell's group, upon which Ingolia draws heavily ( von Dassow et al. 2000 ). Both articles deal with a classical problem in developmental biology, namely, how repeating patterns (such as stripes and segments) are laid down. In the early fruit fly embryo, it is known that a network involving cell-to-cell signaling via the Wingless (Wg) and Hedgehog (Hh) pathways specifies the formation and maintenance of alternating stripes of gene expression and cell identity. This network is clearly complex, in that Wg and Hh signals affect not only downstream genes, but also the expression and/or activity of the components of each other's signaling machinery. Von Dassow et al. (2000) calculated the behaviors of various embodiments of this network over a wide range of parameter values and starting conditions. This was done by expressing the network in terms of coupled differential equations, picking parameters at random from within prespecified ranges, solving the equation set numerically, then picking another random set of parameters and obtaining a new numerical solution, and so forth, until 240,000 cases were tried. The solutions were then sorted into groups based on the predicted output—in this case, spatial patterns of gene expression. When they used a network topology based only upon molecular and generegulatory interactions that were firmly known to take place in the embryo, they were unable to produce the necessary output (stable stripes), but upon inclusion of two molecular events that were strongly suspected of taking place in the embryo, they produced the desired pattern easily. In fact, they produced it much more easily than expected. It appeared that a remarkably large fraction of random parameter values produced the very same stable stripes. This implied that the output of the network is extraordinarily robust, where robustness is meant in the engineering sense of the word, namely, a relative insensitivity of output to variations in parameter values. Because real organisms face changing parameter values constantly—whether as a result of unstable environmental conditions, or mutations leading to the inactivation of a single allele of a gene—robustness is an extremely valuable feature of biological networks, so much so that some have elevated it to a sort of sine qua non ( Morohashi et al. 2002 ). Indeed, the major message of the von Dassow article was that the authors had uncovered a “robust developmental module,” which could ensure the formation of an appropriate pattern even across distantly related insect species whose earliest steps of embryogenesis are quite different from one another ( von Dassow et al. 2000 ). There is little doubt that von Dassow's computational study extracted an extremely valuable insight from what might otherwise seem like a messy and ill-specified system. But Ingolia now argues that something further is needed. He proposes that it is not enough to show that a network performs in a certain way; one should also find out why it does so. Ingolia throws down the gauntlet with a simple hypothesis about why the von Dassow network is so robust. He argues that it can be ascribed entirely to the ability of two positive feedback loops within the system to make the network bistable. Bistability is the tendency for a system's output to be drawn toward either one or the other of two stable states. For example, in excitable cells such as neurons, depolarization elicits sodium entry, which in turn elicits depolarization—a positive feedback loop. As a result, large depolarizations drive neurons to fully discharge their membrane potential, whereas small depolarizations decay back to a resting state. Thus, the neuron tends strongly toward one or the other of these two states. The stability of each state brings with it a sort of intrinsic robustness— i.e., once a cell is in one state, it takes a fairly large disturbance to move it into the other. This is the same principle that makes electronic equipment based on digital (i.e., binary) signals so much more resistant to noise than equipment based on analog circuitry. Ingolia not only argues that robustness in the von Dassow model arises because positive feedback leads to network bistability, he further claims that such network bistability is a consequence of bistability at the single cell level. He strongly supports these claims through computational explorations of parameter space that are similar to those done by von Dassow et al., but which also use strippeddown network topologies (to focus on individual cell behaviors), test specifically for bistability, correlate results with the patterns formed, and ultimately generate a set of mathematical rules that strongly predict those cases that succeed or fail at producing an appropriate pattern. At first glance, such a contribution might seem no more than a footnote to von Dassow's paper, but a closer look shows that this is not the case. Without mechanistic information about why the von Dassow network does what it does, it is difficult to relate it to other work, or to modify it to accommodate new information or new demands. Ingolia demonstrates this by deftly improving on the network topology. He inserts some new data from the literature about the product of an additional gene, sloppy-paired , in Hh signaling, removes some of the more tenuous connections, and promptly recovers a biologically essential behavior that the original von Dassow network lacked: the ability to maintain a fixed pattern of gene expression even in the face of cell division and growth. Taken as a pair, the von Dassow and Ingolia papers illustrate the value of complementary approaches in the analysis of complex biological systems. Whereas one emphasizes simulation (as embodied in the numerical solution of differential equations), the other emphasizes analysis (the mathematical analysis of the behavior of a set of equations). Whereas one emphasizes exploration (exploring a parameter space), the other emphasizes the testing of hypotheses (about the origins of robustness). The same themes can be seen in sets of papers on other topics. For example, in their analysis of bacterial chemotaxis, Leibler and colleagues ( Barkai and Leibler 1997 ) found a particular model to be extremely robust in the production of an important behavior (exact signal adaptation), and subsequently showed that bacteria do indeed exhibit such robust adaptation ( Alon et al. 1999 ). Although Leibler and colleagues took significant steps toward identifying and explaining how such robustness came about, it took a subsequent group ( Yi et al. 2000 ) to show that robustness emerged as a consequence of a simple engineering design principle known as “integral feedback control.” That group also showed, through mathematical analysis, that integral feedback control is the only feedback strategy capable of achieving the requisite degree of robustness. From these and many other examples in the literature, one can begin to discern several of the elements that, when present together, elevate investigations in computational biology to a level at which ordinary biologists take serious notice. Such elements include network topologies anchored in experimental data, fine-grained explorations of large parameter spaces, identification of “useful” network behaviors, and hypothesisdriven analyses of the mathematical or statistical bases for such behaviors. These elements can be seen as the foundations of a new calculus of purpose, enabling biologists to take on the much-neglected teleological side of molecular biology. “What purpose does all this complexity serve?” may soon go from a question few biologists dare to pose, to one on everyone's lips.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC423139.xml
546212	Structure and evolution of the mouse pregnancy-specific glycoprotein (Psg) gene locus	Background The pregnancy-specific glycoprotein ( Psg ) genes encode proteins of unknown function, and are members of the carcinoembryonic antigen ( Cea ) gene family, which is a member of the immunoglobulin gene ( Ig ) superfamily. In rodents and primates, but not in artiodactyls (even-toed ungulates / hoofed mammals), there have been independent expansions of the Psg gene family, with all members expressed exclusively in placental trophoblast cells. For the mouse Psg genes, we sought to determine the genomic organisation of the locus, the expression profiles of the various family members, and the evolution of exon structure, to attempt to reconstruct the evolutionary history of this locus, and to determine whether expansion of the gene family has been driven by selection for increased gene dosage, or diversification of function. Results We collated the mouse Psg gene sequences currently in the public genome and expressed-sequence tag (EST) databases and used systematic BLAST searches to generate complete sequences for all known mouse Psg genes. We identified a novel family member, Psg31 , which is similar to Psg30 but, uniquely amongst mouse Psg genes, has a duplicated N1 domain. We also identified a novel splice variant of Psg16 ( bCEA ). We show that Psg24 and Psg30 / Psg31 have independently undergone expansion of N-domain number. By mapping BAC, YAC and cosmid clones we described two clusters of Psg genes, which we linked and oriented using fluorescent in situ hybridisation (FISH). Comparison of our Psg locus map with the public mouse genome database indicates good agreement in overall structure and further elucidates gene order. Expression levels of Psg genes in placentas of different developmental stages revealed dramatic differences in the developmental expression profile of individual family members. Conclusion We have combined existing information, and provide new information concerning the evolution of mouse Psg exon organization, the mouse Psg genomic locus structure, and the expression patterns of individual Psg genes. This information will facilitate functional studies of this complex gene family.	Background In mammalian pregnancy the interaction between the maternal uterine tissues and foetal trophoblasts is regulated by a wide variety of cellular and endocrinological mechanisms. These mechanisms underpin trophoblastic invasion and remodelling of maternal tissues, placental angiogenesis, and the modulation of maternal immune responses. Central to these processes is the production by trophoblast of a variety of hormones that are found in abundance in the maternal bloodstream during pregnancy [ 1 ]. The pregnancy-specific glycoproteins (PSG) are the most abundant foetal proteins in the maternal bloodstream in late pregnancy [ 2 ]. They are synthesised in the syncytiotrophoblast of the human placenta and giant cells and spongiotrophoblast in the rodent placenta [ 3 - 5 ]. The PSG family of glycoproteins belongs to the carcinoembryonic antigen (CEA) family, which also includes the CEA-related adhesion molecules (CEACAMs). The CEA family is itself part of the immunoglobulin (Ig) superfamily [ 6 ]. The Ig domain structure of the human and rodent PSGs differs. Containing both V-like Ig domains (N), C2-like Ig domains (A and B) and relatively hydrophilic tails (C), domain arrangements in human PSGs are type I (N-A1-A2-B2-C), type IIa (N-A1-B2-C), type IIb (N-A2-B2-C), type III (N-B2-C) and type IV (A1-B2-C) [ 7 ]. In contrast, rodent PSGs are typically comprised of 3, and in a few cases of 5 or 7 N-domains followed by an A-domain [ 8 ]. In the primate / rodent ancestor, the initial duplication of the CEACAM / PSG primordial gene has been estimated to have occurred about 90 Myr ago [ 9 ], approximately at the time of human-rodent divergence. The most probable PSG ancestor in rodents and primates is a CEACAM15-like molecule based on the organisation of N and A domains. CEACAM15 is not classified as a PSG because comparisons of N and A domain sequence identity clearly delineate members of the CEACAM and PSG subfamilies (Roland Zebhauser, WZ, AM, TM, to be published elsewhere). It has been suggested that human and rodent PSG multigene families evolved independently via further gene duplication and exon shuffling events [ 10 ]. There are 11 members of the PSG family in humans that are encoded by genes clustered on chromosome 19q13.2 [ 11 , 12 ]. PSG proteins have a similar domain structure to the CEACAMs, but lack a membrane anchor and are therefore secreted. However, a few variants have been described that are retained within the cell. Conversely, a small number of human and mouse CEACAM variants lack a membrane anchor and are secreted. Membrane-anchored CEACAMs are widely expressed during embryonic development and in adult tissues, and are implicated in carcinogenesis, angiogenesis and regulation of immune functions [ 13 , 14 ]. In contrast, PSGs and some CEACAMs are expressed almost exclusively in trophoblasts of the haemochorial placenta of rodents and primates [ 4 , 5 , 15 ]. The biochemical properties and physiological functions of the PSGs remain to be fully elucidated, although functional experiments and clinical observations are beginning to provide some clues. Low PSG levels in the maternal circulation are associated with threatened abortions, intrauterine growth retardation and foetal hypoxia [ 16 - 19 ]. The importance of PSGs for the maintenance of pregnancy is also underlined by the observation that the application of anti-PSG antibodies or vaccination with PSG induces abortion in mice and monkeys, respectively, and reduces the fertility of non-pregnant monkeys [ 20 , 21 ]. The majority of PSG functional studies have focussed on determining whether PSGs are able to modulate the maternal immune system to prevent rejection of the allotypic foetus. Early studies with complex PSG mixtures isolated from placenta indicated an inhibitory effect on phytohaemagglutinin or allogeneically stimulated lymphocytes [ 22 , 23 ]. In further experiments it was shown that human monocytes secreted anti-inflammatory cytokines in response to PSG exposure. Moreover, recombinant mouse PSG18 was found to induce the production of interleukin (IL)-10 in the mouse macrophage cell line RAW 264.7 [ 24 ]. Human PSG1, PSG6 and PSG11 all induced secretion of IL-10, IL-6 and transforming growth factor (TGF)-β1 [ 25 ]. Whilst IL-10 and TGF-β1 are anti-inflammatory [ 26 ], IL-6 is usually considered to be a proinflammatory cytokine. However IL-6 does have some well-described anti-inflammatory properties [ 27 ]. Furthermore, IL-6 has been shown to indirectly promote trophoblast growth by upregulation of human chorionic gonadotropin (hCG) release by the trophoblast, and induction of granulocyte-macrophage-colony stimulating factor (GM-CSF) [ 28 , 29 ]. Further evidence implicating PSGs in immune modulation arises from PSG mediated suppression of T cells in purulent septic complications of abortion [ 30 ] and elevated circulating PSG levels are correlated with improved symptoms of rheumatoid arthritis [ 31 ]. PSG induction of alternative monocyte activation is of particular importance as it implies a PSG-mediated switching of the immune system from a predominantly T H 1 response to a predominately T H 2 response which is more compatible with a successful pregnancy [ 32 ]. The only PSG receptor identified to date is the integrin-associated CD9 receptor, which was found to bind the N1 domain of both PSG17 [ 33 ] and PSG19 (unpublished data). Additionally, the presence of the conserved tripeptide motif Arg-Gly-Asp (RGD) on a solvent-exposed loop in the N-terminal Ig domain in the majority of human and some lower-primate PSGs implicates a function that involves integrin-related receptors [ 34 ]. Thus it has been speculated that the RGD domain may enable some PSGs to disrupt cell-matrix interactions [ 35 ]. However, no rodent PSG isolated to date possesses an RGD domain. Evidence supporting the hypothesis that the RGD domain may be involved in receptor binding was provided by the discovery that a peptide containing the RGD motif, from human PSG9, bound to a receptor on the surface of a promonocytic cell line [ 36 ]. In common with integrin interactions, this was dependent on the presence of divalent cations and showed sensitivity to cytoskeletal signalling. However, the expected sizes of the receptor subunits differed from those of known integrins, therefore, the identity of the receptor remains elusive. Much current work has focussed on human PSGs due to their possible relevance to disorders of pregnancy. However, the study of rodent PSGs is important because, the evident differences between primate and rodent PSG protein domain structures notwithstanding, there appears to be considerable conservation in terms of expression in trophoblast, independent gene family expansions in mammalian lineages with haemochorial placentation, and postulated immune functions during pregnancy. Moreover, the application of gene targeting and mutagenesis in the mouse is likely to be informative with respect to elucidating the cellular and physiological functions of PSGs. Such experiments will require an accurate genomic map of the mouse Psg locus, which we undertook to produce in the work described herein. It is also pertinent to ask whether the independent expansions of PSG gene families in different mammalian lineages reflect selection for increased gene dosage or for diversification of function mediated through different protein structures or developmental expression patterns. We therefore undertook to examine and correlate protein domain evolution and expression profiles of the various mouse Psg genes to attempt to address this question. Our results suggest that different family members have very different expression levels at different stages of development, which we consider may be supportive of the hypothesis that mouse Psg genes may have evolved divergent functions in mammalian pregnancy. However, mutagenesis of individual family members will be necessary to rigorously test this hypothesis. Results Identification of novel mouse Psg genes For comparative studies of the human PSG family it is relatively easy to compare coding sequences (CDS) and peptide sequences because complete sequence information is available. However the data available for mouse PSGs is not complete, making such analyses difficult. Thus, we firstly collated the currently available public data, and we then attempted to identify sequences for PSGs that were not completely resolved in the databases. Full-length cDNA sequences of Psg17 , Psg18 , Psg19 , Psg21 , Psg23 , Psg28 and Psg30 were identified via basic name searches of the RefSeq RNA database. Their identity was then verified by comparison to cDNA fragment sequences, which were obtained during the course of this work and deposited in GenBank [ 37 ], as misnaming of genes is commonplace in the databases. The cDNA sequence of Psg22 was then identified via BLAST analysis of the mouse RefSeq RNA database using the GenBank partial sequence referenced in Beauchemin et al . [ 37 ]. Psg31 was identified by BLAST analysis of the same database using the full-length Psg30 sequence and found to be the XM_355864.1 predicted transcript. However, there was a discrepancy between the predicted transcript and the sequences of EST clones CK032208 and CN694284. Comparison of these EST sequences with genomic contig NT_039395.2, using pairwise BLAST analysis, revealed that there had been a duplication of the N1 domain exon. We refer to the two N1 domains of Psg31 as N1 and N1* hereafter. The gene and full-length cDNA coding sequences of the remaining mouse genes ( Psg20 , Psg24 , Psg25, Psg26 , Psg27 and Psg29 ) were deduced manually by systematic BLAST analysis of the mouse genome database as described in Methods. None of these predicted cDNAs were observed in the mouse EST database, although all except Psg20 were observed in the Trace Archive EST sequences. A novel splice variant of Psg16 was also found. BLAST analysis of the mouse High Throughput Genomic Sequences (HTGS) database identified contig AC148976.2, which appears to contain the whole Psg16 gene. An alternative exon 1 was discovered upstream of the previously described initiating exon by a pairwise BLAST comparison of this contig with full-length Psg17 coding sequence. The use of this alternative exon 1 produced a transcript that encodes a typical PSG polypeptide complete with a predicted secretory-peptide signal sequence and cleavage site. Multiple hits identified from subsequent BLAST analysis of the mouse EST and Trace Archives EST databases provided evidence that this novel splice variant was placentally expressed in vivo . In contrast, only one hit was obtained by identical analysis using the coding sequence of the brain-specific transcript described in Chen et al . [ 38 ]. This transcript (BC030357) was derived from a retinal cDNA library. The brain-expressed splice variant is generated from an alternative initiation site within exon 2 of the dominant placentally-expressed form of the gene. Alternative promoter usage would explain the brain and placenta-specific expression patterns of these variants of Psg16 . Unlike the brain-specific variant, the placentally-expressed variant possesses a predicted secretory signal peptide at the N-terminus, like most other Psg gene family members. The comparison of the brain derived Psg16 coding sequence with the genomic sequence (AC148976.2) also revealed differences in the encoding of the A-domain. The placental transcript is predicted to be encoded by 5 exons, as are the majority of mouse Psg mRNAs. However, a weak splice donor signal sequence within the fifth exon permits splicing to a strong splice acceptor sequence downstream of the sixth exon, as seen in the brain-expressed transcript. Trace Archive EST data reveals multiple hits to sequences from placental cDNA libraries using the 3' end of the placental Psg16 coding sequence as bait. This confirms the existence of our predicted transcript. Conversely, similar analysis using the brain-expressed variant yielded no hits. The sixth exon is present on a separate randomly ordered gene fragment within the AC148976.2 contig. Psg-ps1 was previously considered to be a pseudogene, based on a point deletion at nucleotide position 30, downstream from the canonical Psg translational start site [ 8 ]. However, despite this frame shift, the open reading frame of this unusual Psg continues 105 bp upstream of the site of the mutation to an alternative ATG. Inspection of the sequence revealed a Kozak consensus, and BLAST analysis of the public EST and Trace Archive EST databases yielded many mRNA clones that contain this region in addition to downstream exons. Hence, this gene is clearly expressed, and we now propose to rename Psg-ps1 as Psg32 hereafter. We note that this mutation and amino terminal extension abolishes the canonical PSG secretory signal and peptide cleavage site. We therefore suggest that if Psg32 is indeed translated, the resulting protein is retained within the cytoplasm. To determine if the deletion observed in BALB/c mice was also present in other murine strains, we amplified and sequenced a 146 bp fragment by PCR using a set of primers specific for the 5'-untranslated region and the leader peptide of Psg32. The deletion observed in the Psg32 cDNA is also present in the genomic DNA of A/J, C57BL6/J, YBR/Ei, and SWR/J inbred mouse strains (data not shown). The nomenclature (past and current) and accession numbers of nucleotide sequences of all the murine PSGs are documented in Table 1 . The genome sequence and predicted CDS and translation products for Psg16 , Psg20 , Psg26 and Psg31 are listed in Additional file 1 . The complete CDS data for all known mouse Psgs (except the brain-specific splice variant of Psg16 ) are listed in Additional file 2 . The complete protein primary sequences for all known mouse Psgs (except the brain-specific splice variant of Psg16 ) are listed in Additional file 3 . Table 1 Summary of mouse PSG nomenclature and sequence accession numbers Current Name Previous Names Accession Number a Comment b Psg16 bCEA AC148976.2 (RC 40000–60000) predicted CDS: join (1878–1941, 4115–4462, 7291–7650, 8750–9109, 11758–12041); bCEA is a splice variant of Psg16 Psg17 Cea2, mmCGM5 NM_007677 Psg18 Cea3, mmCGM6 NM_011963 Psg19 Cea4 NM_011964 Psg20 Cea7 AC079497.1 (113793–127892) predicted CDS: join (1770–1836, 2989–3345, 4997–5356, 6587–6937, 13114–13397) Psg21 Cea8 NM_027403 Psg22 Cea9 NM_001004152.1 Psg23 Cea11 NM_020261 Psg24 Cea12 AC079526 (115000–131000) predicted CDS: join (1648–1696, 2771–3130, 5965–6324, 9351–9710, 10943–11302, 12844–13191, 14196–14479) Psg25 Cea13 NW_000292.1 (RC 890000–910000) predicted CDS: join (4905–4968, 7121–7480, 10406–10765, 11988–12347, 15508–15791) Psg26 Cea14 join (CAAA01217140.1 {RC 1–6315}, CAAA01213459.1 {557–4715}, CAAA01175422.1 {155–2891}) predicted CDS: join (2148–2211, 3292–3651, 5836–6195, 7507–7866, 10823–11106) Psg27 Cea15 AC087156.1 (RC 139366–153050) predicted CDS: join (240–303, 2037–2393, 5271–5630, 6669–7028, 10039–10322) Psg28 Cea16 NM_054063 Psg29 Cea17 AC079526 (183285–194009) predicted CDS: join (1459–1522, 2658–3005, 6275–6634, 8128–8487, 9428–9700) Psg30 XM_145406 GNOMON prediction in NCBI Psg31 AC134475.3 (10000–70000) predicted CDS: join (3923–3986, 5262–5621, 19366–19725, 34382–34741, 36822–37172, 40760–41119, 42413–42763, 47310–47669, 49090–49443, 50473–50756) Psg32 Psg-ps1 XR_000250 GNOMON prediction in NCBI a Where nucleotide start and end positions are shown in parenthesis after accession numbers, they refer to the start and end positions of the genomic sequence excerpt (encompassing the PSG exons) that is included in Additional file 1 . RC indicates that the sequence in Additional file 1 is the reverse complement. b Where we have predicted the full CDS of a PSG (based on common structure and splice sites), the numbers shown refer to exon start and end positions within the excerpted sequence included in Additional file 1 . Domain structure of mouse PSG proteins A schematic representation of the mouse Psg domain structures is shown in Fig. 1 . Of the seventeen mouse Psgs , thirteen encode a common structure of three Ig variable (IgV)-like domains (N-domains) and a single Ig constant (IgC)-like domain (A-domain). Psg24 , Psg30 and Psg31 have an expanded structure created by the duplication of IgV-like domains. An unrooted phylogenetic tree indicates three main branches of IgV-like domain evolution (Fig. 2 ). There is a group consisting of N1 domains, a group of N2 domains and N2-derived domains, and a group of N3 domains and N3-derived domains. Therefore, in agreement with the most common structure observed in Fig. 1 , the ancestral mouse Psg would be expected to have had an N1-N2-N3-A arrangement of domains. The expansion of Psg24 , Psg30 and Psg31 has occurred mostly through duplications of the N2 and N3 IgV-like domains, with the exception of Psg24 N5 and Psg31 N1 domains. Figure 1 Domain organization of mouse PSGs . Mouse PSGs are composed of 3 – 8 IgV-like N domains and one IgC-like A domain. The relative position of potential N-glycosylation sites (consensus amino acid sequence: asparagine-X-threonine / serine; X any amino acid except proline) were identified using the NetNGlyc 1.0 Server online software and indicated by lollipops. Although PSG32 is probably not routed through the endoplasmic reticulum, the putative N-glycosylation sites are shown for comparison. Of the two PSG16 splice variants, only the variant expressed in the placenta is shown. Figure 2 Evolutionary relationships between mouse PSG IgV-like domains . An unrooted evolutionary tree based on ClustalX amino acid sequence alignments showing the relationships between all mouse PSG N-domains. The three main groups N1, N2 and N3 have been ringed for clarity. The scale bar represents 0.1 amino acid substitutions per site. In order to characterise the evolution of the mouse Psg s with expanded domain numbers, Neighbor-Joining (NJ) trees with bootstrap values of 1000 were prepared (Fig. 3A ) and ClustalW amino acid sequence alignments (Fig. 3B ) were studied to identify the origin of the novel IgV-like domains in these three exceptional Psg s. From examination of the data in Fig. 3A(i) it was not apparent from which progenitor domain the Psg24 N5 domain evolved due to lack of confidence in the branch on which it lies. However, using the alignment identities in Fig. 3B(i) it can be seen that, although generally poorly conserved, the best match of 51.2% was obtained by alignment with the N2 domain. Therefore, our evolutionary model assumes that the Psg24 N5 domain arose from an early duplication of the N2 domain. Also, based on agreement of the data in Fig. 3A(i) and 3B(i) , the N2 domain duplicated again more recently to yield the N3 domain. This latter duplication explains why the Psg24 N4 domain is N3-like. The order of these events is shown schematically in Fig. 3C(i) . Figure 3 Domain expansion of Psg24 , Psg30 and Psg31 . A. NJ-trees based on ClustalX amino acid sequence alignments showing: (i) the evolution of PSG24 IgV-like domains compared to those of PSG17; (ii) the evolution of PSG30 IgV-like domains compared to those of PSG17; (iii) the evolution of PSG31 IgV-like domains compared to those of PSG17. The trees were rooted using an outgroup consisting of the N-domain amino acid sequences of human PSG1, PSG2 and PSG3. Alignments were bootstrapped 1000 times yielding node values which are represented as follows < 50%: no mark; 50–74%: marked ; 75–94%: marked ; ≥ 95%: marked *. The scale bar represents 0.1 amino acid substitutions per site. B. The arrangement of domains represented by boxes shaded: cyan for leader (L) peptides; light pink for the N1-domains; dark pink for N2 and N2-like domains; red for N3 and N3-like domains; blue for A-domains. (i) Comparison of Psg17 and Psg24 exon arrangement including identities of amino acid sequence alignments. (ii) Comparison of Psg30 and Psg31 exon arrangements including identities of amino acid sequence alignments. C. Predicted model of IgV-like domain expansion by exon duplications in (i) Psg24 and (ii) Psg30 and Psg31 . Using a similar analysis we propose a model for the expansion of domains within Psg30 and Psg31 (Fig. 3C(ii) ). We suggest that the N4 and N6 domains of Psg30 and Psg31 are derived from a progenitor N2 domain. Similarly, the N5 and N7 domains are derived from a progenitor N3 domain. Expansion is predicted to have occurred in 2 or 3 separate events in a common ancestor of Psg30 and Psg31 . In the first instance the progenitor N3-like and N2-like domains were duplicated, either at different points in evolution or at the same time. The final step was a duplication of both of these daughter domains to create Psg30 and the precursor of Psg31 . The precursor of Psg31 then underwent another duplication, this time of the N1 domain. Expression of Psg genes in mouse placenta at different developmental stages On the basis that all mouse Psg genes originated from a common ancestor, and expanded into a multigene family by duplication and subsequent divergence, the question as to whether the expression patterns have also diversified is relevant to determining the selective forces underlying Psg gene family expansion. As Psg genes are expressed predominantly in the placenta, cDNA was prepared from total RNA extracted from mouse placenta at four stages of development between E10.5 and E17.5. Psg cDNA sequences were then amplified with PCR primers designed to amplify Psg16 – Psg29 inclusive. Size fractionation of PCR products on an ethidium bromide-stained agarose gel, indicates that mouse Psg genes are predominantly expressed from around E15.5, increasing in expression through to at least E17.5 (Fig. 4 ). However, after blotting the products onto nylon membranes and hybridising radiolabelled oligonucleotide probes specific for individual Psg genes (Table 2 ), we observed significant differences in expression profiles of different genes during development. This method is probably semi-quantitative at best but does give some indication of relative expression levels. We observed that Psg16 and Psg26 are weakly expressed at E15.5 but strongly expressed at E17.5. In contrast, Psg17 , Psg18 , Psg21 and Psg23 are expressed strongly at E15.5, further increasing by E17.5. Psg27 shows a similar expression pattern to these four Psgs , but at a relatively low level. Very weak expression was observed on E17.5 for Psg19 , Psg20 , Psg24 , Psg25 and Psg29 , whereas Psg22 and Psg28 were undetectable. Psg30 , Psg31 and Psg32 domain structures had not been finalised and therefore their expression was not analysed in this experiment. Figure 4 Expression of Psg mRNAs during placental development . Total RNA (1 μg) from day 10.5, 12.5, 15.5 and 17.5 BALB/c placentae was reverse transcribed using an oligo (dT) oligonucleotide (reverse PCR primer). After addition of the degenerate Psg-all oligonucleotide (forward PCR primer), which anneals to the cDNA of all known members of the mouse Psg family, Psg cDNAs were amplified by PCR (see schematic diagram depicting generalised mouse Psg cDNA amplification). Aliquots were size-separated by agarose gel electrophoresis. a, PCR products were visualised by ethidium bromide staining. b-o, the amplification products were blotted onto nylon membranes and individual blots were hybridised with single gene-specific 32P-labelled oligonucleotides from the N1 domain regions (Table 2). The location of the primers used for amplification of the Psg cDNAs and the region from which the sequences of the gene-specific oligonucleotides were derived are shown together with a schematic representation of mouse Psg mRNA. The 5'- and 3'-untranslated regions are shown as bold lines. L, leader; N1-N3, IgV-like domains; A, IgC-like domain. Table 2 Oligonucleotides used in this study Oligo Sequence Position Comment Psg17A5' 5'-CTTGCCACACAGCCCGTCAT-3' Psg17 A domain Psg17A3' 5'-TCATCACAGCCAGGATGACT-3' Psg17 A domain mPsg-5' 5'-AWCCTSYTGSYTCCTGC-3' a N1 domain binds to several mouse Psg cDNA sequences mPsg-3' 5'-TGMARGWAYAKGGATGT-3' a N1 domain binds to several mouse Psg cDNA sequences PsgN1-F 5'-GA AGATCT AGCCTCCMTYTTDDCCT-3' a Bgl II intron 1/N1 exon for the amplification of all known Psg N1 exons (except Psg32 ) PsgN1-R 5'-CC ATCGAT TACTTACWGTWSACVTRVA-3' a Cla I N1 exon/intron 2 for the amplification of all known Psg N1 exons (except Psg32 ) Psg32N1-F 5'-GA AGATCT AGCTTTTCTTTTAACCTC-3' Bgl II N1 domain Psg32-exon1 5'-GAGGTGTCCTTGGTGCTTCTC-3' exon 1 Psg32-specific oligo (dT) 5'-TTCTAGAATTCAGCGGCCGC(T) 30 VN-3' a poly(A) tail Psg-all 5'-CCTCCMTYTTDDCCTRCTGS-3' a N1 domain binds to all known Psg cDNA sequences except Psg32 bCEAN/2 5'-GCAAATGTACAGTGGTAG-3' N1 domain Psg16-specific Psg17N 5'-GTGGAATTCTTACCTCCC-3' N1 domain Psg17-specific Psg18N 5'-GGCTGTACTACTATAGTG-3' N1 domain Psg18-specific BK07 5'-AAAGTGCCACCCGGGAA-3' N1 domain Psg19-specific Psg20N 5'-TGCCAAGGTCACTATCCA-3' N1 domain Psg20-specific Psg21N 5'-GCTCTGCATTTTCTGGAC-3' N1 domain Psg21-specific 35N 5'-GTCTGGTATAGAGGGGTG-3' N1 domain Psg22-specific 53N 5'-GCTGTGTATTTACTGGAC-3' N1 domain Psg23-specific 9.3N1 5'-ATAGCAGAGGTGTGACG-3' N1 domain Psg24-specific 11.2N1 5'-ATCTTCTAGGCCTTGCC-3' N1 domain Psg25-specific 189N 5'-CATTCGCTGTACTATAGTG-3' N1 domain Psg26-specific 214N 5'-CGAGTCACCATCCATTCA-3' N1 domain Psg27-specific 2128N 5'-GCACTATAGTTTAACAGCG-3' N1 domain Psg28-specific 9140N 5'-TGCAGTGGTGTCTGACTT-3' N1 domain Psg29-specific Psg-ps1N 5'-TTAGTGCCACCACAAGTG-3' N1 domain Psg32-specific a Standard IUB/IUPAC nucleic acid codes codes have been used to indicate degeneracy where: R = G/A; Y = T/C; K = G/T; M = A/C; S = G/C; W = A/T; B = G/T/C; D = G/A/T; H = A/C/T; V = G/C/A; N = A/C/G/T. To supplement the PCR-based Psg expression studies, we performed 'virtual northern' analysis in silico by screening the public EST database for sequences matching Psg N1 or A-domains and counting the numbers of matches (Fig. 5 ). There was generally good concordance of the virtual data with the RT-PCR data; notably, Psg21 and Psg23 are highly represented in both datasets. However, disagreements were also evident e.g. Psg16 expression was low in the RT-PCR data, but high in the virtual data. A random sample of twenty of the large number of Psg16 EST sequences in the database indicated that all were of placental origin, ruling out contamination with brain-derived sequences as an explanation for the disparity between RT-PCR and virtual analysis. There was also generally good agreement with the results from screening the EST database with N1 and A domain sequences, although the numbers of A-domain hits were 4–5 fold lower than the N1-domain hits. The only exception to this observation was that Psg30 and Psg31 sequences were identified in 2-fold greater abundance when screened with the A domain compared with the N1 domain. Despite some discrepancies, therefore, the combined RT-PCR and virtual Northern data demonstrate that developmental onset of expression, and maximum expression levels, vary considerably within the Psg family. Figure 5 Virtual Northern analysis of the mouse Psg genes . The nucleotide sequences of the Psg exons encoding the N1 or the A domains were used in NCBI-BLAST searches of the GenBank mouse EST database (March 16, 2004) for the presence of Psg transcripts (virtual Northern analysis). A hit was registered when a 100% match for a sequence > 150 bp was observed. Obvious mismatches such as unidentified nucleotides (N) or single nucleotide insertions or deletions (especially at the end of a sequence run) were ignored. Mouse Psg locus genomic organisation The published mouse Psg gene locus is contained on contig NT_039395. However, the complement of Psg genes is incomplete and the majority of gene sequences within the contig are unordered. We therefore decided to determine the organisation of Psg genes within the locus by screening BAC, YAC and cosmid clones using hybridisation with gene-specific oligonucleotide probes. We defined two separate contigs (subclusters) within which the order of Psg genes was determined to the fullest extent possible. The orientation of the two subclusters with respect to each other and the chromosome 7 centromere was determined by fluorescent in situ hybridisation (FISH) analysis. These data are summarised in Fig. 6 . All of the known mouse Psg genes are located within cytobands A1 and A2 on proximal chromosome 7 and are interspersed with other genes, particularly Ceacams , as determined by comparison with the published mouse genomic sequence on contig NT_039395. We did not observe any obvious correlation between the relative positions of the Psg genes at the locus and their domain arrangements or expression patterns. Figure 6 Physical map of mouse Psg gene locus . A. The order of the Psg genes was inferred from the presence of the various genes on overlapping cosmid, BAC and YAC clones. The position of Psgs represented by filled boxes is unequivocal, whereas the position of those represented by open boxes is ambiguous. Arrows between pairs of genes indicate that their order remains unresolved. The distances between individual genes are not shown to scale. Chimeric YACs mapping to separate chromosomes are indicated by stippled and solid lines. The solid lines correspond to chromosome 7 regions containing the Psg genes indicated above. The locations of the non-chromosome 7 regions are not known. Only the sizes of non-chimeric YACs have been determined and are shown (size bar corresponds to 100 kb). The centromere (cen) / telomere (ter) order and the relative orientation of the two Psg gene subclusters were resolved by FISH mapping. B. Two-colour FISH prophase mapping of relative orientation of the two Psg gene subclusters using mouse m5S cells and C57BL/6CrSlc mouse lymphocytes. (i) FISH pattern representative of 38 experiments where BAC 310D2 in subcluster 1, labelled with rhodamine (R), is centromeric to BAC 600E2 from subcluster 2, labelled with fluorosceine (F). (ii) FISH pattern representative of 38 experiments where BAC 310D2 in subcluster 1, labelled with rhodamine, is centromeric to YAC F10104 from subcluster 2, labelled with fluorosceine. (iii) Orientation of subcluster 2 determined by relative positions of BAC 572D4, labelled with rhodamine, which is telomeric to YAC F10104, labelled with fluorosceine. There is a discrepancy with respect to the distance between the two subclusters. The currently poorly resolved data covering this region in the the Ensembl assembly implies the presence of a gap between Psg29 and Psg32 . However, we determined that the subclusters are fused between Psg32 and Psg30/Psg18 . YAC F10104 (which is non-chimeric) is about 460 kb long and contains only two Psg genes which indicates the presence of a non- Psg genomic region. We estimate the gap between the subclusters to be approximately 400 kb based on the size of cosmids containing two Psg genes (ca. 40 kb). Discussion The human PSG genomic data in the public databases are relatively complete. For each PSG gene, there are annotated RefSeq resources comprising information on genomic structure, transcripts and translation products. The nomenclature is also standardised [ 37 ]. Further, there are accurate chromosome 19 locus assignments allowing complete visualisation of the PSG locus and surrounding genes. In contrast, a substantial quantity of mouse Psg genomic data in the public domain is fragmented, incomplete and somewhat unreliable. We sought to collate the existing genomic data, to present novel data to fill in gaps, and to provide a coherent resource of mouse Psg genomic data. To determine whether the existing set of mouse Psg genes was complete we performed systematic BLAST searches of a variety of public DNA sequence databases. This analysis revealed the existence of a novel expressed Psg gene, which we name Psg31 in line with the accepted nomenclature convention [ 37 ]. Psg31 apparently evolved from a duplication of the whole of the Psg30 gene followed by a subsequent internal duplication of the N1 domain. We were also able to predict the complete coding sequences of four Psg genes for which previously only partial fragments were described. The gene, CDS and protein sequences of these predictions, coupled with a complete reference of all known mouse Psg CDS and primary protein sequences are provided in three attached Additional Files. Using the full CDS information obtained for the complete set of mouse Psg gene sequences, domain structures for all family members were predicted. All of the PSG proteins possess previously described arrangements of Ig-like domains. Except for two members, discussed below, all are predicted to encode N-terminal secretory signal sequences. Our predicted novel splice variant of Psg16 has a complete N1 domain and secretory signal peptide sequence. Trace Archive EST database BLAST analysis confirmed that this variant is expressed in the placenta. In contrast, the brain-expressed variant [ 38 ] has only a partial N1 domain and no secretory signal peptide. The previously described Psg-ps1 pseudogene [ 8 ] was found to be expressed in the placenta using Trace Archive EST database BLAST analysis and possesses an excellent Kozak sequence at the predicted translational initiation site. This evidence therefore indicates that this gene, which we rename Psg32 , is not a pseudogene but a bona fide expressed Psg gene family member. The Psg32 transcript may encode a protein that is retained within the cytoplasm. We note that a precedent in the human exists in the form of a non-secreted splice variant of PSG11 [ 7 ]. Psg31 has the unusual N1-N1-N2-N3-N4-N5-N6-N7-A domain structure. This newly characterised Psg gene has evolved from a duplication of the entire Psg30 gene followed by an internal duplication of the N1 domain. There may be functional significance associated with the N1 domain duplication. The complex nature of Psg gene evolution, including putative gene conversion and recombination events between family members [ 34 ], makes it difficult to analyse their evolution. Despite this, the data generated from ClustalX alignments and NJ trees enabled us to generate trees that allow prediction of the order of events of domain duplications in Psg24 , Psg30 and Psg31 . We note that the apparent route to domain number expansion differed between Psg24 , and Psg30 and Psg31 . The extra N-domains of Psg24 are derived from two duplications of the N2 domain. However, for Psg30 and Psg31 , independent duplications of each of the N2 and N3 domains were probably followed by a secondary duplication of the daughter domains, possibly as a single event. Gene expression data from RT-PCR of placental RNA and EST database analysis revealed considerable differences in the expression levels of different Psg genes. In this analysis Psg21 and Psg23 were the most abundant, consistent with a previous report of abundant Psg23 expression [ 39 ]. Whilst there was generally good agreement between the two methods of expression analysis we cannot determine, based on current data, whether Psg gene expression differences reflect selection for divergent functions, or increased gene dosage for enhancement of an existing function, because expression levels were uniformly low for many family members, and there was a general trend of increased expression during gestation. Psg transcripts are found from day 6.5 of embryonic development onward in primary trophoblast giant cells, later (from day 10.5) in spongiotrophoblast cells and, to a lesser extent, in a cell population in the deciduas basalis at day 14.5 [ 40 ]. At present, it is unclear whether the various Psg genes exhibit different cellular expression patterns, which might indicate divergent functions of the various PSGs. There are interesting parallels between the expansion of the Psg gene family and similar expansions of other placentally-expressed gene families such as the prolactin and growth hormone families [ 1 ], and the aspartic and cysteine proteases [ 41 ]. Such duplications may be a manifestation of parent-offspring conflict or inter-sibling rivalry over maternal investment [ 42 ]. Having collated all known mouse Psg gene protein coding sequences and protein domain structures, the mouse Psg genomic locus on chromosome 7 remained to be determined to complete a comprehensive resource for the analysis of Psg function. The NCBI build 32 composite mouse assembly data revealed that only four Psg s had been mapped. Other Psg s on contig NT_039395 are currently unordered. We therefore screened cosmid, YAC and BAC libraries, and orientated Psg -containing clones to identify, where possible, the order of Psg genes within the locus. We were not able to resolve all ambiguities in gene order on our map; however, where public database information is available, our data are in good agreement. We found no clear relationship between gene location and gene expression level suggesting that, within the Psg locus, each Psg gene is autonomously regulated. Conclusions The evolution and physiological functions of the relatively understudied mouse Psg gene family are poorly understood. This is a feature shared with other placentally-expressed, multigene families such as the prolactin and growth hormone genes [ 1 ]. In order to provide a comprehensive resource to facilitate functional studies of mouse Psg genes, including the generation of mouse mutants with modified Psg gene expression profiles, we have collated the entire set of mouse Psg genes, their predicted encoded proteins, and their evolutionary histories. The complete CDS data will enable the cloning, over-expression, and gene targeting of individual or multiple mouse Psg genes. This will facilitate the elucidation of their function and, by extrapolation, their human homologues, which may be involved in diseases of pregnancy. Methods Isolation of cosmid, YAC and BAC clones Cosmid libraries in pWE15 which were made from liver DNA of BALB/c and C57BL/6 mice were obtained from Dr. Edwin N. Geissler, Boston, MA, USA, and Stratagene (Heidelberg, Germany), respectively. They were screened for the presence of Psg gene-containing cosmids using a 32 P-labelled, full-length Psg cDNA (2.1 kb Kpn I/ Xba I fragment of pCea2b [ 8 ]) as a probe. The final wash was in 4x NaCl/Cit (1x NaCl/Cit is 0.15 M NaCl, 0.015 M sodium citrate pH 7.0), 0.1% SDS at 65°C. Psg gene-containing YAC clones were identified by hybridisation of DNA from YAC clones which were spotted at high density onto nylon membranes [ 43 ] with the same Psg cDNA probe under medium stringency conditions with a final wash in 4x SSPE (1x SSPE is 180 mM NaCl, 10 mM sodium phosphate pH7.4, 1 mM EDTA), 0.1% sodium dodecyl sulfate (SDS) at 65°C. The membranes with arrayed YAC clone DNAs were kindly provided by Dr. H. Lehrach, Max-Planck-Institut für Molekulare Genetik, Berlin. Two libraries were screened, 902 and 903, both in the vector pYAC4 [ 44 ] composed of 9216 clones each, containing spleen DNA of C3H and C57BL/6 mice, respectively. Filters with arrayed BAC clones containing genomic DNA from the embryonic stem cell line CJ7 (129/Sv strain) (CloneRanger™ BAC Human CTC, Invitrogen, Karlsruhe, Germany) were screened by hybridisation with a 32 P-labelled probe consisting of the N1 domain exon sequences of 14 mouse Psg genes ( Psg17 – Psg29 and Psg32 ). The N1 domain exons were amplified individually by PCR using the degenerate primer pair PsgN1-F/PsgN1-R or Psg32N1-F/PsgN1-R (4 mM each) for Psg32 (Table 2 ) and cosmid clones (10 ng) with individual Psg genes as template in the presence of 1 U Taq polymerase and 4 mM MgCl 2 in a total volume of 30 ml (annealing: 50°C, 30 s). N1 exons of Psg28 and Psg29 were released by digestion with Sal I and Kpn I from pUC18 (see below). Southern blot analysis of cosmid and YAC DNAs DNA from YAC clones was isolated by CsCl equilibrium density gradient centrifugation in the presence of ethidium bromide, essentially as described [ 45 ], except that spheroblast formation was achieved by incubation for 90 min with 0.17 mg/ml lyticase (approx. 6,000 U/mg) from Arthrobacter luteus (Sigma, Deisenhofen, Germany). Two mg of cosmid or 0.5 mg of YAC DNA were digested with restriction endonucleases, size fractionated by electrophoresis on 1% agarose gels and blotted onto positively charged nylon membranes. To identify N1 and A domain exon-containing DNA fragments, the digested DNAs on the membranes were hybridised with 32 P-labelled N1 (cosmid DNA blots only) and A domain probes from Psg17 and washed under medium stringency conditions (4x SSPE, 0.1% SDS, 65°C). The Psg17 N1 and A domain cDNA fragments used as probes were obtained by PCR (denaturation: 94°C, 15 s; extension: 72°C, 3 min; 30 cycles) using the mPsg-5'/mPsg-3' (annealing: 50°C; 30s) and Psg17A5'/Psg17A3' (annealing: 60°C, 30 s), primer pairs respectively, and the Psg17 cDNA clone pCea2b as template (Table 2 ; [ 8 ]). Identification of new Psg genes from YAC clones N1 exons from unknown Psg genes were amplified by PCR (annealing: 52°C, 30 s; 30 cycles) in a total volume of 100 μl using 200 ng of YAC clone DNA as template, 1 U Taq polymerase, 3 mM MgCl 2 and 4 mM each of PsgN1-F and PsgN1-R degenerate oligonucleotides (Table 2 ) which bind to the N1, but not N2 and N3 exons of all known mouse Psg genes (except Psg32 ). The product was purified by electrophoresis on a 1.8% agarose gel and subcloned into pUC18 after blunt-ending (SureClone ligation kit: Pharmacia, Freiburg, Germany). The N1 exons from two of the 10 newly identified Psg genes ( Psg28, Psg29 ) were analysed by sequencing recombinant plasmids which did not hybridise with oligonucleotide probes specific for known Psg genes (Table 2 ). Mapping of the Psg locus The presence of the different Psg genes within YAC, BAC and cosmid clones was first determined by PCR followed by hybridisation with oligonucleotides specific for individual Psg genes. DNA from Psg -containing YAC (100 ng) and cosmid clones (10 ng) were used to amplify the N1 domain exons of all known Psg genes in a total volume of 60 μl as described above. The N1 exon of Psg32 was amplified in a separate reaction using Psg32N1-F and PsgN1-R (Table 2 ) as primers under the same conditions used for the amplification of the other N1 exons. For the analysis of BAC clones, PCR was performed directly from the BAC-containing bacterial clones according to the supplier's protocol. Aliquots (3.5 μl) from the various PCR reactions were alkali-denatured, dot-blotted onto nitrocellulose and hybridised with individual 32 P-labelled (final concentration: 0.3–1.2 × 10 6 dpm/ml), gene-specific oligonucleotides (Table 2 ) in 0.5 M sodium phosphate pH 7.2, 7% SDS, 1 mM EDTA over night at 40°C. The filters were washed twice for 20 min each in 2x SSPE at room temperature, followed by two washes in 6x SSPE, 0.1% SDS at a temperature 4°C below the calculated melting temperature of the hybrids [ 46 ]. Oligonucleotides containing at least 3 mismatches in comparison with the corresponding sequences of all known Psg and Cea subgroup members were designed using the computer program Primer [ 47 ]. The only exception is the Psg19 -specific oligonucleotide which exhibits only 2 mismatches to the Psg22 sequence. However, the stringency of the post-hybridisation washes only allowed binding of oligonucleotides with a maximum of one mismatch. The specificity of the oligonucleotides and the hybridisation conditions was demonstrated on cosmid DNAs containing individual Psg genes. The identity of the Psg genes was verified by sequencing. No cross-hybridisation with other Psg genes was observed. The size of the YACs was determined by pulsed field gel electrophoresis followed by Southern blot hybridisation with the Psg17 cDNA clone pCea2b (see above) essentially as described previously [ 48 ]. Fluorescence in situ hybridisation (FISH) analyses The chromosomal location and chimerism of YAC clones were determined by FISH analyses, using B1-PCR of YAC DNA for probe preparation essentially as described [ 49 ]. Orientation and order relative to the chromosome 7 centromere and to each other of the two Psg gene subclusters was defined by FISH analysis using probes described in Fig. 6 . FISH was performed essentially as described [ 50 ] on m5S cells [ 51 ] and concanavalin A-stimulated lymphocytes [ 52 ] from the C57BL/6CrS1c mouse strain. RNA isolation, RT-PCR and specific detection of Psg cDNAs BALB/c mice were mated overnight, and the next day plugged females were designated as day 0.5 of gestation. Pregnant females were killed by cervical dislocation and placentae were dissected free of maternal tissue, immediately frozen in liquid nitrogen and stored at -70°C. Total RNA was extracted by the acid phenol method [ 53 ]. The expression of individual Psg genes was studied by RT-PCR followed by hybridisation of the products with gene-specific oligonucleotides. Total RNA (1 μg) from placentae of different gestational stages was reverse transcribed in a total volume of 10 μl by avian myoblastosis virus (AMV) reverse transcriptase (Promega, Mannheim, Germany) in the presence of 6 U/μl RNasin (Promega) using a degenerate oligo (dT) 30 oligonucleotide (1 μM) as primer (Table 2 ). The reaction mix was adjusted to 1x Taq buffer (20 mM Tris-Cl, 16 mM (NH 4 ) 2 SO 4 , pH 8.6), 3 mM MgCl 2 and 0.4 mM dNTPs in a total volume of 100 μl. Amplification of all known Psg cDNAs (except for the cDNA of Psg32 ( Cea6 ), which at the time of the experiment was presumed to be a pseudogene [ 8 ]) was achieved by PCR (denaturation: 94°C, 15 s; annealing: 58°C, 30 s; extension: 72°C, 3 min; 30 cycles) using Taq polymerase after addition of 400 pmoles of Psg-all (Table 2 ) and 50 pmoles of the oligo (dT) oligonucleotide as 5'- and 3'-primer, respectively. Ten μl aliquots each were size fractionated by electrophoresis on a 1% agarose gel, blotted onto a positively charged nylon membrane (Roche Diagnostics, Mannheim, Germany) and hybridised with individual 32 P-labelled, gene-specific oligonucleotides (Table 2 ) as described above. DNA sequence determination Nucleotide sequences were determined on both strands with flanking universal and internal oligonucleotides as primers using a T7 polymerase sequencing kit (Pharmacia) or a Taq Dye Deoxyterminator cycle sequencing kit (PE Applied Biosystems, Foster City, CA, USA). Assessment of availability of full-length mouse Psg sequences in the public databases All bioinformatics searches described below used the online software and databases available at the NCBI . Where fully annotated, Psg cDNA sequences were identified by name searches of the RefSeq RNA database. Attempts were then made to identify remaining known Psg cDNA sequences via BLAST analyses of the mouse RefSeq RNA database using the GenBank partial-sequences referenced in [ 37 ]. Any PSG cDNA sequences that could still not be identified by this method were determined by BLAST analysis of the mouse genome database using known fragments of the sequence to be determined. This identified genomic contigs that could be interrogated for the 'missing' exonic sequences by pairwise BLAST analysis using the Psg17 cDNA sequence, or fragments thereof, as a probe. A similar procedure was applied to situations where the existence of alternatively spliced exons was suspected to reside within in a Psg gene-containing contig. In the cases where Psg mRNA sequences were built from genomic sequence, or splice variants were predicted, evidence for the existence of such mRNA species in vivo was tested by BLAST analysis using the mouse EST and Trace Archive EST databases. Bioinformatic analysis of the mouse PSG Ig domains The coding sequences of the PSG domains were aligned using Clustal X [ 54 ]. For the production of rooted neighbour-joining (NJ) evolutionary trees, alignments were bootstrapped 1000 times. Evolutionary trees were constructed from the alignments using TreeView . Analysis of the Psg32 exon 1 sequence Four hundred nanograms of DNA obtained from four inbred mouse strains (A/J, C57BL6/6J, YBR/Ei, and SWR/J) (Jackson Laboratory, Bar Harbor, Maine, USA) were amplified using the oligonucleotide primers 5'-AAGGAAGGACAGCAAAT and 5'- AGCTGTGAGCAGAAGAC (denaturation: 94°C, 30 s; annealing: 50°C, 30 s; extension: 72°C, 30 s; 30 cycles) with Pfu DNA polymerase (Stratagene) following the manufacturer's instructions. The 146 bp PCR products were subcloned into PCR-Script (Stratagene). Clones that hybridized to a Psg32 -specific oligonucleotide, Psg32 -exon 1, which binds to a sequence internal to the PCR primers were sequenced. Authors' contributions ASM performed bioinformatics relating to Psg locus organisation, phylogenetic analyses, Psg expression studies, and drafted the manuscript. BF characterized Psg genomic clones and performed Psg expression studies. GD isolated the Psg BAC clones and analysed Psg32 exon 1 sequences in mouse strains. MB and FW contributed to Psg expression studies and YAC fragment assembly. TH and KO performed FISH analysis for identification of orientation of Psg subclusters. TM co-conceived the project and coordinated work performed in Cork. WZ co-conceived the project, coordinated all work performed in Germany, and performed bioinformatics relating to Psg phylogeny and expression. Supplementary Material Additional File 1 The gene sequence and predicted CDS and primary protein sequences for Psg16 (placental transcript), Psg20 , Psg24 , Psg25 , Psg26 , Psg27 , Psg29 and Psg31 A basic text file containing the primary genome data (with source HTGS or WGS information). Predicted CDS sequence is included along with translations. Reverse complement is abbreviated 'RC' in the text. Click here for file Additional File 2 Complete set of mouse Psg CDS sequences A basic text file in FASTA format containing CDS sequences for all known mouse Psgs (not including the brain-specific variant of Psg16 ). Click here for file Additional File 3 Complete set of mouse PSG primary protein sequences A basic text file in FASTA format containing primary protein sequences for all known mouse Psgs (not including the brain-specific variant of Psg16 ). Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546212.xml
520815	Experimental evaluation of the relationship between lethal or non-lethal virulence and transmission success in malaria parasite infections	Background Evolutionary theory suggests that the selection pressure on parasites to maximize their transmission determines their optimal host exploitation strategies and thus their virulence. Establishing the adaptive basis to parasite life history traits has important consequences for predicting parasite responses to public health interventions. In this study we examine the extent to which malaria parasites conform to the predicted adaptive trade-off between transmission and virulence, as defined by mortality. The majority of natural infections, however, result in sub-lethal virulent effects (e.g. anaemia) and are often composed of many strains. Both sub-lethal effects and pathogen population structure have been theoretically shown to have important consequences for virulence evolution. Thus, we additionally examine the relationship between anaemia and transmission in single and mixed clone infections. Results Whereas there was a trade-off between transmission success and virulence as defined by host mortality, contradictory clone-specific patterns occurred when defining virulence by anaemia. A negative relationship between anaemia and transmission success was found for one of the parasite clones, whereas there was no relationship for the other. Notably the two parasite clones also differed in a transmission phenotype (gametocyte sex ratio) that has previously been shown to respond adaptively to a changing blood environment. In addition, as predicted by evolutionary theory, mixed infections resulted in increased anaemia. The increased anaemia was, however, not correlated with any discernable parasite trait (e.g. parasite density) or with increased transmission. Conclusions We found some evidence supporting the hypothesis that there is an adaptive basis correlating virulence (as defined by host mortality) and transmission success in malaria parasites. This confirms the validity of applying evolutionary virulence theory to biomedical research and adds support to the prediction that partially effective vaccines may select for increasingly virulent malaria parasite strains. By contrast, there was no consistent correlation between transmission and sub-lethal anaemia, a more common outcome of malaria infection. However, overall, the data are not inconsistent with the recent proposal that sub-lethal effects may impose an upper limit on virulence. Moreover, clone specific differences in transmission phenotypes linked to anaemia do suggest that there is considerable adaptive potential relating anaemia and transmission that may lead to uncertain consequences following intervention strategies.	Background Over the last few decades there has been considerable effort to introduce an evolutionary perspective to biomedical research and examine the extent to which adaptionist argumentation can be used to address questions pertinent to disease management, particularly that of pathogen virulence [ 1 , 2 ]. Contrary to the conventional wisdom that pathogens should evolve to become benign to their hosts (i.e. prudent host exploitation), considerable theoretical work has shown that the extent of pathogen virulence depends on the trade-off between the exploitation of the host (virulence) and the benefits accrued by increasing R 0 , the reproductive rate, either via increasing transmission directly or by reducing host recovery rate [ 3 , 4 ]. Thus, pathogens are expected to evolve a schedule of host exploitation that maximises their transmission and there are an increasing number of experimental studies demonstrating a relationship between transmission and virulence [ 5 - 8 ]. Incorporation of such virulence theory into an epidemiological framework has led to the suggestion that partially effective vaccines targeting pathogen virulence traits (e.g. malaria parasite growth rate) may actually lead to the selection of increased pathogen virulence [ 9 ]. Empirical confirmation of such important theoretical predictions requires the examination of how selection may be acting on the virulence factors underlying the pathogenicity through variability in transmission success [ 10 ]. Identifying the biological determinants governing the constraint between the transmission/virulence trade-off in turn requires an appreciation of the host-pathogen system in question. This has been recently exemplified by the case of the bacteria Neisseria meningitides , where virulence arises through colonisation of atypical tissues with no apparent direct benefit for transmission [ 11 ]. However, such apparently short-sighted evolution has been suggested to be an inadvertent consequence of within-host evolution that enables pathogen survival and transmission in diverse host spp. [ 12 ]. In these situations, there is an indirect trade-off between virulence and transmission that is strongly influenced by the time lag occurring between the beneficial and deleterious outcomes of within-host evolution [ 13 ]. Theoretical advances in virulence theory have revealed the importance of additional properties of host-pathogen systems, such as pathogen population structure and sub-lethal (significantly deleterious for host fitness but not lethal) effects, which can significantly alter the conclusions arising from the simple virulence/transmission trade-off hypothesis [ 14 , 15 ]. Infections are often composed of many simultaneously co-infecting strains that may interact directly or indirectly through their shared host. When parasites share a host, competition for the limited resources is generally expected to favour the most aggressive parasite – the tragedy of the commons [ 16 ]. Many theoretical models incorporating selection into an epidemiological framework show that selection favours increasingly virulent genotypes as the number of co-infecting parasite genotypes increases [ 14 , 17 - 20 ]. The expected levels of virulence, however, depend critically on how parasites interact with their hosts and how co-infecting genotypes affect each other; specifically, optimal levels of virulence depend on the cooperative or competitive nature of host exploitation by co-infecting parasites [ 15 , 21 - 23 ]. Hence, in what way such co-infection and sub-lethal effects potentially influence virulence evolution needs to be addressed explicitly for each pathogen system. In this paper we address the virulence/transmission relationship for the malaria parasite system and thereby examine the premises underlying the theoretical predictions on vaccine efficacy [ 9 ]. Malaria parasites replicate asexually within their vertebrate hosts, but must produce sexual stages (male and female gametocytes) to be transmitted to mosquitoes. Asexual parasites provide the source for gametocyte production and so increasing asexual parasite density is expected to correlate with transmission success. Therefore, the adaptive trade-off hypothesis would predict that faster asexual replication leads both to higher virulence and to higher transmission success [ 24 ]. Although transmission success does broadly increase with gametocyte density [ 25 , 26 ], which largely reflects asexual parasite density, many factors can affect transmission success [ 27 ]. I.e. high density does not always guarantee infection success [ 25 , 28 ]. Malaria parasites demonstrate considerable phenotypic plasticity in the production of sexual transmission stages. This plasticity appears to be linked to the quality of the blood environment: as conditions for asexual replication worsen, the parasite alters its sex allocation strategy by changing the sex ratio of gametocytes [ 29 ] or the proportion of asexual parasites producing gametocytes [ 30 - 32 ]. Thus transmission traits (e.g. gametocyte density and sex ratio) are highly susceptible to environmental changes and may therefore not conform to the simple trade-off between transmission and mortality. Although infection with malaria parasites can lead to host death, non-lethal anaemia is more commonly the result, where the parasitic infection provokes a reduction in haematocrit (red blood cell density) [ 33 ] and thus a reduced quality resource for parasite replication. Thus, anaemia is both a physical manifestation of virulence and a trigger of sex allocation. In this study we examine the trade-offs between transmission and virulence defined by mortality and anaemia in single and mixed clone infections, using the avian malaria system P. gallinaceum in the chicken host. In this specific system, anaemia and parasite transmission are linked physiologically, where the host's response to anaemia (reticulocyte production) triggers an increase in the proportion of gametocytes that are male rather than female [ 29 ]. This has been proposed to be an adaptive parasite response to assure mating success in the face of an immune response (transmission-blocking immunity) that develops simultaneously to the anaemia [ 34 - 37 ]. Experimental abnegation of this immune response can be achieved by injection of a high density parasite inoculum, which results in an uncontrolled infection, host death, and importantly, an absence of both the host reticulocyte response to anaemia and the linked parasite sex allocation response [ 37 ]. Thus, we are able to generate infections known a priori to result in host death or control by the host, enabling independent study of parasite transmission to mosquitoes with respect to either mortality or anaemia. Under both infection conditions, we examine the impact of mixed clone infections on overall virulence and transmission. Thus we assess to what extent there is an adaptive basis relating virulence and transmission under infection conditions of theoretical significance. Results Course of infection – competition between clones We first considered infection by injection of 10 6 infected red blood cells (iRBC) (an inoculum that almost always leads to a lethal infection outcome) of either of two clones of P. gallinaceum (the Thai or SL clone) or an equal mix of the two. The peak densities of asexual parasites did not differ between Thai and SL clone infections (χ 2 1 = 0.3 P = 0.629), but were both significantly higher than those in mixed clone infections (combined single clones vs. mixed clone χ 2 1 = 6.3 P = 0.012). The peak gametocyte density was, however, significantly higher in Thai than in SL clone infections (χ 2 1 = 4.3 P = 0.038), and higher in both the single clone groups than in the mixed infections (χ 2 1 = 4.05 P = 0.044) (Fig. 1 ). The extent of anaemia reflected parasite density, but was greater in the mixed clone group than in either of the two single clones, despite its lower parasite density (parasite density χ 2 1 = 14.1, P < 0.001; Group χ 2 2 = 6.1, P = 0.048). This effect of co-infection was confirmed with a comparison of the mixed clone with the 2 single clone type infections grouped together (parasite density χ 2 1 = 14.2, P < 0.001; combined single clones vs. mixed clone: χ 2 1 = 5.8, P = 0.016). The gametocyte sex ratios (proportion of gametocytes that were male) were found to be different among groups (χ 2 2 = 33.5, P < 0.001), where the Thai clone had a higher sex ratio than the SL clone infections. Closer inspection and analysis of the data revealed similar sex ratios in the SL clone and mixed clone infections (Fig. 2 ) (Haematocrit χ 2 1 = 16.8, P < 0.001; Group χ 2 1 = 25.1, P < 0.001). As previously observed after infections by injection with high inocula [ 37 ], the production of reticulocytes is almost always absent during the growth phase of the infection that is followed either immediately by death or "crisis" when the host destroys all red blood cells (and parasites) and floods its system with reticulocytes. Notably, here one individual displayed a strong reticulocyte response to anaemia during the growth phase of infection, as occurs in natural, low density infections (See Methods); this was predictably associated with a strong increase in sex ratio (see Fig. 2 circled cross) [ 29 ]. Figure 1 P. gallinaceum gametocyte (sexual) parasite density (mean ± S.E.) following intra-muscular injection of 10 6 parasites single clones or 5 × 10 5 of each clone in mixed infections. Thai clone (square, long dotted line); SL clone (open circles, solid line); mixed clones (cross, dotted line). Figure 2 Gametocyte sex ratios (proportion male) of mixed vs . single clone infections compared with the corresponding haematocrit ( r ed b lood c ell density) measured on the same day of infection. Each individual infection was measured daily for both these parameters until either the host died or the infection was cleared by the host (thus until the peak of infection). Infections were initiated by injection of 10 6 parasites for single clone treatments and 5 × 10 5 of each clone in mixed infections. Thai clone (squares); SL clone (open circles); mixed clones (crosses). Circled cross represents one individual that had a high sex ratio associated with a high density of reticulocytes on one particular day (see text). These data strongly suggest that the clones had a negative impact on each other, thus resulting in lower densities of asexual parasites and hence gametocytes. The anaemia (and hence sub-lethal virulence) was greater in the mixed clone infections, as previously observed in mouse models of malaria [ 38 ] and natural infections in humans [ 39 ], but was not accompanied by an increase in parasite densities. Thus, such a virulence outcome likely reflects a more intense host response to infection rather than an adaptive response relating pathogenicity and transmission [ 10 ]. Secondly we considered the outcome of infections initiated by infected mosquito bites (i.e. as with natural infections). These sporozoite-induced infections were initiated by gorging a fixed number of infected mosquitoes on the individual birds: 8 mosquitoes per bird for the single clone infections and either 8+8 or 4+4 for the mixed clone infections. Thus we controlled for total or clone-specific dose in the mixed infections. The experiment was performed twice, with a different batch of infected mosquitoes used for each experiment. The resulting infections were very consistent and no differences were found between the two experimental replicates for any of the analyses. Parasite densities of the Thai clone were significantly higher (χ 2 1 = 4.73, P = 0.029) than the SL and mixed clone infections (4+4 or 8+8), which did not differ from one another. Gametocyte densities followed the patterns of asexual parasite densities, except for the high mixed clone infection (8+8), whose peak gametocyte production paralleled that of the Thai clone (Fig. 3 ). Again, the extent of anaemia reflected parasite densities and treatment group (Interaction between parasite density and group χ 2 3 = 7.9, P = 0.048). Data inspection and model simplification regrouping the treatments into single vs . mixed clone infections revealed that mixed clone infections resulted in greater anaemia for a given parasite density (Parasite density χ 2 1 = 52.8, P < 0.001; mixed vs . single clone infections χ 2 1 = 6.8, P = 0.009). This increased anaemia was particularly evident for the 4+4 mixed infection group and was reflected in the significantly stronger reticulocyte response in this group (mixed vs . single clone infections χ 2 1 = 6.3, P = 0.012). Once again, the gametocyte sex ratio varied with group and anaemia independently (group: χ 2 3 = 11.7, P = 0.008; haematocrit: χ 2 1 = 16.2, P < 0.001) (Fig. 4 ). Data inspection and model simplification showed that Thai /8+8 type infections had a significantly higher sex ratio (χ 2 1 = 10.7, P = 0.001) than SL /4+4 types and both types responded similarly to haematocrit (χ 2 1 = 17.7, P < 0.001) (Fig. 4 ). Figure 3 P. gallinaceum gametocyte (sexual) parasite density (mean ± S.E.) following infection by either or both parasite clones using a fixed number of mosquitoes. Thai or SL clone alone, 8 infectious mosquitoes for either clone per bird; mixed infections were initiated using 8 infectious mosquitoes of each clone per bird (8+8) or 4 infectious mosquitoes of each clone per bird (4+4). 3 birds were thus infected for each treatment type and the experiment was performed twice using different batches of mosquitoes infectious for each clone. Thai clone (squares, thin line); SL clone (open circles, thick line); 8+8 mixed clones (triangles, long dotted line); 4+4 mixed clones (crosses, short dotted line). Figure 4 Gametocyte sex ratios (proportion male) of mixed vs . single clone infections compared with their corresponding haematocrits ( r ed b lood c ell density) measured on the same day of infection. Each individual infection was measured daily for both these parameters until either the host died or the infection was cleared by the host (thus until the peak of infection). Thai or SL clone alone, 8 infective mosquitoes per bird; Mixed infections were initiated using 8 or 4 infective mosquitoes from batches of mosquitoes infected with either of the clones. Thai clone (squares); SL clone (open circles); 8+8 mixed clones (triangles); 4+4 mixed clones (crosses). Thus, again the data support the notion that there is competition between the clones. The greater anaemia in mixed infections was again observed. In addition, the level of anaemia and the reticulocyte response to infection were greatest for the lowest inoculation dose. The extent of anaemia did not, however, seem to affect parasite densities in the mixed infections, where the parasitological parameters measured (e.g. asexual parasite / gametocyte densities and the sex ratio) aligned with those of either one of the 2 clones. Although the clones do appear to compete aggressively, sub-lethal virulence was more strongly influenced by infecting dose. Thus the sub-lethal measure of virulence (i.e. anaemia) appears to be more strongly dependent on the host's response to infection than on parasite density per se . Virulence and transmission success to mosquitoes For the iRBC-induced infections, one set of mosquito transmission studies was carried out on individuals with comparable parasite and gametocyte densities 3 days after inoculation, when parasite densities were low. Transmission success to mosquitoes at this early infection time point, as measured by both geometric mean oocyst load and the percentage of mosquitoes that were oocyst positive, was proportional to the subsequent rate of death (oocyst load: χ 2 1 = 5.6, P = 0.017; percentage infected: χ 2 1 = 4.6, P = 0.032) (Figs. 5a & 5b ). Infection type ( Thai vs. SL vs. mixed clone) was not found to correlate with transmission success (oocyst load: χ 2 2 = 2.3, P = 0.339; percentage infected: χ 2 2 = 3.34, P = 0.17). In addition, neither sex ratio (oocyst load: χ 2 1 = 2.5, P = 0.114; percentage infected: χ 2 1 = 2.1, P = 0.15) nor gametocyte density (oocyst load: χ 2 1 = 0.1, P = 0.75; percentage infected: χ 2 1 = 0.3, P = 0.62) was correlated with transmission success. This is evidence that virulence, as defined by mortality, is positively correlated with transmission success. Figure 5 Transmission success as measured by ( A ) the geometric mean oocyst load per mosquito and ( B ) the percentage of mosquitoes that were positive for at least one oocyst, related to day of death for the 2 single clone and the mixed clone type infections. Three birds in each group (injected with single or mixed parasite clones as in Figs. 1 & 2) were gorged on mosquitoes on the first day of patent infection. Thai clone (squares); SL clone (open circles); mixed clones (crosses). Line represents least squared residual best fit for the relationship between oocyst load and day of death. ( A ) R 2 = 0.68, P = 0.017; ( B ) R 2 = 0.41, P = 0.032. For the sporozoite-induced infections (Figs. 6a & 6b ), we first carried out the same analysis as above, comparing transmission success to mosquitoes at the earliest infection time point with the rate of death, using only those 8 individuals that died. The percentage of mosquitoes infected was proportional to the rate of death (χ 2 1 = 4.54, P = 0.032), however the oocyst load was not (χ 2 1 = 3.36, P = 0.067). By contrast, infection type was not found to correlate with the percentage infected (χ 2 3 = 5.56, P = 0.135), but did with the oocyst load (χ 2 3 = 9.71, P = 0.021); this latter result is most likely the consequence of the small sample size (n = 8), three individuals of which were infected with the Thai clone, died on days 4 and 5 and produced very similar oocyst loads in the mosquitoes. The significant correlation between rate of death and the percentage of mosquitoes infected in both iRBC- and sporozoite-induced infections, despite their small sample sizes, provides consistent evidence that virulence, as defined by mortality, is positively correlated with transmission success. Figure 6 Comparison of lifetime transmission success with day of host death. Lifetime transmission success is defined as ( A ) the mean number of oocysts per mosquito for a given day totalled over the 4 days of acute phase transmission; ( B ) the percentage of mosquitoes with at least one oocyst. Infections were initiated using infectious mosquitoes as in Figs. 3 & 4. Thai or SL clone alone, 8 infective mosquitoes per bird; Mixed infections were initiated using 8 or 4 infective mosquitoes from batches of mosquitoes infected with either of the clones. Mosquitoes were gorged on all birds from the first day parasites were patent in the blood, (a) Thai clone (square); SL clone (open circles); mixed 8+8 clones (triangle); mixed 4+4 clones (crosses). For the sporozoite-induced infections, we then examined the relationship between mortality and parasite lifetime transmission success, as defined by total mean geometric oocyst load and total percentage of mosquitoes infected. For these analyses, individuals were first grouped according to whether they lived or died, and secondly, if they died, whether it was before or after the 4-day acute transmission period. There was no significant difference in either the percentage of mosquitoes infected (χ 2 1 = 1.91, P = 0.16) or the total oocyst load (χ 2 1 = 0.2, P = 0.63) between those individuals that died at any time and those that lived. However, those individuals that survived for the 4-day acute transmission period, irrespective of whether they subsequently died, infected a higher percentage of mosquitoes (percentage infected: χ 2 1 = 12.9, P < 0.001) and with a higher mean oocyst density mosquitoes (χ 2 1 = 18.2, P < 0.001) than those 3 individuals that died rapidly. Excluding these 3 early deaths, we compared the transmission success of those that lived and died following the acute transmission period. There was no significant effect of host mortality on either the percentage of mosquitoes infected (χ 2 1 = 2.37, P = 0.12) or on mean oocyst loads (χ 2 1 = 1.07, P = 0.33). These results suggest that although parasite-induced mortality may incur a transmission cost by reducing the duration of transmission, there is no additional effect of mortality on transmission success. This absence of a positive relationship between virulence and transmission success contradicts the results in the previous paragraphs where the rate of mortality was correlated with a single day measure of transmission success. This may, however, be the consequence of the small sample size (number of individuals that died after completing the acute transmission period, n = 5) and complexity of infection types. Lifetime transmission success was then compared against the level of anaemia at infection peak (Figs. 7a & 7b ). Because mixed clone infections led to increased anaemia independent of parasite density, we first examined the transmission/anaemia relationship in the single clone infections. There was a strong interaction between clone type and haematocrit on lifetime transmission success for both the percentage of mosquitoes infected (χ 2 1 = 12.2, P = 0.0005) and the total mean oocyst load (χ 2 1 = 10.21, P = 0.0015). In the Thai clone infections, anaemia at infection peak was negatively correlated to total mean oocyst load (χ 2 1 = 7.28, P = 0.007) and the percentage of mosquitoes that were infected (χ 2 1 = 8.4, P = 0.004); in the SL clone infections, however, there was no correlation between anaemia and either oocyst load (χ 2 1 = 0.85, P = 0.36) or the percentage of infected mosquitoes (χ 2 1 = 0.56, P = 0.46). When examining the data for all infection types, although the 8+8 mixed clone infections aligned with the Thai clone infections when considering the oocyst load, this was lost when considering the percentage of mosquitoes infected (Figs. 7a & 7b ). Why there is this disparity in mosquito infection parameters for this mixed infection group is unclear, but highlights the importance of considering mixed infections and of careful choice of the measure of transmission success used. Further work on transmission success from mixed infections is clearly required. In conclusion, the data suggest that there are clone-specific effects relating sub-lethal virulence to transmission and importantly that in the case of the Thai clone, that excessive anaemia may be a fitness cost for virulence evolution. Figure 7 Comparison of lifetime transmission success with peak anaemia. Lifetime transmission success is defined as ( A ) the mean number of oocysts per mosquito for a given day totalled over the 4 days of acute phase transmission; ( B ) the percentage of mosquitoes with at least one oocyst. Infections were initiated using infectious mosquitoes as in Figs. 3, 4 & 6. Thai or SL clone alone, 8 infective mosquitoes per bird; Mixed infections were initiated using 8 or 4 infective mosquitoes from batches of mosquitoes infected with either of the clones. Mosquitoes were gorged on all birds from the first day parasites were patent in the blood. Individuals that died before completion of the transmission period are excluded (a) Thai clone (square – thin solid line); SL clone (open circles – thick solid line); mixed 8+8 clones (triangle – long dotted line); mixed 4+4 clones (crosses – short dotted line). Lines are least squared residual best fit. Discussion This study examined the validity of some basic assumptions underlying evolutionary models of virulence for the specific case of malaria parasites, and hence the extent to which such models may be relevant to malaria control. One of the fundamental assumptions is that a parasite's virulence is a necessary consequence of the strategy that it uses to exploit its host in its (evolutionary) attempt to achieve maximal transmission. For the case of malaria, the production of transmission stages (gametocytes) depends on the asexual replication of the parasite clone within the host's blood system. Thus, faster asexual replication would be expected to result in more transmission before the host's immune responses control the infection, but also to cause greater host damage and to increase the risk of the host's death. Hitherto, experimental studies have either used gametocyte density as a proxy for transmission success [ 24 ], or taken a snap-shot count of successfully transmitted parasite stages at a time of estimated maximal transmission [ 7 , 40 , 41 ]. Here we measured transmission success directly by gorging mosquitoes on the infected hosts and counted the number of successfully transmitted parasites either at a time independent of the virulence outcome or throughout the period during which the parasite transmits from the vertebrate host to the mosquito vector. We found (i) that in both iRBC and sporozoite-induced infections, transmission success was positively correlated with host mortality rate but (ii) that there was a total transmission cost to host death only if the host died before the completion of the transmission period; if the host died after completing the acute transmission period there was no impact on total transmission. Finally (iii) there was no clear correlation between transmission success and anaemia (sub-lethal virulence): infections with one of the parasite clones resulted in a negative virulence-transmission relationship whereas with the other clone there was no relationship. These observations suggest that although malaria parasites may conform to the predictions of virulence theory based on parasite-induced host mortality, there appears to be no general trade-off between transmission and sub-lethal virulence. Notably, measurable parasitological parameters such as asexual and sexual parasite densities were predictive of neither mortality rate nor transmission success. This is because the relationships between asexual density and host mortality and between transmission success and the densities of asexual stages and gametocytes are not straightforward and are influenced considerably by developing immune responses. Indeed, despite the appealing simplicity of the idea, it is not obvious that there should be simple measurable parameters, such as asexual parasite density, defining parasite virulence or that these should correlate in a straightforward manner with transmission; this has been discussed in depth for the case Neisseria , as noted in the Background section [ 11 , 12 ]. Irrespective, however, of the precise biological details underlying parasite-induced mortality, malaria parasites do generally confirm one of the most robust predictions of virulence theory. Therefore, natural parasite populations might confirm the prediction that, in the simplest case of single clone infections in a homogeneous host population, intervention strategies targeting asexual replication rate could select for increased virulence [ 9 ]. Previous work in experimental mouse models found there to be positive genetic correlations between virulence, asexual density and transmission [ 24 ]. However, sub-lethal effects, rather than lethal infection outcome, were found to impose an upper threshold to virulence [ 7 ] and transmission was maximised at intermediate levels of host morbidity [ 42 ]. In our system, host death did reduce total transmission success, but only when the mortality rate was very high; mortality per se incurred no cost to transmission. In addition, more virulent infections were generally more infectious prior to host death. These results conform to standard virulence theory where there is a relationship between transmission and virulence, such that intermediate levels of virulence evolve. By contrast, the relationship between transmission success and sub-lethal virulence (anaemia) was less clear. On the one hand, infections with one of the parasite clones ( Thai ) confirm the suggestion from the mouse models [ 42 ] that sub-lethal effects impose selection against virulence. On the other hand, there was no transmission/virulence relationship for the other clone. How to interpret such results? In the mouse models [ 42 ], sub-lethal virulence was measured as a composite parameter including weight loss and anaemia; here we only used anaemia and therefore may have missed additional negative effects of the parasite on the host that might alter our conclusions. Indeed, for a given parasite dose, the level of anaemia induced was the same for the two clones. Thus, there was apparently no between-clone variance in putative parasite traits provoking anaemia, but there was clone-specific variation relating transmission to sub-lethal virulence. Additional clone-specific virulent effects on the host (e.g. weight loss) may therefore be as important as anaemia. An additional consideration is the proximate effect of anaemia on transmission. Previously, clone SL has been shown to adaptively alter its gametocyte sex ratio according to the host response to anaemia [ 29 ]. Notably the gametocyte sex ratio of the Thai clone was different from that of the SL clone, though its response to anaemia appeared similar (sex ratio became less female-biased). If the two clones differ in a transmission trait (sex ratio) that is apparently sensitive to host anaemia, they may be expected to vary in other, less evident traits implicated in transmission. Such variation in transmission traits would result in between-clone variability in the anaemia/transmission relationship. Further exploration of the covariance in sex ratio and the anaemia/transmission relationship is clearly required to ascertain whether sex ratio can be used as a marker phenotype for the complex interactions between the parasite and the host that determine parasite transmission success. In conclusion, the data are not inconsistent with the proposal that sub-lethal effects may impose an upper limit on virulence [ 42 ], but a more detailed understanding of parasite traits implicated in transmission is required. The final consideration concerns how the complexity of infection may alter the virulence/transmission trade-off. Competition between clones in multiple infections would generally be expected to favour the more virulent clone [ 20 ], although sub-lethal effects can considerably reduce the optimum level of virulence [ 15 ]. Malaria parasites are confronted with an unpredictable number of co-infecting clones. Under such conditions, parasites may employ facultative strategies of host exploitation according to co-infecting clone number and virulence would be expected to increase in mixed vs . single clone infections [ 38 ]. Anaemia was increased in all mixed infections, as previously observed in mouse models [ 38 ]. If parasites have evolved transmission strategies correlating with anaemia, such increased anaemia would be expected to result in altered transmission success. Overall transmission success was, however, unaltered and consistent with that observed for the single clone infections. However, the degree of increased anaemia in the mixed clone infections did not seem sufficient to alter the gametocyte sex ratios, previously shown to respond adaptively to changes in anaemia [ 29 ]. Thus, the sub-lethal virulence effect of the mixed infection does not seem to be sufficient to reduce the efficacy of the parasite's transmission strategy, of which facultative shifts in sex ratio is but one manifestation. In addition, although molecular data were not available, transmission phenotypes (e.g. gametocyte density and sex ratio) of mixed clone infections seemed to align themselves with either of the two clones: i.e. there seemed to be competitive dominance by one clone of the other, although the genetic identity of the clones abnegated any objective measurement. Interestingly, the competitive outcome was seemingly resolved early on in the infection. This suggests that there are clone-specific biological features determining the outcome at the very early stages of infection. The importance of the early stages of infection in the host-parasite interaction was further highlighted by the positive correlation between transmission success at the start of infection and host mortality rate, which occurred many days later. An additional surprising outcome of the mixed infections was that the identity of the dominant clone depended upon infecting dose. Infecting dose has previously been shown to influence disease severity in mouse models, but importantly there were consistent clone-specific differences in disease severity across a wide-range of doses [ 43 ]. Here, it is notable that at high infection doses (iRBC-induced infections) there were no differences between peak parasite densities of the two clones whereas at low infection doses (sporozoite-induced), the Thai clone reached a higher peak density. Thus we found dose-dependent differences that may not only alter the potential competitive outcome of a mixed infection, but may also alter the relative virulent nature of the 2 clones: i.e. relative differences in maximum parasite densities of clones may be more influenced by infecting dose than by clone identity. Our data from two clones are clearly limited, but if infecting dose is important, there may be additional subtle consequences of local heterogeneity in transmission intensity (number of infectious mosquito bites per host per unit time; i.e. dose) in addition to increasing R 0 [ 44 , 45 ] and genetic complexity [ 46 ]. Conclusions Research using animal models is generating increasing evidence supporting an adaptive basis to life history traits and virulence of malaria parasites [ 7 , 24 , 42 ]. The studies presented here generally confirm these previous findings from a mammalian model system. Where they are different, probably reflects fundamental life history strategies pertaining to blood use (e.g. type of red blood cell invaded) [ 47 ], which include strategies for transmission stage production [ 30 , 34 , 35 ]. Such life history traits are well-documented (reviewed in [ 47 ]) and may be simple phenotypic markers reflecting the suite of alternative strategies available to the parasite. While the relevance of laboratory models to natural systems is easily criticised, the consistency of infection patterns across both avian and mammalian models suggests that there are general features of Plasmodium- host interactions. Therefore, with careful consideration of the biological details of each system, extrapolation from laboratory model to field data may be justified. Where field data fail to fit predictions, it is likely that we are missing important biological processes underlying the system in question [ 48 ]. One major deficiency of models is the inability to simulate repeated infections. With rare exceptions [ 49 , 50 ], the emphasis is placed on primary infections. The majority of infected humans, however, are not presenting with their first infections and will have confronted parasites for a considerable length of time. The host immune response to primary vs . re-infections is markedly different and it is likely that the host haematological response alters as well. The proportion of individuals confronting parasites for the first time (i.e. primary infections) depends on the transmission intensity, which also determines the parasite population structure and the parasite dose. Thus, interpreting field patterns with respect to evolutionary predictions necessitates stratification according to the epidemiology of the populations in question. However, despite such complexity, that both laboratory and field populations of malaria parasites generally conform to evolutionary theories concerning life history traits (e.g. sex allocation) and virulence is encouraging, and suggests that evolutionary theory can play an important role in predicting consequences of public health intervention strategies [ 9 ]. It remains, however, to be seen whether less conventional, but probably more important, measures of parasite virulence such as anaemia can be considered within an evolutionary framework. Methods Parasite, host and mosquito species Two strains of the chicken malaria parasite P. gallinaceum were used: Strain 8A (originally from Sri Lanka and obtained from D. Kaslow, NIH, Bethesda, USA) and a new Thai strain (obtained from S. Nithiuthai, Chulalongkorn University, Bangkok, Thailand). In this paper, 8A strain is designated as SL and the Thai strain as Thai . Both strains were cloned with limiting dilution in 5-day old chick hosts ( Gallus gallus domesticus ) (INRA, France). The clones were maintained in vivo by inoculation of 1 ml infected blood (20–45% parasitaemia, percent red blood cells (rbcs) infected) into naïve chicken hosts, with frequent passage through the mosquito vector, Aedes aegypti (Liverpool Blackeye strain). For experimentation, different, healthy 3-week old White Leghorn chickens were used for each experiment. All experimental animals were maintained according to European Union guidelines. Parasitaemias were obtained with Giemsa staining of daily blood smears. Parasitaemias and reticulocytes were calculated as percentages observed in a minimum of 100 rbcs; gametocytaemias observed in 10,000 rbcs. Haematocrit (number of rbcs per unit volume) was measured daily for each host using a haemocytometer. Thus, parasite and gametocyte densities are the number of rbcs infected with any parasites or gametocytes per unit volume. Mature male and female gametocytes are distinguishable after Giemsa staining: males stain a pale rose with no distinct nucleus, females stain blue with a distinct red nucleus [ 51 , 52 ]. Sex ratios based on counts of 50–75 gametocytes were found to be representative. We calculated sex ratios from the lesser of 50,000 rbcs or 100 gametocytes. Sex ratios are given as the proportion of males. Ae. aegypti mosquitoes were used in all transmission studies. Mosquitoes were maintained under standard conditions (80% humidity and 26°C). Transmission success was measured by (i) the percentage of mosquitoes positive for oocyst stage parasites and (ii) the mean oocyst density in gorged mosquitoes. Oocysts are the zygote stages of the parasite found on the mosquito stomach wall and which are those parasites that have developed successfully from fertilised female parasite gametes (gametocytes in the vertebrate host give rise to gametes once inside the mosquito bloodmeal which then undergo fertilisation; only fertilised gametes can continue development). Oocyst stage parasite counts were made 7 days post mosquito infection on midguts dissected from 30 gravid females and then stained with 0.5% mercurochrome in 1× Phosphate Buffer solution. Mean oocyst number per mosquito was chosen as an additional measure of transmission success to the number of infected mosquitoes for several reasons. Firstly, oocyst number is a more sensitive measure and is related to the percentage of infected mosquitoes by a simple negative binomial relationship [ 53 ]. Secondly, subsequent transmission from an infected mosquito to a new host may be affected by overall oocyst number through altering the number of sporozoites injected during a bloodmeal (See [ 47 ] for relevant literature). Experimental design a) Infection was initiated by injection of parasitized blood to generate an uncontrolled infection (in this paper, these infections are denoted iRBC-induced). Chickens were inoculated by intra-muscular injection of 10 6 parasites: 6 were inoculated with the SL clone, 6 with the Thai clone and 6 with 5 × 10 5 of each clone. Ae. aegypti mosquitoes were gorged upon 3 infected chickens from each group, chosen for their matching parasite densities on the day P. gallinaceum was detectable by blood smear (>0.1% parasitaemia). Transmission success (mean number of oocysts per mosquito) on day 3 post inoculation, when parasites were visible in the blood smear (>0.1%), was subsequently related to the day at which that individual died. b) Infection was performed using infectious mosquitoes to mimic natural low intensity infections (sporozoite-induced infections). Experimental infections were induced by gorging infected mosquitoes on the chicken hosts. 4 experimental groups (3 chickens per group) were considered: (i) 8 infective bites per host using the SL clone (ii) 8 infective bites per host using the Thai clone (iii) 4 infective bites from the Thai clone and 4 infective bites from the SL clone per host (iv) 8 infective bites from the Thai clone and 8 infective bites from the SL clone per host. The presence of sporozoites in each of these mosquitoes was verified subsequently by dissection of the salivary glands. (Note: the clones do not differ in the gene sequences of the molecular markers available for this species in Genbank (NCBI), so that the clones could not be differentiated and identified, unpubl. data). This experiment was repeated using different batches of mosquitoes infected with either parasite clone and healthy, uninfected chickens. Within each experiment for each parasite clone, each individual was infected by mosquitoes coming from the same cage. Ae. aegypti mosquitoes were gorged upon all infected chickens on the 1st day P. gallinaceum became patent in the blood (<0.1% parasitaemia), predictably day 7 (following the 7 day pre-erythrocytic developmental period), and for the next 3 days at which point the infection reaches its zenith; this is the acute stage of infection. Post-peak infection the parasite is rapidly cleared by the chicken and resurges only intermittently at very low densities. This chronic stage of infection represents the infectious reservoir yielding irregular low levels of transmission to mosquitoes [ 54 ]. The initial acute phase results in very high transmission rates and is here taken to represent the parasite lifetime transmission success (total of the daily mean number of oocysts per mosquito) and is compared against the highest level of anaemia (lowest haematocrit) observed in each individual chicken. Statistical Analyses Statistical analyses were conducted using the statistical package Genstat 6.1. Because each individual chicken was included in the data set many times, we corrected for repeated measures by fitting a generalised linear mixed model (GLMM procedure), nesting day within individual chicken in the random model. Parasite densities and blood cell counts were analysed specifying a Poisson error structure. Transmission success: the percentage of mosquitoes that were infected was analyzed using logistic regression specifying a binomial error structure with a logit link function and the mean oocyst densities were analyzed with a logistic regression specifying a Poisson error structure with a logarithmic link function (which gives the same fit as a negative binomial, [ 55 ]). For the iRBC-induced infections, transmission success was analyzed with respect to day of death, sex ratio and gametocyte density. For sporozoite-induced infections, transmission success on the first day of infection was similarly analyzed with respect to day of death for those individuals that died. In addition, we examined whether host death incurred a transmission cost by comparing transmission success of (i) the bird hosts died or not and (ii) whether birds survived the acute transmission period or not. Between-group comparisons of parasite lifetime transmission success for each individual were also analyzed with respect to the level of anaemia at infection peak. For the sex ratio and transmission analyses the data were over-dispersed and so were corrected for by estimating a dispersion parameter for each analysis. Statistical significance was presented as Wald statistics, which have a χ 2 distribution. When sex ratio was used as an explanatory variable it was arcsine-transformed. Author's contributions RELP carried out the experimental studies, TL assisted in iRBC-induced studies, CMG assisted in experimental design, NS isolated the Thai strain parasite, PTB assisted in experimental design and edited the manuscript and JCK in elaboration of the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC520815.xml
529428	The Intangible Magic of Celebrity Marketing	Drug industry insiders share their tips on using celebrities to market drugs and diseases	As the Hispanic world well knows, the word in Spanish for advertising is ‘propaganda’, its meaning derived literally from the propagation of the faith, the antithesis of science's Enlightenment ideals. The old word somehow seems perfect for describing the new world of drug promotion and its growing use of the famous face. Like the catholic cardinals of the 17th century, many of the feted celebrities of the 21st are now engaged in spreading the word. Now, as then, the religion promises miraculous breakthroughs, wonder cures, and sometimes even eternal life. The difference is that this time around, the stars are earning fat fees from the marketing departments of giant pharmaceutical companies. And if the latest revelations from industry insiders are anything to go by, their hefty investments in celebrity selling are well worth it. Celebrity Selling The epicentre of this phenomenon is of course the United States, where companies routinely hire celebrities to attract attention to the latest drugs and the diseases that go with them. Pfizer famously paid presidential hopeful Bob Dole to promote awareness of erectile dysfunction as sildenafil (Viagra) was hitting the market. Wyeth hired supermodel Lauren Hutton to hawk hormone replacement therapy and menopause. GSK contracted football star Ricky Williams to sell social anxiety disorder, helping make paroxetine (Paxil)—briefly—the world's top-selling antidepressant. Even the dead are raising awareness, with the estate of Errol Flynn now enlisted to help promote cardiovascular disease as a household name [ 1 ]. The celebrity, living or dead, becomes integral to a drug marketing strategy that includes paid advertising and aggressive public relations campaigns that can produce media appearances on the likes of Oprah and The Today Show . According to celebrity brokers, the star's remuneration package, though always confidential, can range from $20,000 to $2 million. Celebrities often hide their conflicts of interest (Illustration: Margaret Shear) ‘A partnership between a celebrity and a brand has an intangible sort of magic’, writes a senior marketing executive at Amgen, in an extremely candid piece published recently in an industry trade magazine [ 2 ]. Amgen is the Californian biotech firm that hired handsome ‘West Wing’ star Rob Lowe to help market an anti-infection drug. Lowe was reportedly paid more than $1 million by Amgen, though there is speculation that part of the fee might flow to charity [ 3 ]. In her report, Amgen's Osnat Benshoshan shares some thoughtful tips with her peers among the pharmaceutical marketing fraternity: ‘use an A-list celebrity’; find a ‘news-hook’ that links the celebrity and your product; develop some simple messages; and make sure the celebrity delivers them at every appearance. Benshoshan then reveals why on-air talk-show appearances on ‘top-tier media venues’ like The Rosie Show can be better forums for celebrities than straight advertisements, which are governed by regulations. ‘The great advantage over advertising is that the airtime is practically free, and there is no fair balance to worry about’ she writes [ 2 ]. The downside with a media interview, she laments, is that compared to a scripted ad, ‘the situation is less controllable. It can be tricky for the celebrity to ensure that all product messages are delivered….’ Her other big tip for drug-makers is to rate your prospective celebrity with a ‘Q score’, a measure of their likeability and recognisability with the public. Apparently Rob Lowe's Q score was high with women over fifty, a key target of the Amgen campaign [ 3 ]. Another recent report from within the industry draws on public opinion survey data to guide drug company marketers on the selection and ‘effective use’ of celebrity spokespersons [ 4 ]. The survey was conducted by a Seattle firm called NexCura Inc., in partnership with the trade magazine that published the study. The major findings echo the insights of the Amgen executive about credibility, and underline the importance of your star being perceived as generally trustworthy, and specifically knowledgeable about the condition on which they are hired to speak. Perhaps not surprisingly, the survey found that people diagnosed as suffering chronic conditions were far more attentive to celebrity messages on health than the general public. The issue of credibility is important, the NexCura Inc. researchers point out, because ‘the credibility rating is used as a surrogate for “buying” behavior’—an intermediate measure of whether the star can persuade people to request the target drug from their doctor. The survey found that Bob Dole was still the most recognisable celebrity marketer with the United States public, but that the skater Dorothy Hamill—currently promoting Merck's arthritis medication rofecoxib (Vioxx)—took the lead in the credibility stakes. Significantly though, almost three-quarters of those surveyed were correctly able to identify Bob Dole with Pfizer's Viagra, despite the fact that the advertisements in which he appeared were ‘unbranded’ ads for erectile dysfunction. The researchers concluded by recommending that drug companies choose a celebrity with personal experience of the target condition; choose someone trustworthy—perhaps a newsreader or sports figure; and choose someone who will promote a single cause or brand rather than multiple ones. Ironically, the NexCura survey also found two-thirds of medical consumers agreed with the proposition that celebrities were ‘just doing it for the money and can't be trusted’. The Trouble with Celebrity Selling The first problem here is that the public is often not even informed whether a celebrity is receiving money from a drug company. In the case of TV star Rob Lowe, there was no mandated requirement for him to disclose his link with Amgen when appearing on media shows watched by millions. According to one industry insider familiar with the case, who did not want to be named, ‘it depended if he remembered to say it, and whether he was asked’. The media's failure to disclose relevant conflicts of interest when covering healthcare is well established [ 5 ]. When Frasier star Kelsey Grammer and his wife were promoting irritable bowel syndrome on top-rating TV shows, viewers thought the pair were speaking on behalf of an independent foundation. In fact the couple's fee had flowed from GSK, which was at that time preparing the market for alosetron (Lotronex), a controversial new drug that carried modest benefits and severe side effects, including possible death [ 6 ]. Equally as serious is the lack of any formal requirement for stars or media outlets to spell out drug side effects along with benefits when celebrities are pushing products or conditions. Lauren Hutton can be quoted, in magazine articles read by millions of readers, as saying, ‘My No. 1 secret is estrogen’ without any need for her, or the magazine, to list the dangers of the hormone replacement therapy made by her sponsor [ 7 ]. But perhaps most troubling is the way celebrities, with their star power, can help to fundamentally shift the public debate about major health problems. While Prince Charles's companion Camilla Parker Bowles takes no money from drug companies, she did choose to make an important public statement about the bone condition osteoporosis at an international conference funded by Lilly, a company promoting a medication for the condition [ 8 ]. Camilla's call for early intervention and greater use of expensive tests and technologies for the primary prevention of osteoporosis drew on materials sponsored by the pharmaceutical industry, and was synchronised with simplistic industry marketing messages. Camilla's high-profile intervention at a drug company sponsored forum, albeit unwittingly, helps keep the focus on biochemical causes of, and biochemical solutions to, the much wider public health problem of fractures. Moreover these simple marketing messages undermine the complexity of the cost-effectiveness arguments that are central to any rational debate about the equitable distribution of health care resources. Other high-profile figures attending the same conference eagerly accepted Lilly money, and one, former Texas Governor Ann Richards, blatantly promoted Lilly's drug during an interview on CNN's Larry King Show just days later [ 8 ]. The Future of Celebrity Selling With pharmaceutical marketing, it is clear that nothing short of a Vatican II-style reform is required, though there are already encouraging signs of change. Scientific journals are slowly disentangling themselves from unhealthy industry influence over what they publish, and public access to clinical trial data is daily a closer reality [ 9 ]. However, a less distorted scientific record about healthcare products is meaningless without regulations on how important science is communicated to the public. Celebrities paid by drug companies to promote drugs, or ‘raise awareness’ about disease, should be subject to the same rules as direct-to-consumer advertising, which would mean prohibition in many nations and much more fulsome disclosure in the United States than is currently the case. At the very least, public disclosure of a product's risks and benefits, and the magnitude of the celebrity's fee, should be mandatory and routine. Let's see what that does to their Q rating.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC529428.xml
551593	A phase I study of dexosome immunotherapy in patients with advanced non-small cell lung cancer	Background There is a continued need to develop more effective cancer immunotherapy strategies. Exosomes, cell-derived lipid vesicles that express high levels of a narrow spectrum of cell proteins represent a novel platform for delivering high levels of antigen in conjunction with costimulatory molecules. We performed this study to test the safety, feasibility and efficacy of autologous dendritic cell (DC)-derived exosomes (DEX) loaded with the MAGE tumor antigens in patients with non-small cell lung cancer (NSCLC). Methods This Phase I study enrolled HLA A2+ patients with pre-treated Stage IIIb (N = 4) and IV (N = 9) NSCLC with tumor expression of MAGE-A3 or A4. Patients underwent leukapheresis to generate DC from which DEX were produced and loaded with MAGE-A3, -A4, -A10, and MAGE-3DPO4 peptides. Patients received 4 doses of DEX at weekly intervals. Results Thirteen patients were enrolled and 9 completed therapy. Three formulations of DEX were evaluated; all were well tolerated with only grade 1–2 adverse events related to the use of DEX (injection site reactions (N = 8), flu like illness (N = 1), and peripheral arm pain (N = 1)). The time from the first dose of DEX until disease progression was 30 to 429+ days. Three patients had disease progression before the first DEX dose. Survival of patients after the first DEX dose was 52–665+ days. DTH reactivity against MAGE peptides was detected in 3/9 patients. Immune responses were detected in patients as follows: MAGE-specific T cell responses in 1/3, increased NK lytic activity in 2/4. Conclusion Production of the DEX vaccine was feasible and DEX therapy was well tolerated in patients with advanced NSCLC. Some patients experienced long term stability of disease and activation of immune effectors	Introduction Vaccine immunotherapy as an approach to cancer treatment has evolved over the last 10 years as the basic biology of the immune response has been elucidated. Tumor-associated antigens that are capable of eliciting cytotoxic T cell responses have been identified. Among the most frequently expressed across many malignancies are the MAGE antigens, originally described in melanoma, but expressed by other tumors including non-small cell lung cancer (NSCLC) [ 1 - 3 ]. Immune responses to MAGE 3 have been correlated with clinical outcome in melanoma patients [ 4 ]. This has lead to many tumor antigen-specific strategies for the treatment of cancer, including the use of an immunodominant peptide alone, protein or peptide-pulsed dendritic cells, and antigen/co-stimulatory fusion proteins expressed from viral vectors. Although each strategy has its proponents, none have achieved the goal of activating significant immune responses that correlate with clinical responses in a majority of patients; therefore, there is a continued need to develop even more effective strategies. Recently, a novel platform for delivering high levels of antigen in conjunction with costimulatory molecules has been described, called exosomes, cell-derived lipid vesicles that express high levels of a narrow spectrum of cell proteins. A variety of cells have been shown to release exosomes including dendritic cells [ 5 ], B lymphocytes [ 6 ], T lymphocytes [ 7 ], mast cells [ 8 ], platelets [ 9 ], and tumor cells [ 10 ]. The small (60–90 mm) vesicles form within the late endosomes or multivesicular bodies and have biologic functions dependent on the cell type from which they were secreted [ 11 - 13 ]. Originally described as vesicles released from reticulocytes containing proteins (transferring receptor) that were no longer required in the mature red blood cell [ 14 ], they have subsequently been demonstrated to play a role in activation of the immune response. Exosomes derived from B lymphocytes were able to stimulate CD4+ T cells in an antigen/MHC class II restricted manner [ 6 ] and have been demonstrated to be the source of MHC class II molecules on follicular dendritic cells [ 15 ]. In addition, tumor cells release vesicles that function in cross-priming by transferring a protein antigen from the tumor cell to a dendritic cell for immune presentation [ 10 , 16 ]. Importantly, dendritic cells release vesicles (named "dexosomes") that have been demonstrated to prime specific T cells in vitro and eradicate established murine tumors [ 5 ]. In vitro, dexosomes have the capacity to present antigen to naïve CD8+ cytolytic T cells and CD4+ T cells [ 17 , 18 ]. Human dexosomes are enriched in the components necessary to function as an antigen-presenting entity. Extensive electron microscopic and protein characterization has revealed that dexosomes contain a specific set of proteins that differentiate them from other plasma membrane derived vesicles (such as apoptotic cells for example), including MHC class I and II molecules and CD1a, b, c, d molecules, as well as the co-stimulatory molecule CD86 and several tetraspan proteins (CD9, CD37, CD53, CD63, CD81, and CD82) [Anosys unpublished data, [ 19 , 20 ]]. Dexosomes have been demonstrated to participate in antigen presentation in the following way [ 21 , 22 ]. After capturing antigens at the periphery, DC incorporate MHC-antigenic peptide complexes in dexosomes with immunostimulating factors. Released dexosomes subsequently transfer MHC-antigenic peptide complexes and associated proteins to antigen-naïve DC in the regional lymph nodes. The latter thereby acquire the ability to stimulate CD4+ and CD8+ T cells. Thus, dexosomes appear to act as a vehicle for disseminating antigen amongst DC, representing a potentially important mechanism of immune response amplification. This hypothesis forms the rationale for the potential use of dexosomes as a therapeutic cancer immunotherapy. Dexosomes have demonstrated significant antitumor activity in a mouse tumor model, suggesting that the use of dexosomes derived from dendritic cells may result in improved efficacy relative to the ex vivo dendritic cell approach for eradication of advanced cancer. Purified dexosomes were shown to be effective in both suppressing tumor growth and eradicating an established tumor in this model. Furthermore, the effect of the dendritic cell-derived dexosome was greater than that of the dendritic cell from which it was produced [ 5 ]. Therefore, we hypothesized that dendritic cell-derived dexosomes would be an effective platform for activating tumor antigen-specific immune responses in humans. We performed this study to investigate the safety, feasibility, and efficacy of administering autologous dexosomes loaded with tumor antigens (subsequently referred to as DEX) to patients with advanced NSCLC. We also evaluated the immunologic responses in selected patients and monitored the clinical outcomes. Methods Patients This phase I clinical protocol was approved by the Duke University Medical Center Institutional Review Board and conducted in compliance with the Helsinki Declaration and under an IND from the United States Food and Drug Administration held by Anosys Corporation. All subjects provided written informed consent. Patients were eligible for enrollment if they had histologically confirmed, unresectable Stage III A or B or Stage IV NSCLC, were HLA A*0201 positive, at least 18 years of age, and had adequate organ function and a Karnofsky performance status of at least 80%. Patients were required to have been treated with at least one prior standard chemotherapy regimen and have measurable disease. In addition, patients were required to have tumor expressing MAGE A3 or MAGE A4. To avoid performing repeat biopsies, this was achieved by detecting MAGE A3 or MAGE A4 expression in peripheral blood tumor cells by RT-PCR using established methods. The main exclusion criteria were: prior therapy within 4 weeks of the leukapheresis, CNS disease, history of autoimmune disease, concurrent use of systemic steroids, presence of HIV infection or acute or chronic viral hepatitis B or C. Pregnant or lactating women were also excluded. Manufacture of DEX Dexosomes were manufactured from peripheral blood mononuclear cells (PBMCs) as previously described [ 23 ]. Briefly, PBMCs were obtained from the patient during a 2-blood volume leukapheresis and shipped overnight to Anosys, Inc., Menlo Park CA. The cells were washed, adhered to plastic to isolate monocytes and placed in a 7-day serum-free culture at 37°C in a humidified 5% CO 2 atmosphere in the presence of 50 ng/mL GM-CSF (Immunex, Seattle, Washington) and 10 ng/mL of IL-4 (Schering-Plough, Kennilworth, NJ). On the 7 th day of culture, the supernatant of the resulting dendritic cell preparation was harvested, filtered, and concentrated. Dexosomes were then isolated by ultracentrifugation on a D 2 O/sucrose cushion. As described in table 1 , the final dexosome product (DEX) consisted of one of three different formulations based on different methods for loading the following peptides onto the dexosomes: MAGE-derived, HLA-A2 restricted Class I peptides KVAELVHFL (MAGE-A3(112–120)), GVYDGREHTV (MAGE-A4(230–239)) and GLYDGMEHL (MAGE-A10(254–262)); MAGE derived HLA-DP04 restricted Class II peptide TQHFVQENYLEY (MAGE-A3(247–258)) [ 24 ]; and the control peptides, the cytomegalovirus (CMV) pp65-derived, HLA-A2 restricted Class I peptide NLVPMVATV and the tetanus toxoid-derived, promiscuous HLA-DR Class II peptide QYIKANSKFIGITE (produced by Multiple Peptide Systems, San Diego, CA). Peptides were loaded either "directly" onto dexosomes (i.e., following purification of dexosomes from the DC culture) or "indirectly" (i.e. onto cultured DCs that are the source of the dexosomes). The quantity of DEX prepared from a single leukapheresis was measured by ELISA as previously described [ 23 ]. The measured number of MHC class II molecules present in the DEX product was utilized for the purpose of dosing. The final DEX product was diluted in 0.9% normal saline for injection, sterile filtered, and stored at -80°C; subsequently the DEX product was shipped overnight to the investigative site, and maintained in its frozen state until 1 hour before use. Table 1 Dose Groups and Product Formulations Dose Cohorts Number of patients in Cohort (Patient number) Peptides loaded/HLA class Peptide loading method and concentration DEX dose (expressed as numbers of MHC class II molecules) A 3 (DU 5, 6, 8) MAGE-A3 (112–120)/class I MAGE-A4 (230–239)/class I MAGE-A10 (254–262)/class I CMV pp65/class II Tetanus toxoid/class II Indirect (10 μg/mL) Indirect (10 μg/mL) Indirect (10 μg/mL) Indirect (10 μg/mL) Indirect (10 μg/mL) 0.13 × 10 14 B 5 (DU 24, 39, 44, 50, 63) MAGE-A3 (112–120)/class I MAGE-A4 (230–239)/class I MAGE-A10 (254–262)/class I CMV pp65/class I MAGE-A3 (247–258)/class II Tetanus toxoid/class II Direct (10 μg/mL) Direct (10 μg/mL) Direct (10 μg/mL) Direct (10 μg/mL) Indirect (10 μg/mL) Indirect (10 μg/mL) 0.13 × 10 14 C 4 (DU 49, 73, 81, 83) MAGE-A3 (112–120)/class I MAGE-A4 (230–239)/class I MAGE-A10 (254–262)/class I MAGE-A3 (247–258)/class II Direct (100 μg/mL) Direct (100 μg/mL) Direct (100 μg/mL) Indirect (10 μg/mL) 0.13 × 10 14 The amino acid sequences of the peptides used for loading the dexosomes are: MAGE-A3(112–120) = KVAELVHFL; MAGE-A4(230–239) = GVYDGREHTV; MAGE-A10(254–262) = GLYDGMEHL; MAGE-A3(247–258) = TQHFVQENYLEY; CMV pp6 = NLVPMVATV; tetanus toxoid = QYIKANSKFIGITE Note: DU39, DU44, and DU83 were not treated. Treatment and Follow-up Schedule Patients were enrolled into three cohorts (A,B,C) that varied in the method of MHC Class I peptide loading and concentration as described in Table 1 . The quantity of DEX administered to the patients in each cohort was identical: 1.3 × 10 13 MHC class II molecules in a volume of 3 mL (divided into twoinjections given at two sites on opposite sides of the body) as a combination of subcutaneous (90% of the volume) and intradermal (10%) injections weekly for 4 weeks. No retreatment was allowed. Vital signs were monitored for 1 hour after each injection. Clinical responses were assessed by RECIST criteria. CT scans of the chest through the upper abdomen were obtained at baseline, 1 month following the last dose of DEX and every 3 months after last dose of DEX for 1 year, but scans to confirm responses were not required in this phase I study. All surviving patients have been followed every 6 months for assessment of vital status. Delayed type hypersensitivity (DTH) testing Prior to the initial leukapheresis and 1 week after the last dose of DEX, the following peptides were injected intradermally, in addition to the standard recall antigen panel of Candida, Mumps, and tetanus: MAGE-A3(112–120), MAGE-A4(230–239), MAGE-A10(254–262), and MAGE-A3(247–258), each at 10 μg in 0.1 mL saline. The diameter of the induration and erythema was measured 48 hours following the peptide injection. ELISPOT testing Immune response was evaluated at baseline and 1 week following last dose of DEX with cryopreserved PBMCs obtained by leukapheresis. The ELISPOT assay was performed by ImmunoSite, Inc. (Pittsburgh, PA) according to previously reported methods [ 25 ] using both direct assessment of thawed PBMCs, and when possible, following in vitro stimulation of PBMCs with autologous DCs pulsed with the MAGE-A3(112–120), MAGE-A4(230–239), and MAGE-A10(254–262). The number of spots (interferon-gamma-secreting T cells) per 20,000 responding PBMC was reported. The background number of spots against an irrelevant antigen was subtracted from the number of spots for the experimental conditions. Natural killer cell activity NK cells were isolated from cryopreserved PBMCs using an NK Cell Isolation Kit (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer's instruction. NK cell purity was checked by flow cytometry using anti-CD3-FITC, anti-CD45-PerCP, and anti-CD56-APC antibodies (BD Bioscience, San Jose, CA). Isolated NK cells and NK cells activated for 40 hours by IL-2 (Proleukin, Chiron) 600 Units/ml were incubated at various effector to target rations with chromium-51 labeled K562 cells, an NK target, for 4 hours at 37°C and cytotoxicity was assessed by the amount of radiolabeled chromium released. Cytotoxicity was calculated as follows: percentage of target cell lysis = 100 × (counts per minute (cpm) of experimental release - cpm of spontaneous release) / (cpm of maximum release - cpm of spontaneous release). Statistics The primary endpoints of this study were safety and feasibility, with secondary endpoints of clinical and immunologic response rates. The incidence, type and severity of adverse events were recorded during the study treatment through 30 days following the last dose of DEX. Descriptive statistics were used to present the data. Adverse events were coded using MEDDRA version 5.0. Survival and time to progression were measured from the date of the first injection to the date of documented disease progression or death. For patients who progressed, the time to disease progression was determined by the interval from the first injection+ 1 day to the last evaluation of disease staging. For patients who did not progress or die during the two year follow up period, the time of disease progression and survival was determined by the interval between the first dose of DEX and the date of last evaluation of disease staging + 1 day and concatenated with the '+' sign. Results Patient Characteristics Thirteen patients, (8 female, 5 male) median age 62 years (range 44–72 years) with unresectable pretreated Stage III or IV NSCLC were enrolled. The median time from original diagnosis to study entry was 9.9 months (range 2–61 months) and the median Karnofsky score of the patient population was 80% (range 80–100). DEX therapy was administered to 9/13 (6 female and 3 male) patients. Of the 9 dosed patients, 5 patients had Stage IV and 4 patients had Stage IIIB disease. Six patients had stable disease and 3 patients had progressive disease at study entry. Two patients had squamous cell carcinoma, 4 patients had adenocarcinoma, 2 patients had large cell carcinoma, and in 1 case the histological type was not reported. All patients had received prior chemotherapy (median number of cycles: 6.5, range: 3–30), 6/9 patients had received prior radiotherapy and 4/9 patients had prior surgery for cancer treatment. Four patients did not receive DEX for the following reasons: manufacturing failure in 2 cases (DU39, DU83), one of whom had received chemotherapy 13 days prior to leukapheresis and one of whom (DU39) also had rapid disease progression at the time of leukapheresis; delay in shipment in one instance (DU14); and, rapid disease progression prior to planed dosing with DEX in one case (DU44). The characteristics of all dosed patients are listed in Table 2 (see separate file for Table 2). Dexosome manufacture The dose of DEX that was selected corresponded to the maximum dose that could be achieved from healthy donors. We confirmed that this dose could be generated in all but two patients with NSCLC. The mean dexosome generation consisted of a total class II number of 3.14 × 10 14 (range 4.1 × 10 12 to 9.1 × 10 14 ). This quantity of dexosomes in our advanced NSCLC patients is similar in quantity to that generated from healthy donors (mean for 111 healthy donors was 3.9 × 10 14 total class II). Toxicity The DEX immunotherapy was generally well tolerated without evidence of serious toxicity. The most frequently reported adverse events causally related to the use of DEX were mild (Grade 1–2) in severity and included: Injection site reactions (erythema, contusion, induration and edema) in 8 patients; flu like syndrome (1 patient); and, peripheral edema and pain in the arm (1 patient). There were no significant organ or laboratory toxicities attributable to the vaccine. No autoimmune reactions were observed. Immunologic Response DTH analysis All 9 dosed patients underwent DTH testing with individual tumor-associated peptides prior to and following all doses of DEX. There was no DTH response to the specific peptide antigens prior to DEX therapy. Three patients (DU06, DU24 and DU49) had a positive response of at least 5 mm erythema or induration in the longest dimension 48 hours after skin testing with one of the MAGE peptides. Specifically, DU06 had 5 mm induration and erythema with MAGE-A4(230–239), DU24 had 6 mm induration and erythema with MAGE-A10(254–262) and DU49 had 5 mm induration and erythema with MAGE-A3(112–120), respectively. In vitro immunologic analysis The peptide-specific immune response to MAGE and CMV was analyzed using ELISPOT in 5 of 9 dosed patients (DU24, DU49, DU50, DU63, DU81). One patient (DU49) exhibited detectable increases in T cell precursor frequency to MAGE-A10(254–262) following in-vitro stimulation (an increase of 12 MAGE-A10-specific cells/20,000 responders). Assays for DU50 and DU63 could not be completed because of poor viability. Robust responses to anti-CD3 and to the control peptide CMV pp65 were observed in DU24, DU81, and DU49, but no MAGE-specific responses were detected. Since most patients did not exhibit a significant increase in antigen-specific T cell activity, we hypothesized that regulatory influences such as CD4+CD25+ regulatory T cell populations might inhibit augmentation of the T cell response. In 2/3 patients who had analyzable specimens available, an increase in CD4+CD25+ T cells as a percentage of CD4+ T cells was observed following completion of DEX therapy when compared with baseline values (DU05: increase from 19.49 to 26.64%, DU08: minimal change from 20.39 to 23.42%, DU50: increase from 17.45 to 31.81%). The small number of samples available for this analysis precludes any conclusions but does suggest that CD4+CD25+ T cell analyses should accompany future studies of DEX immunotherapy. During the study, new data from Escudier et al (manuscript submitted) suggested that the immunologic activity of DEX might be due to activation of NK cells. We therefore explored the hypothesis that NK cells may be activated following DEX therapy. This was not planned as part of the initial analysis and therefore specimens of PBMC were limiting in all but 4 patients (DU05, DU08, DU24, DU50). Although there was no consistent change in NK percentage before and after immunization (Table 1 : DU05: 10.7 to 9.2%; DU08: 6.0 to 5.4%; DU24: 8.8 to 8.5%; DU50: 9.9 to 13.9%), NK activity as determined by the ability to lyse K562 target cells was observed to increase in 2/4 patients following immunizations (Fig 1 -see additional file Fig 1). Short-term culture with IL-2 was required to activate the NK cells in vitro as there was very low activity in the absence of IL-2. Although addition of IL-2 increased the NK cell activity, it did not change the relative pattern of activity, i.e., in no instance did the order of the results change as a result of IL-2 stimulation. Figure 1 Cytolytic activity of NK cells. Cytolytic activity of NK cells isolated from the PBMC of 4 patients (DU05, DU08, DU24, DU50) pre (squares) and post (circles) immunization and cultured with (dark shapes) or without (open shapes) IL-2 was determined. The percentage lysis of the NK target (K562) cells is reported at effector to target ratios of 0.2:1 to 25:1. Clinical Outcomes At approximately 2 years of follow up, survival from the first dose of DEX ranged from 52 to 309 days for cohort A, 280 to 665+ days for cohort B, and 244 to 502 days for cohort C (Table 1 ). In order to obtain preliminary data on response rate, CT scans were obtained prior to immunization and at 1 month and 3 month intervals following completion of the immunizations, but additional scans were not obtained to confirm responses. Of the two patients (DU05, DU08) with disease progression at study entry, DU05 was stable at the end of the immunizations but was felt to have clinically progressed at day 88 shortly before death. DU08 also had stable disease at the end of the immunizations and on every three month follow-up until having progression at day 302. Of four additional patients (DU06, DU24, DU63, DU81) who began the study with stable disease, two (DU24 and DU63) have remained without progression for greater than 12 months. DU06 was stable at the post-immunization CT but subsequently died unexpectedly of unknown etiology and without a follow-up scan, and DU81 was stable at the post-immunization CT but had progressed by the next CT scan at the three month follow-up. The remainder of the patients had progressed at the post-immunization CT scan including DU73 who had disease progression prior to the first dose of DEX. The time until progressive disease, as documented from the first dose of DEX, ranged from 30+ to 302 days for cohort A, 40 to 429+ days for cohort B, and 51 to 166 days for cohort C. Discussion The objective of this study was to show that DEX could be manufactured from NSCLC patients and could be safely administered. We demonstrated the feasibility of producing dexosomes loaded with specific MAGE and other peptides and demonstrated that this form of immunotherapy was well tolerated in patients with advanced NSCLC. Leukapheresis products could be shipped to a central processing facility with good cell viability after transport in a majority of cases, in contrast to other autologous therapies involving tissues where logistics of tissue harvest and processing are complex. The dexosome product was successfully manufactured and loaded with multiple peptides in the majority of patients. This suggests that different panels of tumor antigen-derived peptides could be successfully loaded onto dexosomes. Using multiple peptide panels may allow for targeting various tumor types and larger patient populations, and decreases the likelihood of antigenic escape. We observed increases in systemic immune responses against MAGE by DTH reactivity in 3/9 patients who had no reactivity to the MAGE peptides prior to immunization and activation of NK cells, but found minimal increases in antigen-specific T cell activity in in vitro assays performed circulating PBMCs. Possible explanations include nonoptimized or low-sensitivity assays, inadequate antigen presentation, counter-regulatory mechanisms that dampen immune responses, or the lack of persistence of antigen-specific Tcells in the circulation (i.e., the T cell may have migrated to tumor tissue or lymph nodes). The possible role of negative regulatory mechanisms was suggested by the presence of elevated levels of CD4+CD25+ regulatory T cells following immunization in some patients. An intriguing immunologic observation was the increase in NK activity following immunization in 2/4 patients analyzed. Although DEX are intended to activate antigen-specific, MHC-restricted T cell responses, it is possible that cytokines released in response to DEX therapy could cause activation of NK cells or that DEX could directly activate NK cells. DEX therapy may stimulate both innate and adaptive arms of the immune response and thereby provide a rationale for maximizing the anti-tumor effect of this approach, even in cases where tumors have lost Class I antigens, a common finding as cancers become more advanced [ 26 ]. Indeed, in a phase I study in melanoma patients, DEX loaded with MAGE peptides were well tolerated and associated with both clinical response and increased NK activity (Escudier, manuscript submitted to J. Trans Med). Despite the small sample size and the fact that 3/9 dosed patients had disease progression at the time of initiation of DEX treatment, we observed prolonged disease stabilization in some patients. Large clinical trials in patients with advanced NSCLC have generally reported median time to progression of 3–5 months in patients with advanced NSCLC treated with systemic chemotherapy regimens [ 27 - 30 ]. The lack of toxicity and interesting clinical and immunologic observations support further investigation of DEX immunotherapy as a treatment approach for both advanced and early stage NSCLC and other tumors. Phase II clinical studies in non-small cell lung cancer and other tumor types are planned to continue to explore the efficacy of this novel immunotherapy. Conclusion DEX therapy was well. Immune activation and stability of disease was observed in some immunized patients with advanced NSCLC. Competing Interests Michael Morse received funding from NIH 5R21CA89957-02. Additionally, portions of this study were funded by Anosys, Inc. Nancy Valente, Revati Shreeniwas, Mary Ann Sutton, Alain Delcayre, Di-Hwei Hsu, and Jean BernardLe Pecq held stock and were employees in Anosys, H. Kim Lyerly was a consultant for Anosys, Inc. Authors' contributions MAM was the principal investigator of the study and oversaw all aspects including protocol development, patient management, data collection and analysis, and manuscript preparation. JG enrolled patients to the study and managed their care and participated in data analysis. Takuya Osada performed the NK assays and analyzed the data. SK enrolled patients to the study and managed their care. AH performed in vitro immunologic assays and analyzed the data. TMC oversaw the immunologic analyses performed at Duke University and analyzed the data. NV, RS, and MAS oversaw development of the protocol, data collection and analysis, and manuscript preparation. AD developed and oversaw the MAGE screening for patient eligibility. D-H H oversaw portions of the immunologic analysis and data analysis. J-B L provided scientific direction regarding generation of the dexosomes, protocol development, and data analysis and manuscript preparation HKL provided consultation on immunologic assay development All authors read and approved the final manuscript. Supplementary Material Additional File 1 Table 2 (DOC) presents the remainder of clinical and immunological data from all patients Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC551593.xml
516800	Estimating mutual information using B-spline functions – an improved similarity measure for analysing gene expression data	Background The information theoretic concept of mutual information provides a general framework to evaluate dependencies between variables. In the context of the clustering of genes with similar patterns of expression it has been suggested as a general quantity of similarity to extend commonly used linear measures. Since mutual information is defined in terms of discrete variables, its application to continuous data requires the use of binning procedures, which can lead to significant numerical errors for datasets of small or moderate size. Results In this work, we propose a method for the numerical estimation of mutual information from continuous data. We investigate the characteristic properties arising from the application of our algorithm and show that our approach outperforms commonly used algorithms: The significance, as a measure of the power of distinction from random correlation, is significantly increased. This concept is subsequently illustrated on two large-scale gene expression datasets and the results are compared to those obtained using other similarity measures. A C++ source code of our algorithm is available for non-commercial use from kloska@scienion.de upon request. Conclusion The utilisation of mutual information as similarity measure enables the detection of non-linear correlations in gene expression datasets. Frequently applied linear correlation measures, which are often used on an ad-hoc basis without further justification, are thereby extended.	Background The evaluation of the complex regulatory networks underlying molecular processes poses a major challenge to current research. With modern experimental methods in the field of gene expression, it is possible to monitor mRNA abundance for whole genomes [ 1 , 2 ]. To elucidate the functional relationships inherent in this data, a commonly used approach is the clustering of co-expressed genes [ 3 ]. In this context, the choice of the similarity measure used for clustering, as well as the clustering method itself, is crucial for the results obtained. Often, linear similarity measures such as the Euclidean distance or Pearson correlation are used in an ad-hoc manner. By doing so, it is possible that subsets of non-linear correlations contained in a given dataset are missed. Therefore, information theoretic concepts, such as mutual information, are being used to extend more conventional methods in various contexts ranging from expression [ 4 - 8 ] and DNA sequence analysis [ 9 , 10 ], to reverse engineering [ 11 ] and independent component analysis [ 12 , 13 ]. Also aside the bioinformatics field, mutual information is widely utilised in diverse disciplines, such as physics [ 14 ], image recognition [ 15 ], speech recognition [ 16 ], and various others. In extension to other similarity measures, mutual information provides a general measure of statistical dependence between variables. It is thereby able to detect any type of functional relationship, extending the potentialities of linear measures as illustrated in Figure 1 . In this work, we discuss mutual information as a measure of similarity between variables. In the first section, we give a short introduction into the basic concepts including a brief description of the commonly used approaches for numerical estimation from continuous data. In the following section, we then present an algorithm for estimating mutual information from finite data. The properties arising from this approach are compared to previously existing algorithms. In subsequent sections, we then apply our concept to large-scale cDNA abundance datasets and determine if these datasets can be sufficiently described using linear measurements or if a significant amount of non-linear correlations are missed. Mutual information Mutual information represents a general information theoretic approach to determine the statistical dependence between variables. The concept was initially developed for discrete data. For a system, A , with a finite set of M possible states { a 1 , a 2 , ... , }, the Shannon entropy H ( A ) is defined as [ 17 ] where p ( a i ) denotes the probability of the state a i . The Shannon entropy is a measure for how evenly the states of A are distributed. The entropy of system A becomes zero if the outcome of a measurement of A is completely determined to be a j , thus if p ( a j ) = 1 and p ( a i ) = 0 for all i ≠ j , whereas the entropy becomes maximal if all probabilities are equal. The joint entropy H ( A, B ) of two systems A and B is defined analogously This leads to the relation H ( A, B ) ≤ H ( A ) + H ( B ) (3) which fulfils equality only in the case of statistical independence of A and B . Mutual information MI ( A, B ) can be defined as [ 17 ] MI ( A, B ) = H ( A ) + H ( B ) - H ( A, B ) ≥ 0 (4) It is zero if A and B are statistically independent and increases the less statistically independent A and B are. If mutual information is indeed to be used for the analysis of gene-expression data, the continuous experimental data need to be partitioned into discrete intervals, or bins. In the following section, we briefly review the established procedures; a description of how we have extended the basic approach will be provided in the subsequent section. Estimates from continuous data In the case of discrete data the estimation of the probabilities p ( a i ) is straightforward. Many practical applications, however, supply continuous data for which the probability distributions are unknown and have to be estimated. In a widely used approach [ 7 ], the calculation of mutual information is based on the binning of data into M discrete intervals a i , i = 1... M A . For experimental data consisting of N measurements of a variable x u , u = 1... N , an indicator function Θ i counts the number of data points within each bin. The probabilities are then estimated based on the relative frequencies of occurrence with For two variables the joint probabilities are calculated analogously from a multivariate histogram. Additionally it has been suggested [ 14 ] to adaptively choose the sizes of the bins, so that each bin constructed nearly has a uniform distribution of points. In a different approach, kernel methods are used for the estimation of the probability density of Eq. (5) [ 18 - 20 ]. Entropies are then calculated by integration of the estimated densities. Recently, an entropy estimator was suggested [ 21 ] and showed in an extensive comparison to other commonly used estimators to be superior. Results Fuzzy mutual information In the classical binning approach, described above, each data point is assigned to one, and only one, bin. For data points near to the border of a bin, small fluctuations due to biological or measurement noise might shift these points to neighbouring bins. Especially for datasets of moderate size, the positions of the borders of the bins can thereby strongly affect the resulting mutual information [ 18 ]. In a manner analogous to kernel density estimators (KDE), we now present a generalisation to the classical binning in which we aim to overcome some of the drawbacks associated with the simple approach. Within our algorithm, we allow the data points to be assigned to several bins simultaneously. For this, we extended the indicator function Θ( x ) to the set of polynomial B-spline functions. Here, we do not provide the mathematical details for these functions since they have been discussed extensively in the literature [ 22 - 24 ], but rather focus on the practical applicability. Within the B-spline approach, each measurement is assigned to more than one bin, i , with weights given by the B-spline functions B i,k . The spline order k determines the shape of the weight functions and thereby the number of bins each of the data points is assigned to. A spline order k = 1 corresponds to the simple binning, as described in the previous section: Each data point is assigned to exactly one bin (Figure 2 , left). For k = 3, each data point is assigned to three bins, with the respective weights given by the values of the B-spline functions at the data point (Figure 2 , right). B-spline functions The first step in the definition of the B-spline functions is the definition of a knot vector t i for a number of bins i = 1... M and one given spline order k = 1... M - 1 [ 22 ] where the spline order determines the degree of the polynomial functions. The domain of the B-spline functions lies in the interval z ∈ [0, M - k + 1]. To cover the range of the variables, the new indicator function based on the B-spline functions needs to be linearly transformed to map their range. The recursive definition of the B-spline functions are as follows [ 22 ] An important property of B-spline functions is the implicit standardisation of coefficients: All weights belonging to one data point sum up to unity. Algorithm Input • Variables x and y with values x u and y u , u = 1... N • Bins a i , i = 1... M x and b j , j = 1... M y • Spline order k Output • Mutual information between variable x and y Algorithm 1. Calculation of marginal entropy for variable x (a) Determine with (b) Determine M x weighting coefficients for each x u from (c) Sum over all x u and determine p ( a i ) for each bin a i from (d) Determine entropy H ( x ) according to Eq. (1) 2. Calculation of joint entropy of two variables x and y (a) Apply steps 1 (a) and (b) to both variables x and y , independently (b) Calculate joint probabilities p ( a i , b j ) for all M x × M y bins according to (c) Calculate the joint entropy H ( x,y ) according to Eq. (2) 3. Calculate the mutual information MI ( x,y ) according to Eq. (4) Example We show the estimation with the standard binning and our approach ex-emplarily on two artificial variables x = 0.0,0.2,0.4,0.6,0.8,1.0 and y = 0.8,1.0,0.6,0.4,0.0,0.2 for M = 3 bins, spline order k = 2, and the logarithm to basis two. Simple binning For both variables, each of the three histogram bins contains two values p ( a 1 ) = p ( a 2 ) = p ( a 3 ) = , analogously for p ( b i ) due to the symmetry of data H ( x ) = H ( y ) = = log 2 3 ≈ 1.58. For the calculation of the joint probability, three of the nine two dimensional bins contain two values each p ( a 1 , b 3 ) = p ( a 2 , b 2 ) = p ( a 3 , b 1 ) = resulting in H ( x, y ) = log 2 3 and MI ( x, y ) = log 2 3. B-spline approach The modified indicator function is determined to B i,k (2 x ) according to Eq. (9) (rule 1(a)). For each value x u three weighting coefficients are determined (rule 1(c)) and probabilities are calculated (rule 1(d)) (Table 1 ). The analogous procedure is applied to variable y and the single entropies are calculated to H ( x ) = H ( y ) = Iog 2 (10) - 0.61og 2 (3) - 0.41og 2 (4) ≈ 1.57. Both, H ( A ) and H ( B ), are slightly smaller than the entropies calculated from the simple binning. The joint probabilities are p ( a 1 , b 1 ) = p ( a 3 , b 3 ) = 0, p ( a 1 , b 2 ) = p ( a 2 , b 1 ) = p ( a 2 , b 3 ) = p ( a 3 , b 2 ) = 0.56/6, p ( a 1 , b 3 ) = p ( a 3 , b 1 ) = 1.24/6, p ( a 2 , b 2 ) = 1.28/6 (rule 2 (b)) resulting in H ( x,y ) = 2.7 and MI ( x,y ) = 0.45. In the next sections, we discuss some of the properties arising from the utilisation of B-spline functions for the estimation of mutual information and compare our approach to other commonly used estimators. We support this discussion using examples for which the underlying distributions and thereby the true mutual information is known. Size of data It has been discussed elsewhere [ 25 - 28 , 20 ] that the estimated mutual information is systematically overestimated for a finite size of N data points. For the simple binning approach, the mean observed mutual information can be calculated explicitly as the deviation from the true mutual information As can be seen for an example of artificially generated equidistributed random numbers (Figure 3 , left), mutual information calculated from the simple binning scales linearly with 1/ N , with the slope depending on the number of bins M in accordance with Eq. (12). Figure 3 shows that this scaling is preserved for the extension to B-spline functions, while the slope is significantly decreased for k = 3, compared to the estimation with the simple binning ( k = 1). Mutual information calculated from KDE does not show a linear behaviour but rather an asymptotic one with a linear tail for large datasets. The values are slightly increased compared to the ones from the B-spline approach. The entropy estimator gives values comparable to the ones obtained from the B-spline approach. More importantly, a similar result also holds for the standard deviation of mutual information. As shown in Figure 3 (right), the standard deviation of the mutual information estimated with the simple binning ( k = 1) scales with 1/ N for statistically independent events [ 26 , 29 ]. For the B-spline approach ( k = 3), this scaling still holds, but the average values are decreased significantly. For the KDE approach, an asymptotic run above the values from the B-spline approach is observed, again with linear tail for large datasets. shows a linear scaling slightly below the simple binning. The spline order The interpretation of any results obtained from the application of mutual information to experimental data is based on testing to see if the calculated results are consistent with a previously chosen null hypothesis. By following the intuitive approach that the null hypothesis assumes the statistical independence of variables, mutual information is tested against a surrogate dataset, which is consistent with this null hypothesis. As discussed previously in more detail [ 20 ], one way of generating such a surrogate dataset is by random permutations of the original data. From the mutual information of the original dataset MI ( X,Y ) data , the average value obtained from surrogate data < MI ( X surr , Y surr ) >, and its standard deviation σ surr , the significance S can be formulated as For each S the null hypothesis can be rejected to a certain level α depending on the underlying distribution. With increasing significance the probability of false positive associations drops. In the following, we address the influence of the spline order and the number of bins on the estimation of mutual information. Based on 300 data points of an artificially-generated dataset drawn from the distribution shown in Figure 1 , we calculate the mutual information for M = 6 bins and different spline orders k = 1... 5 (Figure 4 , left). From 300 shuffled realisations of this dataset, the mean and maximum mutual information are shown with the standard deviation as error-bars. For all spline orders the null hypothesis can be rejected, in accordance with the dataset shown in Figure 1 . To estimate the strength of the rejection, we calculate the significance according to Eq. (13) (Figure 4 , right). It can be observed that the largest change in the significance of the mutual information occurs in the transition from k = 1 (simple boxes) to k = 2 with an increase by roughly two-fold. Using more sophisticated functions ( k ≥ 3) does not further improve the significance. Similar findings have been reported in the context of kernel density estimators [ 19 ]. The major contribution leading to this increase of the significance is given by the distribution of surrogate data which becomes more narrow for k > 1 leading to smaller standard deviations σ surr . The same dataset is used to show the dependency of mutual information on the number of bins for two spline orders k = 1 and k = 3 (Figure 5 ). Mutual information estimated from data as well as from surrogate data shows a robust run without strong fluctuations within the range of bins shown. From this we can conclude that the choice of the number of bins does not affect the resulting mutual information notably as long as it is chosen to be within a reasonable range. Again, the significance is calculated (Figure 6 ) and compared to the significances obtained from the KDE approach and the estimator. It can be observed that the significance of the mutual information calculated with B-spline functions increased roughly by two-fold compared to the simple binning. The significance obtained from KDE is not depending on M and was determined to be similar to the significance estimated from the B-spline approach. The numerically expensive integration of KDE, however, limits the size of utilisable datasets. The KDE run time requirements were (10 4 ) times higher than the ones from the B-spline approach. Strategies to simplify the integration step were proposed [ 20 ] but have to be used with caution since they assume particular properties of the distribution of experimental data that are in general not fulfilled. The recently introduced entropy estimator produces intermediate significances between the ones from the binning and the B-spline approach for higher bin numbers. For low bin numbers, the significances are relatively poor. Application on data We now turn to the analysis of experimentally measured gene expression data. As shown previously, the application of mutual information to large-scale expression data reveals biologically-relevant clusters of genes [ 7 , 30 ]. In this section, we will not repeat these analyses, but determine if the correlations detected using mutual information are missed using the established linear measures. Among the most frequently used measures of similarity for clustering co-expressed genes are the Euclidean distance and the Pearson correlation coefficient R [ 3 ]. If correlations are well described by the Pearson correlation and the distribution of data is approximately Gaussian like, the relationship between the mutual information and the Pearson correlation given by [ 32 ] is expected to be fulfilled. Therefore, we calculated both, the mutual information and the Pearson correlation, for two large-scale gene expression datasets (Figure 7 ). For each pair of genes X and Y we plot the tuple ( MI ( X,Y ), R ( X,Y )). In order to address significance, we additionally calculate all tuples from shuffled data. The first dataset contains cDNA measurements for S. cerevisiae for up to E 1 = 300 experiments [ 31 ]. To avoid numerical effects arising from different numbers of defined expression values (missing data points) for each gene, we exclusively utilised genes that are fully defined for all experimental conditions resulting in G 1 = 5345 genes. Analysis on this dataset using mutual information has been done before [ 20 , 32 ] on rank-ordered data. The rank-ordering lead to homogeneously distributed data and thereby enabled the application of a simplified algorithm for the numerical estimation from kernel density estimators. The utilisation of our B-spline approach allows us to extend this analysis to non rank-ordered data thereby keeping the original distribution of experimental data. In contrast to the previous studies we find for non rank-ordered data that the theoretical prediction of Eq. 14 is no longer a lower bound for the comparison. Many tuples with high Pearson correlation but low mutual information can be detected arising from outlying expression values (Figure 8A ). However, pairs of genes with high mutual information and low Pearson correlation, thus indicating a non-linear correlation, are not observed. The only remarkable tuple (marked with an arrow in Figure 7 and shown in Figure 8B ) also arises from outlying values. The second dataset contains cDNA measurements for E 2 = 102 experiments on G 2 = 22608 genes derived from 20 different human tissues [ 33 ]. In contrast to the first dataset, tuples with low Pearson correlation but high mutual information are indeed detected. For two exemplary chosen tuples (Figure 8C and 8D ), clusters of experimental conditions can be clearly detected by eye. Such type of correlations are missed by analyses based exclusively on linear measures, such as the the analysis done in the original publication of this dataset. For both datasets, tuples calculated from shuffled data (Figure 7 , blue data points) result in small values for both similarity measures. Thereby, they indicate a high significance of the original associations. Peaks with high Pearson correlation in the first dataset arise from gene-gene associations with outlying values. Significance values for the exemplarily chosen pairs of genes of the second dataset (Figure 8C , and 8D ) were explicitly calculated (Figure 9 ). They show high significance values for the two examples of observed non-linear correlations on the basis of the mutual information. Compared to this, the significances calculated from the Pearson correlation are poor. In summary, our analysis confirms for the first dataset that the Pearson correlation does not miss any non-linear correlations. As a side effect we are able to detect gene-gene pairs containing outlying values. For the second dataset, however, a substantial amount of non-linear correlations was detected. Gene-gene pairs exemplarily chosen from this fraction show a clustering of data points (experiments) with a high significance. Even though such patterns can be easily found by eye, computational methods need to be applied for the inspection of several hundred million comparisons. Discussion and conclusion After a brief introduction into the information theoretic concept of mutual information, we proposed a method for its estimation from continuous data. Within our approach, we extend the bins of the classical algorithm to polynomial B-spline functions: Data points are no longer assigned to exactly one bin but to several bins simultaneously, with weights given by the B-spline functions. By definition, the weighting coefficients for each data point automatically sum up to unity. Though our algorithm is reminiscent of kernel density estimators [ 18 ], it keeps the basic idea to associate data points to discrete bins. In this way, we are able to avoid time-consuming numerical integration steps usually intrinsic to estimates of mutual information using kernel density estimators [ 20 ]. To show that our approach improves the simple binning method and to compare it to KDE and the recently reported estimator , we provided a systematic comparison between all these algorithms for artificially generated datasets, drawn from a known distribution. We found that mutual information, as well as its standard deviation, scales linearly with the inverse size of a dataset for the standard binning method, for the B-spline approach, and for . For the KDE approach we find an asymptotic behaviour with a linear tail for large datasets. Moreover, the discrimination of correlations from the hypothesis of statistical independence is significantly improved by extending the standard binning method to B-spline functions, as shown by a two-fold increase of the significance. Compared to KDE, the B-spline functions produce similar significances. However, due to the numerical expenses of the KDE, an application of this algorithm is limited to datasets of mod-erate size. The application of leads to significances in-between the standard binning and the B-spline approach for reasonable bin numbers. Linear correlation measures are among the most applied measures of similarity in the literature. Often, they are used on an ad-hoc basis and it is unclear whether a considerable number of non-linear correlations are missed. Here, we asked the question whether previous analyses, based on linear correlations, sufficiently described the correlations within gene expression datasets or whether mutual information detects additional correlations that are not detected by linear measures, such as the Pearson correlation. For data that is well described by the Pearson correlation, we can give the relation of the Pearson correlation to the mutual information explicitly [ 32 ]. Both measures were then applied to publicly available large-scale gene expression datasets [ 31 , 33 ]. We aimed to verify whether non-linear correlations shown as deviations from this relation can be detected. Our findings show that the first dataset is fairly well described by the given relation of the Pearson correlation to the mutual information. No data points with high mutual information and low Pearson correlation are detected. Comparisons of genes containing outlying values, however, result in deviations with low mutual information and high Pearson correlation. From this, it follows that previous analyses on this dataset, based on Pearson correlation, did not miss any non-linear correlations. This presents an important finding since it is by all means supposable that the regulations inherent in the genetic network under consideration might show more complex behaviour than the observed linear ones. Even for one of the largest expression datasets at hand, insufficient data might complicate the detection of such complex patterns of regulation. Alternatively, the biological mechanisms which underlay the regulatory networks might not lead to non-linear correlations. It also has to be considered that the experimental methods applied for the generation of this dataset may make non-linear correlations difficult to detect. The second dataset, in contrast, reveals highly significant tuples with high mutual information and low Pearson correlation. Detailed gene-gene plots of such tuples show that the expression values of the contributing genes fall into groups of experimental conditions. Without attempting to draw conclusions about the biological context of such clusters here, they might reflect interesting situations worth to be analysed in detail. Authors' contributions Most of the manuscript text was written by CD and edited by all authors. CD carried out the calculations and produced the figures. RS strongly contributed to the theoretical background of entropy and mutual information. The implementation of the C++ program was carried out by SK. JS and SK supervised this work. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC516800.xml
521074	A high-throughput cell migration assay using scratch wound healing, a comparison of image-based readout methods	Background Cell migration is a complex phenomenon that requires the coordination of numerous cellular processes. Investigation of cell migration and its underlying biology is of interest to basic scientists and those in search of therapeutics. Current migration assays for screening small molecules, siRNAs, or other perturbations are difficult to perform in parallel at the scale required to screen large libraries. Results We have adapted the commonly used scratch wound healing assay of tissue-culture cell monolayers to a 384 well plate format. By mechanically scratching the cell substrate with a pin array, we are able to create characteristically sized wounds in all wells of a 384 well plate. Imaging of the healing wounds with an automated fluorescence microscope allows us to distinguish perturbations that affect cell migration, morphology, and division. Readout requires ~1 hr per plate but is high in information content i.e. high content. We compare readouts using different imaging technologies, automated microscopy, scanners and a fluorescence macroscope, and evaluate the trade-off between information content and data acquisition rate. Conclusions The adaptation of a wound healing assay to a 384 well format facilitates the study of aspects of cell migration, tissue reorganization, cell division, and other processes that underlie wound healing. This assay allows greater than 10,000 perturbations to be screened per day with a quantitative, high-content readout, and can also be used to characterize small numbers of perturbations in detail.	Background When wounded or scratched, cell monolayers respond to the disruption of cell-cell contacts and an increased concentration of growth factors at the wound margin by healing the wound through a combination of proliferation and migration [ 1 - 3 ]; these processes reflect the behavior of individual cells as well as the properties of the cell sheet as a surrogate tissue. To perform a wound healing assay, a wound is typically introduced in a cell monolayer using an object such as a pipette tip or syringe needle and the assay is performed on an individual coverslip or in a multiwell plate. The monolayers recover and heal the wound in a process that can be observed over a timecourse of 3–24 hrs. The wound heals in a stereotyped fashion – cells polarize toward the wound, initiate protrusion, migrate, and close the wound. Progression of these events can be monitored by manually imaging samples fixed at timepoints or by time-lapse microscopy. Wound healing assays are a classic and commonly used method for studying cell migration and the biology underlying it [ 4 ]. They have been used with multiple cell types and, as the monolayers heal the wound in a characteristic manner, they have been used to study cell polarization, matrix remodeling, cell migration, and numerous other processes [ 5 - 7 ]. Wound healing assays have been used both for detailed cell biological studies and for the discovery and validation of small molecule leads and other perturbations that affect cell migration [ 8 - 11 ]. The role of the Rho family GTPases, Rac, Rho, and Cdc42, in the establishment of polarity and the regulation of actin cytoskeletal structures has been studied using wound healing [ 12 - 14 ], as has the role of p53 in migration [ 15 ], and orientation of the microtubule organization center (MTOC) and the Golgi apparatus [ 16 - 18 ]. The assay has also been used as a proxy for angiogenesis, metastasis, and other physiological and pathophysiological processes [ 19 - 24 ]. In order to perform high throughput screening of cell migration, we developed a wound healing assay in a 384 well plate format that does not require expensive reagents, provides consistently shaped wounds, can provide detailed information on numerous processes involved in cell migration, and provides a quantitative, information-rich readout. We use multiple imaging technologies to assay the results and compare their relative merits. Results Adaptation of wound healing to a 384 well format For the development of a high-throughput wound healing assay we chose to use BS-C-1 cells, a cell type with a classic wound healing response on glass coverslips [Figure 1A ]. BS-C-1 cells were seeded in clear-bottom 384 well plates at high density and allowed to form monolayers overnight. We found that wound healing was observable between 3 and 24 hrs after wounding with a pipette tip or syringe needle. Significant cell migration could be seen at 3 hrs with lamella and protrusions at the wound margin. After 7 hrs, cell migration could be observed easily with a low magnification (4×) objective, and after 24 hrs wounds were completely healed [Figure 1B , Additional file: 1 ]. Because of the ease of distinguishing phenotypes at 7 and 24 hrs, we have used these timepoints for the assay. To adapt this assay for parallel screening, we needed a method for introducing uniformly sized wounds in the same position of each well. We used a 96 well floating-pin transfer device – a tool primarily used for the transfer of solutions between plates. A floating pin array, with foam padding placed between the top plate and the pins, provides an adaptive stop to pin height and overcomes problems with plate planarity. We have adapted a 24-channel aspirator in the same manner for small-scale work. To wound all 384 wells in a plate, the 96 well pin array is placed in the corner of a well, pushed down to engage all pins with the surface of the plate, and then moved laterally to produce the wound. This is then repeated in three neighboring wells to cover the plate and produce uniform wounds throughout [Figure 1C ]. After the cells were wounded, we introduced perturbations to individual wells (in our case, small molecules). We then allowed the cells to recover for 7 hrs or 24 hrs before fixing, staining, and imaging each plate. Comparison of imaging technologies We used four different imaging technologies to analyze the results of this assay, each with distinct advantages and drawbacks [summarized in Table 1 ]. We will discuss these approaches in order of image resolution, from highest to lowest: automated fluorescence microscopy after 7 hrs recovery, fluorescence and transmitted-light scanners after 7 hrs, and a fluorescence macroscope after 24 hrs. Automated microscopy For highest resolution imaging of the assay we chose an end point of 7 hrs, when migration can be clearly seen, and used an automated fluorescence microscope to image individual wells after fixing and staining for filamentous actin and DNA [Figure 1D ]. The microscope is a standard inverted fluorescence instrument. Augmented with an x-y stage, it moves between plate wells and a piezoelectric z-motor on the objective gives a focused image. Capturing images with a 4× objective provided sufficient resolution to determine the extent of migration and the morphology of the cells at the wound margin. From these data, we defined four distinct phenotypes [Figure 1D ]. A control well shows polarization of the cells toward the wound and concerted migration of the cell sheet, with neighboring cells connected and moving together into the wound. Wells showing decreased migration or aberrant morphology are readily apparent by visual inspection as are wells showing an increase in the number of mitotic cells, which manifest as bright spheres (in the actin channel) in an otherwise intact and adherent monolayer [Figure 1D ]. Phenotypes that cause disruption of the monolayer, are considered toxic though we have not shown them to be. Using automated microscopy, the image resolution is relatively high and the time required to image an entire plate is relatively long. Imaging one 384-well plate takes ~1 hr at 4× magnification. At 10× magnification, ~1.5 hrs are required per plate because at higher magnification two images per well must be taken to ensure that the wound edge is captured. We found it more informative to observe the entirety of the wound at 4× rather than parts of it at 10×, despite the higher resolution in the latter case. Scanners A fluorescence scanner can be used to determine the extent of cell migration at the 7 hr timepoint. By staining filamentous actin and using a fluorescence scanner with a 42 μm/pixel resolution setting, we can observe consistent differences between normal and inhibited migration. Control wells show a veil of less-densely stained, migrating cells that extend into the wound with a concomitant decrease in wound width [Figure 2A ]. Titration of a compound that blocks cell migration (the actin inhibitor cytochalasin D) shows complete inhibition of migration at 1 μM, as seen by a sharply delineated wound edge and a wider wound width [Figure 2A ]. A simple transmitted-light scanner, normally used for scanning documents and costing less than $1,200, can also be used to monitor wound healing at 7 hrs and only requires that cells are stained with a dye. Figure 2B shows images of wells stained with Coomassie Brilliant Blue. Wells treated with cytochalasin D show inhibition of wound healing similar to that observed with the fluorescence scanner. Inhibited wells stain darkly at the wound margin while normal migration can be seen by a more diffuse wound margin, denoting migrating cells [Figure 2B ]. In both cases, scanners do not provide specific information on cell morphology or other subtle effects in the 7 hr assay; however, acquisition time is greatly decreased. For the fluorescence scanner, acquisition time is 26 minutes at the resolution and image quality shown (42 μm/pixel and medium quality). For the conventional scanner, acquisition time is 8.5 minutes (at 10 μm/pixel, 2400 dpi) and cell staining with Coomassie Brilliant Blue takes only 10 minutes. Macroscope – Tundra or LeadSeeker The lowest resolution imaging technology that we tested, a fluorescence macroscope had a resolution of ~100 μm/pixel. Detecting wound healing at this resolution required an incubation time of 24 hours after wounding as differences are not readily seen with this method at the 7 hr timepoint. Cells are wounded and allowed to recover for 24 hours before fixing, staining for filamentous actin, and imaging. The macroscope captures an image of the entire plate. At this magnification and resolution, an unhealed wound is seen as a non-staining, black streak within the monolayer [Figure 3 ]. In contrast, wounds that have healed completely are seen as lower-intensity, grey streaks. The time required for acquisition of images using this technique, is limited only by the fluorescence signal and was typically ~5 seconds per plate. Automated image analysis of wound healing images With all of the imaging techniques discussed here, we initially scored the assay by visual inspection. Visual inspection is fast, information-rich, and can distinguish subtle effects. This method proved useful during adaptation and optimization of the assay in high-throughput format. However, screening large numbers of perturbations by visual inspection is limited by subjectivity, operator fatigue, and the lack of quantifiable metrics. Thus, we developed an automated image analysis routine to provide a rapid and quantitative measurement of the wound healing assay with images captured using the automated microscope (automated methods for the other readouts could also be developed). Using images captured by automated microscopy with a 4× objective, we are able delineate measurable characteristics of the wound. After 7 hrs recovery, actin staining defines the extent of healing as well as the morphology of the cells and, because of cell polarization, DNA staining defines the approximate starting point of the migrating cells [Figure 4A ]. Applying a standard set of image processing filters to threshold the image, we defined the area of the wound not staining for nuclei and the area not staining for actin. The difference between these two areas defines an annulus containing the lamella of the cells at the wound front (hereafter, lamellar region) [Figure 4A ]. With the lamellar region defined, we can extract several metrics that characterize migration. Measurements that have been useful in quantifying the phenotypes of migration inhibition and aberrant morphology include: area of lamellar region, width of lamellar region, and the smoothness of the wound margin. Automated analysis and visual inspection are complementary approaches. To illustrate this, we compare the two approaches using the results from one 384 well plate of a screen for small molecules that affect wound healing [Figure 4B ]. The automated analysis values for each well are plotted and the average lamellar width from control wells is marked as a black line, with three standard deviations above and below that average marked as red dashed lines. Visual inspection of this same plate is illustrated on the graph by color-coding each well. Wells where migration is inhibited or morphology affected are shown in green; compounds that affect the well in other ways – wells with fewer cells, increased mitotic index, or disrupted monolayers – are shown in red; and compounds that show no significant deviation from control are shown in blue. As can be seen, the automated analysis picks up 4 wells that show inhibition of wound healing with lamellar widths more than 3 standard deviations from the mean and with less stringent bounds (2 standard deviations) 7 compounds are found to inhibit. Other compounds identified by visual inspection were not picked up by automated analysis. Validation of wound healing as a high-throughput assay We have used methods described in this paper (notably, automated microscopy) to screen ~1,000 bioactive and ~20,000 random drug-like compounds. We were able to reproducibly identify compounds that affected wound healing with different effects including: inhibiting migration, affecting morphology, blocking completion of mitosis, and disrupting the cell monolayer. The details of these results will be presented elsewhere. Discussion We present a cell migration assay in a 384 well plate format through the adaptation of tissue culture cell wound healing. We have also compared the readouts provided by four distinct imaging techniques technologies – automated microscopy, fluorescence and transmitted-light scanners, and a fluorescence macroscope – for their relative acquisition speed, image resolution, and information content. High-content image-based screening is often performed at high magnification, however, we have found that low magnification images are information-rich and can be sufficient for observation of detailed phenomena. New screens utilizing imaging technologies are often explicitly developed to image plates at the highest resolution possible given time constraints (10× or higher), rather than at the lowest resolution required to discern differences between wells by visual inspection or automated analysis. Increased magnification further requires acquisition of a greater number of images in order to guard against sampling error. The more sophisticated imaging technologies also carry a hefty price tag. We show here that images taken at magnifications of 4× and lower still allow us to easily discern wells in which wound healing is inhibited and morphology affected. Low magnification imaging is an underutilized method and may be particularly effective for screens that monitor gross changes in protein localization such as nuclear transport, or protein transport from Golgi to plasma membrane; processes in which changes in localization might manifest as changes in the image texture of the stained cell monolayer. While the choice of imaging technologies represents a tradeoff between resolution and information, we think that they should be used together rather than suggesting that they are mutually exclusive. One can iteratively image a given assay plate and take advantage of the best aspects of each technique. For example, a screen for inhibitors of cell migration could initially be performed using a lower resolution technique, taking advantage of the speed to identify interesting wells before imaging just those wells at higher resolution. Related to this, one can re-stain assay plates with different molecular markers. Plates initially screened with a marker that most easily defines the phenotype of interest can be subsequently re-stained with markers that provide different information. In this way, one can consider an assay plate a resource to which one can return to ask new questions. Similar to the tradeoffs seen with different imaging technologies, different image processing techniques like automated image analysis and visual inspection have their distinct advantages but also provide complementary approaches. Our scheme for automated analysis of wound healing allows us to rapidly and easily identify wells in which wound healing is significantly inhibited or morphology clearly affected. However, the automated analysis missed a number of interesting wells that were identified by visual inspection. In part this is because the analysis is based on only a few measures – lamellar width, area, and smoothness. Any automated analysis based on measurable parameters will be limited to phenotypes that manifest along those parameters and might miss more subtle or complex phenotypes that "jump off the screen" when viewed by eye. Automated analysis can also be confounded by artifacts within the plate, dust particles, precipitates, or fluorescent small molecules. Instead of being thought of as a complete answer, automated analysis can enrich for wells that differ from the norm. We don't address automated analysis with the other imaging techniques, but from what can be seen by eye, automated approaches that measure the intensity of staining at the wound edge, or the steepness to which the intensity at the wound edge drops off should provide a good measure of wound healing at very low magnification and resolution. As with any complex phenotypic assay, cell migration during wound healing can be inhibited by effects on global cellular processes – e.g. inhibition of protein translation, disruption of metabolism, disruption of ion homeostasis, etc. – and validation of the specificity of new perturbations is required. Even if the molecular target of the perturbation is unknown, this assay can be adapted or used to distinguish compounds that specifically inhibit or potentiate cell migration. For example, partial inhibition of migration with low concentrations of cytochalasin D could create a sensitized screen for cell migration and used to find suppressors or enhancers. Other complementary approaches include comparative screens across multiple cell lines that migrate differently and the use of specific trophic factors and cognate cell lines such as VEGF and endothelial cells. In our studies, we used still another approach: a series of counter screens that eliminated from our pool of hits, toxic compounds and compounds that do not inhibit processes of interest (Yarrow et. al. unpublished results). This approach has worked well and results will be reported elsewhere. Conclusions The adaptation of a wound healing assay to a 384 well format facilitates the study of aspects of cell migration, tissue reorganization, cell division, and other processes that underlie wound healing. This assay allows greater than 10,000 perturbations to be screened per day with a quantitative, information-rich readout, and can also be used to characterize small numbers of perturbations in detail. Methods Tissue culture and 384 plate preparation BS-C-1 (ATCC CCL-26) cells were grown in DMEM, 10% FCS, and antibiotics. Cells were plated in black 384 well plates with clear bottoms (Corning Costar 3712) at a density of 8500 cells/well in a volume of 50 μl using a liquid dispenser (Labsystems Multidrop). Plates were spun briefly at 500 rpm for ~30 s to in a tabletop centrifuge (Sorval RT7 plus). Cells were incubated overnight (37°C 5% CO 2 ) and wounding was preformed 12 hours later. Wound healing assay, cell fixation, and staining Wound healing was performed using a 96 well floating-pin transfer device with a pin diameter of 1.58 mm coming to a flat point at the tip with a diameter of 0.4 mm (VP Scientific VP-408FH). Foam backing was inserted between the plates of the pin array to provide a resistive stop and the external guide pins were bent to allow greater movement in the z-axis. The pin array was placed in the top corner of a well, pushed down into the plate to engage all pins, and then pulled toward the user. This was repeated in the three neighboring wells to cover all 384. Plates were returned to the tissue culture incubator for 7 or 24 hours before fixation. Cells were fixed after removal of the media with a wand aspirator (VP scientific VP-186L) used along with the Labsystems Multidrop for all liquid handling. Fixation solution (100 mM K-Pipes pH 6.8, 10 mM EGTA, 1 mM MgCl 2 , 0.2% Triton X-100, 3.7% Formaldehyde) was added as 30 μl and incubated for 15 min. Wells were aspirated and washed 2× with TBS with 0.1% Triton-X 100 (TBS-Tx) and stained. For experiments involving the automated microscope and macroscope, cells were stained in TBS-Tx with TRITC-phalloidin (Sigma P1591) 0.5 μg/ml and Hoechst (Sigma B2261) 0.1 μg/ml as 15 μl per well for 15 minutes. Wells were washed 2× with TBS-Tx and imaged. For experiments using the fluorescence plate scanner, after fixation, cells were incubated in TBS-Tx with 2% BSA (AbDil) for 30 minutes, incubated with mouse anti-actin antibody (Chemicon MAB1501) at 1:10,000 in AbDil for 45 minutes, washed 2× with TBS-Tx, incubated with secondary antibodies appropriate for the plate scanner (Molecular Probes A-21057), washed 2× with TBS-Tx and imaged. For experiments using the transmitted-light scanner, after fixation, cells were incubated with SDS-Page gel staining solution (0.25% Coomassie Brilliant Blue R-250, 50% methanol, 10% acetic acid) for 10 minutes, washed 2× with TBS and imaged. Imaging Automated microscopy We used a NikonTE300 inverted fluorescence microscope with filter wheel (Sutter Lamda10-2), x-y stage (Prior H107N300), and piezoelectric-motorized objective holder (Physik Instrumente P-723.10). Images were captured on a CCD camera (Hammamatsu OrcaER). Metamorph software (Universal Imaging Corporation) running the "Screen Acquisition" drop-in allowed coordination of software-based auto-focusing, movement between wells, imaging, and image evaluation. Images were acquired using a 4× or 10× objective with 2 × 2 binning. Exposure times were ~300 ms for TRITC-phalloidin and ~10 ms for Hoechst. An individual plate took ~1 hr to image at 4× and ~1.5 hrs at 10×. Individual black and white actin images (.tif) were compiled as a .stk file and scrolled through using keystrokes to visually annotate. Visual inspection of 1 plate of images (384) took 10 minutes. Transmitted light scanner An Epson 1680 scanner was used with the positive film setting. A scanning resolution of 1200 dpi (equivalent to 20 μm/pixel) gave an acceptable image with a read time for one plate of 3.5 minute. The image shown in Figure 2 was 2400 dpi (10 μm/pixel) and read time was 8.5 minutes. We also found that using the document scanning mode (with no transmitted light attachment) worked well. In this case, a lamp with a paper diffuser (placed on top of the plate) was used for even illumination. Fluorescence scanner An Odyssey scanner (LiCor) was tested at all combinations of scanning resolution and image quality. Scanning at a resolution setting of 42 um with medium image quality was found to be the optimal balance between speed and image quality. The scanner was used as per the manual, with no modifications. Fluorescence macroscope A Tundra macroscope, (Imaging Research – now available as the Leadseeker from Amersham [ 25 ]) a 12 cm telecentric lens with a N/A of 0.45, mounted with a -50°C cooled, thinned, back illuminated CCD camera for image capture, and a motorized stage to hold the plates – all enclosed in a light-tight box. The software was used as per the manual. To image the underside of plates, they were sealed (Corning Costar 6570) while containing TBS and inverted. Exposure times were ~5 seconds. Automated image analysis Wound healing images were analyzed using software written using Visual Basic 6.0 (Microsoft) and Halcon 6.0.1 (MVTec Software) but could be implemented with most basic image analysis software. This software iterates the analysis over images specified by the Metamorph .HTS file and returns values to an Excel spreadsheet. Non-DNA staining region ( Additional file: 2A ) Hoechst images are convolved with a Laplacian-of-a-Gaussian (σ = 2 pixels) kernel. The resulting image is thresholded for pixels of value 0 and binarized. After a binary morphological opening (an erosion followed by a dilation) with a disc of radius 2.5 pixels, the largest contiguous region that does not touch the image edge is defined, and holes within this region are filled. Non-actin staining region ( Additional file: 2B ) A Kirsch edge detection filter is applied to actin images and the resulting image is thresholded at a manually set value (changed when needed to account for variation in staining intensity). The largest contiguous region that does not touch the image edge is defined and holes within this binary region are filled. Definition of measures (in pixels) lamellar area: area of annulus defined by the difference between the non-actin staining area and the non-DNA staining area; lamellar width: (lamellar area)/(perimeter length of the non-DNA staining region); lamellar smoothness ( Additional file: 2C ): (Perimeter of the non-actin staining region)/(perimeter of the morphological closing (dilation followed by erosion) of the non-actin staining region with a disc of radius 10 pixels). Authors' contributions JCY developed the assay and analysis. ZEP wrote the software for automated analysis. NJW initiated a project for in vivo actin cytoskeleton small molecule inhibitors. TJM provided support and enthusiasm for the project. Supplementary Material Additional file 1 Wounds generated with the 96 well floating-pin array heal in a characteristic and measurable manner. (A) Wounds generated with the 96 well floating-pin array healed for 0, 3, 7, 12, or 24 hours, were processed as above, and wound regions were measured using automated analysis. Normalized average values and standard deviations of both the area and median width of the non-actin staining region and the non-DNA staining region (see Methods for definitions) are shown. The non-DNA staining region at the 24 hr timepoint was automatically not measured (asterisks) because the non-actin staining region was 0. At least 24 wells of each condition were measured. Click here for file Additional file 2 Sample images illustrating automated analysis of wound healing images. (A) Processing of DAPI image to generate non-DNA staining region. (B) Processing of actin image to generate non-actin staining image. (C) Comparison of lamellar wound region with its morphological closing allows assessment of lamellar smoothness. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC521074.xml
406391	Early Myocardial Function Affects Endocardial Cushion Development in Zebrafish	Function of the heart begins long before its formation is complete. Analyses in mouse and zebrafish have shown that myocardial function is not required for early steps of organogenesis, such as formation of the heart tube or chamber specification. However, whether myocardial function is required for later steps of cardiac development, such as endocardial cushion (EC) formation, has not been established. Recent technical advances and approaches have provided novel inroads toward the study of organogenesis, allowing us to examine the effects of both genetic and pharmacological perturbations of myocardial function on EC formation in zebrafish. To address whether myocardial function is required for EC formation, we examined s ilent heart (sih −/− ) embryos, which lack a heartbeat due to mutation of cardiac troponin T ( tnnt2 ), and observed that atrioventricular (AV) ECs do not form. Likewise, we determined that cushion formation is blocked in cardiofunk (cfk −/− ) embryos, which exhibit cardiac dilation and no early blood flow. In order to further analyze the heart defects in cfk −/− embryos, we positionally cloned cfk and show that it encodes a novel sarcomeric actin expressed in the embryonic myocardium. The Cfk s11 variant exhibits a change in a universally conserved residue (R177H). We show that in yeast this mutation negatively affects actin polymerization. Because the lack of cushion formation in sih- and cfk- mutant embryos could be due to reduced myocardial function and/or lack of blood flow, we approached this question pharmacologically and provide evidence that reduction in myocardial function is primarily responsible for the defect in cushion development. Our data demonstrate that early myocardial function is required for later steps of organogenesis and suggest that myocardial function, not endothelial shear stress, is the major epigenetic factor controlling late heart development. Based on these observations, we postulate that defects in cardiac morphogenesis may be secondary to mutations affecting early myocardial function, and that, in humans, mutations affecting embryonic myocardial function may be responsible for structural congenital heart disease.	Introduction The genetic programs and developmental processes that lead to organ formation are still poorly understood. We are currently witnessing an expansion in research that aims to identify the genes responsible for the structural development of organs and their later function. Among the organs of the body, the heart is unique because it begins to function mechanically before structural development is complete, begging the important question of whether myocardial function is required for the morphogenetic events that occur after the heart begins beating. One of the late steps of heart development is the formation of the endocardial cushions (ECs), which are tissue swellings that develop in characteristic locations along the anterior–posterior (AP) extent of the heart tube and contribute to valves and, in four-chambered hearts, to septae. Because of the clinical significance and prevalence of EC defects in humans ( Hoffman 1995 ), an understanding of the genetic and epigenetic factors controlling cushion and valve formation is critical. During the process of EC and valve development, specific endocardial cells undergo multiple poorly understood specification, differentiation, and migration events en route to becoming functional heart valves. The genes involved in one substep of this process, epithelial–mesenchymal transformation (EMT), are gradually being identified. Analysis of EMT during cardiac-cushion development has implicated molecules such as Fibronectin, Transferrin, ES-130, hLAMP-1, TGF-β2, TGF-β3, BMP-2 (reviewed in Nakajima et al. 2000 ), Alk-3 ( Gaussin et al. 2002 ), and hyaluronic acid ( Camenisch et al. 2000 ) as being required for this process. However, many other molecules are likely to be involved in the complex process of EMT, and little is known about the events leading to heart-valve formation that precede or follow EMT, prompting us to take a forward genetic approach to this question ( Stainier et al. 2002 ). To examine the role of epigenetic factors involved in EC formation, Hove et al. (2003 ) surgically manipulated zebrafish embryos and put forth the hypothesis that shear stress on endocardial cells is required for EC development. In our study, we use both genetic and pharmacological approaches to test the importance of myocardial function in EC development. Results sih −/− Embryos Lack ECs In zebrafish, an initial step of EC development is the formation of the endocardial ring, a structure generated by the clustering of endocardial cells at the atrioventricular (AV) boundary. This process is easily visualized at 48 h postfer-tilization (hpf) by examining the endocardial cells expressing green fluorescent protein (GFP) under the control of the mouse tie2 promoter ( Motoike et al. 2000 ; Walsh and Stainier 2001 ) ( Figure 1 A). We have previously reported the early cardiac phenotype of silent heart (sih −/− ) embryos, which establish neither a heartbeat nor blood flow due to mutation of cardiac troponin T (tnnt2) , but do undergo looping morphogenesis ( Sehnert et al. 2002 ). We crossed the sih mutation into the tie2::GFP line and found that although sih −/− embryos are otherwise morphologically normal, they fail to form an endocardial ring at the AV boundary ( Figure 1 B). Because sih −/− embryos do not have a heartbeat and therefore no blood flow, it is unclear from this observation whether the defect in EC formation is due to lack of myocardial function or lack of shear stress on endocardial cells. Figure 1 Embryos with Defective Myocardial Function Do Not Form AV ECs (A–C) Fluorescence micrographs of embryos carrying a tie2::GFP transgene, visualized at 48 hpf. In (A), the endocardial ring is visible as a collection of GFP-positive cells at the AV boundary in wild-type (wt) embryos (red arrow). In (B), sih −/− embryos fail to form an AV ring at 48 hpf. In (C), cfk −/− embryos fail to form an AV ring at 48 hpf. (D and E) Cushion development remains defective in cfk −/− embryos. In (D), a 5 μm hematoxylin and eosin-stained plastic section shows the initial stages of cushion development at the AV boundary (red arrows) in a 72 hpf wild-type embryo, with the ECs being two to three cell layers thick at this stage. In (E), a cfk −/− embryo at 72 hpf shows dilation of both chambers of a blood-filled heart with no evidence of cushion formation at the AV boundary (red arrows). (F and G) cfk −/− embryos fail to form ECs at late stages. Embryos were visualized at identical magnification after counter-staining with rhodamine phalloidin. Red blood cells (RBCs) are seen in the atria of the hearts. In (F), confocal microscopy of a 96 hpf wild-type heart from a tie2::GFP line shows triangular ECs at the AV boundary (blue arrows). In (G), cfk −/− embryos at 96 hpf lack cushion formation and clustering of GFP-positive cells at the AV boundary (blue arrows). (H) At 72 hpf, wild-type embryos have narrow hearts with forward blood flow through the embryo. (I) At 72 hpf, cfk −/− embryos have dilated hearts filled with blood that regurgitates freely from the ventricle to the atrium. (J and K) The initial phenotype in cfk −/− embryos is cardiac dilation at 36 hpf. In (J), wild-type embryos have a narrow ventricle and generate pulsatile flow at 36 hpf. In (K), cfk −/− embryos have an increased end-diastolic diameter (on average 1.18× wild-type, p < 0.01) and do not generate blood flow at 36 hpf. (L and M) Increased bmp-4 expression at the AV boundary (red arrow) is observed in wild-type (L) and cfk −/− (M) embryos at 42 hpf in anticipation of endocardial ring formation. (N) Orientation of the embryos shown in (L) and (M). cfk −/− Embryos Lack ECs Subsequent to Impaired Function Through genetic screens ( Alexander et al. 1998 ; Stainier et al. 2002 ) we have identified a new mutant, cardiofunk (cfk) , that fails to accumulate tie2::GFP-positive cells at the AV boundary by 48 hpf ( Figure 1 C). We further examined EC development in cfk −/− embryos histologically. By 72 hpf, wild-type embryos have developed AV ECs that are more than one cell-layer thick ( Figure 1 D), while cfk −/− embryos show no evidence of cushion formation ( Figure 1 E). Examination by confocal microscopy further shows that wild-type embryos have well-developed cushions by 96 hpf ( Figure 1 F), whereas cfk −/− embryos still show no evidence of an endocardial ring or cushion formation ( Figure 1 G). Thus, EC formation appears to be defective from an early stage, and not simply delayed, in cfk −/− embryos. The lack of EC development in these mutants leads to toggling of the blood between the atrium and ventricle and its accumulation inside the heart by 48 hpf ( Figure 1 H and Figure 1 I; see also Video S1 ). Interestingly, the earliest observable phenotype in cfk −/− embryos is cardiac dilation, as evidenced by an increase in ventricular end-diastolic diameter at 36 hpf ( Figure 1 J and Figure 1 K), at which time cfk −/− embryos have failed to establish a circulation ( Video S2 ). The cardiac dilation and lack of circulation in cfk −/− embryos are nearly fully penetrant phenotypes, while the failure of EC development occurs in about 50% of cfk −/− embryos. This observation raises the possibility that the lack of EC formation may result secondarily from some other defect, such as changes in myocardial function or shear stress on endocardial cells. AV Boundary Specification Is Not Affected in sih −/− and cfk −/− Embryos To determine more precisely which step of EC formation is affected in sih −/− and cfk −/− embryos, we examined the expression of bmp-4 , a gene implicated in EC morphogenesis ( Eisenberg and Markwald 1995 ). In wild-type embryos, bmp-4 is initially expressed throughout the anterior–posterior (AP) extent of the heart tube before becoming localized to the AV boundary at 42 hpf ( Walsh and Stainier 2001 ) ( Figure 1 L). In sih −/− and cfk −/− embryos, bmp-4 expression similarly becomes restricted to the AV boundary by 42 hpf ( Figure 1 M) (data not shown) . Therefore, the lack of endocardial ring formation in sih −/− and cfk −/− embryos does not appear to be due to a defect in specification of the AV boundary, as assessed by bmp-4 expression, nor general arrest of cardiac development. The wild-typelike restriction of bmp-4 expression in these mutant embryos contrasts with the situation in jekyll −/− embryos, which also lack ECs, but in which bmp-4 expression does not become restricted to the AV boundary ( Walsh and Stainier 2001 ). cfk Encodes a Sarcomeric Actin To understand better the molecular basis for the cfk phenotypes, we isolated cfk by positional cloning. cfk was initially localized to a region of LG13 by bulk-segregant analysis; fine-mapping placed cfk between Z9289 and Z10582 ( Shimoda et al. 1999 ). Analysis of 2,034 meioses allowed us to perform a chromosomal walk that restricted cfk to bacterial artificial chromosome (BAC) zC202M22 ( Figure 2 A). Following the sequencing of zC202M22, a portion of the BAC insert was assembled as a 52 kb contiguous sequence, and cfk was genetically localized to this region. Using GenScan ( Burge and Karlin 1997 ) and BLAST analysis, we identified four genes in this 52 kb region: abc-b10 , rab4a , actin ( zeh0631 ), and fv49b10 ( Figure 2 B). Comparison of this region of the BAC to fugu and human sequences identified regions of conserved synteny on fugu scaffold 2777 ( Figure 2 C) and human Chromosome 1q42.13 ( Figure 2 D) and confirmed our GenScan analysis. The four zebrafish genes in the critical interval were analyzed by sequencing from cfk −/− mutants and reverse genetic techniques using morpholino antisense oligonucleotides. Injections of morpholinos against each of the four coding sequences showed no obvious cardiac phenotype, indicating that we might not be working with a loss-of-function mutation or that cfk has overlapping function with another gene. Subsequently, sequencing of the actin gene ( zeh0631 ) showed a change of arginine 177 to histidine. The three embryos recombinant at either end of the 52 kb region were homozygous at the site of the R177H lesion, suggesting that we had isolated cfk . Figure 2 Positional Cloning of cfk (A) The locus on zebrafish LG13 containing cfk is shown, with the number of recombinants indicated below each marker. The one recombinant at the Sp6 end of BAC zC202M22 and the two recombinants at the T7 end of the same BAC define the critical region. (B) The first 52 kb of zebrafish BAC zC202M22 is shown, including the Sp6 end. One of the two recombinants from the T7 end of the BAC is still present at a marker at 52 kb, narrowing the critical region to this span. GenScan and BLAST analyses identified four coding sequences in this span, including abc-b10 (green), actin (red), a novel EST fv49b10 (black), and rab4a (blue). (C) Each of the three known genes identified on BAC zC202M22 has a homologue on scaffold 2777 of Fugu rubripes , in the same orientation (green, abc-b10 ; red, actin; blue, rab4a ). (D) The same three genes lie in proximity to each other and in the same order on human Chromosome 1q42.13 (green, abc-b10 ; red, actin; blue, rab4a ). Units in black are genes, predicted genes, or ESTs, which are unique in this region to that particular organism. The cardiac dilation caused by the cfk mutation supported our hypothesis that cfk corresponded to an actin gene. Analyses of Cfk indicate that it is a sarcomeric actin by virtue of its homology to zebrafish α-cardiac and α-skeletal actins and lack of homology to zebrafish cytoplasmic actin, as well as its conservation of synteny with the human skeletal actin. Further evidence that cfk encodes a sarcomeric actin and not a cytoplasmic actin includes the presence of an extra residue at the N-terminus of Cfk, which is seen in all sarcomeric actins but no cytoplasmic actins, as well as the presence of residues that are stereotypic for sarcomeric actins at all 20 locations where sarcomeric and cytoplasmic actins have unique amino acids ( Khaitlina 2001 ) ( Figure 3 A). Searches of GenBank databases identified both zebrafish α-cardiac– and α-skeletal–actin genes that are well represented in the zebrafish expressed sequence tag (EST) collection and are related to, but clearly distinct from, cfk (see Figure 3 A). Analysis of multiple actins from several species revealed that the R177 residue, which is mutated to histidine in the cfk s11 allele, is universally conserved ( Figure 3 B). In vitro mutations at the R177 residue usually affect actin polymerization or function, although the biochemical effect of an R117H transition has not been tested. Figure 3 Sequence and Expression Analysis of cfk (A) cfk encodes a sarcomeric actin highly homologous to zebrafish α-cardiac and α-skeletal actins. Cfk differs from zebrafish α-cardiac actin at six residues and from zebrafish α-skeletal actin at four residues, but from zebrafish β-actin at 28 residues. Residues in red are those that differ from Cfk. Dots above the sequence indicate residues that universally distinguish sarcomeric from cytoplasmic actins. The arrow at R177 indicates the location of histidine in Cfk s11 . (B) The arginine at position 177 is universally conserved in all actin proteins examined. (C and D) cfk is expressed in the myocardium during development. In (C), whole-mount in situ hybridization on a cmlc2::GFP embryo at 36 hpf shows that cfk is expressed throughout the AP extent of the heart tube. Blue staining indicates areas of cfk expression; green is the region of cmlc2 expression. The red arrow indicates the heart tube. In (D), a plastic section of stained embryo shows cfk expression in the myocardium of the heart (blue arrow), but not in the endocardial cells (red arrow). From the onset of its expression in the heart region (around the 16-somite stage), cfk does not appear to be expressed in endothelial and endocardial cells. Weak cfk expression is also seen in the somites (data not shown). To determine the expression pattern of cfk , we performed in situ hybridization with a probe corresponding to its 3′ untranslated region (UTR), which is distinct from the 3′ UTR sequences of zebrafish α-cardiac and α-skeletal actins. These data showed that cfk is expressed in the myocardial but not endocardial cells of zebrafish embryos from 24 to 48 hpf ( Figure 3 C and Figure 3 D), suggesting a nonautonomous role in EC formation. Taken together, the tight linkage between cfk and this actin gene, the presence of a significant genetic lesion in the actin gene, and the expression profile of the actin gene indicate that cfk corresponds to the actin gene previously identified by the EST zeh0631 . The R177H Mutation Alters Actin Polymerization Previous biochemical studies have revealed a critical role for R177 in actin polymerization. An R177A yeast variant is heat sensitive and does not grow in 0.9 M NaCl ( Wertman et al. 1992 ; Drubin et al. 1993 ), while an R177D mutation in chick β-actin leads to severely altered polymerization properties in vitro ( Schuler et al. 2000 ). These results suggest that a positive charge at position 177 is important for actin function in vivo ( Wriggers and Schulten 1999 ). The R177H cfk s11 mutation represents a less severe change in charge than the previously mentioned yeast or chick mutations. To assess the effect of the R177H mutation on actin function, we used site-directed mutagenesis to generate a haploid yeast strain in which the R177H mutant actin was the only actin expressed in the cell. These cells were readily obtained and showed no significant altered morphology or growth characteristics on normosmolar complete medium at 30°C. However, the R177H mutation produced a severe growth defect, similar to the R177A mutation, when the cells were incubated in hyperosmolar complete medium containing 0.9 M NaCl ( Figure 4 C). Figure 4 An Arginine to Histidine Change at Position 177 of Cfk Alters the Location of Positive Charges in the Nucleotide Binding Cleft of Actin (A) The position of R177 (red) in the structure of the yeast-actin monomer. The bound ATP is shown in green. The structure shown is based on a crystal structure of actin (PDB:1YAG). (B) Magnification of the cleft region showing the hydrogen bonding (green dashed lines) involving R177, which will be disrupted in H177. (C) The R177H mutation restricts yeast growth under hyperosmolar stress. Wild-type (wt) and mutant cells were grown to a density of 3 × 10 6 per milliliter, and aliquots were plated either on YPD or YPD plus 0.9 M NaCl at different dilutions of the culture. The cells were then incubated at 30°C for 72 h. The normosmolar and hyperosmolar experiments were done at different times. (D) Purified R177H actin has a higher critical concentration of polymerization and a delayed nucleation phase. Wild-type or mutant actin was purified from yeast cultures and polymerization mea-sured by light diffraction. Symbols and abbreviations: circles, 5 μM wild-type actin; triangles, 5 μM R177H actin; diamonds, 7.5 μM R177H actin; squares, 10 μM R177H actin; ATP, adenine triphosphate; wt, wild-type. To determine whether the R177H mutation affected actin polymerization, we purified the actin from R177H cells ( Cook et al. 1991 ) and assessed the extent of polymerization by the increase in light scattering as a function of time. The mutant actin exhibited two distinct differences in comparison to a similarly prepared sample of wild-type yeast actin ( Figure 4 D). First, the extent of polymerization of a quantity of actin equal to that of the wild-type sample was significantly decreased as judged by the final plateau. Second, there was a prolonged apparent nucleation phase not seen with wild-type actin, even at concentrations of mutant actin that produced more F-actin than the wild-type sample. Samples of the polymerization solution were negatively stained with uranyl acetate and examined by electron microscopy to confirm that the increase in light scattering was caused by F-actin formation and not merely aggregation (data not shown). To define better the apparent difference between R177H and wild-type actins in regard to the critical concentration necessary for polymerization, we assessed the extent of polymerization of different amounts of actin relative to the amount of total actin using a light-scattering assay. Based on determinations with two independent preparations of actin, the critical concentration is 0.3 μM for wild-type actin and 1.9 μM for the mutant actin, confirming that this apparently mild mutation exerts a drastic effect on actin-filament stability. During our work with cfk s11 , we observed that some embryos showing the cfk phenotype were cfk +/− ( Figure 5 A), suggesting that the R177H mutation can exert a partially dominant effect. Indeed, subsequent experiments ( Wen and Rubenstein 2003 ) demonstrate that the presence of the mutant actin in a solution of wild-type actin exerts a partially dominant effect on actin polymerization. The partial dominance of cfk s11 and the yeast data support a model whereby substantial copolymerization of Cfk s11 actin monomers with wild-type monomers (either Cfk or α-cardiac actin) leads to unstable filaments, occasionally giving rise to a phenotype in heterozygous embryos. Figure 5 Lack of EC Formation Is Secondary to Defective Myocardial Function (A) Embryos were observed for defective myocardial function and lack of blood flow at 36 hpf and lack of endocardial ring formation at 48 hpf. Approximately half of the embryos with defective myocardial function at 36 hpf did not develop endocardial rings. All embryos with a lack of endocardial rings previously demonstrated a myocardial function phenotype at 36 hpf. Further analyses including the genotyping of a subset of embryos showed that a small percentage of cfk −/− embryos were unaffected at 36 hpf and that they subsequently developed ECs. (B) Embryos treated from 24 to 48 hpf with 2,3-BDM to decrease myocardial force failed to develop endocardial rings in a dose-dependent manner. Loss of ring formation was not linked to loss of blood flow—14% of embryos treated at 4 mM 2,3-BDM did not form rings despite the presence of blood flow, and 58% of embryos treated at 6 mM did form rings despite the absence of blood flow. (C) Example of an embryo treated with 10 mM 2,3-BDM that failed to develop an endocardial ring at the AV boundary (red arrow). cfk Affects EC Development through Its Effect on Myocardial Function Because most cfk −/− embryos have a dilated heart and lack blood flow and because subsequently approximately 50% of these embryos fail to form ECs, we wondered whether these two phenotypes were causally related or whether lack of EC formation could occur independently of poor early myocardial function. To address this question, 379 embryos from multiple cfk clutches were assayed for cardiac dilation and lack of blood flow at 36 hpf and for lack of endocardial ring formation at 48 hpf. Embryos that were phenotypically wild-type at 36 hpf invariably developed endocardial rings over the next 12 h, whereas all embryos that failed to form endocardial rings had previously shown a functional phenotype at 36 hpf (see Figure 5 A). Importantly, the few cfk −/− embryos that were phenotypically wild-type at 36 hpf remained so at 48 hpf. These data strongly suggest that lack of EC formation occurs secondarily to poor myocardial function or lack of blood flow at 36 hpf. Myocardial Function, Not Shear Stress, Is Likely Required for EC Formation Because cfk −/− and sih −/− embryos each exhibit both a myocardial phenotype (dilation and silence, respectively) and fail to generate blood flow, it is impossible to conclusively state whether EC formation is affected in these embryos directly as a result of poor myocardial function or indirectly as a result of the perturbation in blood flow and shear stress caused by poor myocardial function. Hove et al. (2003 ) attempted to address a similar question by inhibiting blood flow mechanically without affecting myocardial function. However, other aspects of cardiogenesis were disturbed in their experiments, leaving open the question of which effects of their manipulations were primary and which were secondary. We chose an alternate approach to analyze the respective roles of myocardial function and blood flow in EC development by finding doses of an inhibitor of myofibril function that would or would not affect blood flow. We treated tie2::GFP embryos with various concentrations of 2,3-butanedione monoxime (2,3-BDM), which blocks myofibrillar ATPase in a dose-dependent manner ( Herrmann et al. 1992 ) and decreases myocardial force. As the treatment concentration of 2,3-BDM increased, the percentage of embryos that formed endocardial rings at 48 hpf decreased ( Figure 5 B). Importantly, blood flow was abolished in all embryos treated with 2,3-BDM at 6 mM or higher, yet 58% of them ( n = 74) formed an endocardial ring, indicating that blood flow is not required for the initial steps of cushion formation. When myofibril function was further decreased by treatment with 10 mM 2,3-BDM, the percentage of embryos with an endocardial ring decreased to 13% ( n = 68). Studies with the anesthetic tricaine confirmed the observation that ECs may form in the absence of blood flow and that the likelihood of forming cushions is inversely proportional to the concentration of tricaine (data not shown). In summary, the results from the 2,3-BDM and tricaine treatments suggest that it is poor myocardial function, and not lack of blood flow, which is primarily responsible for the loss of EC formation in cfk −/− and sih −/− embryos. Discussion The hearts of sih −/− embryos fail to beat yet undergo looping morphogenesis and AV boundary specification. The specific absence of EC formation in these embryos clearly demonstrates that myocardial function is required for EC formation. Precisely how myocardial function is required for EC formation is unclear. Prior work has demonstrated a requirement for AV boundary myocardium in the formation of ECs, and several signaling molecules emanating from the myocardium at the AV boundary have been identified (reviewed in Eisenberg and Markwald 1995 ). Although we have shown that the expression of bmp-4 is not affected by myocardial function defects, it is possible that myocardial function is somehow required for another aspect of this signaling event. Hove et al. (2003 ) recently argued for the importance of intracardiac hemodynamics as a key epigenetic factor affecting embryonic cardiogenesis. In their experiments, blood flow into the heart was eliminated by surgical placement of a bead at the inflow tract. One of the resulting phenotypes in these embryos was lack of EC formation, leading them to hypothesize that a reduction in shear stress on endocardial cells caused this phenotype. However, it is possible that the lack of EC formation observed by Hove et al. (2003 ) in surgically manipulated embryos was secondary to either the lack of looping observed in these same embryos, an effect on myocardial function by the surgical manipulations, or a more general arrest of cardiac development. In contrast, sih −/− embryos undergo looping morphogenesis, making them perhaps a more appropriate model to analyze, and these embryos also lack ECs. Therefore, myocardial function and/or blood flow appear to play a role in EC development. Data from 2,3-BDM and tricaine treatments, in which many embryos without blood flow still formed ECs, suggest that endocardial shear stress may not be key to EC morphogenesis. In mouse, a number of studies indicate that mutation of a single gene, including tbx5 ( Bruneau et al. 2001 ), nkx2.5 ( Schott et al. 1998 ; Biben et al. 2000 ), and has2 ( Camenisch et al. 2000 ), affects both myocardial function and cardiac morphogenesis. These mutations have been thought to affect cardiac structural development and function in parallel pathways or in a causal pathway with the abnormal structure of the hearts leading to abnormal function. Based on our analyses of cfk, we would like to propose an alternative interpretation, namely that some of these mutations may primarily perturb early myocardial function, which then disrupts subsequent steps of heart morphogenesis. This model is supported by studies of the ncx1 null mouse, in which the heartbeat is eliminated and the EMT associated with EC development appears to be defective ( Koushik et al. 2001 ). The ncx1 studies, together with ours, indicates that EC development may be particularly sensitive to perturbations in myocardial function. However, because of the difficulty in completely unlinking heart function and blood flow, further studies will be required to elucidate the exact contributions of myocardial function and endothelial shear stress on cushion and valve morphogenesis. Our data show that mutations in two different sarcomeric genes lead to EC defects in zebrafish. Similar mutations affecting myocardial function may thus cause EC defects in humans. While most infants born with EC defects exhibit relatively normal myocardial function, actin-gene expression is known to undergo significant changes during development ( Cox and Buckingham 1992 ). Thus, a mutation in an actin gene expressed in the heart during embryogenesis, when control of actin-gene expression appears to be less specific than in the mature animal, could lead to an EC phenotype without apparent effects on myocardial function at birth. As shown by the cfk mutant, it is possible that genetic lesions in actin genes expressed only transiently in myocardial cells could cause embryonic cardiovascular phenotypes, particularly if those lesions act in a dominant manner. Therefore, as we seek to identify the genes responsible for human cardiac malformations, those primarily involved in myocardial function should also be considered. Materials and Methods Zebrafish Zebrafish were maintained and staged as described ( Westerfield 2000 ). We used the sih tc300b allele ( Chen et al. 1996 ) and the cfk s11 allele identified in a screen in our laboratory ( Alexander et al. 1998 ). Animal protocols were approved by the Committee on Animal Research of the University of California, San Francisco (San Francisco, California, United States). In situ hybridization We carried out whole-mount in situ hybridization as described elsewhere ( Alexander et al. 1998 ). We used antisense RNA probes to bmp-4 ( Nikaido et al. 1997 ) and the 3′ UTR of cfk . Visualization of GFP-positive cells after in situ hybridization was performed by treating embryos with a rabbit anti-GFP-IgG followed by a fluoresceinated mouse anti-rabbit-IgG. Morpholino antisense “knock-down.” We designed morpholino oligonucleotides (Gene Tools, Philomath, Oregon, United States) to bind to the initiation codon and flanking sequences of the following zebrafish genes: sih ( Sehnert et al. 2002 ), rab4a (5′-GTCTCTGACATGACTGACGCTGCGT-3′), abc-b10 (5′-TCATTCGCAACATTGTCCCATACAT-3′), fv49b10.y1 (5′-CCGACCTAATTCGCTTGGT-CACCAT-3′), and cfk (5′-CATCTTGATGTATTCTTTCTCTGCT-3′). The morpholinos were injected at 4 ng and 8 ng into tie2::GFP embryos at the one-cell stage and examined at 24 to 72 hpf for myocardial function and EC phenotypes. Morpholinos against abc-b10 , rab4a , and fv49b10.y1 did not affect myocardial function or EC formation. The cfk morpholino caused no phenotype, an expected result given the ability of actin genes to compensate for each other when down-regulated ( Kumar et al. 1997 ; Crawford et al. 2002 ) and the redundancy of cfk with α-cardiac actin. Genetic mapping We genotyped diploid mutant embryos from a cfk s11 -AB/SJD hybrid strain using SSLP with various CA repeat markers and SSCP with various ESTs. SSCP was performed by denaturing PCR products at 95°C for 10 min in 0.7 mM EDTA and 36 mM NaOH and placing on ice before loading on nondenaturing acrylamide gels. BAC library filters for the CHORI-211 (zC) library were obtained from P. de Jong at BACPAC (Oakland, California, United States). Linkage of the R177H mutation to the cfk s11 allele (through genotyping of mutant embryos) was accomplished by performing PCR across the mutation (forward primer = 5′-ATCGTGCTGGACTCTGGTG-3′ and reverse primer = 5′-GAAAGAATAACCGCGCTCAG-3′) and digesting with NsiI, which cuts only the H177 allele. Synteny analysis was performed through http://genome.jgi-psf.org/fugu3/fugu3.home.html (fugu) and http://genome.ucsc.edu (human). Mutation detection We used a pool of 25 cfk s11 mutant embryos to extract mRNA (Trizol, GIBCO–BRL, Gaithersburg, Maryland, United States) and synthesize cDNA (SuperScript First Strand, Invitrogen, Carlsbad, California, United States). PCR was then performed from the 5′ UTR to 3′ UTR and multiple clones sequenced. To confirm that the amino acid change we saw was a mutation and not a polymorphism in the strain used for mutagenesis, we examined the DNA of four F1 females from the screen ( Alexander et al. 1998 ) in which cfk was identified, one of which led to the cfk line and thus should be mosaic for the mutation. This female, but not her three sisters, was indeed mosaic for the mutation (data not shown). Pharmacological treatment of embryos Clutches of dechorionated embryos were placed in 2,3-BDM (Sigma B-0753, Sigma–Aldrich, St. Louis, Missouri, United States) at 24 hpf at concentrations between 2 and 20 mM. This concentration range has been shown to affect myofibrillar ATPase in a dose-dependent manner, with 25% of activity remaining at 10 mM ( Herrmann et al. 1992 ). 2,3-BDM had a rapid onset of action—at 2 and 4 mM, the embryos had visibly weakened heartbeats within minutes, and at 6 mM or greater the myocardial force was weakened enough to eliminate blood flow almost immediately. During the next 24 h, the embryos were examined periodically to ensure that the pharmacological effect remained constant over time. At 48 hpf the presence or absence of the endocardial ring was assayed. For tricaine treatment, dechorionated embryos were placed in a solution of 0.4 mg/ml ethyl 3-aminobenzoate methanesulfonate salt (Sigma A-5040, MS-222, 886–86-2) from 24 to 48 hpf and assayed for myocardial function during that time and for EC formation at 48 to 54 hpf. Yeast-actin biochemistry Site-directed mutagenesis of the yeast-actin coding sequence in a centromeric plasmid and construction of haploid cells producing only the mutant actin were carried out as described previously ( Cook et al. 1992 ). Wild-type and mutant actins were purified using a DNase I agarose/DEAE-cellulose-based procedure as described previously ( Cook et al. 1991 ). G-actin was stored at 4°C in G-buffer (10 mM Tris–HCl [pH 7.5], containing 0.2 mM ATP, 0.2 mM CaCl 2 , and 0.1 mM DTT) and used within 2 d. Actin polymerization was induced by the addition of MgCl 2 and KCl to final concentrations of 2 mM and 50 mM, respectively, and polymerization was assessed by the increase in light scattering as a function of time at 25°C in a 120 μl volume in a thermostatted cuvette using a SPEX Fluorolog 3 fluorimeter with excitation and emission wavelengths set at 360 nm. Supporting Information Video S1 Phenotype of Wild-Type and cfk −/− Embryos at 72 hpf Wild-type hearts propel blood through the vasculature, whereas cfk −/− hearts have complete regurgitation of blood from the ventricle to the atrium and therefore no forward blood flow. (2.91 MB MOV). Click here for additional data file. Video S2 Phenotype of Wild-Type and cfk −/− Embryos at 36 hpf Wild-type embryos have a narrow, strongly contracting heart and generate effective blood flow. cfk −/− embryos (two different embryos shown) have dilated hearts with weak contractions and are not capable of generating circulation. (2.9 MB MOV). Click here for additional data file. Accession Numbers The GenBank ( http://www.ncbi.nlm.nih.gov/Genbank/index.html) accession numbers discussed in this paper are for cfk (AY222742), zebrafish α-cardiac actin (AF116824), zebrafish α-skeletal actin (AF180887), and zebrafish cytoplasmic actin (AF057040).	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC406391.xml
548940	Exploring the equity of GP practice prescribing rates for selected coronary heart disease drugs: a multiple regression analysis with proxies of healthcare need	Background There is a small, but growing body of literature highlighting inequities in GP practice prescribing rates for many drug therapies. The aim of this paper is to further explore the equity of prescribing for five major CHD drug groups and to explain the amount of variation in GP practice prescribing rates that can be explained by a range of healthcare needs indicators (HCNIs). Methods The study involved a cross-sectional secondary analysis in four primary care trusts (PCTs 1–4) in the North West of England, including 132 GP practices. Prescribing rates (average daily quantities per registered patient aged over 35 years) and HCNIs were developed for all GP practices. Analysis was undertaken using multiple linear regression. Results Between 22–25% of the variation in prescribing rates for statins, beta-blockers and bendrofluazide was explained in the multiple regression models. Slightly more variation was explained for ACE inhibitors (31.6%) and considerably more for aspirin (51.2%). Prescribing rates were positively associated with CHD hospital diagnoses and procedures for all drug groups other than ACE inhibitors. The proportion of patients aged 55–74 years was positively related to all prescribing rates other than aspirin, where they were positively related to the proportion of patients aged >75 years. However, prescribing rates for statins and ACE inhibitors were negatively associated with the proportion of patients aged >75 years in addition to the proportion of patients from minority ethnic groups. Prescribing rates for aspirin, bendrofluazide and all CHD drugs combined were negatively associated with deprivation. Conclusion Although around 25–50% of the variation in prescribing rates was explained by HCNIs, this varied markedly between PCTs and drug groups. Prescribing rates were generally characterised by both positive and negative associations with HCNIs, suggesting possible inequities in prescribing rates on the basis of ethnicity, deprivation and the proportion of patients aged over 75 years (for statins and ACE inhibitors, but not for aspirin).	Background The aim of this paper is to provide data on the equity of general medical practitioner (GP) prescribing rates for coronary heart disease (CHD) drugs. Since CHD represents a major cause of premature mortality in the Western world, it is vital that those populations with the highest need for CHD drugs actually receive them. Whilst there is a large literature on inequities in the provision of a number of other health care services and treatments, the equity of GP practice prescribing has received little attention. Therefore, this study was an attempt initiate the development of an evidence-base and to provide data on the equity of GP practice prescribing rates. Conceptualisation and definition of the equity of prescribing There are large literatures around how to define, operationalise and measure equity in relation to health care services [ 1 - 3 ], although equity is generally taken to mean 'fair' or 'just'. Equity has been divided into three domains: equal access to health care for people in equal need; equal treatment for people in equal need; and equal outcomes for people in equal need [ 1 ]. Whilst this is a simplification of the nature of equity, it is useful in delineating the various domains in which inequities may arise. The current paper is focussed around the equal prescribing (i.e. equal treatment ) for people in equal need. Using the example of the current study, an analysis of equity would assess the differences in prescribing rates provided to the population of one GP practice compared to another GP practice, weighted to take account of the levels of need for CHD drugs in their patient populations. Therefore, it would be equitable to have higher prescribing rates for populations with higher levels of health care need and lower prescribing rates for populations with lower levels of health care need. However, it would be inequitable to have higher prescribing rates for populations with lower levels of health care need and lower prescribing rates for populations with higher levels of health care need. The identification of populations where prescribing was deemed inequitable could then be targeted for further resources aimed at redressing the balance between prescribing rates and health care need. Previous research on the equity of GP practice prescribing A recent paper by the authors questioned the equity of GP practice prescribing rates for a range of CHD drugs[ 4 ] and highlighted the contemporary relevance of the 'inverse care law'[ 5 ] in the context of GP prescribing. That paper presented the findings of bivariate correlations between prescribing rates and healthcare needs indicators (HCNIs). One of the inherent problems with bivariate analysis is that prescribing rates are likely to be associated with a number of HCNIs. Therefore, the purpose of this paper is to present the findings of multivariate regression analyses between prescribing rates and the HCNIs and ultimately to examine the independent associations between prescribing rates and HCNIs. In doing so, this paper sets a benchmark for future studies aimed at assessing the effectiveness of the National Service Framework for CHD in developing CHD services commensurate with healthcare need [ 6 ]. There is a growing body of research which has highlighted large variations in overall prescribing rates between GP practices, which are only partially explained by factors other than health care need [ 4 , 7 - 11 ]. Statin prescribing has been shown to vary between health authorities and GPs [ 12 - 15 ] and between patients on the basis of gender [ 13 , 16 - 18 ], demographics [ 13 , 19 ], ethnicity [ 20 ] and deprivation [ 21 ]. Prescribing rates of beta-blockers have also been found to be lower in patients aged over 75 years and in minority ethnic groups [ 22 , 23 ]. Studies attempting to 'explain' the variation in statin prescribing rates have been modest, with most studies explaining around 20 per cent of the variation [ 13 , 15 , 21 , 24 ]. In addition, one study explained around 40 per cent of the variation in prescribing rates for all cardiovascular drugs [ 21 ]. The prevalence of CHD explained 12 per cent of the variation in statin prescribing in men, and 7 per cent in women [ 13 ], deprivation explained 14 per cent [ 15 ], and a combination of nitrate prescribing rates and population aged between 35 and 74 years explained 18 per cent [ 24 ]. In addition, the indicative prevalence of CHD was moderately correlated with prescribing rates for statins, and was the most important variable in the multiple regression model [ 21 ]. However, characteristics of GP practices such as their training status, the number of GPs, or their single-handed status (i.e. whether or not they are 'lone' GPs or work in a multi-partner practice) have been found to have no relationship with prescribing rates for statins [ 15 , 24 ]. Therefore, the majority of variations in statin prescribing rates in addition to prescribing rates for other CHD drugs remain unexplained. Context and setting for the study The planning and provision of health care to local populations in England is now the role of primary care trusts (PCTs). Essentially, PCTs are organisations whose main responsibilities are around developing, commissioning and providing services, which are targeted to the needs of local people, and ultimately to improve the health (and reduce health inequalities) of local people [ 25 , 26 ]. PCTs have taken over these responsibilities from health authorities, which no longer exist [ 27 ] and are responsible for spending 75 per cent of the overall NHS budget in England [ 28 ]. This study was undertaken in 4 PCTs in England (referred to as PCT1, PCT2, PCT3, and PCT4 throughout this paper), which included 132 GP practices (PCT1 had 50 GP practices, PCT2 had 24, PCT3 had 31 and PCT4 had 27). In terms of patient populations, we excluded patients aged under 35 years, since prevalence of CHD is particularly low in this age group. In total, there were 353,897 registered patients aged over 35 year across all 4 PCTs. Methods In order to analyse the equity of prescribing for CHD drugs, we firstly needed to gather data on, and then develop rates for both GP practice prescribing and health care need (called health care needs indicators (HCNIs) in this paper). The data sources and methods used to develop prescribing rates and HCNIs have been outlined in previous papers by the authors [ 4 , 29 , 30 ], although a brief précis will be presented here. Local Research Ethics Committee approval was sought and granted for this study. Developing prescribing rates When an NHS prescription is dispensed in primary care, the prescription form (FP10) is sent to the Prescription Pricing Authority (PPA) for processing. The PPA collates these data and provides them to GP practices and PCTs in the form of Prescribing Analysis and Cost (PACT) data. PACT data are available for all GP practices in England, and allow detailed interrogation in terms of drugs prescribed along with their dosages, pack sizes and formulations. For example, for a specific time period, we can collect data on which statins were prescribed by a GP practice in addition to the dosages and pack sizes. This allows for a complex and timely analysis of PACT data. Useful critiques of PACT data can be found elsewhere [ 31 , 32 ]. PACT data were obtained for all GP practices in the 4 PCTs for the 12-month period October 1999 to September 2000. These data were collected for statins, ACE-inhibitors, beta-blockers, aspirin, and bendrofluazide (a full list of drugs obtained are listed in Appendix A). These drug groups were chosen because they represent major drug groups recommended for the prevention (primary and secondary) of coronary heart disease (CHD) in the United Kingdom (UK) [ 6 ]. Using prescribing rates for aspirin was potentially difficult, since it can also be purchased over the counter (OTC) in community pharmacies and therefore may not represent the totality of aspirin use within populations [ 33 , 34 ]. It has also been found that non-prescription aspirin use was higher in men aged under 65 years, and also in more affluent areas [ 35 ]. Therefore, PACT data may underestimate actual aspirin use within the community. Nevertheless, given the importance of aspirin within the management and prevention of CHD [ 17 , 36 , 37 ], it was postulated that prescribing rates for aspirin may reflect CHD prevalence within GP practice populations, at least as well or even better than prescribing rates for the other CHD drugs. The numerator in all prescribing rates was based on a measure of prescription volume, as opposed to prescription cost. The validity of using the number of prescription items or total cost as a measure of prescribing volume has been called into question [ 38 , 39 ] since it does not specify the quantity of prescription medication (e.g. number and/or dosage of tablets). Therefore, a measure of prescription volume which calculates the total number of grams prescribed is much more useful. The main options available are defined daily doses (DDDs) [ 40 , 41 ] and Average Daily Quantities (ADQs) [ 40 , 42 , 43 ]. The Prescribing Support Unit website provides up to date lists of DDDs and ADQs for all drugs for which they have been developed. Within this study, total ADQs were used as the unit of analysis since they represent prescribing practices in the UK, as opposed to DDDs which represent prescribing practices internationally. The denominator was the total registered (and resident) patient population aged over 35 years. This age group was chosen since the prevalence of CHD is particularly low in people aged less than 35 years [ 44 ]. Developing health care needs indicators (HCNIs) In total, 24 HCNIs were developed for each GP practice in this study, and all of these were entered into multiple regression models (all HCNIs are outlined in Appendix B). These HCNIs reflect patient demographics, ethnicity, socio-economic status, long term limiting illness, CHD mortality and CHD morbidity. However, the only HCNIs discussed here are those included in the final regression models (see Table 1 ). Table 1 Descriptions of statistically significant variables in final regression models Descriptor used in text and Table 4 Description CHD HES rate 6-year crude rate of CHD hospital procedures per 1000 patients (i.e. proxy for CHD morbidity). Data source was the General Practice Research Database. % patients aged 55–74 Proportion of patients in GP practice population aged 55–74 years. Data source was the individual GP practices. % patients aged >75 Proportion of patients in GP practice population aged over 75 years. Data source was the individual GP practices. LISI score Low Income Scheme Index (LISI) score which is the proportion of prescriptions in GP practices which were exempt from payment due to low income (i.e. proxy for deprivation). Data source was the individual GP practices. Ethnicity Proportion of patients in GP practice population defined as South Asian. Data source was the 1991 Census. In order to assess the equity of prescribing for CHD drugs, we needed to develop a set of variables which measured (or acted as proxies for) the health care need in GP practice populations. In terms of coronary heart disease (CHD), there are a number of identifiable risk-factors, all of which increase the risk of CHD morbidity and/or mortality. However, readily available data are not available for a number of these risk-factors, and therefore could not be explored within this study. Nevertheless, data were available on the following risk-factors: age, gender ethnicity, socio-economic status. Evidence on the importance of these risk-factors for assessing health care need is presented below. Rates of CHD related morbidity and mortality increase dramatically with age [ 45 , 46 ]. For example, the prevalence of angina in England was 1 per cent in people aged 16–44 years, 3 per cent in those aged 45–54 years, although this increased to 10 per cent in those aged 55–64 years, 16 per cent in those aged 65–74 years and 18 per cent in those aged over 75 years [ 47 ]. Rates of CHD mortality and morbidity in the UK are generally seen to be higher for people from South-Asian groups, than from white-English people, and much lower for people born in the Caribbean. It has been suggested that the mortality rate for South Asians is 50 per cent higher than for the general population [ 48 ]. In the UK, the standardised mortality rates for CHD in South-Asian men was 146, compared to 100 for all men and just 46 for men born in the Caribbean [ 49 ]. It is also recognised that there are differences within South Asian groups, especially between Indians, who generally experience better health, and Pakistanis or Bangladeshis, who have generally worse health [ 50 - 53 ]. However, reliable data on CHD morbidity or mortality down to these more complex ethnic groupings are not readily available, and any estimates are, at best, imprecise [ 50 ]. Nevertheless, data on the ethnic minority profiles of GP practice populations may prove very useful in determining the need for CHD drugs. There is an extensive literature on socio-economic inequalities in CHD mortality and morbidity [ 54 - 57 ]. Whilst rates of CHD have been declining in the UK for almost 20 years, they have not been falling as fast as countries such as Australia and the United States [ 46 ]. Declining CHD mortality rates are only partially explained by reductions in established cardiovascular risk factors [ 58 - 60 ] and it is possible that general social and economic improvement over time has contributed to this trend [ 61 ]. However, it is noteworthy that these benefits have not been observed by all of the socio-economic groups within the UK [ 62 ]. For example, deaths from CHD in men in the highest social class have halved in the past 20 years, but remain almost unchanged among men in the lowest social class [ 60 ]. Between 1986 and 1992, people from the highest social classes had a rate of 160 CHD deaths per 100,000 population, whereas the rate for the lowest social classes was 266 per 100,000 [ 49 ]. In addition, a 22-year follow-up study on the relationship between socioeconomic status and CHD in middle-aged men [ 63 ] found that irrespective of length of follow-up, lower social classes had a clearly increased risk of fatal CHD – after 8 years they had a 69 per cent increased risk, which dropped to 67 per cent after 15 years and 59 per cent after 22 years. Therefore, socio-economic status is significantly related to CHD mortality and morbidity, and as such, represents an important indicator of health care need for CHD drugs. Overall, the age, ethnicity and socio-economic status are all important factors in shaping the epidemiology of CHD in the UK, and as such, are important variables for use in developing the HCNIs in this study. The actual development of HCNIs relating to these risk factors are outlined below, in addition to additional HCNIs related more directly to CHD morbidity and mortality. Some HCNIs were collected directly from GP practice lists (proportion of patients aged over 75 years and proportion of patients aged 55–74 years), some have been calculated for GP practices (Low Income Scheme Index (LISI) score) [ 64 ], and others were calculated from data directly attributable to GP practices (data based on hospital episode statistics). However, due to the lack of information available from GP practices on variables such as socio-economic status and ethnicity (in this study, this refers to South Asian groups) of patients, a number of HCNIs were developed which estimated this from data such as the 1991 Census and Local Authority statistics. The method of patient weighted attribution was used to develop these estimates using data at enumeration district level (small geographical areas), whereby the postcodes of patients were linked to enumeration districts. Data on the enumeration districts of all registered patients were then aggregated for each GP practice and divided by the total registered population in order to provide a patient-weighted score. In this way, data from the Census were directly applied to GP practice populations. Further information and critiques of this method can be found elsewhere [ 65 - 67 ]. Whilst 1991 Census data may be regarded as rather old now (and has since been superseded by 2001 Census data), these were the only data available at the time of the study, since GP practices do not routinely collect data on the ethnicity of patients. Nevertheless, the HCNI relating to ethnicity may be seen with caution in this paper. Demographic HCNIs were developed directly from GP practice list data, and these relate to the proportion of patients aged 55–74 years, and the proportion aged over 75 years. Both of these demographic groups are indicators of health care need for CHD drugs [ 44 ]. The Low Income Scheme Index (LISI) was used as a proxy for low income since it represents the proportion of prescriptions which are exempt from prescription charges due to low income [ 64 ]. The proportion of patients from South Asian groups was estimated for each GP practice using data from the 1991 census. Data were also obtained from hospital episode statistics (HES) on specific hospital procedures (coronary artery bypass graft (CABG), percutaneous transluminal angioplasty (PTCA), and coronary angiogram) and diagnoses (primary diagnosis of CHD at discharge). Although HES relate to the supply, as opposed to need for health care services [ 68 ], it was hypothesised that in the absence of other CHD morbidity data, they may represent a useful proxy of CHD morbidity in GP practice populations. Due to the low numbers of procedures and diagnoses within a GP practice population, data were aggregated for 6 years (1995 to 2000 inclusive). Crude rates (per 1000 patients aged over 35 years) were calculated for CHD HES, which represents the CHD procedures + CHD diagnoses. As outlined in Appendix B, a number of other variables relating to CHD mortality and morbidity were entered into the regression models, although these were not statistically significant. Indeed, as outlined in a previous paper by the authors [ 4 ], bivariate analyses revealed that associations between prescribing rates and standardised mortality rates for CHD were often not statistically significant and in some cases had a negative association. Therefore, prescribing rates and CHD mortality rates do not exhibit a strong relationship, adding to the suggestion of an inequity in prescribing rates. Data analysis Multiple linear regression modelling was undertaken for each drug group in each PCT in addition to the combined dataset. The dependent variable in each model was the respective prescribing rate and the independent variables were all 24 HCNIs developed in the study. The forward-stepwise approach was taken but not slavishly adhered to since considerations of model coherence and plausibility were taken into account when seeking the most parsimonious models. The final regression models only contained the HCNIs which had statistically significant (p < 0.05) independent associations with the dependent variable, and each model was checked for collinearity and normality of residuals. Overall, for each multiple regression model, all 24 HCNIs were entered as independent variables, and the final model included only those variables that were statistically significant and added to the fit of the model. Results Descriptive findings will be presented first, in order to contextualise the findings from the multiple regression analyses. Variations in prescribing rates and health care needs indicators (HCNIs) Table 2 provides details of variations between primary care trusts (PCTs) in prescribing rates for each of the study drug groups. PCT1 generally had the highest median prescribing rates, followed by PCT2, with the lowest prescribing rates occurring in PCT3 and PCT4. The median ADQs per patient over 35 years for all study CHD drugs combined were 105 in PCT1, 93 in PCT2, 90 in PCT3, and 84 in PCT4. Table 2 Variation in prescribing rates by primary care trust (PCT) Drug group PCT1 PCT2 PCT3 PCT4 Statins Min 11.05 9.56 6.86 6.44 Max 47.23 32.12 46.60 45.76 Median 22.88 19.65 18.08 17.36 IQR 8.20 11.20 8.40 7.24 Aspirin Min 14.62 15.34 4.94 9.6 Max 44.97 46.40 46.63 45.65 Median 29.27 27.35 25.33 30.87 IQR 7.40 16.5 7.60 18.85 Beta-blockers Min 5.04 6.38 3.94 3.43 Max 23.92 29.04 23.91 26.19 Median 12.84 12.38 12.32 12.64 IQR 4.61 7.70 7.41 10.11 ACE inhibitors Min 15.47 9.85 4.20 4.69 Max 48.40 39.25 32.97 35.69 Median 26.43 19.76 16.59 19.19 IQR 8.82 11.17 9.37 11.29 Bendrofluazide Min 4.19 1.63 1.63 0.60 Max 35.61 29.55 32.90 22.91 Median 14.42 10.85 11.04 9.01 IQR 6.98 8.15 12.70 7.28 All study CHD drugs Min 71.70 57.28 25.89 24.76 Max 156.82 154.21 136.64 150.68 Median 105.37 92.78 90.33 83.65 IQR 28.25 57.07 32.08 51.18 PCT1 had the highest median prescribing rates for all drugs except aspirin, where PCT4 had the highest median rate. PCT3 had the lowest median prescribing rates for aspirin, beta-blockers and ACE inhibitors, and PCT4 had the lowest median prescribing rates for statins, bendrofluazide, and all study CHD drugs combined. The difference between the median prescribing rates for PCT1 and PCT4 for bendrofluazide, statins and all CHD drugs was around 4, 5, and 21 ADQs per patient over 35 years. Therefore, an average GP practice in PCT1 with 3000 registered patients over 35 years, prescribed 12000 more ADQs for bendrofluazide, 15000 more ADQs for statins, and 63000 more ADQs for all study CHD drugs. In addition, a similar GP practice in PCT1 also prescribed almost 30000 more ADQs for ACE inhibitors than a comparably sized GP practice in PCT3. An initial overview of the health care needs of the PCTs suggests that PCT4 had the highest levels of CHD related health care needs within the study, whereas PCT1 had the lowest health care needs. PCT4 was the most deprived of all PCTs, had the highest proportions of patients aged over 75 years, and the highest median score for standardised mortality rates (SMR) for CHD. In contrast, PCT1 may be seen as the 'least needy' of all PCTs on the basis of the HCNIs developed in this study. PCT1 was the least deprived, had the lowest proportions of South Asian groups, had the lowest median SMR and the lowest median rate of CHD hospital procedures. These overall descriptive findings are indicative of inequitable prescribing rates, whereby PCT1 had high prescribing rates and low comparative health care needs, whereas PCT4 had low prescribing rates and high comparative health care needs. However, these are purely descriptive and do not take into account the multiple risk factors for CHD or the interactions between HCNIs. Therefore, findings from the multiple regression analyses are presented below. Multiple regression analysis Table 3 presents the percentage of variation explained in each of the models (i.e. R 2 * 100), Table 1 presents a description of the HCNIs included in the final regression models and Table 4 presents details for the combined dataset in terms of the HCNIs included in each model, their standardised beta coefficients and their contribution (in terms of percentage variation explained) to the total R 2 of the model. Table 3 Percentage of variation in prescribing rates explained by HCNIs PCT1 PCT2 PCT3 PCT4 Combined Statins 24.5 58.3 31.3 50.5 24.9 Aspirin 44.2 88.2 62.3 51.7 51.2 ACE inhibitors No Model 55.3 26.6 45.8 31.6 Beta-blockers 28.3 42.4 46.1 41.1 27.1 Bendrofluazide 15.6 53.9 40 54.5 24.1 Table 4 Regression models for combined dataset Drug group R 2 Needs indicator Beta coefficient % explained Statins .249 CHD HES rate .350 14.7 % patients aged >75 -.240 4.2 Ethnicity -.233 3.1 % patients aged 55–74 .199 2.9 Aspirin .512 CHD HES rate .579 32.9 LISI score -.261 10.4 % patients aged >75 .347 8.2 ACE inhibitors .316 Ethnicity -.395 25.3 % patients aged >75 -.170 3.4 % patients aged 55–74 .248 2.9 Beta-blockers .271 CHD HES rate .411 19.7 % patients aged 55–74 .274 7.4 Bendrofluazide .241 LISI score -.261 13.1 CHD HES rate .171 6.6 % patients aged 55–74 .274 4.4 All CHD drugs .454 LISI score -.303 22.7 CHD HES rate .461 20.0 % patients aged 55–74 .242 2.7 Regression models generally explained much more variation in prescribing rates in PCT2 than any other PCT. For example, in PCT2, modelling explained around 55 per cent of the variation in prescribing rates for ACE inhibitors and bendrofluazide, almost 60 per cent for statins and almost 90 per cent of variation in aspirin prescribing rates. Multiple regression models explained around 25 to 55 per cent of the variation in prescribing rates in the combined dataset, around 25 to 60 per cent in PCT3, and around 40 to 55 per cent in PCT4. PCT1 had the lowest R 2 values in general, and no model was derived for ACE inhibitors. Given the reduced explanatory power of the regression models in PCT1, when regression modelling was undertaken for the combined dataset, the R 2 values were much lower than those for PCT2, and generally lower than those for both PCT3 and PCT4. In general, regression modelling explained more variation in prescribing rates for aspirin than for the other drug groups. Not including PCT1, regression modelling explained between 50 and 90 per cent of the variation in aspirin prescribing rates. Comparative percentages were between 30 and 60 per cent for statins, 30 and 45 per cent for beta-blockers, and 25 and 55 per cent for ACE inhibitors and bendrofluazide. Table 4 presents details of the regression models derived for each drug group in the combined dataset. Many of the models in Table 4 are quite similar in terms of the HCNIs included within them, with all models (except for ACE inhibitors) including the rate of CHD HES (positive association) and all models (except for aspirin) including the proportion of patients aged between 55 and 74 years (positive association). In addition, the LISI score was included in the models for aspirin, bendrofluazide and all CHD drugs (negative association). The models for statins, aspirin and ACE inhibitors also included the proportion of patients aged over 75 years although this indicator had a negative association with prescribing rates for statins and ACE inhibitors, and a positive association with prescribing rates for aspirin. This suggests that GP practices with higher proportions of patients aged over 75 years have lower prescribing rates for statins and ACE inhibitors, and higher prescribing rates for aspirin. The CHD HES rate varied considerably between models, in terms of the amount of variation explained. For example, it explained 33 per cent of the variation in aspirin prescribing rates, 20 per cent of the variation in prescribing rates for all CHD drugs and beta-blockers, 15 per cent of the variation in statin prescribing rates, and 7 per cent of the variation in prescribing rates for bendrofluazide. The LISI score explained 23 per cent of the variation in prescribing rates for all CHD drugs, 13 for bendrofluazide and 10 per cent for aspirin. In addition, the proportion of patients aged between 55 and 74 years explained between 2.5 and 7 per cent of the variation for statins, ACE inhibitors, beta-blockers, bendrofluazide, and all CHD drugs. Discussion The rate of CHD hospital procedures (CHD HES) explained some of the variation in prescribing rates and therefore, prescribing rates in these instances were positively related to health care need (i.e. equitable). These findings concur with other studies whereby positive relationships were found between rates of PTCAs and both aspirin prescribing [ 36 ] and statin prescribing [ 36 , 69 , 70 ]. These may suggest that prescribing rates are positively related to healthcare need, although as already outlined, HES may actually reflect supply or demand, rather than need per se. Nevertheless, within this study, HES were the best data available as proxies for CHD in the GP practices, and therefore, we suggest that these prescribing rates reflect CHD prevalence, and hence healthcare need. GP practices in the UK are now developing CHD registers, which may be used in the future as proxies of CHD prevalence, although again, these will only reflect treated CHD, as opposed to the true prevalence of CHD in the community. The proportion of patients aged 55–74 years was generally positively related to prescribing rates, suggesting that prescribing rates are related to health care need, since people in this age group have a higher prevalence of CHD [ 49 ]. A number of studies have found that statin prescribing is higher in this age group, than in older age groups [ 13 , 16 , 17 , 70 ], with one study finding that the proportion of patients aged 35–74 years explained just 5 per cent of the variation in statin prescribing rates between GP practices [ 24 ]. Whilst the previous finding is not unexpected given the higher CHD prevalence in this age group, the negative relationship between the proportion of patients aged over 75 years and some prescribing rates should be of great interest. The proportion of patients aged over 75 years had a negative relationship with statins and ACE inhibitors and a positive association with aspirin. A number of studies found that older patients were much less likely than younger patients to undergo a CABG or a PTCA[ 71 , 72 ] or to receive a prescription for a statin [ 13 , 16 , 17 , 70 ], which may result from the lack of research evidence on the efficacy of statins in elderly populations or judgements about likelihood of benefit based on age. Therefore, the negative associations between the proportion of patients aged over 75 years and rates of ACE inhibitor and statin prescribing are in line with other seemingly inequitable health care. Although the proportion of patients aged over 75 years is a proxy for CHD prevalence, it may not represent a useful proxy of the potential to benefit from statins. However, it may be suggested that prescribing rates for aspirin are higher in this age group since they are more cost-effective than statins [ 73 ]. Therefore, these drugs may be more readily prescribed to patients for whom the benefits (both clinical and cost-effectiveness) of statins have not been studied. Conclusion Overall, this study has found that prescribing rates may be explained (to differing degrees between primary care trusts (PCTs) and between drug groups) by a mixture of health care need indicators (HCNIs). Whilst prescribing rates were generally positively related to the rates of hospital procedures and diagnoses, they were also negatively associated with proxies of deprivation and ethnicity and with the proportion of patients aged over 75 years. However, the relatively low R 2 values reveal a large amount of unexplained variation in prescribing rates. Indeed, even with the models for aspirin and all CHD drugs, there is still around 50 per cent of unexplained variation in prescribing rates. This study cannot be used to infer inequitable prescribing by GPs, since the lower prescribing rates in areas with high proportions of South Asian and deprived groups may be due to lower utilisation of primary health care services due to social, economic or cultural barriers [ 74 , 75 ]. Therefore, further work needs to be undertaken in areas of deprivation and with high proportions of minority ethnic groups in order to understand the reasons for the low prescribing rates and ultimately to make CHD prescribing commensurate with healthcare need. In addition, the HCNIs used in this study are proxies for healthcare need, and may not reflect actual healthcare need. Therefore, future studies may attempt to verify our findings by using either morbidity or mortality data gathered from GP practices. Whilst the clinical and epidemiological data on these CHD drugs has allowed for the development of evidence-based guidelines and evidence-based prescribing, this paper suggests that in practice, actual prescribing rates may not be related to health care need. Further research needs to concentrate on verifying or falsifying these claims on a more micro-level analysis (eg clinical audit in specific GP practices identified in the study) and on exploring the reasons why such a relationship exists (e.g. qualitative studies with GPs, practice nurses and patients in the identified GP practices). In addition, more work is required to understand the differences between PCTs in terms of the explanatory power of the regression models (i.e. much more of the variation in prescribing rates were explained in PCT2 than any other PCT). Such a strategy may enable educational tools to be developed which would facilitate more evidence-based prescribing, but may also identify particular patient groups who do not present symptoms of CHD (ie unmet need) and therefore may require educational outreach or targeted screening in order to increase their consultations and ultimately prescribing to these groups. Although we have focussed on drugs used in the prevention of CHD, a similar approach may be taken across different therapeutic groups of drugs, in different health care settings and in different countries in order to generate more rounded, grounded and extensive evidence on the equity of prescribing in general. Appendix A – List of drugs used in this study • Aspirin (75 mg) • Bendrofluazide 1 (2.5 mg) • Statins (Atorvastatin, Cerivastatin, Fluvastatin, Pravastatin, Simvastatin) • ACE inhibitors 2 (Captopril, Enalapril, Lisinopril, Ramipril, Trandolapril) • Beta-blockers 3 (Atenolol, Co-tenidone 4 ) 1 In some countries, this may also be known as (among other names) Neo-NaClex, Bendroflumethiazide, Aprinox, Berkozide, Naturetin, Pluryl, Polidiuril, Salural, Urizde. 2 The 5 ACE inhibitors represent the majority of prescribing for all ACE inhibitors 3 Atenolol represents the majority of prescribing of all beta-blockers 4 Co-tenidone is a combination product containing both a beta-blocker (atenolol) and a diuretic (chlorthalidone). Appendix B – List of health care needs indicators (HCNIs) developed during the study • Proportion of patients aged between 55 and 74 years • Proportion of patents aged over 75 years • Proportion households with no car • Proportion males who are economically inactive • Townsend Score • Proportion of households receiving council tax benefits • Proportion unemployment • Index of Multiple Deprivation • Income Deprivation Index • Low Income Scheme Index (LISI) score • Standardise mortality rate (SMR) for CHD under 75 years • 6-year crude rate of coronary artery bypass grafts (CABGs) per 1000 patients • 6-year crude rate of percutanious transluminal angioplasty (PTCAs) per 1000 patients • 6-year crude rate of coronary angiograms per 1000 patients • 6-year crude rate of CHD hospital procedures (CABGs + PTCAs + angiograms) per 1000 patients • 6-year crude rate of CHD hospital diagnoses per 1000 patients • 6-year crude rate of CHD prevalence (diagnoses + procedures) per 1000 patients • Regionally specific prevalence, age and sex standardised prescribing units per patient over 35 years (PASS-PUs) • Proportion of population defining themselves as 'non-white' • Proportion of population defining themselves as 'South Asian' • Proportion of population over 30 with a limiting long-term illness (LLI) • Health Deprivation Index • Proportion of households with more than 2 cars • Access Deprivation Index Competing interests The author(s) declare that they have no competing interests. Authors' contributions PW was awarded funding for this study, was involved in the conception, design and managed the day to day running of the study, undertook all data collection and analysis, and wrote the paper. PN and ASL were involved in the conception and design of the study, made active contributions in project meetings, were involved in an advisory capacity in all aspects of the project and advised on drafts of the paper. PW is the guarantor.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC548940.xml
555939	Searching QTL by gene expression: analysis of diabesity	Background Recent developments in sequence databases provide the opportunity to relate the expression pattern of genes to their genomic position, thus creating a transcriptome map. Quantitative trait loci (QTL) are phenotypically-defined chromosomal regions that contribute to allelically variant biological traits, and by overlaying QTL on the transcriptome, the search for candidate genes becomes extremely focused. Results We used our novel data mining tool, ExQuest, to select genes within known diabesity QTL showing enriched expression in primary diabesity affected tissues. We then quantified transcripts in adipose, pancreas, and liver tissue from Tally Ho mice, a multigenic model for Type II diabetes (T2D), and from diabesity-resistant C57BL/6J controls. Analysis of the resulting quantitative PCR data using the Global Pattern Recognition analytical algorithm identified a number of genes whose expression is altered, and thus are novel candidates for diabesity QTL and/or pathways associated with diabesity. Conclusion Transcription-based data mining of genes in QTL-limited intervals followed by efficient quantitative PCR methods is an effective strategy for identifying genes that may contribute to complex pathophysiological processes.	Background Understanding the molecular etiology of disease processes is a pressing goal of 21 st century medicine. Completion of the mouse genome holds considerable promise in the discovery of genes responsible for genetically determined complex diseases. Quantitative trait loci (QTL) are allelically variant regions detected by virtue of their contribution to the overall complex disease phenotype and thus are "experiments in nature", which mark chromosomal intervals carrying genes with a proven disease involvement. Since gene expression is a key link between the genome and the plethora of phenotypic traits exhibited, tools that permit the analysis of the tissue expression pattern of genes in their chromosomal context provides a bridge between QTL and the genes responsible. Global microarray gene expression technologies offer a promising, unbiased approach toward this goal in that they reveal gene expression changes, which can be correlated with the disease phenotype. However, such global methods of analysis are not routine analytical tools and can suffer from incomplete gene coverage, as well as lack of sensitivity. Because only a small fraction of the transcriptome is typically involved in any given etiopathological process, bioinformatic data mining tools that allow for the intelligent prioritization of genes would make it possible to employ more routine and sensitive expression technologies, such as quantitative polymerase chain reaction (QPCR). ExQuest [ 1 , 2 ] organizes pre-existing information-rich expression databases in a way that quantitative tissue and/or developmental gene expression patterns can be extracted and displayed within the context of whole chromosomes. By overlaying ExQuest "chromosomal expression maps" with QTL coordinates, one can search within defined genomic intervals for candidate genes with enhanced expression in tissues consistent with disease pathology. Screening candidates by QPCR will determine whether or not they exhibit expression changes in genetically resistant versus susceptible mice. Type II diabetes (T2D), also referred to as non-insulin dependent diabetes mellitus, is characterized by insulin resistance of target cells combined with insufficient insulin production and/or abnormal secretion by pancreatic β-cells, which eventually results in chronically elevated glucose levels in the circulation. Obesity is a major risk factor for T2D, and the term "diabesity" has been coined to collectively describe these overlapping conditions [ 3 ]. This paper describes an ExQuest compilation of novel candidate genes from mouse diabesity QTL intervals mapped to chromosomes 1, 4, 10, 17, 19 and X. This is followed by QPCR analysis of the ExQuest candidates, in addition to known diabetes and metabolism genes selected from the literature, using tissues derived from the newly defined T2D model, Tally Ho (TH) [ 4 ]. By selecting genes from QTL intervals derived from multiple diabesity models and using the TH mouse as a susceptible strain to test for differential expression, we have identified a number of genes whose altered expression may be involved in the development of diabesity. Results Diabesity-relevant candidate gene selection To identify candidate genes that contribute to diabesity, we first established the local boundaries of 29 QTL from the LocusLink database that contribute to body weight, adiposity or T2D, and map to mouse chromosomes 1, 4, 10,17, 19 and X (Table 1 ). Using ExQuest, we were able to cluster publicly available ESTs to genomic fragments, extract and normalize tissue information from EST datasheet records, and display quantitative expression information linearly along the six chromosomes for over 70 tissues. We then narrowed the chromosomal search by overlaying the in silico expression data with QTL intervals. We were particularly interested in genes with known or potential metabolic function whose expression patterns were biased towards high expression in three diabesity relevant-tissues, pancreas, liver and adipose tissue, for which there was good EST library representation. (Skeletal muscle was not included due to the lack of available mouse EST libraries.) An example is illustrated in Figure 1A , in which a region of genes with expression strongly biased toward pancreatic tissue is found centered over the 1-LOD 95% confidence interval of the TH QTL Tafat [ 4 ]. By using ExQuest's zoom-in capability, resolution down to the expression of individual exons revealed that the expression bias was explained by a single gene, Elastase 2 ( Ela2 ), with a pattern of expression that includes pancreas, stomach and tongue (Fig. 1B ). A second example is illustrated in Figure 1C , in which a cluster of pancreatic expressing genes is localized to two chromosome 19 T2D QTL, T2dm2 [ 5 ] and Tanidd1 [ 4 ]. The expanded chromosomal view revealed that this cluster contained pancreatic lipase ( Pnlip ) and two paralogs, Pnliprp1 and Pnliprp2 . From all QTL intervals, we chose the 71 "best expression" candidates (Table 1 ) out of a possible 4,013 genes as predicted by Ensembl's gene annotation. From PubMed searches, we selected an additional 24 genes considered to be highly relevant to diabesity and/or metabolism, as well as the standard normalizer 18S RNA, bringing our total to 96, a convenient number for QPCR profiling (for gene list and oligonucleotide primer sequences, see Additional file 1 ). Table 1 Genes selected for expression analysis by ExQuest. Chr QTL 1 Genes picked via ExQuest Total Genes in Region Interval Range (Mb) 1 Obq2 Gsta3 73 13.7 Obq7 , Wt6q1 , Insq2 , Insq6 Aox1 , Fn1 , Pecr , Igfbp2 , Plcd4 , Scg2 , IRS-1 , Inpp5d 401 44.1 Nidd6 Qscn6 85 9.4 Obq9 Fmo1 , Fmo3 , Apoa2 117 11.7 Wt6q2 Hsd11b1 63 7.3 4 Bwq1 Decr1 64 12.1 Triglq1 , Bglq4 Ttpa , Bhmt , Baat , Aldo2 40 9.8 Nidd1 , Dob1 Lepr , Dio1 , Scp2 , Faah , Usp1 350 32.4 Tafat Ela3b , Ela2 , Dvl1 319 18.0 10 Insq9 , Igfbp3q2 Ass1 , Ftcd , Col18a1 , Itgb2 , Agpat3 , Ndufs7 , Oaz1 , Pah , Igf-1 , Nr1h4 , Kitl 605 55.1 Bgeq8 Rdh7 , Hsd17b9 257 19.8 17 Obq4 , Wta4 Plg , Acat2 , Actl , Hagh , Igfals , Decr2 , Clps , Tff1 , Tff2 , Tff3 , Apom 554 27.2 Hdl4 Gnmt , Lrg , Sepr 161 10.7 Insq5 Abcg5 69 10.0 19 Iba4 Aldh1a1 , Aldh1a7 , Vldlr , Plce1 190 23.3 Afw8 Pi4k2a , Cpn1 , Elovl3 159 8.8 Nobq2 , Bglq13 Ins1 , Gpam , Facl5 39 8.4 Tanidd1 , T2dm2 Gfra1 , Pnlip , Pnliprp1 , Pnliprp2 40 3.6 X Bw1 Rgn 149 51.5 Bw3 Rab9 278 30.0 1 QTL name, abbreviation: Obesity, Obq ; Body weight, 6 weeks, Wt6q ; Insulin, Insq ; Non-insulin-dependent diabetes mellitus, Nidd ; Body weight, Bwq ; Triglyceride, Triglq ; Body growth late, Blgq ; Dietary obesity, Dob ; Tally ho associated mesenteric fat pad weight, Tafat ; Insulin-like growth factor binding protein 3, Igfbp3q ; Body growth early, Bgeq ; Weight adult, Wta ; High density lipoprotein level, Hdl ; Induction of brown adipocytes, Iba ; Abdominal fat weight, Afw ; New Zealand obese, Nobq ; Tally ho associated non-insulin dependent diabetes mellitus, Tannid ; Type 2 diabetes mellitus, T2dm ; Body weight, Bw . Figure 1 Mirrored tissue-specific expression (red) and absolute EST representation (green) for whole ExQuest chromosomes. Maximum scales set for specific or absolute expression is within the circle at the centromere. (a) Pancreas expression for mouse chromosome 10 showing peak position of the LOD for the QTL Tafat over a pancreas expression peak that contains the gene Ela2 . (b) Quantitative tissue specific expression profile for the gene Ela2 . Results are derived from the number of Ela2 -aligned EST hits normalized for library density. (c) Pancreas expression for mouse chromosome 19. Genes in the Tinidd1 and T2dm2 region are pancreas specific (red) and highly expressed (green). Arrow shows a 20X zoom of 3 individual genes Pnlip (1810018F18Rik), Pnliprp1 and Pnliprp2 , which contribute to the large pancreatic histogram peak. Real-time QPCR profiling The newly-described TH T2D mouse model develops obesity, hyperinsulinemia, hyperlipidemia, and male-limited hyperglycemia [ 4 ]. To test the validity of our candidate gene selection and determine a quantitative molecular signature for this mouse model, we extracted adipose, liver and pancreas RNA from three early diabetic (as verified by glycemic index) 7-week-old male TH mice, as well as three age and sex matched B6 controls. Real-time QPCR was then independently performed for all 96 genes on the three biological replicates, and the results were processed using GPR software (for raw Ct values and GPR Reports, see Additional files 2 , 3 and 4 ). Twenty-one of 96 genes exhibited significant expression changes between TH and control B6 mouse tissues (Table 2 ). Fifteen were among the 71 candidates chosen by ExQuest, while six were from the 24-gene PubMed pool (Table 3 ). Table 2 GPR report of B6 vs. TH gene expression comparisons. 1 Gene # Total hits 2 GPR Score 3 Fold change 4 Adipose Pparg 44 0.92 -14.9 Lep 40 0.83 5.3 Alb1 39 0.81 -9.4 Gsta3 38 0.79 -10.6 Glut2 37 0.77 -28.1 Fmo1 35 0.73 -4.8 Scp2 25 0.52 -2.4 Agpat2 24 0.50 -2.9 Agpat3 19 0.40 -2.1 Pancreas Ela2 39 0.95 -55.8 Pnliprp2 39 0.95 -21.8 Kitl 29 0.71 -4.3 Col18a1 18 0.44 -1.7 Ela3b 16 0.40 -1.8 Liver Tff3 63 1.00 -593.0 Vldlr 58 0.92 15.6 Hsd17b9 57 0.91 9.5 Pparg 56 0.89 -11.5 Igfbp2 46 0.73 -2.7 Kitl 39 0.62 -2.4 Decr2 28 0.44 -1.3 Lpl 25 0.40 -1.2 Pi4k2a 25 0.40 -1.1 1 Complete GPR report and raw CT data in Supplementary Data. 2 Number of qualified normalizer genes against which test gene ΔC T values are statistically significant at p = 0.05. 3 # of total hits / # of qualified normalizers. A GPR score = 0.40, meaning that the test gene ΔC T comparisons differed at p= 0.05 compared with at least 40% of the qualified normalizers, is considered most reliable. 4 Relative to 18S RNA. Table 3 Medline citations to genes showing altered expression in TH mice. Gene name Diabesity-relevant articles General function Genes chosen known to be involved in diabetes with expression change Alb1 5548 Lipid binding and carrier activity Leptin 3170 Hormone activity Lpl 775 Lipid catabolism or fatty acid metabolism Pparg 702 Lipid metabolism and steroid hormone receptor activity Glut2 142 Glucose transport and carbohydrate metabolism Agpat2 5 Phospholipid biosynthesis ExQuest chosen genes with expression change, known or unknown to be involved in diabetes Col18a1 1 7 Potent antiangiogenic protein, structural molecule Vldlr 2 7 Lipid transport Scp2 4 Sterol carrier activity, lipid binding, and steroid biosynthesis Fmo1 3 Electron transport and oxidoreductase activity Pnliprp2 1 Lipid catabolism KitL 0 Signal transduction Igfbp2 0 Insulin-like growth factor binding Ela2 0 Proteolysis and peptidolysis Ela3b 0 Proteolysis and peptidolysis, and cholesterol metabolism Gsta3 0 Biosynthesis of steroid hormones Agpat3 0 Phospholipid metabolism Hsd17b9 0 Steriod biosynthesis and oxidoreductase activity Decr2 0 Peroxisome organization and biogenesis, and oxidoreductase activity Tff3 0 Epithelium healing Pi4k2a 0 Inositol/phosphatidlinositol kinase activity Functional summary of genes with expression changes 3 Lipid related gene activity 8 Hormone related gene activity 6 Inflammation or response to injury 4 Oxidoreductase activity 3 Signal transduction 3 Proteolysis and peptidolysis 2 Carbohydrate related gene activity 1 1 Hsd17b9 medline search: there are 88 articles for the homolog "11-beta hydroxysteroid dehydrogenase", and 157 with "hydroxysteroid dehydrogenase". 2 Vldlr MEDLINE search: 4818 diabesity relevant articles found for Vldl . 3 Genes with dual functions are counted twice. Novel genes potentially implicated in the progression or response to diabetes To assess the novelty of the differentially expressed ExQuest-selected genes in regard to diabesity, we performed an extensive PubMed search for articles associated with both the gene name (or its alternative nomenclature as defined by LocusLink gene name aliases), and diabetes or obesity. The number of articles retrieved were typically one to two orders of magnitude lower for the differentially expressed ExQuest selected genes in comparison to the genes analyzed with known diabesity or metabolism involvement, with many having no literature citations (Table 3 ). QTL-based ExQuest expression mining thus revealed a number of novel diabesity candidate genes with little to no prior association with this disease. Discussion Our results suggest that in silico data mining focused on gene expression in diabesity-affected tissues and limited to intervals containing diabesity-specific QTL (and thus enriched in genes that contribute to diabesity) is an efficient method to identify genes whose expression is altered in T2D-susceptible mice. This analysis not only demonstrated expression alterations in genes known to be associated with the development of diabesity, but also identified a number of novel genes whose expression changes may contribute to the development of diabesity. Expression changes in genes mapping within TH QTL could be considered as candidates for transcriptional polymorphisms contributing to the associated TH QTL. However, genes selected on the basis of other diabesity QTL that showed transcriptional differences in the TH/B6 comparison are more likely minor QTL in the TH model or symptomatic effects of diabesity rather than a major genetic cause of the TH disease. As other diabesity studies commonly observe, there were highly significant changes in adipose tissue expression of Pparg and Lep . This included a ~5-fold increase in leptin expression in TH adipose with no accompanying increase in expression of the leptin receptor ( Lepr ). Since leptin is responsible for the regulation of food intake [ 6 ], this increase is most likely in response to incipient weight gain. In contrast, the observed decrease in the expression of peroxisome proliferator activated receptor gamma ( Pparg ), both in TH adipose tissue and liver, is consistent with a decrease in adipocyte differentiation often observed in obese states [ 7 , 8 ]. In addition, Pparg agonists are generally associated with promoting insulin sensitization in the context of obesity [ 9 ]. Other genes previously associated with diabesity and showing differential expression in TH vs. B6 tissues included a 28.1-fold decrease of glucose transporter 2 ( Glut2 ) in adipose, a 2.9-fold decrease in 1-acylglycerol-3-phosphate O-acyltransferase 2 ( Agpat2 ) in adipose, and a 1.2-fold decrease in lipoprotein lipase ( Lpl ) in liver. Their reduced expression in the TH model might be in response to disease onset. Serum albumin levels are decreased in T2D patients (reviewed in [ 10 ]). The primary source of serum albumin is liver. Albumin synthesis is decreased in both diabetic humans and in rat T2D models[ 11 , 12 ] and in liver cells deprived of insulin [ 13 ]. We failed to detect any alteration in the normally high levels of Alb1 expression in liver, but TH mice showed a 9.4-fold decrease in Alb1 expression in adipose tissue. Whether this fat-specific decrease in Alb1 expression is an early manifestation of subsequent hypoalbuminemia remains to be established. Altered expression of a number of ExQuest-selected genes was also found. TH livers showed a 15.6-fold expression increase compared with B6 for the very low-density lipoprotein receptor ( Vldlr ). As a deficiency in Vldlr has been reported to reduce adipocyte size and obesity in the ob/ob mouse model [ 14 ], this expression alteration may contribute to TH diabesity. The expression of hydroxysteroid (17β) dehydrogenase 9 ( Hsd17b9 ), which localizes to the early body weight QTL Bgeq8 QTL, was elevated 9.5-fold in TH livers. While the function of this gene is not described in the literature, its paralog 11beta-hydroxysteroid dehydrogenase type 1 ( Hsd11b1 ) bidirectionally catalyzes the conversion of cortisol to the inactive metabolite cortisone. Over 150 articles describe an association of Hsd11β1 with diabesity, and an intronic Hsd11b1 polymorphism is associated with obesity and insulin resistance in children [ 15 ]. However, whether the elevated expression of the Hsd17b9 paralog in TH liver is a candidate for Bgeq8 remains to be established. Exocrine secretion of pancreatic lipases are known to hydrolyze triglycerides to free fatty acids in the small intestine. Pancreatic lipase ( Pnlip ) and two paralogs, Pnliprp1 and Pnliprp2 , were expressed preferentially in pancreas and mapped very close to the peak of the LOD values for the overlapping diabesity QTL, T2dm2 and Tannidd1 (Fig. 1b ). T2dm2 and Tannidd1 are responsible for increased insulin levels [ 5 ] and elevated plasma glucose [ 4 ], respectively. Long-term high fat feeding, leading to glucose intolerance, occurs with a simultaneous decrease in mRNA expression of Pnlip [ 16 ]. While no expression alteration in the normally high levels of expression of Pnlip or its closely related Pnliprp1 paralog were found, Pnliprp2 expression was decreased 21.8-fold in TH pancreas when mice were fed a 4% fat diet. By facilitating fat storage, the consequence of which is hyperglycemic diabetes, the selective decrease in Pnliprp2 in TH mice may explain Tanidd1 , and potentially, the genetically overlapping T2dm2 QTL. Trefoil 3 (Tff3 ), which maps to the Obq4 obesity QTL, was the most significantly changed of all genes analyzed. It is quite transcriptionally active in control B6 liver but virtually undetectable in TH liver. Trefoils are small, stable secretory proteins expressed in goblet cells in the gastrointestinal mucosa where they stabilize the mucus layer and promote epithelial healing [ 17 ]. Mice deficient in Tff3 are highly susceptible to colon damage [ 18 ], which is not commonly associated with T2D. While the colon is the primary tissue of expression, Tff3 has also been reported to be expressed in the bile ducts of normal human liver and is upregulated in diseased livers [ 19 ]. However, a potential association of the remarkable reduction of Tff3 expression in TH liver with diabesity remains to be established. While not detectably expressed in the liver or fat tissue in B6 or TH mice, Ela2 was ranked by GPR as the gene most significantly changed in the pancreas, with a 55-fold reduction in TH vs. B6 mRNA (Table 2 ). Originally cloned from the pancreas [ 20 ], Ela2 is located within 75 kb of the peak LOD score for the TH QTL Tafat . ExQuest expression profiling suggests that Ela2 expression is biased strongly towards digestive tissues (pancreas, stomach and tongue) of normal mice (Fig. 1A ). Both whole pancreas and Islet of Langerhans EST libraries show high levels of Ela2 expression, suggesting that the endocrine pancreas actively transcribes it. Ela2 (alias polymorphonuclear neutrophil elastase) encodes a serine protease present in neutrophil endosomal granules, which are known to be important in myelopoiesis (reviewed in [ 21 ]). The substantially decreased expression in TH pancreas may result in secretory granules deficient in this serine protease. Conclusion We have tested the concept that transcription-based ExQuest data mining of genes in QTL-limited intervals is an effective method to identify genes that contribute to the complex genetic disease, diabesity. By limiting the in silico candidate genes to those with expression biased to tissues normally affected by this disease, we have shown that sensitive, high-throughput QPCR methods reveal expression changes in novel genes in the TH model. The search within additional diabesity QTL using this technique may facilitate the identification of a limited number of genes that comprise a 'complete' diabesity molecular phenotype. Moreover, this general approach may be an efficient method to identify genes that contribute to complex pathophysiological processes. Methods QTL analysis Obesity and T2D QTL on mouse chromosomes 1,4,10,17,19, and X, were identified by a LocusLink search [ 22 ]. Information regarding individual QTL boundaries was gathered through the Mouse Genome Informatics site (MGI) [ 23 ]. Simple sequence length polymorphism (SSLP) markers extracted from MGI were queried at the Ensembl website [ 24 ] to determine the exact chromosomal location. In silico expression analysis ExQuest [ 1 , 2 ] is a gene expression program that uses public EST databases to create a comprehensive transcriptome map overlaid with tissue specific expression. In brief, chromosomal sequence was downloaded from Ensembl and stringently masked using the RepeatMasker program. ESTs were clustered to the chromosomes using the MegaBlast algorithm on a 32-node, 64 cpu Beowulf cluster. Tissue or library source information was extracted from EST datasheet records for each 10,000 base pair genomic fragment and normalized based upon library density and EST hit frequency. Whole chromosomal plots display total EST hits for each tissue as well as normalized data, which is a measure of a specific tissue expression level compared to all other tissues available in dbEST. In this way, the expressional bias of a genomic region towards a specific tissue is determined. QTL intervals were overlaid onto the chromosomal plots displaying adipose, liver and pancreas expression and these regions were then scanned for areas that exhibited high tissue specificity. ESTs clustering to genomic regions showing high tissue specificity were linked to Unigene [ 25 ], to determine if the particular EST aligned to a known gene cluster from which mRNA sequence could be extracted for primer design. Tissue procurement, RNA preparation and cDNA synthesis Three 7 week old, male TH mice (nonfasting blood glucose levels 295 to 433 mg/dl) and their respective T2D-resistant, sex- and age-matched C57BL/6J (B6) controls, were maintained on a 4% fat diet [ 4 ]. Following CO 2 asphyxiation, pancreas, adipose, and liver tissues were collected. Each biological replicate sample was processed in parallel. Whole pancreas tissue was collected, placed in 3.5 mL lysis/binding solution from the RNAqueous ® -4PCR Kit (Ambion # 1914), homogenized and stored in an ethanol/dry ice bath until the RNA could be extracted. Approximately 25 mg of adipose and liver tissue were collected and stored in 1 mL of RNALater (Ambion # 7020) until RNA was extracted. For pancreatic RNA extraction, we prepared a 5.7 M CsCl solution containing 0.05 M EDTA, pH 7.0. The solution was made RNase free using RNASecure (Ambion # 7005), conforming to manufacturer's recommendations. 5.0 mL SW 55.1 ultracentrifuge tubes were made RNase free by treatment with RNAZap (Ambion # 9780) for five minutes followed by washing with Nuclease-Free water (Ambion # 9932). 2.0 mL of the 5.7 M CsCl solution was added to each tube. 3.0 mL of pancreatic solution was layered carefully on top. The tubes were centrifuged at 36,000RPM for 16–20 hours at 25°C in a SW 55.1 rotor. The supernatant was carefully removed 1.0 mL at a time, using fresh pipette tips. When approximately 1.0 mL remained, the tubes were quickly inverted and dried with filter paper to prevent residual RNases from contact with the RNA pellet at the bottom of the tube. The RNA was resuspended in 2 portions of 150 μL of RNA resuspension solution (Ambion #7010), conforming to manufacturer's recommendations. Adipose and liver tissues stored in RNAlater were processed using the RNAqueous ® -4PCR Kit, conforming to the manufacturer's recommendations. All RNA was subsequently digested with DNAse in accordance with the aforementioned protocol and analyzed for purity using the Agilent Bioanalyzer 2100. RNA concentrations were determined using a NanoDrop ® ND-1000 Spectrophotometer. The collected RNA was converted into cDNA via MessageSensor™ RT Kit (Ambion # 1745), conforming to manufacturer's recommendations. The adipose tissue cDNA reaction synthesis contained 350 ng of RNA while pancreas and liver samples utilized 1000 ng of RNA. All RNA was stored at -80°C when it was not being processed. Primer design Sequences for the candidate genes were extracted from GenBank and imported into Primer Express software v2.0 from Applied Biosystems, Inc (ABI). All primers were prepared in accordance with universal thermocycling parameters as described for real-time PCR on the ABI 7900HT. The primer sequences were then blasted [ 26 ] to ensure specificity of the primers. Forward and reverse primers (MWG Biotech) were combined in a 96-well master plate at a final concentration of 50 μM. Short amplicons of approximately 75 bp were generated to ensure a high level of sensitivity. Dissociation curves confirmed that only single amplicons were generated. Amplicons were TA cloned (Invitrogen # K204040) and bidirectionally sequenced to confirm sequence identity for instances in which genes exhibited significant expression. For gene names and primer sequences, see Additional file 1 . Real-Time QPCR Each reaction consisted of 5.0 μl of 2X SYBR Green Master Mix (ABI # 4309155), 3.0 μL of dH 2 O, 1.5 μL of 0.5 μM forward and reverse primer solution, and 0.5 μL of cDNA. Pancreas and liver cDNA was diluted 1:10 before addition to the master-mix, while adipose cDNA was diluted 1:3. A 384-well plate format was utilized such that 4 samples × 96 genes were amplified per plate. The plate was sealed with Optical Adhesive Covers (ABI # 4311971) and centrifuged. The samples were assayed on the ABI Prism 7900HT Signal Detection System v2.0 using default conditions, and baseline range values were set from 3 to 10 cycles. The data were then analyzed using Global Pattern Recognition (GPR) analytical software [ 27 , 28 ]. In typical QPCR experiments, the comparative expression of all genes is based on single gene normalizer, whose expression is assumed to be invariant. In contrast, the Global Pattern Recognition (GPR) algorithm employs a global normalization feature in which the expression data from each gene are normalized against that of every other gene, thus eliminating the reliance on single gene normalization. GPR's ranking is based on biological replicate consistency, and is thus not skewed by fold change in the magnitude of expression. For raw CT values and GPR Reports, see Additional files 2 , 3 and 4 . Abbreviations used in this paper QPCR, quantitative polymerase chain reaction; QTL, quantitative trait loci; GPR, global pattern recognition; ExQuest, expressional quantification of ESTs Authors' contributions ACB and DCR conceived of the study and were primarily responsible for its coordination and design. ACB and WIO were responsible for maintenance and execution of software algorithms as well as candidate gene selection. WIO and DJS preformed all tissue extraction and QPCR. CJD and MEM provided supercomputer and database management. JKN provided all mice, animal husbandry, and diabesity technical expertise. ACB, WIO and DCR drafted the manuscript and figures. Supplementary Material Additional File 1 Gene names and primer sequences used for QPCR. Click here for file Additional File 2 Provides adipose tissue raw Ct values and GPR report. Click here for file Additional File 3 Provides liver raw Ct values and GPR report. Click here for file Additional File 4 Provides pancreas raw Ct values and GPR report. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC555939.xml
544834	Inhibition of HIV-1 replication in primary human monocytes by the IκB-αS32/36A repressor of NF-κB	Background The identification of the molecular mechanisms of human immunodeficiency virus type 1, HIV-1, transcriptional regulation is required to develop novel inhibitors of viral replication. NF-κB transacting factors strongly enhance the HIV/SIV expression in both epithelial and lymphoid cells. Controversial results have been reported on the requirement of NF-κB factors in distinct cell reservoirs, such as CD4-positive T lymphocytes and monocytes. We have previously shown that IκB-αS32/36A, a proteolysis-resistant inhibitor of NF-κB, potently inhibits the growth of HIV-1 and SIVmac239 in cell cultures and in the SIV macaque model of AIDS. To further extend these observations, we have generated NL(AD8)IκB-αS32/36A, a macrophage-tropic HIV-1 recombinant strain endowed to express IκB-αS32/36A. Results In this work, we show that infection with NL(AD8)IκB-αS32/36A down-regulated the NF-κB DNA binding activity in cells. NL(AD8)IκB-αS32/36A was also highly attenuated for replication in cultures of human primary monocytes. Conclusions These results point to a major requirement of NF-κB activation for the optimal replication of HIV-1 in monocytes and suggest that agents which interfere with NF-κB activity could counteract HIV-1 infection of monocytes-macrophages in vivo .	Background HIV-1 infection is characterized by a long period of clinical latency followed by the development of acquired immunodeficiency syndrome, AIDS. During latency and when viral replication is being controlled in patients treated with antiretroviral therapy, HIV-1 is present in cellular reservoirs and continues to replicate, with each ensuing round of replication giving rise to escape mutants, which further replenish viral reservoirs [ 1 , 2 ]. This grim picture calls for novel targeted therapies for eradicating virus-infected cells and for preventing new infections. Initial infection in vivo by HIV-1 is thought to occur in CD4-positive, CCR5-positive lymphocytes and monocytes. Accordingly, when HIV-1 envelope protein in its oligomerized g160 form contacts the cell surface receptor a signalling cascade is triggered that results in transcriptional activation of specific gene arrays, such as the inflammatory cytokines IL-1 β, IL-6, IL-8, TNF-α, TGF-β; these cytokines, in turn, function to enhance the transcriptional activity of the proviral long terminal repeat (LTR) promoter [ 3 , 4 ]. This cytokine-driven inflammatory-like setting is mediated molecularly by the NF-κB family of transcription factors [ 5 , 6 ]; thus, it serves to reason that preventing NF-κB activation would attenuate HIV-1 replication. Indeed, the LTR of HIV-1 does contain two tandem NF-κB sites [ 7 ] and three repeated Sp1 sites [ 8 ] upstream of the TATAA box with an additional NF-κB site located in the 5' untranslated region of viral genome [ 9 ]. Both sets of NF-κB sequences enhance HIV-1 transcription in response to various signals [ 9 ]. However, the Sp1 sites and TATAA box can redundantly sustain the Tat-mediated transactivation of the HIV-1 LTR in the absence of NF-κB sites [ 10 ]. It is controversial whether NF-κB cellular factors are required for the HIV-1 replication. Mutant HIV-1 carrying deletions or base-pair substitutions in the NF-κB enhancer in the LTR have been shown to be either competent or incompetent for replication [ 11 - 13 ]. These divergent observations are likely explained by differing cellular contexts, such as primary cells versus immortalized cell lines, and varying levels of cellular activation. IκB inhibitors regulate NF-κB activity [ 14 ]. In response to activating stimuli, IκB proteins become phosphorylated, ubiquinated and degraded by proteasomes. This releases cytoplasmic-sequestered NF-κB to enter the nucleus to activate the transcription of responsive genes [ 14 ]. The mutant IκB-αS32/36A is defective for serine 32- and serine 36-phosphorylation and is resistant to proteolysis. IκB-αS32/36A acts as a potent inhibitor of the NF-κB-dependent gene transcription, including those from the HIV-1 genome [ 15 ]. To verify the requirement of NF-κB in the replication of HIV-1 in primary cells, we previously designed HIV-1 and SIV molecular clones containing the IκB-αS32/36A cDNA positioned into the nef region of the respective viral genome [ 16 , 17 ]. We found that these recombinant viruses were highly attenuated for replication in T cell lines as well as in human and simian PHA-activated peripheral blood mononuclear cells, PBMCs [ 16 , 17 ]. These findings supported an interpretation that in these cellular contexts NF-κB is required for efficient viral replication. We also showed that a recombinant SIV which expressed IκB-αS32/36A inhibitor was also highly replication attenuated in vivo in rhesus macaque [ 17 ]. Here, we have extended our analysis of IκB-αS32/36A function in HIV-1 replication to primary monocytes. We report that a macrophage-tropic derivative of NL4-3 strain that expresses the proteolysis-resistant IκB-αS32/36A inhibitor of NF-κB replicated poorly in cultured primary human monocytes. Results Construction of pNL(AD8)IκB-αS32/36A To generate a macrophage-tropic HIV-1 expressing the IκB-αS32/36A cDNA fused to the FLAG epitope, the CXCR4-tropic envelope of pNLIκB-αS32/36A [ 16 ] was replaced with the CCR5-tropic envelope from pNL(AD8) [ 18 ]. Briefly, the 2.7 Kb EcoR1-BamH1 fragment of pNL(AD8) was religated to the 13.1 Kb EcoR1-BamH1 fragment of pNLIκB-αS32/36A or pNLIκB-antisense, thus generating pNL(AD8)IκB-αS32/36A and pNL(AD8)IκB-antisense, respectively (Fig. 1A ). Both molecular clones are Nef-minus because our cloning strategy deleted the first 39 amino acids from the N terminus of Nef and engineered a translational frameshift into the remaining Nef-encoding codons [ 16 ]. The respective molecular clones were transfected into 293T cells to analyse for the expression of HIV-1 proteins and IκB-αS32/36A polypeptide by immunoblotting (Fig. 1 B, C ). As expected the IκB-αS32/36A-FLAG protein was expressed by pNL(AD8)IκB-αS32/36A (Fig. 1C , lane 4). Figure 1 Genome structure and expression of recombinant pNL(AD8)IκB-αS32/36A and pNL(AD8)IκB-antisense molecular genomes. Panel A shows the structure of pNL(AD8) derivatives that carry the IκB-αS32/36A-FLAG insert into the nef region in sense (pNL(AD8)IκB-αS32/36A) or antisense (pNL(AD8)IκB-antisense) orientations. Panel B shows the immunoblot analysis using hyperimmune AIDS patient serum of total extracts (10 μg) from 293T cells 24 hours after transfection with the indicated viral plasmids (10 μg). Panel C shows the immunoblot analysis using an anti-FLAG monoclonal antibody of total extracts (10 μg) from 293T cells 24 h after transfection with the indicated viral plasmids (10 μg). Inhibition of NF-κB activity by pNL(AD8)IκB-αS32/36A To assess the functional impact of IκB-αS32/36A expressed from the recombinant NL(AD8) genome, 293T cells were transfected individually with pNL(AD8), pNL(AD8)IκB-αS32/36A or pNL(AD8)IκB-antisense, and the respective nuclear extracts were evaluated for NF-κB (Fig. 2A ) and Sp1 DNA binding activity (Fig. 2B ). A significant reduction in NF-κB DNA binding activity was observed upon transfection of pNL(AD8)IκB-αS32/36A (Fig. 2A , lane 5) as compared to the other viral transfections (Fig. 2A , lanes 3,4). The specificity of the IκBαS32/36A-mediated inhibition of NF-κB was verified by the demonstration that Sp1 binding to DNA was unaffected (Fig. 2B ). These results support the interpretation that IκBαS32/36A expressed from the recombinant viral genome functionally inhibited NF-κB activity. Figure 2 Reduced NF-κB DNA binding activity in cells transfected with pNL(AD8)IκB-αS32/36A. Panel A shows the NF-κB binding activity of nuclear extracts (5 μg) from 293 T cells transfected with the indicated viral plasmids (10 μg) or were mock-transfected. Panel C shows the Sp1 binding activity of the same nuclear extracts as in panel A. Binding competitions were performed with 100-fold molar excess of the respective unlabelled oligonucleotide. Attenuation of pNL(AD8)IκB-αS32/36A in primary monocytes We next analyzed the replication properties of the recombinant HIV-1 genomes in cultured human monocytes from different individuals. Based on normalized amounts of input virus, we found that NL(AD8)IκB-αS32/36A was highly attenuated for replication when compared to NL(AD8) and NL(AD8)IκB-antisense (Fig. 3 A-B ). Accordingly, virus-induced syncitium formation was also strongly inhibited in monocytes infected with NL(AD8)IκB-aS32/36A (Fig. 4 A, B ). Taken together, our results underscore a critical contribution of NF-κB to HIV-1 growth in monocytes. Figure 3 Attenuated replication of NL(AD8)IκB-αS32/36A in primary human monocytes. Panels A and B show the growth NL(AD8), NL(AD8)IκB-antisense and NL(AD8)IκB-αS32/36A in cultures of primary human monocytes. Cells (10 5 ) were infected with equal amounts of viruses normalized based on RT counts of 10 6 cpm (A) or 10 5 cpm (B). A representative experiment of three independent infections of monocytes from different individuals is shown. Figure 4 Reduced syncitia formation by NL(AD8)IκB-αS32/36A in infection of primary human monocytes. Panel A shows the kinetics of syncitia generation upon infection of primary human monocytes with 10 5 cpm RT activity of the indicated viral stocks. The average of syncitia observed per optical field is reported. Panel B shows the picture of primary human monocytes at 14 days post-infection with 10 5 cpm RT activity of the indicated viral stocks (original magnification × 430). Discussion Substantial numbers of monocytes are preserved in infected individuals even at later clinical stages of AIDS, when T cell numbers are dramatically reduced. Consistently, in animal models of HIV-1 infection, monocytes are the major reservoir after acute depletion of CD4-positive T cells [ 19 , 20 ]. This indicates that these cells are long lasting infected moieties that shuttle from mucosal sites to lymph nodes and could function as a major HIV-1 reservoir in vivo . In addition, monocytes are programmed to produce a large amount of inflammatory cytokine, including IL1-β, IL-6, TNF-α, which are strong inducers of HIV-1 replication [ 5 ]. Indeed, HIV-1 envelope binding to CCR5 receptor activates an intracellular signalling cascade that promotes high levels of transcription factors, including NF-κB, which sustain the initial rounds of viral replication and induce the production of inflammatory cytokines which activate surrounding cells to become more susceptible to virus infection [ 3 , 4 ]. Based on the published literature, the role of NF-κB in HIV-1 replication has been controversial [ 13 , 16 , 21 ]. For instance, the deletion of NF-κB binding sites from HIV-1 and SIV LTRs [ 22 ] has suggested that NF-κB activity may not be required for HIV-1 LTR-directed transcription. Moreover, deletion of NF-κB sequences in the LTR has also been reported not to affect HIV-1 replication in defined cellular settings [ 11 , 12 ]. These latter studies relied on short-term infections of immortalized cells that may not express a physiologic concentration of transcription factors. To address this issue, we have developed a novel HIV-1 strain, NL(AD8)IκB-αS32/36A, which was engineered to express a proteolysis-resistant IκBαS32/36A, and is a strong inhibitor of NF-κB activity. This recombinant virus expresses the envelope of the AD8 strain, a macrophage-tropic virus. Our findings show that NL(AD8)IκB-αS32/36A replication profile is different from that of the NL(AD8)IκB-antisense control. NL(AD8)IκB-αS32/36A failed to produce a productive infection in primary monocytic cells over a thirty-days acute infection (Fig. 3 ). These results were correlated with a strong inhibition NF-κB activity in NL(AD8)IκB-αS32/36A-infected cells (Fig. 2 ), indicating that in the setting of HIV infection of primary monocytes NF-κB plays a non-redundant role. These results are in agreement with the evidence that IκB-αS32/36A negatively affected the replication of HIV and SIV in PBMC cultures and in monkeys [ 16 , 17 ]. Because IκB-αS32/36A constitutively inhibits NF-κB [ 15 ], the potent inhibition of HIV/SIV replication could be due to repression of the NF-κB-dependent activation of HIV/SIV transcription. However, additional mechanisms might explain the potent inhibition of HIV/SIV replication by IκB-αS32/36A. In this regard, IκB-α regulates the transcriptional activity of NF-κB-independent genes by interacting with nuclear co-repressors, histone acetyltransferases and deacetylases [ 23 , 24 ]. Further studies are required to clarify novel activities of IκB-α in the modulation of the transcriptional machinery. Our results underscore a central role for IκB-α as a potent inhibitor of the replication of HIV-1 in both T cells [ 16 ] and monocytes (this study), and point to the NF-κB/IκB network as a suitable target for therapeutic intervention of AIDS. Conclusions In this study we have addressed the role of NF-κB/IκB proteins in the replication of HIV-1 in primary human monocytes. We show a strong attenuation in the replication of a macrophage-tropic HIV-1 strain expressing the IκB-αS32/36A repressor of NF-κB in primary cultures of human monocytes. These results are consistent with previous evidence of HIV/SIV inhibition by IκB-αS32/36A in PBMCs and in macaques [ 16 , 17 ]. In addition, these findings further support a role of NF-κB inhibitors in blocking HIV-1 replication and suggest novel strategies for the development of anti-viral therapy that targets NF-κB factors. Methods Transfections and Viral stocks 293T cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% v/v heat-inactivated fetal bovine serum and 3 mM glutamine. Viral stocks were produced by transfecting 293T cells (10 6 ) with viral plasmids (10 μg) using calcium phosphate. Forty hours later, the cell culture supernatant was passed through a 0.45-μm filter and measured for RT activity as previously described [ 16 ]. Immunoblotting analysis 293T cells were transfected with viral plasmids (10 μg) and lysed in RIPA buffer (150 mM NaCl, 1 % Nonidet P-40, 0.5 % sodium deoxycholate, 0.1% sodium dodecyl sulfate, 50 mM Tris-HCl pH 8.0) 24 hours later. Proteins (10μg) were separated by electrophoresis in 10% SDS-polyacrylamide gel and transferred to Immobilon-P (Millipore). Filters were blotted with an AIDS patient serum or with anti-FLAG monoclonal antibody by using Western-Light Chemiluminescent Detection System (Tropix, Bedford, MA). Electrophoretic Mobility Shift Assays Nuclear extracts and gel retardation assays were performed as described previously [ 9 ]. Briefly, cells were harvested, washed twice in cold phosphate-buffered saline, and resuspended in lysing buffer (10 mM Hepes, pH 7.9, 1 mM EDTA, 60 mM KCl, 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride, 0.2% v/v Nonidet P-40) for 5 min. Nuclei were collected by centrifugation (500 × g , 5 min), rinsed with Nonidet P-40-free lysing buffer, and resuspended in 150 μl of buffer containing 250 mM Tris-HCl, pH 7.8, 20% glycerol, 60 mM KCl, 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride. Nuclei were then subjected to three cycles of freezing and thawing. The suspension was cleared by centrifugation (7000 × g , 15 min), and aliquots were immediately tested in gel retardation assay or stored in liquid phase N2 until use. The HIV-1 NF-κB oligonucleotide probe was 5'-CAAGGGACTTTCCGCTGGGGACTTTCCAG-3'; the Sp1 oligonucleotide probe was 5'-GGGAGGTGTGGCCTGGGCGGGACTGGGGAGTGGCG-3'. The probes were end-labelled with [γ- 32 P]ATP (Amersham Int., Buckinghamshire, UK) using polynucleotide kinase (New England Biolabs, Beverly, MA). Equal amounts (5 μg) of cell extracts were incubated in a 20 μl reaction mixture containing 10% glycerol, 60 mM KCl, 1 mM EDTA, 1 mM DTT, and 2 μg of poly [d(G-C)] (Boehringer Mannheim, Germany) for 5 min on ice. One μl of [γ 32 P]-labelled double-stranded probe (0.2 ng, 5 × 10 4 cpm) was then added with or without a 100-fold molar excess of competitor oligonucleotide. The reactions were incubated at room temperature for 15 min and run on a 6% acrylamide:bisacrylamide (30:1) gel in 22.5 mM Tris borate, 0.5 mM EDTA. Gels were dried and autoradiographed. Monocytes cultures and infections Human monocytes were isolated from PBMC by elutriation, cultured in RPMI, 10% FCS and GMCSF (20 ng/ml) for 48 hours. Infections were performed with viral stocks measured by reverse-transcriptase (RT) activity [ 16 ]. Usually, cell cultures (10 5 cells) were infected with 10 5 - 10 6 cpm of RT activity. The cell culture supernatants were collected every two days and replaced with fresh medium. The viral production was measured as RT activity in the culture supernatants as previously described [ 16 ]. The syncitia formation in cell cultures was evaluated by calculating the average number of syncitia in at least six optical fields. List of abbreviations used NF-κB, nuclear factor kappa B IκB, inhibitor of nuclear factor kappa B IL-1, interleukin-1 IL-6, interleukin-6 IL-8, interleukin-8 TNF-α, tumor necrosis factor alpha TGF-β, transforming growth factor-beta cpm, counts per minute FCS, fetal calf serum GMCSF, granulocyte-macrophage colony-stimulating factor Competing interests The author(s) declare that they have no competing interests. Authors' contributions CP carried out the analysis of viral growth and DNA band-shift assays. FT was responsible for cell cultures. AP performed the immunoblotting analysis. GF produced the viral plasmids and viral stocks, and performed the artwork of the paper. GS participated in the design of the study and discussion of results. IQ designed this study and edited the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544834.xml
368161	Interaction Networks in Yeast Define and Enumerate the Signaling Steps of the Vertebrate Aryl Hydrocarbon Receptor	The aryl hydrocarbon receptor (AHR) is a vertebrate protein that mediates the toxic and adaptive responses to dioxins and related environmental pollutants. In an effort to better understand the details of this signal transduction pathway, we employed the yeast S. cerevisiae as a model system. Through the use of arrayed yeast strains harboring ordered deletions of open reading frames, we determined that 54 out of the 4,507 yeast genes examined significantly influence AHR signal transduction. In an effort to describe the relationship between these modifying genes, we constructed a network map based upon their known protein and genetic interactions. Monte Carlo simulations demonstrated that this network represented a description of AHR signaling that was distinct from those generated by random chance. The network map was then explored with a number of computational and experimental annotations. These analyses revealed that the AHR signaling pathway is defined by at least five distinct signaling steps that are regulated by functional modules of interacting modifiers. These modules can be described as mediating receptor folding, nuclear translocation, transcriptional activation, receptor level, and a previously undescribed nuclear step related to the receptor's Per–Arnt–Sim domain.	Introduction The aryl hydrocarbon receptor (AHR) is a ligand-activated transcription factor found in a variety of vertebrate species. The AHR is a prototype member of the Per–Arnt–Sim (PAS) superfamily of signaling molecules. Members of this superfamily regulate cellular responses to a variety of environmental stimuli, including pollutants, hypoxia, and external light cues ( Gu et al. 2000 ). Our initial interest in AHR biology arose from its pivotal role in mediating the adaptive metabolic response to both polycyclic aromatic hydrocarbons (PAHs) and the toxic effects of more potent agonists like the halogenated dioxins ( Schmidt and Bradfield 1996 ; Whitlock 1999 ). More recently, it has been observed that the AHR plays an important role in normal vascular development, suggesting the existence of an endogenous ligand ( Lahvis et al. 2000 ). From the broader perspective, the AHR can be viewed as a prototype of all PAS protein signaling. That is, what we learn about AHR biology will have a direct influence on how we think about PAS-mediated hypoxia, circadian, and developmental pathways. An initial understanding of AHR signal transduction has resulted from the biochemical and molecular studies that have been performed over the past two decades ( Schmidt and Bradfield 1996 ; Whitlock 1999 ). The resultant model holds that the unliganded AHR resides in the cytoplasm, where it is associated with a dimer of the chaperone protein Hsp90 and cochaperones such as ARA9/XAP2 and p23 ( Pongratz et al. 1992 ; Carver and Bradfield 1997 ; Ma and Whitlock 1997 ; Meyer et al. 1998 ; Kazlauskas et al. 1999 ). Upon binding ligands, the cytoplasmic AHR translocates to the nucleus, where it dimerizes with another PAS protein known as ARNT. The AHR–ARNT heterodimer then binds to specific dioxin-responsive enhancers (DREs) and transactivates a battery of genes encoding xenobiotic-metabolizing enzymes, most notably CYP1A1 , CYP1A2 , and CYP1B1 ( Schmidt and Bradfield 1996 ; Whitlock 1999 ). Transactivation of target genes has been shown to be mediated through a variety of histone acetyltransferases (HATs) and SWI/SNF coactivators, such as SRC, p300/CBP, and BRG-1 ( Kobayashi et al. 1997 ; Beischlag et al. 2002 ; Wang and Hankinson 2002 ). Although the initial model of AHR signaling provides a valuable framework, its completeness has not yet been assessed. That is, we have no estimates of the total number of gene products involved in AHR signaling, nor can we be sure we have identified all the important steps. Without these estimates, it is difficult to gauge how much or how little we understand about this pathway. In an effort to address these issues, we employed the comprehensive set of gene deletions available in a yeast model system to systematically identify gene products that influence AHR function. We then employed a protein interaction network (PIN) strategy to provide a framework to describe AHR signaling. By coupling both computational and experimental annotations, we were able to deduce the minimum number of genetic loci and signaling events required for AHR signaling. Results Rationale A number of laboratories have demonstrated that the yeast Saccharomyces cerevisiae is a valuable model system for the study of signaling by mammalian nuclear receptors ( Garabedian and Yamamoto 1992 ; McEwan 2001 ). Although there is no yeast ortholog of the AHR, it has been also shown that AHR signaling can be recapitulated in yeast and that this system can be used to identify novel players in AHR biology ( Carver et al. 1994 ; Whitelaw et al. 1995 ). The experimental advantages of S. cerevisiae as a tool to study AHR signaling are related to the yeast's fundamental similarities with mammalian systems, the more thorough characterization of its smaller genome, and the availability of its specific genomic tools, such as arrayed deletions of each individual open reading frame (ORF) and large-scale databases describing protein and genetic interactions ( Winzeler et al. 1999 ; Resnick and Cox 2000 ; Kennedy 2002 ; Mewes et al. 2002 ; Xenarios et al. 2002 ). These convenient genomic tools allowed us to employ a systematic approach to identify gene products involved in the AHR pathway and to interpret them in the context of a protein interaction network. Owing to a lack of corresponding reagents/databases, such an approach is not yet feasible for the study of AHR signaling in more complex eukaryotic systems such as human or mouse. Identification of AHR Modifiers by a High-Throughput Deletion Array Screen In earlier attempts to identify AHR modifiers in yeast, it was demonstrated that genetic screens can be performed more efficiently by using an AHR construct that is fused to the DNA-binding domain of the bacterial LexA protein (AHR–LexA) ( Carver et al. 1994 ; Whitelaw et al. 1995 ). This chimeric system removes the requirement for ARNT and allows our screens to be more specific for those mutations/modifiers that directly influence AHR function. Using this system, we set out to identify gene products that play important roles in AHR signaling ( Figure 1 A). Figure 1 High-Throughput Deletion Array Screen for AHR Modifiers (A) The flow chart of the deletion array screen. Each individual deletion strain was transformed with the AHR–LexA chimera and LacZ reporter constructs using a 96-well microtiter plate transformation approach. The AHR-dependent reporter activity of each deletion strain was examined with a 384-well plate-based fluorescence assay method. A total of 92 deletion strains were identified that displayed AHR signaling significantly different from the wt control. (B) Identification of “AHR-specific” modifiers. The effect of modifier deletions on the AHR pathway was compared with their effect on a Gal4TAD control pathway. It was found that 54 deletions influenced AHR signaling specifically, whereas 38 deletions corresponded to general factors. See text for details. To accomplish this screen, we employed the yeast deletion strains made available by the Saccharomyces Genome Deletion Project ( Winzeler et al. 1999 ). We developed a high-throughput approach to efficiently transform each deletion strain with two plasmids, one harboring the AHR–LexA chimera (pCEN-AHR) and the other, a LexA operator-driven LacZ reporter. Of the 4,695 available deletion strains, 4,507 (96%) were successfully transformed with the complete AHR signaling system (i.e., both plasmids). In the primary screen, we selected transformants that exhibited a 4-fold or greater change in AHR response as compared to the wild-type ( wt ) BY4742 strain ( p < 10 –6 ). To minimize false positives, we selected clones that influenced signaling at no less than two of the six concentrations of agonist tested. In addition, we retested each positive strain in a secondary screen with another AHR system containing the same LacZ reporter and a high-copy AHR–LexA chimera (pAHR) ( Carver 1996 ). By these criteria, 92 deletion strains were identified that reproducibly displayed a significant change in AHR signaling as compared to the wt strain ( Table S1 ). To eliminate those deletions that influenced the AHR pathway in a nonspecific manner, each of the 92 deletion strains was examined with a control plasmid pGal4TAD (see Materials and Methods ). This construct harbors the transcriptional activation domain (TAD) of Gal4p fused to the LexA DNA-binding domain and was cotransformed into each deletion strain with the LacZ reporter ( Figure 1 B). Of the 92 deletions, 38 were observed to also influence pGal4TAD signaling. We concluded that these deletions either represented general players in both pathways or exhibited nonspecific effects through their influence on, e.g., the common LexA domain, plasmid maintenance, or cell growth rate. Therefore, the inclusion of the pGal4TAD control led us to eliminate 38 nonspecific factors and identify 54 deletions that appeared to influence the AHR pathway in a specific manner. Of these “AHR-specific” factors, Hsc82p and Cpr7p were previously described AHR modifiers, and the other 52 were novel ( Carver et al. 1994 ; Whitelaw et al. 1995 ; Miller 2002 ) ( Table S2 ). The analysis of the annotated function of these AHR modifiers revealed that they were associated with a great variety of cellular functions ( Table S3 ). For many of these annotations, their direct association with AHR signaling appeared elusive. Therefore, in order to appreciate the function of identified modifiers in the AHR pathway, an information framework was required to put them in context. Portrayal of the AHR–PIN Recent experiments from a number of laboratories have provided data to support the idea that protein interaction network (PIN) can be used to portray the workings of complex biological systems ( Schwikowski et al. 2000 ; Ge et al. 2001 ; Ideker et al. 2001 ; Tong et al. 2002 ). To investigate how identified modifiers and their interactions influence AHR signaling, we constructed a modifier network (AHR–PIN) based on known protein and genetic interactions derived from the DIP and MIPS databases ( Mewes et al. 2002 ; Xenarios et al. 2002 ). Our AHR–PIN map is comprised of “nodes” and “links.” A “node” is a graphic depiction of a protein or locus, and a “link” is a line between two nodes in the map that depicts the known interaction between them. As yeast protein–protein interactions identified to date are still far from saturating and are heavily biased towards proteins of high abundance, genetic interactions were also included in the network building as a complement ( Tong et al. 2002 ; von Mering et al. 2002 ). In the AHR–PIN, protein interactions are depicted with black lines, and genetic interactions are labeled in red. In addition, nodes also come in two types, “M-nodes” and “I-nodes.” We refer to the protein or locus that has an identified effect on the AHR pathway as the “M-node,” or modifier node, and refer to the nonmodifier node that is required on a path to connect two M-nodes as the “I-node,” or intervening node. In an effort to determine the most informative PIN, we examined how the structure and complexity of the map was influenced by the choice of the maximally allowed number of links between any two M-nodes (we refer to this value as D max ). One common feature of AHR–PINs with D max values greater than 1 was that the majority of M-nodes were interconnected in a single large network with no breaks ( Figure 2 A– 2 C). For convenience, we refer to this single large network simply as the AHR–PIN in following discussions. When D max was set at low stringency (D max ≥ 3), the representation of M-nodes in AHR–PIN was high. For example, at D max = 3, 46 of 54 M-nodes were included. However, AHR–PINs resulting from these inclusive, yet low-stringency conditions exhibited high complexity, which made it impossible to assess the interactions visually ( Figure 2 A and 2 B). When D max was set at higher stringency (D max = 2), the resultant AHR–PIN now comprised 34 closely interconnected M-nodes and was much easier to visualize ( Figure 2 C; Table S4 ). Further simplification of the AHR–PIN with D max = 1 was of little utility because it resulted in a large proportion of isolated M-nodes, with the largest cluster containing only three M-nodes ( Figure 2 D). Figure 2 AHR–PIN versus Random PINs (A–D) AHR–PINs at various D max levels. AHR modifiers are highlighted with bigger green nodes. A total of 48, 46, 34, and three AHR modifiers are interconnected in the AHR–PINs with D max values of 4, 3, 2, and 1, respectively. (E–H) Distribution of random PINs at various D max levels in histogram. Each distribution graph represents 5,000 randomly generated PINs. The density estimation curve (in red) is plotted on top of the histogram where applicable. The number of M-nodes in the AHR–PIN and the average number of M-nodes in random networks are marked in each distribution graph. See text for details. The AHR–PIN Is Distinct from Random PINs To examine the statistical significance of the AHR–PINs, we tested whether they could have been generated by random chance. If the AHR–PIN represents a valid description of the AHR pathway, it should comprise significantly more interconnected M-nodes than would be interconnected by random chance. To test this idea, a Monte Carlo simulation was conducted by generating 5,000 random PINs at each D max setting. Each of these test PINs was constructed based on 54 mock M-nodes randomly selected from genes contained in the entire deletion set. To estimate the statistical significance of the AHR–PIN, the random graph was defined as the null distribution, and the p value for the AHR–PIN at each D max was calculated from the fraction of trials with a higher number of interconnected M-nodes ( Figure 2 E– 2 H). The AHR–PIN at D max = 1 was not statistically significant compared to those generated at random chance ( p < 0.08; Figure 2 H). However, at D max = 2, D max = 3, and D max = 4, the number of interconnected M-nodes in the AHR–PIN was significantly larger than that of random PINs ( p < 10 –4 , 10 –3 , and 3 × 10 –3 , respectively; Figure 2 E– 2 G). These observations were consistent with the idea that AHR–PINs at these settings provide a biologically meaningful description of AHR signaling. For further exploration, we chose to focus on the network with the greatest statistical significance, i.e., the PIN generated at D max = 2. In this AHR–PIN, 63.0% of the M-nodes (34/54) are interconnected, while in corresponding random PINs with mock M-nodes, this number drops to 18.5% (10/54). Although the AHR–PINs at D max = 3 and D max = 4 also exhibited statistically significant differences from random PINs, these AHR–PINs were not considered further for two reasons. First, these networks were visually complex and could not be simply annotated in two dimensions. Second, the ratios of interconnected M-nodes in these AHR–PINs to those of random PINs were quite low (1.3 and 1.2 for D max = 3 and D max = 4, respectively). This observation suggests a much greater potential for displaying false positive interactions at these settings as compared to the AHR–PIN at D max = 2, where this ratio was 3.4 (34/10). Modular Organization of AHR–PIN as Revealed by Network Clustering Our next objective was to use the PIN to enumerate and define steps in AHR signaling. It has been suggested that PINs exhibit a modular nature, with each module comprising highly interconnected proteins of related cellular functions ( Hartwell et al. 1999 ; Schwikowski et al. 2000 ). Our hypothesis was that functional modules in the AHR–PIN would correspond to discrete steps in the mechanism of signaling. To test this idea, we attempted to define the functional modules using a number of computational and experimental annotation approaches. As a strictly computational approach, we attempted to identify the functional modules in the AHR–PIN by a network-clustering method ( Rives and Galitski 2003 ). In brief, an all-pairs-shortest-path distance matrix was generated for every pair of nodes within the AHR–PIN (D max = 2). Each distance ( d ) in the matrix refers to the length of the shortest path between a pair of nodes in the full network space of yeast genomic PIN and was transformed into an “association” value ( 1/d 2 ). The resultant pairwise association matrix was used to identify network clusters in the AHR–PIN by a hierarchical average-linkage clustering algorithm ( Eisen et al. 1998 ; Rives and Galitski 2003 ). The cluster boundaries were delimited by using a similar “tree-depth threshold” that was set low enough to separate the largest cluster from others ( Figure 3 A) ( Rives and Galitski 2003 ). If we define a network cluster to include at least two M-nodes, ten such clusters can be identified ( Figure 3 A). Consistent with the modular PIN hypothesis, we found that these clusters overlapped with ten local areas (modules) in the AHR–PIN, with each module comprised of two to six M-nodes ( Figure 3 B). Figure 3 Functional Modules Identified by Network Clustering (A) Network clustering of AHR–PIN. Protein nodes in the AHR–PIN (D max = 2) were clustered by a hierarchical clustering algorithm. A tree-depth threshold was set to delimit cluster boundaries ( Rives and Galitski 2003 ). Clusters with at least two M-nodes are shown. See text for details. (B) Overlay of the network clusters on the AHR–PIN. The ten network clusters correspond to ten local areas in the AHR–PIN. Each network cluster (local area) is labeled with its significant functional enrichment as calculated using the FunSpec program ( Robinson et al. 2002 ). Color scheme. Nodes: modifier deletions that incurred down- and up-regulation of AHR signaling are marked in green and red, respectively. For intervening nodes, essential genes are marked in gray and nonessential genes in white. Links: physical interactions are labeled in black and genetic interactions in red. If both interactions are available for a given link, only the physical interaction is shown. This color scheme is also applied to Figures 4–7. In an effort to define the function of these proposed network modules, we asked whether each individual module could be best described by a particular annotation. A module is considered to be enriched for a given annotation if the number of components known to have that function within the module exceeds the number that could be expected from random chance. It has been proposed that the degree of enrichment for a given annotation can be measured by its hypergeometric distribution ( Tavazoie et al. 1999 ). Using this approach, we calculated the annotation enrichment for each of the ten protein modules in the AHR–PIN with the FunSpec program ( Table S5 ) ( Robinson et al. 2002 ). As shown in Figure 3 B, it was found that the AHR–PIN is organized by protein modules that perform distinct cellular functions (e.g., protein folding and chromatin modification). Functional Modules as Revealed by Their Influence on Different AHR Domains In an effort to test the predicted modules and define how they influence AHR signaling, we annotated the AHR–PIN using a number of independent functional tests. First, we examined whether functional modules could be identified based upon their influence on different domains of the AHR. To this end, we examined the influence of each modifier on the signaling of a partial-deletion mutant, pAHRΔPASB, which contains the AHR's transcriptionally active domain but is missing those domains responsible for ligand binding and Hsp90 interaction ( Figure 4 A). Of the 53 modifier deletions successfully transformed with the pAHRΔPASB system, we found that 25 deletions affected both the parent AHR and the deletion mutant. This observation indicated that these 25 modifiers had an influence on the shared C-terminal TAD region and not on the PASB domain ( Figure 4 A). These modifiers were referred to as the “TAD influence group.” The remaining 28 deletions, which required the PASB domain for their effect, were referred to as the “PASB influence group.” Figure 4 Functional Modules Identified by the “Domain Influence” (A) Identification of domain influencing groups. The effects of modifier deletions on the signaling of AHR and AHRΔPASB were compared in parallel. It was found that 28 modifiers were required for the function of the PASB domain (i.e., their deletions affected the AHR, but not the AHRΔPASB). The other 25 modifiers were found to be required for the shared TAD region (i.e., their deletions affected the signaling of both AHR and AHRΔPASB). (B) Overlay of the “domain influence” layer (blue boundary) and the network-clustering layer (shadowed) on the AHR–PIN. The PASB influence group corresponds to a central region in the AHR–PIN. The TAD influence group corresponds to two peripheral areas. Occasional outlier nodes are marked with their corresponding module names. When the AHR–PIN was annotated according to the domain influence of each modifier, it was found that modifiers from the same domain influence group closely interacted in the map. That is, the PASB influence group resided in a single connected region, whereas the TAD influence group occupied two peripheral regions ( Figure 4 B). Interestingly, the PASB module was found to overlap with the computationally identified clusters 1, 3, 5, 8, 9, and 10. For the two TAD modules, one overlapped with cluster 6, and the other with clusters 4 and 7. This overlap supported both the computational and experimental annotations. For example, the “chromatin modification cluster,” 6, identified and annotated computationally, was found to be associated with the TAD influence group, defined experimentally. Similarly, the “protein folding cluster,” 5, was associated with the PASB domain influence group. The PASB domain is known to interact with the chaperone protein Hsp90, which plays a significant role in the folding of the mammalian AHR ( Pongratz et al. 1992 ; Carver et al. 1994 ; Whitelaw et al. 1995 ). Functional Modules as Revealed by Their Effect on AHR Pharmacology To further annotate the AHR–PIN, each of the 54 modifiers was tested for its influence on AHR signaling (pAHR system) at various agonist concentrations, times, and temperatures, as well as after exposure to two distinct AHR agonists, α-naphthoflavone (αNF) and β-naphthoflavone (βNF). The relationship between each modifier and signaling was then examined using a hierarchical average-linkage clustering algorithm ( Eisen et al. 1998 ) ( Figure 5 A). It was found that the five major clusters corresponded to five closely intraconnected local areas in the map, designated A, B, C, D, and E ( Figure 5 B). Among them, modules A and C exhibited significant functional enrichment of protein folding and transcriptional control, respectively (data not shown). When the clustering result was overlaid upon the previous maps, it was found that modules A, D, and E corresponded to the PASB influence module, and modules B and C corresponded to the TAD influence module ( Figure 5 B). Figure 5 Functional Modules Revealed by Effect on AHR Pharmacology (A) Cluster analysis of the effect of modifier deletion on AHR pharmacology. AHR signaling was examined at various doses, timepoints, and temperatures, and with the two AHR agonists βNF and αNF. The influence of modifier deletion on the dose-response of the AHR was analyzed by a hierarchical clustering algorithm. Rows in the clustering diagram represent modifier deletions. Columns correspond to experimental conditions. Green and red indicate down- and up-regulated AHR signaling, respectively. Color brightness is proportional to fold change. Black indicates wt signaling. Sparse gray boxes represent missing datapoints. (Insert) Diagram of corresponding dose-response curves of the wt strain and the average of cluster C. (B) Overlay of the “pharmacology clustering” layer (shadowed, black boundary) and “domain influence” layer (blue boundary) on the AHR–PIN. The major pharmacology clusters are coincident with five local areas in the AHR–PIN. In addition, clusters A, D, and E correspond to the PASB influence module, and clusters B and C correspond to the TAD influence module. Functional annotations determined by pharmacology clustering are indicated in black, and those derived from domain influencing are indicated in blue. Occasional outlier nodes are marked with their corresponding module designation. See the legend of Figure 3 for the color scheme of the nodes and links. Functional Modules as Revealed by Their Influence on AHR Localization Lastly, we examined each modifier's influence on AHR's subcellular localization. This was accomplished using an AHR–GFP fusion protein (pAHRGFP). When the wt strain was transformed with the plasmid pAHRGFP, it was found that the fusion protein was evenly distributed in the cell in the absence of AHR agonist. In the presence of the agonist βNF, the AHR–GFP protein translocated to the nucleus ( Figure 6 A). To examine the influence of each modifier on this translocation process, the pAHRGFP construct was transformed into each of the 54 modifier deletion strains and its localization was examined by fluorescence microscopy in the presence of agonist. Four localization phenotypes were identified ( Figure 6 B). About 50% of the deletion strains exhibited AHR translocation similar to that observed in the wt strain (group I). Approximately 30% of the strains were found to contain a marked reduction in the level of AHR protein in the cell (group II). Approximately 10% of the deletion strains displayed receptor aggregates in the cell (group III). The final 10% of the deletion strains displayed a normal level of AHR protein, but the receptor failed to translocate into the nucleus in the presence of agonist (group IV). When overlaid with the previously determined experimental layers, group I was found to overlap with the modules of C and D, and groups II, III, and IV corresponded to modules B, A, and E, respectively ( Figure 6 C). According to this overlap, module B can be further described as being associated with the regulation of receptor level in the cell, and module E is associated with the regulation of nuclear translocation of the AHR ( Figure 6 C). Figure 6 Functional Modules Identified by the “Localization Influence” (A) The AHR–GFP fusion protein translocates to nucleus in the presence of agonist βNF. Nucleus position in the cell was confirmed by DAPI staining (data not shown). Dimethyl sulfoxide (DMSO) is a vehicle control for βNF. (B) Classification of modifier deletion strains according to AHR–GFP phenotype (with βNF). Group I displays wt phenotype. Group II contains decreased level of receptor protein. Group III contains aggregated misfolded receptor. Group IV displays the AHR that is not capable of translocating to the nucleus. (C) Overlay of “localization influence” layer (shadowed, red boundary) and the “pharmacology clustering” layer (black boundary) on the AHR–PIN. Group I corresponds to modules C and D. Groups II, III, and IV overlap with modules of B, A, and E, respectively. Functional annotations determined by localization influence are indicated in red, and those derived from pharmacology clustering and domain influencing studies are indicated in black. Occasional outlier nodes are noted with their corresponding module designation. See the legend of Figure 3 for the color scheme of the nodes and links. Discussion Modifier Identification Our initial objective was to identify the number of loci that are required for AHR signal transduction. In this regard, our high-throughput deletion screen identified 52 novel and two known AHR modifiers. Although this is a surprisingly large number of modifiers for the function of a single protein, it is probably an underestimate since the deletion screen cannot identify modifiers that are encoded by essential genes. Moreover, our criteria of including only strong modifiers (influence of 4-fold compared to control) may have caused us to miss some important modifiers of this pathway. Nevertheless, the number of AHR modifier loci reported here is approximately 10-fold greater than what has been reported using mammalian cell culture and animal models ( Schmidt and Bradfield 1996 ; Whitlock 1999 ). Once we identified these AHR modifiers in yeast, we sought a way to position and characterize them in the context of the AHR pathway. Given the idea that PINs can be used to portray the cellular workings, we attempted to use our deletion data to generate and annotate an AHR–PIN ( Hartwell et al. 1999 ; Schwikowski et al. 2000 ; Ge et al. 2001 ; Ideker et al. 2001 ; Tong et al. 2002 ). To construct the AHR–PIN, the yeast genomic PIN was decomposed by extracting those nodes/links relevant to AHR modifiers. To test the utility of the resultant AHR–PIN, a series of Monte Carlo simulations were carried out. It was demonstrated that when D max was set at 2, 3, or 4, the resultant AHR–PIN was of a complexity that could not have resulted from random chance. Furthermore, the comparison of various simulations at different D max settings guided us to select the linking parameter at D max = 2. This setting of intervening links resulted in the highest level of statistical significance, displayed the lowest potential for false positive interactions, and decreased the map's visual complexity to a level that was readily understood in a two-dimensional map. The Modular Structure of AHR–PIN Reveals Five Discrete Steps in Signaling Our analysis of the AHR–PIN revealed an underlying modular structure. That is, there are areas in the AHR–PIN that display high interconnectedness of nodes, and these regions represent functionally related modifiers. The modularity of AHR–PIN was revealed by both computational and functional tests. In our initial computational approach, a total of ten clusters were identified, and the functional enrichment of each cluster was calculated by hypergeometric distribution ( Tavazoie et al. 1999 ; Robinson et al. 2002 ). Although the computational approaches of module identification and annotation were useful in hypothesis generation, they did not provide a direct description of AHR signaling. Therefore, we set out to annotate the AHR–PIN with a number of functional tests. In our first annotation experiment (“domain influence”), we found that the AHR–PIN could be divided into three discrete functional modules (i.e., one module that influenced the PASB domain and two modules that influenced the C-terminal domain we referred to as TAD). Additionally, each of these modules was found to overlap with one to several network clusters (see Figure 4 ). This tight overlay of functional data with highly interconnected regions in the AHR–PIN also held true when we applied annotations for pharmacological clustering and subcellular localization studies (see Figures 5 and 6 ). Given the overlay of these annotations derived from both functional and computational tests, we conclude that the AHR–PIN provides a biologically meaningful representation of the regulatory network of AHR signaling ( Figure 7 A). Moreover, based upon the combined annotations for each individual module, we propose that AHR signal transduction is regulated at five discrete steps: (1) receptor folding, (2) receptor translocation, (3) receptor transcriptional activation, (4) receptor level, and (5) a previously undescribed signaling event related to the PASB domain ( Figure 7 B). Figure 7 Regulatory Network of AHR Signaling (A) The summary map of AHR–PIN. Functional modules were determined by the overlapped annotations from three experimental layers (domain influence, pharmacology clustering, and localization influence) as well as from network clustering. For each functional module, the main “stacking pattern” of experimental layers is noted in italics. Modifiers initially left outside the single large cluster of the AHR–PIN were assigned to corresponding functional modules by sharing the similar stacking pattern where applicable. See the legend of Figure 3 for the color scheme of the nodes and links. (B) An expanded model of AHR signaling. The AHR signaling pathway is regulated by at least five functional modules that are involved in the control of receptor folding, nuclear translocation, transcriptional activation, receptor level, and a PASB-related nuclear event. Within each functional module, modifers intially enclosed in the single large cluster of the AHR–PIN are highlighted in bold. Known human homologs of the modifiers are noted at the side with a smaller font ( Costanzo et al. 2001 ) . ARNT is dimmed because modifiers were identified in this study from an “ARNT-free” chimeric AHR system. See text for details. The AHR Folding Module A module that regulates AHR folding was identified by the known activities of its constituents, as well as the appearance of receptor aggregates when these modifiers were absent (see Figure 6 B, group III). Given that AHR folding has been well studied over the past 15 years, examination of this module provided insight into the fidelity of our screen and the transference of our observations to the mammalian system. For example, two known modifiers were identified by our high-throughput screen: Hsc82p (homolog of human Hsp90) and Cpr7p (homolog of human Cyp40) ( Pongratz et al. 1992 ; Miller 2002 ). In addition, we identified a previously unknown player in the AHR folding pathway, the chaperone protein Sti1p (homolog of human p60/HOP). Sti1p/p60 has been shown to be an essential component of the glucocorticoid receptor signaling pathway, where it is required to form an Hsp90 chaperone complex ( Chang et al. 1997 ; Dittmar and Pratt 1997 ). By analogy, we propose that Sti1p/p60 is involved in the formation of an Hsp90·cochaperone complex that is essential for the proper folding of the AHR. Finally, our analysis of this module suggests that a number of proteins not known to be chaperones are involved in receptor folding. These proteins include Sec28p and possibly Rpl19b. The AHR folding module can also be used to explain the existence of I-nodes within a functional module. Given their “linker” position and the observation that they often share similar annotated function with their neighboring M-nodes (data not shown), it is a logical prediction that I-nodes play a role in AHR signaling that is functionally similar to their modifier neighbors. We propose that I-nodes most commonly arise as the result of their essential gene nature (gray nodes in the figure; nontestable in the deletion screen) or because they represent a redundant gene product (white nodes in the figures). We offer two examples that support this idea. First, one essential gene I-node in the folding module, Cns1p, has recently been reported to be involved in AHR signaling ( Miller 2002 ). Second, the possibility that white nodes may often result from redundancy is supported by what we know about Hsp90. The Hsc82p and Hsp82p proteins are yeast orthologs of human Hsp90, a well-studied chaperone required for proper AHR folding ( Pongratz et al. 1992 ; Carver et al. 1994 ; Whitelaw et al. 1995 ). Under normal growth conditions, Hsp82p and Hsc82p account for 7% and 93% of the total “Hsp90 level,” respectively ( Borkovich et al. 1989 ). Thus, it is not surprising that Hsp82p was not identified as a modifier, since its deletion would have had little effect on the total Hsp90 level in the cell ( Figure 7 A). Finally, white I-nodes can also arise from weak modifiers that influenced AHR signaling by less than 4-fold, e.g., Sba1p (ortholog of human AHR modifier p23) ( Kazlauskas et al. 1999 ). In this regard, although a choice of 4-fold was somewhat arbitrary, we found that lowering the cutoff greatly increased the network complexity without enhancing the statistical significance of the AHR–PIN (as compared with random PINs; data not shown). The AHR Employs a Multistep Transcriptional Mechanism The composition of the transcriptional activation module suggests that the AHR activates target genes via the coordination of histone acetylation, ATP-dependent chromatin remodeling, and direct recruitment of basal RNA polymerase II transcriptional apparatus (see Figure 7 ). We base this idea on the observation that this functional module is composed of components of the histone acetyltransferase SAGA complex (homolog of the mammalian PCAF complex)—Gcn5p, Spt3p, and Spt8p; components of the SWI/SNF chromatin-remodeling complex—Snf12p and Swi3p; and a subunit of the Srb–mediator complex—Srb2p ( Grant et al. 1998 ; Myers et al. 1998 ; Peterson et al. 1998 ). This interdependent requirement of three distinct classes of transcriptionally relevant proteins is consistent with observations from mammalian cells, where the involvement of both HAT and SWI/SNF coactivators in AHR signaling has been reported, as has the direct interaction of the AHR with basal transcriptional factors TBP, TFIIF, and TFIIB ( Rowlands et al. 1996 ; Kobayashi et al. 1997 ; Swanson and Yang 1998 ; Beischlag et al. 2002 ; Wang and Hankinson 2002 ). These collective data support the idea that AHR transactivation is mediated by a multicomponent, synergistic process. Nuclear Translocation of the AHR Our network analysis has also identified a functional module that regulates the ligand-dependent translocation of the AHR (see Figure 7 ). This nuclear translocation module appears to be associated with the PASB domain, which is known to play roles in both ligand binding and interaction with chaperones (see Figure 4 A). This observation is consistent with the idea that ligand exposure releases the AHR from the cytosolic chaperone anchors ( Kazlauskas et al. 2001 ; Petrulis et al. 2003 ). Although the mechanism for this translocation event remains unclear, it is interesting to note that the “translocation module” overlaps with a protein degradation cluster, cluster 10 (see Figure 7 A). This observation suggests that the underlying control of subcellular localization of the AHR might be related to the selective degradation of certain tethering factors by ubiquitination, possibly mediated by Doa1p and other members in this module ( Hochstrasser and Varshavsky 1990 ). Regulation of AHR Expression A module that regulates the amount of receptor protein was also identified in our AHR–PIN (see Figure 7 ). This module is associated with the C-terminal domain of the AHR (see Figure 4 A). Although we have commonly referred to this region as the TAD domain, these data suggest that other functions are also encoded here. We base this assessment on two observations. First, members of this module are not known to play direct roles in transcription (see Table S4 ). Second, this module influences receptor level in a manner that is upstream of the AHR's activity as a transcription factor. Our interpretation of this module is that these modifiers are associated with a domain that is proximal to or overlaps with the receptor's TAD and that this domain plays a role in the regulation of receptor level (see Figure 4 A). At the present time it is not clear whether this module influences the AHR at its mRNA or protein level. A Novel Step Defined by the PASB Module A novel PASB-dependent step in AHR signaling appears to have been revealed by this network analysis (see Figure 7 , PASB-related module). Given that corresponding deletions of this PASB-related module did not impair the receptor's nuclear translocation (see Figure 6 , group I), we conclude that this module must influence either a downstream nuclear event or some cytosolic event that is not revealed until the receptor is within the nuclear compartment. On the other hand, this module did not appear to be involved in the final transactivation step, as it was distinct from the transactivation module according to our functional annotations (see Figures 4 and 5 ). Taken in sum, there must exist a PASB-dependent event that is posttranslocation and pretransactivation. Such an event could be related to the receptor's dimerization, DNA binding, or an as-yet-undefined nuclear event, such as the unfolding of a transcriptionally active domain ( Sun et al. 1997 ; Heid et al. 2000 ). Interestingly, the existence of this PAS-related signaling is consistent with the previous observation that the DNA binding ability of the AHR can be impaired by a point mutation within its PAS domain ( Sun et al. 1997 ). Lastly, the fact that this PASB-related module overlaps with multiple network clusters (1, 2, 8, 9) suggests a cooperative mechanism that involves more than one cellular function (see Figure 7 A). Conclusion We began this study with the objective of defining the AHR signal transduction pathway in a manner that would allow us to quantify the number of loci and enumerate the steps involved in signaling. By integrating our deletion screen with the PIN framework and through subsequent computational and experimental annotations, we were able to identify modifier modules that regulate five distinct AHR signaling steps. In this regard, we found that the integration of multiple annotation approaches is vital for the reconstruction of the final picture by connecting and cross-validating individual information pieces. As interaction datasets become more fully developed and annotated, such a map will steadily improve and provide more accurate description of AHR signaling. Lastly, the systematic strategy that we developed in this work should be readily applicable to the study of most mammalian proteins to reconstruct corresponding modifier networks that regulate their signaling. Materials and Methods Strains and plasmids A set of deletion derivatives of S. cerevisiae strain BY4742 (MATα, his3Δ1, leu2Δ0, lys2Δ0, ura3Δ0) was used in this study. This deletion set was obtained from Research Genetics (now a part of Invitrogen, Carlsbad, California, United States) in a 96-well arrayed format. The plasmid pCEN-AHR (PL1605) was constructed by replacing the TRP1 autotrophic marker of PL883 ( Hogenesch 1999 ) with a HIS3 marker using a “marker swap” method ( Cross 1997 ). This CEN-based plasmid contains the LexA–AHR chimera cDNA (LexA-AHRNΔ166) under the control of an alcohol dehydrogenase I ( ADH1 ) promoter. LexA-AHRNΔ166 is a chimeric AHR, with its amino acid residues 1–166 replaced by residues 1–202 of bacterial repressor LexA, and is referred to in the Results section simply as “AHR” for convenience. The reporter plasmid pSH18–34 (PL623) (Clontech, Palo Alto, California, United States) is a 2μ-based, URA3 -selectable vector that contains the bacterial LacZ gene, as a reporter, under the control of eight LexA-binding sites. The plasmid pEG202 (Clontech, Palo Alto, California, United States) is a 2μ-based, HIS3 -selectable plasmid containing the LexA 1–202 sequence under the control of the ADH1 promoter. The plasmid pAHR (PL700) has been described previously ( Carver 1996 ). This plasmid contains the AHRNΔ166 sequence inserted into the EcoRI site of pEG202. The pGal4TAD control plasmid (PL1573) (Display Systems Biotech, now NeuroSearch A/S, Ballerup, Denmark) contains the transcription activation domain of yeast GAL4 inserted into the EcoRI site of pEG202. The control plasmid pAHRΔPASB (PL1799) is the same as pAHR except for the removal of the C-terminal half of the PAS domain. This pAHRΔPASB plasmid was constructed by subcloning the EcoRI fragment of PL248 ( Carver et al. 1998 ) into the EcoRI site of pEG202. The plasmid pAHRGFP (PL1890) was constructed as follows: the GFPS65T cassette ( Heim et al. 1995 ) was amplified by PCR from pRSETBGFPS65T (PL1803) (a generous gift from Dr. Catherine Fox, University of Wisconsin–Madison) using primers OL4125 (5′-ACAGCTCTGAAATTCCAGGTTCTCAGGCATTCCTAAGCAAGGTGCAGAGTGGTCGGGATCTGTACGACGAT-3′) and OL4126 (5′-TTAGCTTGGCTGCAGGTCGACTCGAGCGGCCGCCATGGTCGACGGATCCCACCAGCTGCAGATCTCGAGCT-3′). The amplicon was cloned into the DraIII - digested pAHR by a gap repair method ( Lundblad and Zhou 1997 ). The resulting plasmid was designated PL1855. The coding sequence for amino acids 1–166 of the AHR was amplified by PCR from PL65 ( Dolwick et al. 1993 ) using primers OL4176 (5′-GCTATACCAAGCATACAATCAACTCCAAGCTTGAATTAATTCCGGGCGGAATGAGCAGCGGCGCCAACAT-3′) and OL4177 (5′-CCTTGTGCAGAGTCTGGGTTTAGAGCCCAGTGAAGCTGGCGCTGGAATTCCGCCCGGTCTTCTGTATGGA-3′). The amplicon was cloned into the PmeI/MluI-digested PL1855 by gap repair. The resultant plasmid was designated pAHRGFP (PL1890). High throughput yeast deletion array transformation A high-throughput protocol was developed for 96-well transformation based on work previously described ( Chen et al. 1992 ). Unless otherwise noted, all steps were performed with a Hydra 96-channel dispenser (Robbins Scientific, Sunnyvale, California, United States) and a vortex mixer with a microwell plate adaptor (#12-812 and #12-812C, Fisher Scientific, Hampton, New Hampshire, United States). Deletion strains were stored in a stack of 96-well plates (–80 °C). For transformation, each stock plate was thawed and cells were gently resuspended by vortexing. About 0.5 μl of each strain culture was transferred to a 96-well round bottom target plate (Costar #3795, Corning Inc., Acton, Massachusetts, United States) containing 96 μl per well of yeast extract–peptone–dextrose (YPD) medium plus G418 (200 mg/l). This transfer was accomplished with a 96-pin disposable replicator (GenomeSystems, now Incyte Genomics, Palo Alto, California, United States). The inoculum was incubated at 30°C without shaking until the OD 600 absorbance of individual wells reached 0.2–0.7 (approximately 18 h). The OD 600 was measured using a SpectraMax 250 microplate reader (Molecular Devices, Sunnyvale, California, United States). Cells were then subjected to centrifugation at 3,500 rpm for 8 min, and the supernatant was decanted. The 96-well plates were placed upside-down on a stack of paper towels for 10 min to drain residual medium. For transformation, each plate was vortexed at maximal speed for 15 s before dispensing 22 μl of DNA in “OneStep” buffer (V 1M LiAc :V 50% PEG 3350 = 1:4, with BME added to 0.77% V before use) into each well. To make the DNA in “OneStep” buffer, one volume of DNA (5 μg/μl ssDNA, 0.1 μg/μl each plasmid DNA) was mixed vigorously by vortexing with ten volumes of “OneStep” buffer. After DNA was dispensed, the plate was quickly vortexed again at maximal speed for 10 s to resuspend the cells, followed by incubation at 45°C for 40 min. After this “heat shock” step, 5 μl of the transformation mix from each well was inoculated into a fresh 96-well flat-bottomed plate containing 96 μl per well of dropout medium without Trp, Ura, and His (dropout minus TUH medium) plus G418. The inoculum was gently mixed by vortexing and incubated at 30°C for about 4 d until transformants grew out. The 384-well fluorescence assay for LacZ reporter To perform the LacZ reporter assay, transformants from the 96-well plates were rearrayed into 384-well stock plates containing 30 μl per well of dropout minus TUH medium. The inoculum was incubated at 30°C for 2–3 d to allow cell growth. For the LacZ reporter assay at each agonist concentration, 0.5 μl of cell culture was transferred from the 384-well stock plate (30°C) into a clear-bottomed/black-walled 384-well assay plate (Falcon #353962, Becton Dickinson, Franklin Lakes, New Jersey, United States) using a disposable 384-pin replicator (GenomeSystems/Incyte Genomics). In the 384-well assay plate, each well contained 18 μl of dropout minus TUH medium (diluted 1:4 in water) plus agonist at the tested concentration. The plates were then incubated at 30°C for 48 h to allow all strains to reach stationary phase. Cell growth was monitored by measuring the OD 600 of each well using a SpectraMax Plus 384 microplate reader (Molecular Devices). To initiate the fluorescence assay, 18 μl of lysis/assay buffer was added to each well. Lysis/assay buffer contained a mixture of CUG substrate (#F-2905, Molecular Probes, Eugene, Oregon, United States), 10% SDS, 1 M NaPO 4 , and 25× TAE in the ratio of 1:1.4:350:17.5. For assays with pCEN-AHR transformants, no TAE was required. Plates were vortexed at medium speed for 1 min and left at room temperature for 20 min. The reaction was stopped by dispensing 6.5 μl of 25× TAE to each well and vortexing at medium speed for 1 min. The fluorescence emission of each well was detected using a Wallac “VICTOR V” microplate reader (Perkin-Elmer, Boston, Massachusetts, United States). The fluorescence reading was normalized to the corresponding OD 600 value to obtain the LacZ reporter activity of each deletion strain. In vivo microscopic analysis of AHR–GFP localization Selected deletion strains were transformed with the plasmid pAHRGFP. Transformants were incubated in a 96-well microtiter plate containing 100 μl per well of dropout minus TH medium at room temperature. Given that we have observed that small temperature shifts can affect AHR's localization, we found it more convenient to both grow and examine cells at the same temperature. For some samples, assays were repeated at 30°C using a heating chamber attached to the microscope. Such results were found to be comparable to those obtained at room temperature. For strains that reached early log phase, 0.5 μl of culture was mounted on a glass slide, and the AHR–GFP subcellular localization was examined using a Zeiss (Oberkochen, Germany) Axiovert 200M microscope (α Plan-FLUAR 100× objective). Images were captured using an AxioCam HR digital microscope camera (Zeiss). To stain the nucleus in living cells, 4,6-diamidino-2-phenylindole (DAPI) was added to the dropout minus TH medium to a final concentration of 5 μg/ml. Modifier identification and network analysis To identify deletions that modify AHR signaling, the LacZ reporter activity of each deletion strain was compared to the average of wt BY4742 strain controls included in the same plate, and the fold change was obtained and log 2 transformed. These data-processing steps, as well as subsequent modifier selection, were performed automatically using Perl scripts written “in house.” In brief, for the primary screen involving 4,507 deletion strains with low-copy pCEN-AHR system, a stringent cutoff of 4-fold change over wt control was chosen for selecting a pool of most significant AHR signaling mutants. This cutoff corresponds to a p value of less than 10 –6 at all six assessed concentrations (null distribution: wt control). The initial positives were subject to validation and characterization in secondary screens with high-copy pAHR and control systems. The cutoffs for control pathways pGal4TAD and pAHRΔPASB in the secondary screens were chosen at 2-fold change over wt control, which corresponds to p values of 3.3 × 10 –2 and 5.6 × 10 –4 (null distribution: wt control), respectively. For PIN construction, the main physical interaction table was downloaded from the DIP database ( http://dip.doe-mbi.ucla.edu ) and the genetic interaction table from the MIPS database ( http://mips.gsf.de/proj/yeast/ ). Perl scripts, written “in house,” were used to search the combined physical and genetic interaction database and identify all valid paths (less than or equal to D max ) that linked each pair of modifiers. The network graph was rendered using the Graphviz tool kit ( http://www.research.att.com/sw/tools/graphviz/ ) ( Ellson et al. 2004 ). Within experimental annotation layers of the AHR–PIN, the region corresponding to each functional module was outlined by a closed line (boundary) drawn manually on the network map. This boundary was delineated to include the maximal number of modifier nodes that are members of the corresponding functional module and the minimal number of modifier nodes that are nonmembers. This boundary was also defined in such a way that all enclosed modifier nodes were interconnected via paths within the enclosed region or through at most one modifier node outside. When defining functional modules in the summary AHR–PIN, the highest weight was given to the results from the localization influence experiments because these results provided the most direct indication of a modifier's effect on AHR signaling, and the lowest weight was given to the pharmacology clustering result because this result was highly sensitive to the choice of clustering algorithm. Supporting Information Table S1 Significant AHR Modifiers This table contains all of the ORFs whose corresponding deletion strains reproducibly displayed a significant change in AHR signaling compared to wt BY4742 strain. Also shown are their known gene names, products, gene descriptions, and Gene Ontology (GO) annotations ( Ashburner et al. 2000 ; Issel-Tarver et al. 2002 ). (35 KB XLS). Click here for additional data file. Table S2 AHR-Specific Modifiers This table contains all of the ORFs that were observed to influence the signaling of the AHR but not the pGal4TAD control. Also shown are their known gene names, products, gene descriptions, and GO annotations ( Ashburner et al. 2000 ; Issel-Tarver et al. 2002 ). (27 KB XLS). Click here for additional data file. Table S3 YPD Annotation of AHR Modifiers This table summarizes the annotation on cellular functions of AHR modifiers. The annotation was derived from the YPD database, as of May 2002 ( Costanzo et al. 2001 ) . (23 KB XLS). Click here for additional data file. Table S4 M-Nodes in the AHR–PIN This table contains all of the AHR modifiers that were interconnected in the AHR–PIN (D max = 2). Also shown are their known gene names, products, gene descriptions, and GO annotations ( Ashburner et al. 2000 ; Issel-Tarver et al. 2002 ). (24 KB XLS). Click here for additional data file. Table S5 Functional Enrichment of Network Clusters This table summarizes the functional enrichment of each network cluster as calculated by the hypergeometric distribution of MIPS and GO annotations. For each cluster, the functional enrichment is determined by using M-nodes alone and both M- and I-nodes, respectively. In each case, the annotation that corresponds to the largest number of nodes in the cluster and the smallest p value is shown ( k , number of genes from the query cluster in the given category; f , total number of genes in the given category). (22 KB XLS). Click here for additional data file.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC368161.xml
545055	Years of Life Lost due to exposure: Causal concepts and empirical shortcomings	Excess Years of Life Lost due to exposure is an important measure of health impact complementary to rate or risk statistics. I show that the total excess Years of Life Lost due to exposure can be estimated unbiasedly by calculating the corresponding excess Years of Potential Life Lost given conditions that describe study validity (like exchangeability of exposed and unexposed) and assuming that exposure is never preventive. I further demonstrate that the excess Years of Life Lost conditional on age at death cannot be estimated unbiasedly by a calculation of conditional excess Years of Potential Life Lost without adopting speculative causal models that cannot be tested empirically. Furthermore, I point out by example that the excess Years of Life Lost for a specific cause of death, like lung cancer, cannot be identified from epidemiologic data without assuming non-testable assumptions about the causal mechanism as to how exposure produces death. Hence, excess Years of Life Lost estimated from life tables or regression models, as presented by some authors for lung cancer or after stratification for age, are potentially biased. These points were already made by Robins and Greenland 1991 reasoning on an abstract level. In addition, I demonstrate by adequate life table examples designed to critically discuss the Years of Potential Life Lost analysis published by Park et al. 2002 that the potential biases involved may be fairly extreme. Although statistics conveying information about the advancement of disease onset are helpful in exposure impact analysis and especially worthwhile in exposure impact communication, I believe that attention should be drawn to the difficulties involved and that epidemiologists should always be aware of these conceptual limits of the Years of Potential Life Lost method when applying it as a regular tool in cohort analysis.	Introduction The most common epidemiological exposure-disease effect measures are based on exposure or disease frequency statistics, like risks or odds. Such frequency statistics focus on the question whether an exposure or disease occurred in a population. This information is used to measure the effect of exposure on disease by comparisons of such statistics. Although these measures have been proven by practice and theory to be useful for this purpose, these frequency statistics are unable to reflect all causal effects of exposure in general (Greenland and Robins 1988 [ 3 ], Robins and Greenland 1989 [ 4 ]). One reason stems from the fundamental fact that exposure and disease are processes in time. In particular, if time plays a major role in the link between exposure and disease which is certainly true for long-term exposures and chronic diseases, the question when a disease occurs becomes of paramount relevance. It is important to note, although not widely recognised, that the temporal shift of the onset of disease caused by exposure falls beyond the grasp of conventional statistics based on risks or odds, at least in part. And this shortcoming is even true, albeit perhaps counter-intuitive at first glance, when time-dependent incidence rates are analysed by applying sophisticated time-related statistical procedures like Cox modelling with or without adjustment for time-dependent covariates (Rothman and Greenland 1998 [ 5 ], Greenland 1999 [ 6 ], Morfeld and Piekarski 2001 [ 7 ]). An illustrative example of a Cox analysis in which the true probability of causation can not be derived correctly from the hazard ratio estimate due to an incompletely reflected temporal shift of the disease onset is given as an endnote (see endnote 1). Therefore, alternative measures that focus more directly on the time-shift of events or the time-shift of frequency statistics would be most welcome. One such approach aims at Years of Life Lost (YLL). Interestingly, even in the title of one of the very first articles about Years of Life Lost, Dempsey [ 8 ] expressed the opinion that important aspects are missed by frequency statistics that could be well covered by Years of Life Lost methodology. An overview of different explications of the concept of Years of Potential Life Lost (YPLL) was given by Gardner and Sanborn 1990 [ 9 ]. Moreover, the authors presented a unifying conceptual framework for all these explications of YPLL. In the past the method of Years of Potential Life Lost was mainly used to describe the impact of different causes of death on the survival of a population. This concept was developed further trying to estimate the health effects of specific exposures like smoking (Quellet et al. 1979 [ 10 ], Centers for Disease Control 1989 [ 11 ]). For this purpose, excess Years of Potential Life Lost due to exposure (e-YPLL) were calculated in two steps: first, for each age group the number of excess deaths among the exposed was multiplied by the expected remaining years of life at age at death, given no exposure, and second, these products were summed over all age categories. Recently, this approach was extended by Park et al. 2002 [ 2 ] from SMR-based calculations to YPLL-estimates based on Poisson regression models. The authors illustrated the method with the Colorado Plateau Uranium Miners Cohort (Lundin et al. 1971 [ 12 ]) estimating Years of Life Lost per years worked as a uranium miner, in particular focusing on premature lung cancer deaths. A further application was published by Bailer et al. 2003 [ 13 ] in an analysis of occupational fatal injuries. However, in a letter to the editor, Morfeld 2003 [ 14 ] claimed that the proposed method of estimating excess Years of Life Lost due to exposure is potentially biased in certain settings. The arguments presented were based on the fundamental but abstract article written by Robins and Greenland 1991 [ 1 ] who investigated the estimability of expected Years of Life Lost due to a hazardous substance. The critique raised in Morfeld 2003 [ 14 ] focused on the unavoidable non-identifiability of the true excess Years of Life Lost due to exposure after conditioning on age. However, the train of thoughts was presented in a rather condensed way, and obviously, the arguments were not easy to digest (compare the response by Park et al. 2003 [ 15 ]) Here I try to explain the reasoning in detail. I expand the critique demonstrating additionally that the e-YPLL method is unjustified when applied to specific causes of death like lung cancer – the main topic of Park et al. 2002 [ 2 ]. First, a general framework of causal thinking in epidemiology is developed. This provides a background to analyse the validity of estimation procedures. In the second Chapter the most elementary setting is used to introduce the concepts of excess Years of Life Lost (e-YLL), Years of Potential Life Lost (YPLL) and excess Years of Potential Life Lost (e-YPLL). Then I show that e-YPLL is an unbiased estimator of e-YLL when considering all causes of death. In the third Chapter I prove the potential bias of the e-YPLL estimator when conditioning on age. The proof is given in an elementary setting. Moreover, I illustrate this bias by a realistic life table example. The next Chapter deals with the potential bias of e-YPLL when applied to specific causes of death. This potential distortion is proved in a simple epidemiological setting and demonstrated by a realistic life table calculation. All life tables presented are designed to discuss the analysis published by Park et al. 2002 [ 2 ]: their potentially biased applications of the Years of Potential Life Lost method specific for lung cancer or conditioned on periods of age. These tables are provided in pdf-format as Additional files 1 , 2 and 3 . They are based on two spreadsheets that are available as Additional files 4 and 5 . The final Chapter deals with generalized scenarios, alternative estimation procedures and a brief discussion of the counterfactual approach to causality, even including a hint at its important link to ontological concepts in quantum mechanics. To summarize, the main focus of this paper is first, on a new and simple proof of the potential bias of the e-YPLL method, and second, on an illustration of the degree of potential bias involved by realistic life table examples. These examples are constructed to critically discuss the e-YPLL analysis of a cohort of US uranium miners presented by Park et al. 2002 [ 2 ]. The demonstrated limitation of the e-YPLL approach was first proven rigorously in Jamie Robins's and Sander Greenland's 1991 [ 1 ] breakthrough article on the estimability of expected years of life lost due to a hazardous exposure. My didactic intention is to prove and illustrate the problems involved on a much simpler level of argumentation. Analysis 1. Causal concepts: setting out a cohort gold standard The aim of this paper is to scrutinise the validity of an exposure effect statistic. Therefore, a gold standard is needed against which the statistic can be evaluated to identify and to measure potential biases. A comprehensive cohort investigation is the most natural and purposeful epidemiologic approach towards causality when leaving aside all practical obstacles linked to this study design (Rothman and Greenland 1998 [ 5 ]). To approximate the theoretical gold standard I suppose such a cohort study as being free of selection biases, information errors, losses to follow up, and inferential problems. However, to define an optimal study scenario, I need to assure additionally that contrasts between response statistics of differently exposed sub-cohorts measure correctly the true effect of exposure. This "no confounding" assumption can be visualised by exchangeable sub-cohorts, the most elementary one consisting of only one subject each. These differently exposed but exchangeable subjects are assumed to be like ideal twins: if exposure status had been interchanged between both twins the health response of the subject exposed to exposure level A would have been exactly the response of his/her twin exposed to level B and vice versa. This scenario of totally exchangeable twins is at the bottom of the so called "counterfactual" approach to causality (Lewis 1973 [ 8 ], Rubin 1974 [ 18 ], Maldonado and Greenland 2002 [ 23 ]). It idealises the experimental approach often applied in science by carefully preparing test objects as similary as possible. The aim of this preparation is to measure the effect of an independent variable on a dependent variable in a series of experiments as unconfounded as possible. A mathematical outline of this approach was developed by Neyman 1923 [ 16 ] and even extended to non-experimental studies by Simon and Rescher 1966 [ 17 ], Rubin 1974 [ 18 ] and Holland 1986 [ 19 ]. Overviews are given by Rosenbaum 1995 [ 20 ] and Pearl 2000 [ 21 ]. In addition, Robins 1997 [ 22 ] applied the counterfactual approach to conceptualise and analyse scenarios with interdependent exposure variables, response variables and other covariates that develop in time. I will come back to a critical discussion of this fundamental approach to causality within the Discussion section. At the moment it is sufficient to understand that the gold standard for epidemiology is set out here as an ideal cohort study where exposure levels are allocated to n-tuples of exchangeable subjects. Hence, a gold standard scenario with a binary point exposure assumes that each exposed subject has an ideal unexposed twin so that the exposed sub-cohort consists of all the exposed twins and the unexposed sub-cohort comprises all the unexposed twins. Each causal comparison between the sub-cohorts is based on the pairwise comparison of the differently exposed but perfectly exchangeable twins, at least indirectly (Maldonado and Greenland 2002 [ 23 ]). Obviously, even randomisation of exposure is unnecessary to gain an unbiased estimate of the exposure effect in such a gold standard study. Note that the sub-cohort of unexposed twins constitutes an ideal reference population. It is ideal with respect to 1) being an optimal reference group to define the response of the exposed cohort had it not been exposed (comparison on the group level). Over and above this optimality criterion, the reference is ideal with respect to 2) being an optimal reference on the individual level: the response of each exposed subject can be compared to the response of his/her twin. Note that optimality criterion 2) entails optimality criterion 1). Whereas an external reference population as often chosen in concrete epidemiological studies (Rothman and Greenland 1998 [ 5 ], Breslow and Day 1987 [ 24 ]) can be supposed to fulfil criterion 1), at least approximately, the optimality criterion 2) will not be guaranteed. The reason is that the external reference population cannot provide an unexposed control partner on the individual level. Thus, statistics like SMRs tailored for analysis of such data do not rely on individual level comparison information. It is important to note that these statistics, due to their definition, could not make use of such information even if it were available. Chapter 2 demonstrates that the e-YPLL method can be applied unbiasedly given optimality criterion 2) when applied to all causes of deaths and without any stratification. Chapters 3 and 4 show that e-YPLL stratified on age or specified for certain endpoints, like lung cancer, may be potentially biased even if criterion 2) is assumed. This is so because e-YPLL ignores the information about individual matching. However, this information is crucial as I will demonstrate in Chapters 3 and 4. It follows that a stratified or specified version of e-YPLL is potentially biased if it is used to analyse cohort studies in comparison to an external reference population. 2. Years of Life Lost in an ideal study 2.1 Elementary situation with one pair of twins only In this section I define excess Years of Life Lost (e-YLL) and excess Years of Potential Life Lost (e-YPLL). Then I analyse their relationship in the most simple setting of an ideal study: a pair-matched cohort study with binary point exposure and a death process as outcome. The cohort is assumed to consist of only two subjects, both being differently exposed but totally exchangeable twins. Figure 1 illustrates the scenario. Figure 1 Two exchangeable twins 1 and 2. Birth at b(1) = b(2); allocation of exposure at t(1) = t(2): twin 1 exposed, twin 2 unexposed; death at d(1) ≤ d(2) ; excess years of life lost due to exposure e-YLL = d(2) - d(1); potential death time under no exposure pd(1) = d(2); pd(2) = d(2). Years of potential life lost YPLL(i) = pd(i) - d(i), i = 1,2; excess years of potential life lost e-YPLL = 1(pd(1) - d(1))-0(pd(2)-d(2)) = e-YLL. Both twins (1 and 2) were born on the same day, b(1) = b(2). For simplicity I rescale the time axis and set b(1) = b(2) = 0. Thus, time is identical with age throughout this paper. The point exposure was allocated at an age of t(1) = t(2) > b(1) = b(2) = 0: twin 1 was exposed, twin 2 was never exposed. The twins died at ages d(1) > t(1) and d(2) > t(2), respectively, and deaths are understood as deterministic responses on the individual level. I assume the exposed twin dies earlier or at the same date as the unexposed control twin, d(1) ≤ d(2). Since both subjects are perfectly exchangeable exposure was neutral or detrimental but not preventive. Exposure caused an advancement of death by d(2) - d(1) years. Thus, the true excess Years of Life Lost due to exposure are e-YLL = d(2) - d(1). Next, I introduce potential death times pd(i), i = 1, 2. If both subjects had been unexposed, I would have expected them to die at pd(i) when relying upon the sub-set of the reference population surviving at least to d(i). Since the ideal reference population consists of the unexposed subject only, it immediately follows pd(1) = d(2) and pd(2) = d(2). For this reason I set pd(1) and d(2) equal in Figure 1 . In general, methods like those proposed by BEIR IV 1988 [ 25 ] may be applied to the life table of the reference population to estimate pd(i). I now calculate life expectancy at age at death, YPLL(d(i)), based on the reference population. These Years of Potential Life Lost are defined as YPLL(d(i)) = pd(i) - d(i). Hence, YPLL(d(1)) equals pd(1) - d(1) = d(2) - d(1) = e-YLL in this elementary setting. Obviously I have YPLL(d(2)) = 0. Note that the identity of YPLL(d(1)) and e-YLL relies on the elementary scenario and on the exchangeability condition (together with other usual conditions necessary for study validity). Next, the effect estimate based on YPLL is introduced. The effect estimate of exposure on health is defined as the excess Years of Potential Life Lost, e-YPLL, which is the sum of excess deaths among the exposed (i.e., the observed exposed deaths minus the expected exposed deaths) times Years of Potential Life Lost at age at death, where the summation runs over all death times t (Park et al. 2002 [ 2 ]): e-YPLL = Σ (observed(t) - expected(t)) YPLL(t). Since the number of events is discrete this can always be understood as e-YPLL = Σ observed(d(i)) YPLL(d(i)) - Σ expected(d(j)) YPLL(d(j)), where the first sum is running over all death times d(i) among the exposed twins and the second sum is running over all death times d(j) among the unexposed twins. The number of expected deaths among the exposed at age t, expected(t), is calculated as the product of the risk among the unexposed at t and the number of exposed at t. This definition is analogous to the calculation of exposed cases in an SMR analysis (Breslow and Day 1987 [ 24 ]): baseline risk at t = # unexposed events at t / # unexposed twins at t expected(t) = (baseline risk at t) (# exposed twins at t). If there are no unexposed twins at age t, I define expected(t) to be zero. Two risk sets are given in the simple scenario discussed here: one exposed death occurs at d(1) and one unexposed death at d(2). At d(1) two subjects are under risk: one exposed, the other unexposed. Since the unexposed subject does not die at d(1) we get expected (d(1)) = (0/1) (1) cases among the one (1) exposed at d(1). At d(2) no exposed but one unexposed subject is under risk. Since the unexposed dies at d(2) this leads to expected(d(2)) = (1/1) (0) cases among the zero (0) exposed at d(2). Because one exposed death is observed at d(1) and zero exposed deaths at d(2), i.e., observed(d(1)) = 1 and observed(d(2)) = 0), it follows that e-YPLL = (1 - (0/1)1) YPLL(d(1)) + (0 - (1/1)0) YPLL(d(2)) = YPLL(d(1)) = e-YLL. Of course, using the second representation of e-YPLL given above one also gets e-YPLL = 1(d(2) - d(1)) - 0(0) = d(2) - d(1) = e-YLL. Hence, e-YPLL measures e-YLL unbiasedly, given the elementary scenario and the exchangeability condition (together with other usual conditions necessary for study validity). 2.2 Scenario with two pairs of twins I have proven that e-YPLL is identical to e-YLL in an elementary setting of a cohort consisting of two exchangeable subjects given appropriate conditions. But does this hold if the cohort consists of two or more pairs of exchangeable twins? For didactic reasons I will study a cohort of two pairs of twins first. If we assume that exposure is never preventive, we are left with only two principal scenarios when investigating the situation with two pairs of twins. Indexing the pairs with a and b, the first principal scenario supposes the death times of the four subjects to be ordered as d a (1) ≤ d a (2) ≤ d b (1) ≤ d b (2). Obviously e-YLL = d a (2) - d a (1) + d b (2) - d b (1). Now I will calculate the Years of Potential Life Lost under this scenario 1. For the sake of clarity, I add a second argument to the YPLL function denoting the pair under consideration (a or b). To simplify the notation of YPLL, I drop the letter d that is superfluous since we analyse a scenario without censoring (all subjects die under observation) (i.e., YPLL(a,1) = YPLL(d a (1))). The same simplified notation is used to denote observed and expected numbers of cases. Following Gardner and Sanborn 1990 [ 9 ] and using the ideal unexposed twins as the reference population the Years of Potential Life Lost at d a (1) are given as YPLL (a,1) = d a (2) - d a (1) + 0.5 (d b (2) - d a (2)). This follows because all unexposed subjects survive at least to d a (1) and half of the unexposed subjects survive at least to d a (2). Analogous considerations yield YPLL (a,2) = d b (2) - d a (2) YPLL (b,1) = d b (2) - d b (1) YPLL (b,2) = 0. The observed cases among the exposed are observed (a,1) = 1 observed (b,1) = 1 and the expected, based on the unexposed, are given as expected(a,2) = (1/2)(1) = 0.5 expected(b,2) = (1/1)(0) = 0. Finally, this leads to e-YPLL = d a (2) - d a (1) + 0.5 (d b (2) - d a (2)) - - 0.5 (d b (2) - d a (2)) + + d b (2) - d b (1) = e-YLL. Therefore, the equality of e-YPLL and e-YLL is proven for scenario 1. Next, I will investigate the other principal scenario. In contrast to scenario 1 the second principal scenario assumes the ordering d a (1) ≤ d b (1) ≤ d a (2) ≤ d b (2). Given this scenario 2 we have e-YLL = d a (2) - d a (1) + d b (2) - d b (1) and YPLL(a,1) = d a (2) - d a (1) + 0.5(d b (2) - d a (2)) YPLL(a,2) = d b (2) - d a (2) YPLL(b,1) = d a (2) - d b (1) + 0.5(d b (2) - d a (2)) YPLL(b,2) = 0. Moreover, it follows expected(a,2) = (1/2)0 = 0 expected(b,2) = (1/1)0 = 0. Therefore, e-YPLL = (1)YPLL(a,1) + (1)YPLL(b,1) - (0)YPLL(a,2) - (0)YPLL(b,2) = d a (2) - d a (1) + d a (2) - d b (1) + d b (2) - d a (2) = d a (2) - d a (1) + d b (2) - d b (1) = e-YLL. If we assume exposure to be never preventive all possible configurations of the death times of both twins can be mapped onto these two principal scenarios by renaming the subjects accordingly. Thus e-YPLL = e-YLL is always valid in such an ideal study with two pairs of twins, although e-YPLL does not use the individual matching information. It is important to note the following: the true excess Years of Life Lost can be calculated for each single pair as "observed cases times potential years of life lost" minus "expected cases times potential years of life lost" only in scenario 1. This calculation is obviously invalid under scenario 2. Only the total e-YPLL equals the total e-YLL in both scenarios. Thus, scenario 2 indicates that we have to be careful when a specified version of e-YPLL is applied to estimate the excess Years of Life Lost given specific conditions. Such an analysis is possible without bias if first, optimality criterion 2 is fulfilled as introduced in the last paragraph of Chapter 1, and second, if this individual matching information is used in the analysis. However, this kind of application is not justified if only criterion 1 holds. I will deal with this important issue of potential biases in Chapters 3 and 4. 2.3 Scenario with n pairs of twins A general proof of e-YPLL = e-YLL can be constructed using the principle of mathematical induction. Hence, this Chapter is a little bit technical. However, it may be skipped on first reading because it is not necessary to understand this proof to follow the arguments on potential biases developed in Chapters 3 and 4. For the sake of clarity I introduce an additional functional argument denoting the number of pairs in the study: e-YLL(n) denotes the total excess Years of Life Lost due to exposure in a study with n pairs of twins. To start the inductive argument let us assume that e-YPLL = e-YLL has already been proven to be valid in all studies with n pairs of twins: e-YPLL(n) = e-YLL(n). This is the so called induction assumption. Now I add another pair of twins to the study and ask whether e-YPLL(n+1) = e-YLL(n+1). If I can prove this equality then it follows from the principle of mathematical induction that e-YPLL = e-YLL in all studies. This extension is true because I already have demonstrated that e-YPLL = e-YLL if n = 1 (start of induction). To simplify the argument I now suppose the additional pair to have an unexposed twin with minimal survival among the unexposed n+1 subjects in the extended study. This simplication can always be assumed without any loss of generality for the following reason. If the added pair is not the one with the minimal death time among the unexposed I focus on a pair that fulfils that condition. Of course, such a pair always exists. Let us denote this pair with x and the death times of its twins with d x (1) and d x (2) accordingly. Regarding the other n pairs we know that e-YPLL(n) = e-YLL(n) due to the induction assumption. Hence it suffices to show that e-YPLL(n+1) = e-YLL(n+1) holds in the scenario with the added pair being pair x. This will prove the equation e-YPLL = e-YLL convincingly for all studies with n+1 pairs. Since we assume throughout that exposure is never preventive we know that d x (1) ≤ d x (2). This means that the survival function of the exposed is the same in the extended study group with n+1 pairs as in the study group with n pairs, given we restrict the analysis to all exposed subjects with age at death greater than or equal to d x (2). Moreover, because d x (2) is minimal among the death times of all unexposed subjects the survival function of the unexposed is unchanged too. It follows that the Years of Potential Life Lost (i.e., conditional life expectancy) calculated from the unexposed reference population remains the same for all exposed subjects with age at death greater than or equal to d x (2). In addition, the expected number of cases among the exposed calculated from the unexposed remains unchanged. Therefore, the difference between e-YPLL(n+1) and e-YPLL(n) can only stem from events occurring before d x (2). In the next step I will calculate the change in e-YPLL caused by these events occurring before d x (2). I assume that k of the n exposed in the study with n pairs may die before d x (2), 0 ≤ k ≤ n. Thus, there are n+1 - (k+1) = n - k exposed subjects of the extended study group under risk at d x (2), and it follows expected(x,2)=(n-k)/(n+1). The factor 1/(n+1) is to apply since the first event among the n+1 unexposed subjects occurs at d x (2). The amount of change in e-YPLL due to the additionally expected deaths at d x (2) is therefore given by -[(n-k)/(n+1)]YPLL(x,2). This is the first kind of contribution to the change in e-YPLL I have to consider. The second kind of contribution stems from the new exposed case dying at d x (1). This amount of change is simply the same as the Potential Years of Life Lost for this case: YPLL(x,1) = d x (2) - d x (1) + [n/(n+1)]YPLL(x,2) where n/(n+1) is the probability to survive d x (2) given the unexposed reference population. The third kind of contribution to the difference between e-YPLL(n+1) and e-YPLL(n) stems from the k exposed cases dying before d x (2). Their potential survival is shortened after introducing the new additional death at d x (2). Note that the reference-based probability to die at d x (2) is 1/(n+1) for each of these exposed cases. Therefore, the Years of Potential Life Lost is reduced for each of these cases by the amount [1/(n+1)]YPLL(x,2). This sums to [k/(n+1)]YPLL(x,2) for all k exposed cases dying before d x (2). Now I can calculate how e-YPLL changes when the pair x is added to the n pairs: e-YPLL(n+1) = e-YPLL(n) + d x (2) - d x (1) + + [n/(n+1) - (n-k)/(n+1) - k/(n+1)] YPLL(x,2) = e-YPLL(n) + d x (2) - d x (1) = e-YLL(n) + d x (2) - d x (1) = e-YLL(n+1). The second to last equation follows from the induction assumption. The last equality is based on the obvious fact that the true excess Years of Life Lost increases by d x (2) - d x (1) if the pair x is added to the n pairs I started with. Hence, I have proven that e-YPLL equals e-YLL in all ideal studies consisting of pairs of exchangeable twins given that exposure is never preventive. Note that the measures are always calculated with respect to deaths from all causes. It is remarkable that the equality e-YPLL = e-YLL holds although e-YPLL does not make use of the individual matching information. 3. Non-identifiability of e-YLL conditional on age at death I have shown in Chapter 2 that the total e-YLL of the whole study group can be measured accurately by the total e-YPLL provided the assumptions of an ideal study hold including optimality criterion 2 (cf. last paragraph of Chapter 1) and provided the analysed response is death from all causes. However, this identity of e-YLL and e-YPLL no longer holds if I am interested in e-YLL conditional on age at death. To see why consider the scenarios illustrated in Figure 2 . Figure 2 Two scenarios of different exposure effects in two pairs of twins. The pairs are indexed by a and b. As before, d a (1), d b (1) are death times of the exposed and d a (2), d b (2) are death times of the unexposed twins. Scenario 1 assumes causal effects of the same degree in both pairs, d a (2) - d a (1) = d b (2) - d b (1) = n years. Scenario 2 supposes no effect in pair b, d b (1) - d b (2) = 0 years, and a large one in pair a, d a (1) - d a (2) = 2n years. In both scenarios (Fig. 2 ) the total e-YLL's are identical and thus, are also the total e-YPLL's. However, the e-YLL conditional on age at death varies with the scenarios: the Years of Life Lost due to exposure at d a (1) are smaller in scenario 1 than in scenario 2 whereas it is vice versa at d b (1). Irrespective of the scenario, the e-YPLL are always calculated in the same way. Therefore, we get in both scenarios the same e-YPLL at d a (1) as well as the same e-YPLL at d b (1). Note that the calculation of e-YPLL does not make any use of the individual matching information. Hence, the true excess Years of Life Lost conditional on age at death are not identifiable without having access to a perfect control twin or without supposing a specific mechanism for how exposure causes death. Note that criterion 1 alone – as defined in the last paragraph of Chapter 1 and hopefully fulfilled in usual cohort data – does not render the excess Years of Life Lost stratified on age at death identifiable. Consequently, estimates based on e-YPLL conditional on age at death are potentially biased in all usual settings. Next, I analyse a theoretical birth cohort of 100,000 men to demonstrate this potential bias by a realistic life table example (Table 1, see Additional file 1 ). I assume the death rates of the male US population as tabulated in BEIR IV 1988 [ 25 ] as reference rates. The example is restricted to the unexposed men surviving at least to an age of 30 years. Life expectation at death is calculated from the size of the unexposed population and the number of unexposed deaths, presuming that all deaths occur at the midpoint of age categories. Numbers of exposed deaths include a certain advancement of age at death without changing the totals. Excess deaths are calculated as the difference between the number of exposed deaths and the expected deaths among the exposed. The latter is determined by multiplying the size of the exposed population by the fraction of deaths among the unexposed population conditional on age. Finally, e-YPLL is derived according to Park et al. 2002 [ 2 ] by multiplying the number of excess deaths by the life expectation at death in each age category (see Chapter 2.1 for details). The total number of excess Years of Life Lost due to exposure sums up to 255,094 years. Table 2 (see Additional file 2 ) describes the impact of two different causal exposure-response mechanisms, both compatible with the data presented in Table 1 (see Additional file 1 ). Mechanism 1 assumes two different response types of subjects. One type is immune and does not react under exposure in comparison to no exposure (no advancement of age at death). The other type shows an adverse effect of exposure because his death is advanced by five years. In contrast, mechanism 2 supposes three different reaction types: an immune one and two types adversely affected by exposure. The affected types differ in their degree of response (advancement by five years or ten years). Tables 1 and 2 are based on a spreadsheet that is supplied as an additional data file (see Additional file 4 ). Note that the total numbers of deaths, unexposed and exposed, as well as the total excess Years of Life Lost are identical in both sub-tables of Table 2 (see Additional file 2 ) although the sub-tables reveal a different causal impact of the varied mechanisms. Note further that the distribution of unexposed deaths as well as the distribution of exposed deaths are constant across both sub-tables and Table 1 (see Additional file 1 ). Moreover, the total e-YLL calculated in the sub-tables of Table 2 (see Additional file 2 ) agrees with the total e-YPLL derived in Table 1 (see Additional file 1 ). But obviously, the excess Years of Life Lost conditional on age differ between mechanism 1 and 2, and moreover, diverge from the conditional excess Years of Potential Life Lost calculated in Table 1 (see Additional file 1 ). Figure 3 contrasts the distributions. Figure 3 Excess Years of Potential Life Lost e-YPLL and true excess Years of Life Lost e-YLL1 and e-YLL2, assuming two different causal mechanisms, conditional on age at death. Basic data, exposure impact, and mechanism producing e-YPLL, e-YLL1, and e-YLL2 are given in Table 1 and Table 2. Lost Years, true and estimated, sum up to 255,094 years always. The excess Years of Potential Life Lost differ remarkably from the true excess Years of Life Lost. The difference depends on the chosen mechanism. In this example e-YPLL overestimates the true excess Years of Life Lost up to an age of about 65 years and underestimates the impact at higher ages. The zigzag increase of e-YPLL may reflect a real phenomenon since mechanism 2 produces a similar structure. However, the somewhat surprising zigzag increase of e-YPLL may be totally artificial because mechanism 1 causes a nearly smooth incline. Thus, it is not justified to interpret the form of the increase of e-YPLL across age as a true phenomenon without assuming untestable mechanistic assumptions about how exposure causes death. Note that such naive interpretations can be grossly misleading. Hence, as illustrated by this example, the true excess Years of Life Lost conditional on age at death cannot be identified from cohort data like that presented in Table 1 alone (survival curves, see Additional file 1 ). The estimator e-YPLL conditional on age is shown to be potentially biased even if optimality criterion 2 (as defined in the last paragraph of Chapter 1) is fulfilled. The reason is that e-YPLL conditional on age does not make use of the relevant individual matching information. 4. Non-identifiability of Years of Life Lost due to a specific cause of death The equality of total e-YPLL and total e-YLL was proven for deaths from all causes given the assumptions mentioned in Chapter 2. However, such an equality does not hold for a specific cause of death like lung cancer. To see why, consider the simple scenario presented in Figure 4 . Figure 4 Two exchangeable twins 1 and 2. The exposed twin died from lung cancer at age d(1), the unexposed twin 2 died from heart attack at d(2) > d(1). Figure 4 illustrates an elementary gold standard cohort study performed with two twins, one exposed and the other unexposed. Since the subjects are assumed to be exchangeable, exposure caused a premature death with an overall e-YLL = d(2) - d(1). However, if we are interested in the exposure impact on age at death for lung cancer we are left with an identification problem. Obviously, if twin 2 had died from lung cancer also the causal effect on age at death from lung cancer would have been d(2) - d(1). But in case of the scenario illustrated in Figure 4 , one can only conclude that the effect on age at death from lung cancer must be at least d(2) - d(1). Therefore, as a simple but important consequence, the causal effect on age at death from lung cancer is not identifiable even in an optimal study comprising exchangeable twins as long as competing causes of death exist. Since this causal effect cannot be identified even in an optimal study (i.e., fulfilling condition 2 defined in the last paragraph of Chapter 1) it does not make sense to estimate it without assuming untestable mechanistic assumptions about how exposure causes death from lung cancer. Note, however, that this causal effect is not the e-YLL due to exposure and manifested by lung cancer, e-YLL(lung cancer) for short. In contrast to the causal effect, e-YLL(lung cancer) comprises exactly that part of e-YLL from all causes of death that is due to the occurrence of lung cancer deaths among the exposed. In the example (Figure 4 ) these e-YLL(lung cancer) are d(2) - d(1) which is the exposure effect on the age at death from all causes, e-YLL(overall) for short. The reason ist that there are no other causes of death among the exposed besides lung cancer. Whereas the causal effect of exposure on age at death from lung cancer cannot be identified in an optimal study with pairs of twins fulfilling optimality criterion 2 (defined in the last paragraph of Chapter 1) without adopting further untestable assumptions we may ask whether at least e-YLL(lung cancer) can be estimated unbiasedly assuming criterion 2. To give an answer we must investigate the situation in more detail. The concept presented supposes that each subject has a set of hypothetical (deterministic) death times for each combination of level of exposure and cause of death. Just one of these death times is effective, the others are latent. For the exposed twin 1 lung cancer is effective as a cause of death and heart attack is latent. It is vice versa for the unexposed twin 2. Both causes of death compete within each twin for the effective position. Of course, the result of this competition depends on the exposure conditions applied. This result motivates to introduce a second (competing) exposure into the scenario, in which I deal with two competing responses (lung cancer and heart attack). Figure 5 expands on this concept accordingly. Figure 5 Effective and latent causes of death in a quadruplet of exchangeable subjects. Each subject is supposed to have been allocated to a different combination of two binary exposures, illustrated here by asbestos and smoking. Without any exposure and as well as after asbestos exposure the subjects are assumed to have died from lung cancer, LC (effective cause). Lung cancer was latent in the smoker not exposed to asbestos while heart attack, HA, became the effective cause in this situation. Asbestos exposure had no causal effect on age at death from heart attack but from lung cancer. Smoking affected age at death from heart attack and lung cancer, the latter to the same amount as asbestos did. Under joint exposure a synergistic effect on age at death from lung cancer is assumed, but not for heart attack. In Figure 5 both exposures are assumed to be never preventive for all endpoints of interest (LC = lung cancer, HA = heart attack). However, when contrasting the two situations without asbestos exposure smoking obviously leads to a decrease in the number of lung cancer deaths, and correspondingly, asbestos exposure causes a decrease in the number of deaths from heart attack given smoking. This spuriously preventive effects of both exposures are an outcome of the competing causes of death structure. In the following, I prove that this structure leads to a potential bias of e-YPLL(lung cancer). Note that the misleading impression of a preventive effect is produced by relying on a risk or rate statistic only. As I emphasized in the Introduction section, these statistics count accurately how many responses of what kind occurred ("whether"), but unfortunately, mainly ignore the specific causal structure in time ("when"). From Figure 6 I conclude Figure 6 The effective data of Figure 5 analysed in an ideal cohort study consisting of two pairs of twins which are supposed to form a quadruplet of exchangeable subjects, cmp Figures 1 and 2. The pairs are indexed by a and b: the twins in pair a are assumed to have been smokers, those in pair b non-smokers; d a (1), d b (1) are death times of the asbestos exposed and d a (2), d b (2) are death times of the twins not exposed to asbestos. The effective cause of death, lung cancer LC or heart attack HA, are denoted at age at death, accordingly. e-YLL(overall) = e-YLL(lung cancer) = e-YPLL(overall) = d a (2) - d a (1) + d b (2) - d b (1). The first equation holds because the only cause of death among all asbestos exposed is lung cancer and the second because I have shown that e-YLL(overall) = e-YPLL(overall) is true in an ideal study given the aforementioned assumptions. The third equation follows from a simple evaluation of e-YLL(overall). Next, I calculate e-YPLL(lung cancer) according to Park et al. 2002 [ 2 ]: e-YPLL(lung cancer) = Σ (observed lungcancer (t) - expected lungcancer (t)) YPLL(t). The only change in comparison to the formula of e-YPLL introduced in Chapter 2.1 is the specification of the observed and expected cases. The important point here is that the procedure, as defined by Park et al. 2002, only uses information at age at death from lung cancer. For the YPLL-terms involved I get YPLL (a,1) = d a (2) - d a (1) + 0.5 (d b (2) - d a (2)) YPLL (b,1) = d b (2) - d b (1), from which I derive e-YPLL(lung cancer) = d a (2) - d a (1) + 0.5 (d b (2) - d a (2)) + + d b (2) - d b (1) = e-YLL(lung cancer) + 0.5 (d b (2) - d a (2)), since no expected lung cancer deaths are to be subtracted. Hence, e-YPLL(lung cancer) is biased upward by 0.5 (d b (2) - d a (2)) which is half of the causal effect of smoking on age at death from all causes (overall) according to Figure 5 . This overestimate is due to the fact that no lung cancer deaths are expected to occur among the exposed during follow-up. This empirical shortcoming of e-YPLL(lung cancer) could only be overcome in general if we were able to identify the matched partner of the lung cancer case not dying from lung cancer and if we use this information in the analysis. Since this is impossible when relying on usual cohort data (survival curves) I have proven the non-identifiability of the excess Years of Lost Life due to exposure for a specific cause of death, given the natural situation of competing causes for death and usual cohort data. I demonstrate the potential bias of e-YPLL(lung cancer) by a life table example additionally (Table 3, see Additional file 3 ). This table relies on a spreadsheet which is also supplied (see Additional file 5 ). Table 3 (see Additional file 3 ) uses the same basic data as Table 2 (see Additional file 2 ) to describe the unexposed population but is extended by an additional column presenting the number of lung cancer deaths by age. The impact of exposure on lung cancer death is defined by an advancement of age at death by five years in all situations. Thus, factual (effective cause) age at death as well as hypothetical (latent cause) age at death is assumed to occur five years earlier under exposure. Furthermore, it is supposed that the advancement of latent lung cancer deaths leads to an excess of 50% among unexposed lung cancer deaths in each age category, if exposed. This advancement means that 3367 (= 10102 - 6735) subjects died from lung cancer after exposure who are assumed to have died from other causes if unexposed. A mixture of advancements greater or equal than 0 years, differing in amount between other causes of death than lung cancer, is assumed to produce the distribution of the number of deaths from all causes among the exposed. Hence, the overall excess Years of Life Lost for lung cancer are less than the number of exposed lung cancer deaths times five years (i.e., 50,509 years). This number is an upper bound because the additional lung cancer deaths occurring under exposure are related to a loss of life in the individual of at most five years while the cause of death changes from, say, heart attack to lung cancer. Applying the procedure proposed by Park et al. 2002 [ 2 ] I calculate a total e-YPLL(lung cancer) of 86,373 years. Thus, the estimate is biased upward by more than 35,000 years which corresponds to a relative bias of at least 70%. In addition, the life table analysis demonstrates that the e-YPLL method is also biased when used to estimate the causal effect of exposure on age at death from lung cancer among the exposed lung cancer cases, which exactly amounts to 50,509 years. Moreover, the causal effect of exposure on the lung cancer death times among all subjects was defined to be 5(92,219) years = 476,095 years, again clearly different from e-YPLL(lung cancer) = 86,373 years. 5. Discussion I have shown that the total excess Years of Life Lost (e-YLL) for death from all causes can be accurately determined by calculating the corresponding excess Years of Potential Life Lost (e-YPLL) provided conditions hold that are also often cited for general study validity. However, the equality of e-YLL and e-YPLL does not hold in general under certain conditions of interest because under these certain conditions we need to assure criterion 2: each exposed subject has to have an ideal unexposed control partner so that the effect of exposure can be measured on the individual level (see also the last paragraph of Chapter 1). The equality also requires that this information about individual matching is used in the analysis. I pointed out that the e-YPLL conditional on age at death are potentially biased as they were published in some analyses (see e. g., Figure 1 in Park et al. 2002 [ 2 ]). Furthermore, the excess Years of Potential Life Lost calculated for a specific cause of death are potentially biased also. Hence, estimates of Years of Life Lost presented for chronic diseases like lung cancer (see for example the detailed lung cancer Poisson model analysis by Park et al. 2002 [ 2 ]) are unjustified without referring to untestable assumptions about the causal mechanism involved. I should add, just for the sake of clarity, that in all my examples presented that cast doubt on the validity of e-YPLL, exposure was always assumed to be never preventive. Therefore, preventive effects of exposure are not the reason for these problems. The deeper reason of these validity problems of e-YPLL is a non-identifiability of excess Years of Life Lost due to exposure in an ideal but unmatched cohort study. Note further that the calculation of e-YPLL will ignore this important matching information even if such data are available. This structural problem of non-identifiability was already clearly described, analysed and discussed on an abstract level by Robins and Greenland 1991 [ 1 ], even under the generalized scenario of stochastic responses on the individual level. For further theoretical discussions of the problems involved and for an investigation into untestable assumptions that render identification possible, I therefore like to refer to this important publication. I simply remark here that mechanisms 1 and 2 (Table 2, see Additional file 2 ) do not preserve ranks. It is important to note that the validity problems of e-YPLL cannot be resolved by applying the rare disease assumption or by restricting the discussion to a situation with a constant relative risk across age. Rather, the problems are structural ones. For example, if the additional lung cancer cases under exposure are calculated in Table 3 (see Additional file 3 ) by applying an increasing factor with age at death instead of using the constant value of 1.5, the SMR's for lung cancer can be kept almost constant across age. However, e-YPLL still overestimates the true excess Years of Life Lost due to lung cancer by a considerable amount. A substantial upward bias, nearly independent of the increasing factor, can be demonstrated by simulations that assume an isolated and homogeneous detrimental effect of exposure on lung cancer. This describes a scenario where other endpoints are only changed due to exposure because they are competing with lung cancer. Again using the data from BEIR IV 1988 [ 25 ] (Table 3, see Additional file 3 ) I find an upward relative bias of at least 30% if the additional lung cancer deaths (that take over the effective position from competing causes if exposed) amount to about half of the lung cancer deaths shifted to a younger age at death (under exposure) without changing the endpoint. The examples I studied show that the relative bias is at least 70% if both groups are of the same size. In agreement with the direction of bias indicated by the abstract quadruplet analyses of total excess Years of Life Lost for lung cancer (Figures 5 and 6 ) I show an upward distortion of e-YPLL(lung cancer) in the life table example (Table 3, see Additional file 3 ). This upward bias is due to an inappropriate mixture of life expectancy at age at death from lung cancer, based on data from all later deaths, with lung cancer excess deaths only, while ignoring the corresponding excess deaths from competing causes among the non-exposed that are produced as a side-effect of the causative exposure (against the interpretation of Park et al. 2003 [ 15 ]). However, total e-YPLL for a specific cause of death can be used to assess total e-YLL for this cause unbiasedly if despite potentially competing causes no change of the cause of death occurred within the whole cohort after a different exposure level of interest had been allocated counterfactually. Unfortunately, this assumption is not provable. When excess cases of the cause of death of interest are observed among the exposed – a situation quite natural in occupational cohort studies – this assumption definitely does not hold. My analysis reported so far on problems with ideal studies with complete follow-up. Obviously, if a constant end of follow-up is introduced into the scenarios of Figure 5 and 6 so that the number of lung cancer deaths are affected, the calculated e-YPLL can change whereas the true e-YLL always remain constant. Hence, the degree of bias depends additionally on the censoring structure even if left censoring is totally uninformative. This dependence means that censoring is not taken into account adequately by the e-YPLL method when applied to specific endpoints. One important point to discuss is whether the constructed life table examples are realistic ones in comparison to the data and results published in Park et al. 2002 [ 2 ]. All tables presented here are based on the mortality experience of the male US population as published by BEIR IV 1988 [ 25 ], Table 2A-10, p. 139. Since the paper of Park et al. 2002 [ 2 ] analysed the mortality of white US uranium miners, hired during 1950 - 1963 and followed until 1990, the choice of this reference population appears to be appropriate. Park et al. 2002 [ 2 ] described the total excess Years of Life Lost per cohort member as 3.1 years (Table IV, page 6) and an effect of similar magnitude was chosen here: 2.7 years per cohort member (n = 95,219) according to Tables 1 or 2 (see Additional file 1 and 2 ), and 3.3 years per cohort member according to Table 3 (see Additional file 3 ). In Park et al. 2002 [ 2 ] the lung cancer SMR dropped from 7.6, age period 30–34, to 5.4, age band 45–49, and even further to SMR = 2.6, age period 65–69 (Table III, page 5). The example given in Table 3 (see Additional file 3 ) tries to mimic this downward trend starting with 7.0 in age category 30–34, then decreasing to 3.0 in age band 45–49 and to 1.7, age period 65–69. Thus, the overall downward trend is somewhat more pronounced here than in Park et al. 2002 [ 2 ]. However, I try to simulate a trend of SMR's across age that is quite similar to the relative decrease observed within the main body of data in Park et al. 2002 [ 2 ] (age bands 65–69/45–49): 2.6 / 5.4 = 48% (Table III, Park et al. 2002 [ 2 ]) and 1.7 / 3.0 = 57% (Table 3 of this paper, see Additional file 3 ). The overall lung cancer SMR can be calculated from Table 3 (see Additional file 3 ) as 1.74, less than the value of 3.8 observed by Park et al. 2002 [ 2 ] (Table III, page 5). A lower SMR was chosen here to avoid an exaggeration of potential biases due to overstating the effect of exposure in the constructed example. Accordingly, the e-YPLL for lung cancer was presented as 1.47 years per cohort member by Park et al. 2002 [ 2 ] (Table IV, p. 6) but only 0.91 years per cohort member (n = 95,219) can be calculated from Table 3 of this paper (see Additional file 3 ). Hence, the life table examples presented in Tables 1, 2 and 3 (see Additional file 1 , 2 and 3 ) are realistic ones and the magnitude of potential biases identified appears to be defensible. The potential biases derived are fairly extreme: the true total Years of Life Lost are overstated by 100% (mechanism 1) or 33% (mechanism 2) in age band 50–54 but understated by more than 40% in age period 75–79 (Figure 3 and Tables 1 and 2, see Additional file 1 and 2 ). Moreover, a relative upward bias of more than 70% is demonstrated for the e-YPLL related to lung cancer (Table 3, see Additional file 3 ). These findings render the central results published by Park et al. 2002 [ 2 ] scientifically unreliable (i.e., the age-specific analysis of excess Years of Life Lost due to exposure as well as all results published on excess Years of Life Lost in relation to lung cancer). Note that the latter was the main topic of Park et al. 2002 [ 2 ]. Due to its definition (Breslow and Day 1987 [ 24 ]) the average lung cancer SMR is calculated as observed / expected = exposed / (exposed - excess) which leads to 10,101.74 / (10,101.74 - 4,286.88) = 1.74 (Table 3, see Additional file 3 ). For the sake of clarity I should note that this value does not coincide with the average causal case ratio (exposed / unexposed = 10,101.74 / 6,734.49 = 1.5). The discrepancy is due to the fact that the construction of all life tables in this paper strictly follows counterfactual principles: the elevated death rates in the exposed population affects the person-years yielding an expected number of lung cancer cases smaller than the true observed number of lung cancer cases in the unexposed population (cp. Keiding and Vaeth 1986 [ 26 ], Greenland 1996 [ 27 ]). Statistical measures that focus on the advancement of disease onset are assumed to be helpful in communicating risk factor impact on disease (e.g., Peto 1980 [ 28 ]). Fischhoff et al. 1993 [ 29 ], Jardine and Hrudey 1997 [ 30 ] as well as Yamagishi 1997 [ 31 ] pinpointed problems in risk perception and communication when frequency statistics are solely presented as descriptors of health effect. Weinstein et al. 1996 [ 32 ] demonstrated additionally a considerable effect of framing risk levels on the perception of risk. Statistics conveying information about the advancement of disease onset are therefore helpful in exposure impact analysis and especially worthwhile in exposure impact communication. However, attention should be drawn to the difficulties involved and that epidemiologists should always be aware of the conceptual limits of the Years of Potential Life Lost method when applying it as a regular tool in cohort analysis. Aside from Years of Life Lost, other approaches are available to convey information on the impact of a risk factor on the onset of a disease and may thus facilitate communication of epidemiological findings. One such concept is the risk and rate advancement period (RAP) introduced by Brenner et al. 1993 [ 33 ] which could easily be calculated from standard software output. Risk and rate advancement periods are the time periods by which the risk or rate of disease is advanced among exposed subjects given a disease-free survival to some baseline age. An application of RAPs in assessing the impact of risk factors on myocardial infarction was published by Liese et al. 2000 [ 34 ]. However, the RAP statistic cannot be applied if disease rates do not increase strictly with age. Another approach describing the exposure impact on disease onset, complementary to RAP, was developed by Boshuizen and Greenland 1997 [ 35 ]. The authors estimated the time shift in average age at first occurrence of disease due to exposure as a measure of disease advancement. Although especially valuable when the background incidence of the disease is high, a technical drawback of this procedure stems from the necessity to correct regression model likelihoods for left truncation, an option that is not offered in standard statistical packages. Of course, none of these procedures can be applied sensibly if, according to study design, cases and controls or exposed and unexposed were matched on age. A rigorous causal approach in estimating the shrinkage or extension of time to an event, like lung cancer death, due to exposure was developed by Jamie Robins and is called G-estimation (for a brief overview see Rothman and Greenland 1998 [ 5 ], p. 424, 425). Since this method has a profound logical basis and can be applied even in complicated longitudinal scenarios with interrelated time-dependent covariates, and in addition, is programmable in standard software like SAS or STATA it should become a regular tool in cohort analysis, in particular when Years of Life Lost are to be estimated. The main conclusions about non-identifiability of e-YLL are derived as an application of a causal theory based on counterfactuals (for a review see Greenland 2000 [ 36 ]). Some authors seem to believe that this approach has no clear logical and philosophical backing, judging Greenland's reasoning as purely academic (Armstrong and Thieriault 1996 [ 37 ]) or Robin's counterfactual based methods like G-estimation as not applicable in occupational epidemiology (Steenland et al. 1996 [ 38 ]). The arguments and examples given in this article prove the applicability and practical relevance of the counterfactual approach in occupational epidemiology. However, this does not help to eliminate a more fundamental philosophical resistance I experienced in a number of controversial discussions (e. g., with epidemiologist from NIOSH, Cincinnati). Thus, some explanation of the philosophical background of counterfactual reasoning should be given here. The ideas of counterfactual reasoning can at least be traced back to philosophers in the eighteenth and nineteenth centuries, like Hume and Mill. Even the oldest clearly structured theory of causality, developed by Aristotle, has some similarities to counterfactual reasoning due to its manipulative four-causes concept (Vorländer 1979 [ 39 ]). In its most simple form, the counterfactual approach assumes the existence of potentially different hypothetical response variables for the same subject depending on differently assumed exposure conditions. At most one of these conditions and responses can become true, the others remain hypothetical, so counterfactual. A causal comparison is understood as a comparison of the responses due to different exposure conditions within the same subject. Indeed, this is the rigorous version of ideal twins I have used throughout to derive statements about causality (see Figures 1 , 2 , 4 , 5 , 6 ). In their famous statistical textbook, Cox and Oakes 1985 [ 40 ], p. 64 implicitly used this approach to define the stronger version of the accelerated failure time model: "any individual having survival time t under z = 0 would have survival time t/Ψ under z = 1". The counterfactual nature of this statement is reflected by the tense used (conditional II). Note that modern physics also uses counterfactual reasoning, expressed in conditional II, to define causes in the framework of the general theory of relativity: "We say that an event, let us call A, is in part the cause of another event, B, if A was necessary for B to occur. If A had not occurred, B could not have. In this case I can say that A was a contributing cause of the event B" (Smolin 2001 [ 41 ]). And note further that also everyday causality is often expressed by statements in conditional II (i.e., by statements that describe situations that contrary to fact did not occur – "If the train had not stopped for such a long time in front of the railway station I would have reached the connecting train"). A detailed philosophical discussion of the counterfactual causal model is given in Lewis 1973 [ 42 ]. Some authors resist a counterfactual approach and argue against the speculative and metaphysical background of counterfactual worlds (Dawid 2000 [ 43 ]): How can one object have two different contradictory properties like being exposed (factually) and simultaneously not being exposed (counterfactually)? In addition, the units of observation are assumed to exist in different states simultaneously in the counterfactual world if the properties have more than two distinct values. This duality appears to be even more puzzling. However, exactly such weird statements are made by quantum mechanics about objects like electrons or Fulleren molecules which are part of the real world we live in (Mittelstaedt 1972 [ 44 ], Feynman 1985 [ 45 ], Zeilinger 2003 [ 46 ]). It is important to understand that this superposition principle ("superposing" contradicting properties simultaneously on the same object) is no marginal phenomenon but lies at the center of modern physics (Dirac 1967 [ 47 ]). Note further that the mathematically consistent analytical treatment of causal questions by counterfactual theory is obviously related to the so called multiverse approach (Everett 1957 [ 48 ]). The latter is a suggestive ontological interpretation of abstract Hilbert space theory that was introduced into modern physics to represent quantum states (Weberruβ 1998 [ 49 ], Penrose 1994 [ 50 ], Deutsch 1997 [ 51 ]). A new mathematical approach was necessary because – due to the superposition principle – quantum states are much richer than classical states and therefore fall beyond the grasp of Newtonion and Einsteinian terminology and theory (Hughes 1999 [ 52 ]). Critics of counterfactual logic, like Dawid 2000 [ 43 ], overlook this deep connection to quantum mechanics that renders the approach empirically plausible and mathematically as consistent as modern physics. In other words, a fundamental philosophical attack of counterfactual reasoning leads inevitably into an attack of quantum physics, which has survived successfully all criticism raised, empirically and theoretically, during the last century (Feynman 1985 [ 45 ], Hawking 1996 [ 53 ], 2001 [ 54 ]). In a discussion of the so called null measurements in quantum mechanics Roger Penrose described the link of quantum mechanics to counterfactuals as follows: "It is quite extraordinary that quantum mechanics enables you to test whether something might have happened but didn't happen. It tests what philosophers call counterfactuals. It is remarkable that quantum mechanics allows real effects to result from counterfactuals!" (Penrose 2000 [ 55 ], p. 67). In addition, this link to quantum mechanics disproves the repeatedly made statement by Dawid (cp. the discussion following Maldonado and Greenland 2002 [ 23 ]) that counterfactual reasoning were subject to a deterministic or 'fatal' world view. On the contrary, quantum mechanics is the realm of pure chance and statistics: counterfactuals are therefore clearly consistent with indeterminism and stochastic principles. Finally, critics of counterfactual logic, like Dawid 2000 [ 43 ], have not presented an appealing and logically consistent alternate mathematical theory of causality. Whereas the counterfactual approach can help to clarify terminology and substance of causal relations, it points simultaneously at some ambiguities when discussing competing causes of death. In this scenario, it is unclear what kind of action should be taken to cause a suppression of competing risks (Greenland 2002 [ 56 ]). Hence, the logical and causal background of the example presented with competing causes of death (Figure 4 , 5 , 6 and Table 3, see Additional file 3 ) appears to need further elaboration. Conclusion In conclusion, the excess Years of Potential Life Lost estimates the excess Years of Life Lost due to exposure unbiasedly if we are interested in a) death from all causes and b) total excess Years of Life Lost summed up across the whole cohort. However, the method of calculating excess Years of Potential Life Lost due to exposure is potentially biased if it is applied 1) to estimate the impact of exposure on specific causes of death, like lung cancer, in the presence of competing causes or 2) to estimate the impact of exposure (overall or cause specific) conditional on age at death. These potential biases can be rather severe in published analyses (e.g., Park et al. 2002 [ 2 ]). Rigorous causal thinking (Greenland et al. 1999a [ 57 ], Greenland et al. 1999b [ 58 ]) can help to identify and avoid such empirical shortcomings. Fortunately, as an important outcome of these causal considerations, methods are available (Robins 1997 [ 22 ]) and readily applicable in occupational epidemiology (Witteman et al. 1998 [ 59 ]) to estimate the cause-specific reduction in life span due to occupational exposures unbiasedly given a specified failure time model and the assumption of no unmeasured confounders. These methods are valid even under the complicated but often realistic conditions of dependent censoring, survivor biases and intermediate confounding (Keiding et al. 1999 [ 60 ], Morfeld et al. 2002 [ 61 ]). Competing Interests The author(s) declares that he has no competing interests. Endnote 1 The following example is based on the counterfactual framework presented in Chapter 1. It demonstrates that neither relative risks nor relative rates can be used in general to estimate the probability of causation unbiasedly. In particular, I show that an estimate of the attributable hazard derived from a Cox model (Breslow and Day 1987 [ 24 ]) fails to describe the causally attributable fraction among the exposed cases. The example demonstrates that this shortcoming is due to the fact that an advanced onset of disease is not reflected completely by risk or rate statistics. As in Chapter 1, a binary exposure indicator is assumed for simplicity in the following. The cohort is supposed to comprise four subjects: A, A's twin, B, and B's twin. In addition, it is assumed that the cohort is followed up in mortality until the fixed censoring date t end >0. The time scale can be chosen arbitrarily as age, calendar time or time since start of follow-up without affecting the following arguments. I assume that the exposed subject A may experience the event (death) during the follow-up period at t 1 years (t 1 > 0), if unexposed counterfactually (A's twin) at t 2 > t 1 , t 2 < t end . No event may occur in subject B and his/her twin during follow-up (i.e., neither under exposure nor when unexposed). Thus, among the two exposed subjects, A and B, only one case occurs and this case is causally affected by exposure since t 1 < t 2 . It follows that the probability of causation (i.e., the percentage of exposed cases occurring during follow-up which are causally affected by exposure) is exactly 100%. In contrast, the incidence proportion (Rothman and Greenland 1998 [ 5 ]) is 0.5 among both, the exposed and unexposed, leading to a relative risk of 1 and an attributable risk among the exposed of 0. Hence, the attributable risk calculated from incidence proportions fails to reflect the causal impact of exposure on subject A's event time (true probability of causation among the exposed = 100% >> attributable risk among the exposed calculated from incidence proportions = 0%). I'd like to emphasize that the reason for this shortcoming is the fact that relative risks reflect excess cases only but no advanced cases. Note that a change from risks to rates does not overcome the problem. The rate among the exposed is 1/(t 1 + t end ) and among the unexposed is 1/(t 2 + t end ). Consequently, the rate ratio is (t 2 + t end )/(t 1 + t end ) > 1 because t 2 > t 1 . The attributable rate among the exposed can be determined as (rate ratio -1) / rate ratio = 1 - 1/rate ratio. Thus, I conclude 0 < 1 - (t 1 + t end )/(t 2 + t end ) < 1, again proving a systematic underestimate of the true probability of causation among the exposed by the attributable rate ratio. Note that the rare disease assumption is of no help. Assuming n exposed subjects (n>>1) that neither react nor do their n unexposed twins, we get an attributable rate among the exposed of 1 - (t 1 + (n)(t end )) / (t 2 + (n)(t end )) < 1. If n approaches infinity the attributable rate among the exposed decreases to zero whereas the true probability of causation remains always constant at 100%. Hence, the discrepancy is even sharpened under the rare disease assumption. Last, I try to escape these problems by performing a Cox analysis. Due to its construction, the whole cohort (exposed and unexposed) comprises 4 subjects and 2 risk sets. One set is generated by the exposed case (A), the other set by the unexposed (A's twin). Therefore, a Cox analysis of the cohort yields the following. Assuming a relative hazard (rate) in the Cox model of λ / λ 0 = exp (β exposure), exposure = binary exposure indicator we get the partial likelihood which is maximized at b = ln 2/2. Therefore, the estimated hazard ratio is exp (b) = √ 2 yielding an estimated attributable hazard among the exposed of 1 - (√2) -1 < 1 = probability of causation. Note that the rare disease assumption is again of no help to overcome this discrepancy because the estimated attributable hazard among the exposed approaches 0 when the number of controls is rising indefinitely. Greenland 1999 [ 6 ] emphasizes correctly that this methodological error (i.e., using attributable risks or rates among the exposed to measure the probability of causation) has already become a social problem. Compensation of workers who developed a disease after occupational exposure is often decided by impact measures based on this erroneous identification of probability of causation and attributable risk or rate among the exposed. Usually a threshold value of 50% is chosen in Germany, which corresponds to the so called doubling of risks (i.e., a relative risk or rate of 2). Thus, compensation is usually withheld if the attributable risk or rate is below 50% (Morfeld and Piekarski 2001 [ 7 ]). Similar procedures are applied in other countries (Greenland 1999 [ 6 ], Armstrong and Theriault 1996 [ 37 ]). However, as demonstrated in this endnote, such an argument based on the attributable risk or rate is not justified. Discussions have started about how the discrepancy between the calculated attributable risk among the exposed and the intended probability of causation can be accounted for in amended compensation schemes in Germany (Morfeld und Piekarski 2001 [ 7 ]). These amended compensation schemes should reflect the advancement of disease onset due to exposure more accurately than the conventional ones. Supplementary Material Additional File 1 Life table analysis with calculation of excess Years of Potential Life Lost e-YPLL according to Park et al. 2002. Basic data (unexposed) from BEIR IV (1988), Table 2A-10, p. 133: death rates of the male US population, surviving at least 30 years, applied to a birth cohort of 100,000. Exposure impact: advancement of certain fractions of deaths. For details of assumed mechanism see Table 2 (Additional file 2). Click here for file Additional File 2 Two different exposure-response mechanisms compatible with the life table analysis in Table 1. Fractions of the unexposed deaths are advanced by 0 yr or 5 yr according to mechanism 1 and by 0 yr, 5 yr or 10 yr according to mechanism 2. The age distribution of all exposed deaths is identical under both mechanisms. The distribution of the true excess Years of Life Lost e-YLL differs between mechanisms and both diverge from e-YPLL (Table 1, see Additional file 1), whereas the totals agree. (advcm = advancement) Click here for file Additional File 3 Life table analysis with calculation of excess Years of Potential Life Lost e-YPLL and true excess Years of Life Lost e-YLL for overall and lung cancer mortality (ICD9-162). Basic data (unexposed) from BEIR IV (1988), Table 2A-10, p. 133: overall and lung cancer death rates of the male US population, surviving at least 30 years applied to a birth cohort of 100,000. Exposure impact: advancement of factual and hypothetical lung cancer deaths by 5 years, mixture of advancements among deaths from all causes. It is assumed that the advancement of hypothetical lung cancer deaths leads to an excess of 50% of lung cancer deaths among exposed in each age category. The overall e-YLL for lung cancer are less than the number of exposed lung cancer deaths times 5 years. The e-YPLL for overall death and lung cancer death are determined according to Park et al. 2002. For all deaths e-YPLL must equal e-YLL, but e-YPLL is obviously biased for lung cancer death. Click here for file Additional File 4 Excel sheet explaining the calculations in Additional files 1 and 2. Click here for file Additional File 5 Excel sheet explaining the calculations in Additional file 3. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545055.xml
523850	What can we learn from noncoding regions of similarity between genomes?	Background In addition to known protein-coding genes, large amounts of apparently non-coding sequence are conserved between the human and mouse genomes. It seems reasonable to assume that these conserved regions are more likely to contain functional elements than less-conserved portions of the genome. Methods Here we used a motif-oriented machine learning method based on the Relevance Vector Machine algorithm to extract the strongest signal from a set of non-coding conserved sequences. Results We successfully fitted models to reflect the non-coding sequences, and showed that the results were quite consistent for repeated training runs. Using the learned models to scan genomic sequence, we found that they often made predictions close to the start of annotated genes. We compared this method with other published promoter-prediction systems, and showed that the set of promoters which are detected by this method is substantially similar to that detected by existing methods. Conclusions The results presented here indicate that the promoter signal is the strongest single motif-based signal in the non-coding functional fraction of the genome. They also lend support to the belief that there exists a substantial subset of promoter regions which share several common features including, but not restricted to, a relative abundance of CpG dinucleotides. This subset is detectable by a variety of distinct computational methods.	Background Since the publication of draft sequences for the human [ 1 ] and mouse [ 2 ] genomes, several groups have run large-scale comparisons of the sequences to detect regions of conserved sequence. An initial survey of these was published along with the draft mouse genome [ 2 ], with additional comparisons appearing since then [ 3 ]. Briefly, protein coding genes are – as we might expect – among the most strongly conserved regions, but homologous sequences can be found throughout the genome. In total, it is possible to align up to 40% of the mouse genome to human sequence [ 4 ], but it seems likely that at least some of this is just random "comparative noise" – regions of sequence which serve no particular purpose but which, purely by chance, have not yet accumulated enough mutations to make their evolutionary relationship unrecognisable. However, it is widely accepted that some of the noncoding-but-similar regions, especially those with the highest levels of sequence identity between the two species, are preferentially conserved because they perform some important function. It has been estimated that around 5% of the genome is under purifying selection [ 2 ], indicating that mutations in these regions have deleterious effects: a strong suggestion of some important function. Here, we apply the Eponine Windowed Sequence (EWS) sequence analysis method method which uses a Relevance Vector Machine (RVM) [ 5 ] to extract a minimal set of short motifs which are able to discriminate between two sets of sequences: in this case, a positive set of conserved non-coding sequences and a negative set of randomly picked non-coding sequences. The EWS model is an adaption of the Eponine Anchored Sequence (EAS) model, first applied for transcription start site prediction in [ 6 ] and subsequently used to predict a range of additional biological features including translation start sites and transcription termination sites [A. Ramadass, unpublished] While EAS is designed to classify individual points in a sequence – a feature which allows the model to predict precise locations for features such as transcription start sites – EWS classifies complete blocks (windows) of sequence. The basis functions (inputs) of the RVM are sums of position-weight matrix scores [ 7 ] across the whole window. Results We considered a set of alignments made by the blastz program [ 4 ] between release NCBI33 of the human genome and release NCBIM30 of the mouse genome. Since unprocessed blastz aligns around 40% of human sequence to the mouse genome, we chose to focus on the 'tight' alignments. These are a subset of alignments which are rescored and thresholded using a set of parameters given in [ 4 ], and cover only around 5.6% of the human genome – a proportion much closer to the fraction of bases thought to be under purifying selection [ 2 ]. In total, the tight blastz set contained 787173 blocks of sequence with high-scoring alignments between the two genomes. We considered only those blocks assigned to human chromosome 6, a 170 Mb chromosome which has recently undergone manual annotation of gene structures and other features [ 8 ]. This chromosome included 44105 (5.6%) of the total alignments. These varied in length from 34 to 9382 bases, with a length distribution skewed towards relatively short alignments, as shown in figure 1 . Since we were interested in non-coding features of the genome, we ignored all regions where an alignment overlaps an annotated gene structure. This removed 20.8% of aligned bases. It is possible that some genes, and especially pseudogenes, have been missed by the annotation process, so we also removed portions covered by ab initio gene predictions from the Genscan program [ 9 ]. This eliminated an additional 4.3% of aligned bases. Finally, repetitive sequence elements annotated by the programs RepeatMasker [ 10 ] and trf [ 11 ] (5.9%) were removed from the working set. The remainder of the aligned regions were split into non-overlapping 200 base windows, ignoring any portions less than 200 bases. This gave a set of 13925 sequences which are well-conserved between human and mouse – and therefore likely to be functional – but which are very unlikely to be part of the protein-coding repertoire. These formed the positive training set for our machine learning strategy. A negative training set of equal size was prepared by picking 200-base windows at random from the non-coding, non-repetitive portions of chromosome 6, using the same criteria to define repeats and coding sequence. While it is probable that this set also included some functional sequences, we would expect them to be represented at a substantially lower level than in the conserved set. These two sets of sequence were presented to the Eponine Windowed Sequence machine learning system, as described in the methods section. Randomly chosen 5-base words were used as seed motifs, and three independent training runs were performed, each for 2000 cycles. The set of motifs used in model 1 is shown in table 1 . While the exact set of motifs used in the model varied somewhat from run to run, testing pairs of models on non-overlapping windows from a 1 Mb region of human chromosome 22 and plotting the scores showed that the model outputs were highly correlated ( e.g . figure 2 ). We calculated the Pearson correlation coefficient for all pairs, and in all cases this was greater than 0.96. From this strong correlation, we concluded that any variations in the model were simply the result of the trainer picking one representative from a group of motifs which provide similar information. We scanned genomic sequences using these models at a range of thresholds, and examined the results on the Ensembl genome browser [ 12 ] using a Distributed Annotation System [ 13 ] server. Visual inspection showed that many of the highest-scoring regions were localised near the start of genes. This prompted us to look at the distribution of high-scoring sequences with respect to the starts of a set of well-annotated genes. We considered the GD_mRNA genes from version 2.3 of the human chromosome 22 annotation. These are confidently annotated genes with experimental evidence as described in [ 14 ], which confirms at least the approximate location of the ends of the transcripts, and are independent from the chromosome 6 training data. Figure 3 shows the density of predictions with EWS scores ≥ 0.90 relative to the annotated 5' ends of these genes. This shows a strong peak of predictions close to the annotated starts, demonstrating that the model is predicting some sequences commonly located around the transcription start site of genes. Combining this observation with the fact that the model was trained from conserved (and therefore presumed functional) sequences, we believe that it is detecting signals found in the promoter regions of genes. Evaluation of promoter-prediction methods on a large scale is a difficult exercise, since there are no large pieces of genomic sequence for which we can be certain we know the complete set of transcribed regions, and even in the case of well-known genes we often do not know the precise location at which transcription begins. In [ 6 ], we developed a pseudochromosome, derived from release 2.3 of the chromosome 22 annotation. As described above, this includes a subset of 284 experimentally verified gene structures. The pseudochromosome was constructed to include these genes while omitting all other annotated genes (which could be substantially truncated). We considered predictions (groups of one or more overlapping windows which all have scores greater than some chosen threshold) to be correct if they lie withing 2 kb of an annotated gene start, and false otherwise. Plotting accuracy (proportions of predictions which are correct) against coverage (proportion of transcript starts which are detected by one of the correct predictions) gives a Receiver Operating Characteristic (ROC) curve. Using this criterion, a totally random set of predictions would be given an accuracy of around 0.07. ROC curves are plotted for the three independently trained models in figure 4 . Firstly, this shows that predictive performance for all three models is rather similar. It also shows that they can function as accurate promoter predictors, with accuracy rising to a plateau of around 0.7, much higher than expected for random predictions. We picked model 1 for further study. Using a score threshold of 0.91, this gives an accuracy of 0.68 and a coverage of 0.31. We compared the set of genes correctly detected by this model to two other methods: firstly, the EponineTSS predictor described in [ 6 ], and secondly, the published results from the PromoterInspector program [ 15 ]. PromoterInspector results were mapped to pseudochromosome coordinates using the procedure described in [ 6 ]. Figure 5 shows how the set of promoters detected by these three distinct methods overlaps. There are clearly strong correlations between all three methods. In particular, at this threshold the EWS homology model detects 98 promoters which were found by at least one of the other methods, but only 4 novel promoters. We investigated the robustness of the signal learned by this process by retraining models with a variety of seed word sizes, from 2 to 6 bases. During training, motifs can be trimmed to lengths shorter than that of the seed words (down to a minimum of 2 bases) but can never grow longer than the seed word size. When evaluated on the pseudochromosome, the resulting models always showed a preference for regions around gene starts, regardless of word length, as shown in figure 6 . However, the accuracy was reduced when using short seed words – particularly words of length of 2. The best accuracy was seen for a seed word length of 5, and decreased somewhat for words of length 6. This suggests that a large fraction (but not all) of the information learned by these models can be encoded in dinucleotide frequencies. It is well known that many transcription start sites are close to regions of relatively high CpG dinucleotide composition (CpG islands) [ 16 ]. To investigate the contribution that CpG dinucleotides make to our models, we deleted all CpG dinucleotides from the training data, then re-evaluated the resulting models on the pseudochromsome (also with CpG dinucleotides removed), as shown in figure 7 . Perhaps not surprisingly, dinucleotide models now show very little tendency to detect gene starts. However, as the word size increases, the preference for gene starts gradually increases, until a seed size of 6 gives an accuracy comparable to that see when CpG dinucleotides are included, although the maximum coverage before accuracy begins to drop rapidly is somewhat lower. Broadly similar results are seen if CpG dinucleotides are randomly replaced with other dinucleotides. Conclusions We have shown here that, when presented with a set of non-coding sequences which are strongly conserved between human and mouse, a simple motif-oriented machine learning system consistently builds models which are able to detect a substantial fraction of human promoter regions with good accuracy. This strongly suggests that this promoter signal represents the most widely used motif-based signal in functional non-coding sequence. While the model learned here can clearly be applied for the purpose of genome-wide promoter annotation, in practise existing methods offer better coverage and (in the case of the EponineTSS predictor) predictions for the precise location of the transcription start site. It is interesting that the promoter model learned by this technique detected substantially the same set of promoters as found by the EponineTSS and PromoterInspector methods. It has previously been remarked that these two methods detect similar sets [ 6 ], but this could perhaps be explained by the fact that both methods were initially derived from similar sets of known promoter sequences (in both cases, training data was extracted from the EPD database [ 17 ]. In the case of the homology models described here, there is no connection with EPD, or any similar set of known promoters: the training data was picked purely on the basis of its high similarity to corresponding portions of the mouse genome. These results therefore support the alternate view that there is a particular 'easily detected' subclass of promoter sequences. One distinct group of promoters, which previous results show may correspond to this easily detected family, is the set of promoters associated with CpG islands [ 16 ]. However, while a number of the motifs listed in table 1 are G/C rich and/or contain the CpG dinucleotide, by no means all of the motifs match this description, and indeed one motif containing CpG has a negative weight in the linear model – its presence in a sequence will reduce the model's output score – while some A/T rich motifs have positive weights. We therefore believe that the signals detected here are significantly more complex than a simple over-representation of CpG dinucleotides. Experiments with smaller seed-word sizes support this assumption: while dinucleotide-based models were also able to predict promoter regions, the accuracy was lower than for models including longer motifs. Finally, we show that while the predictive capacity of dinucleotide models is largely eliminated once CpG dinucleotides are removed from the sequence, models including longer words are still able to make correct promoter predictions in many cases. So while CpG dinucleotides are an important contribution to the promoter signal, they are clearly not the only component. Methods Genomic sequence and annotation Human genome sequence release NCBI33 and mouse genome release NCBIM30 were extracted from Ensembl databases [ 12 ], which also contained gene predictions from Genscan [ 9 ] and repeat data from RepeatMasker [ 10 ] and trf [ 11 ]. Curated annotation of gene structures on human chromosome 6 was obtained from the Vega database [ 18 ]. Vega and Ensembl data was extracted directly from the SQL databases using the BioJava toolkit with biojava-ensembl extensions [ 19 ]. Genome alignments Human-mouse genome alignments were generated by the blastz alignment program. These were subsequently re-scored and filtered to give a 'tight' set of high-confidence alignments, as described in [ 4 ]. We downloaded the tight alignment set from the UCSC genome website [ 20 ]. Pseudochromosome for testing promoter-finding methods A 16.3 Mb pseudochromosome sequence was produced based on version 2.3 of the curated annotation for human chromosome 22. This includes all the experimentally-validated gene structures and their upstream regions, while omitting regions containing genes that are predicted but not fully verified. In the case of a pair of divergent genes where one has been verified and the second has not, their shared upstream region was cut at the midpoint. More information about pseudochromosome construction is given in [ 6 ]. Eponine Windowed Sequence learning The Eponine Windowed Sequence (EWS) model is designed by analogy to the Eponine Anchored Sequence model first described in [ 6 ], but rather than targeting individual points in the sequence, it is designed to classify small regions or windows of a sequence, based purely on their own sequence content. The EWS model uses the Relevance Vector Machine [ 5 ] algorithm to drive the training process. Relevance Vector Machines solve classification and regression problems by building Generalised Linear Models (GLMs) as weighted sums of a "working set" of basis functions. During the training process, those basis functions which are not informative are given weights close to zero and eventually discarded from the working set. To explore very large sets of possible basis functions, it is possible to add extra basis functions during the course of the training process [ 6 ]. The "sensors" of the EWS model are DNA position-weight matrices [ 7 ], which make convenient models of short sequence motifs. When using weight matrices to analyse sequence windows, we sum the weight matrix probability scores for all possible positions within the sequence. Normalising for the length of the sequence being inspected and the size of the PWM, the basis functions of the model take the form: where W ( s ) is the probability that sequence s was emitted by weight matrix W , \| S \| is the sequence length, \| W \| is the weight matrix length, and denotes a subsequence from i to j . An initial set of basis functions is proposed by taking all possible DNA motifs of a specified length (typically 5) and generating weight matrices which preferentially recognise these motifs. As the relevance vector machine trainer removes non-informative basis functions from the working set, they are replaced by applying one of the following sampling strategies to a basis function picked randomly from the working set: • Generate a new weight matrix in which each column is a sample from a Dirichlet distribution with its mode equal to the weights in the corresponding column of the parent weight matrix. • Generate a new weight matrix one column shorter than the parent by removing either the first of the last column. By using these sampling rules, the trainer is able to explore motif space. The process of generating candidate motifs using these rules then selecting the most informative using the RVM can be seen as a form of genetic algorithm. Authors' contributions TD and TH conceived and designed this study, and analysed results. TD implemented the Eponine machine learning system and drafted the manuscript. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC523850.xml
546010	ACRATA: a novel electron transfer domain associated to apoptosis and cancer	Background Recently, several members of a vertebrate protein family containing a six trans-membrane (6TM) domain and involved in apoptosis and cancer (e.g. STEAP, STAMP1, TSAP6), have been identified in Golgi and cytoplasmic membranes. The exact function of these proteins remains unknown. Methods We related this 6TM domain to distant protein families using intermediate sequences and methods of iterative profile sequence similarity search. Results Here we show for the first time that this 6TM domain is homolog to the 6TM heme binding domain of both the NADPH oxidase (Nox) family and the YedZ family of bacterial oxidoreductases. Conclusions This finding gives novel insights about the existence of a previously undetected electron transfer system involved in apoptosis and cancer, and suggests further steps in the experimental characterization of these evolutionarily related families.	Background A family of vertebrate proteins containing a six transmembrane domain (6TM) has been recently implicated in apoptosis and cancer. Up to date, this family contains four vertebrate members: STEAP, STAMP1/STEAP2, TIARP and TSAP6/pHyde. STEAP (six-transmembrane epithelial antigen of the prostate) was the first described member of this family and identified as a prostate-specific cell-surface antigen overexpressed in cancer, located at the cell-cell junction of the secretory epithelium of prostate, and found as well in both colon and bladder cancer cell lines [ 1 , 2 ]. STAMP1 (six transmembrane protein of prostate 1), also known as STEAP2, also overexpressed in prostate cancer, has been located in the trans-Golgi network and shuttles to plasma membranes, which suggest a role in the secretory/endocytic pathways [ 3 , 4 ]. TIARP (Tumor necrosis factor-alpha-induced adipose-related protein) is a cell surface protein induced by TNF-α and IL-6, probably implicated in resistance to insulin [ 5 , 6 ]. TSAP6 (tumor suppressor activated pathway-6), also known as pHyde [ 7 - 10 ], is a p53 inducible protein which regulates apoptosis and the cell cycle via direct interaction with Nix (a pro-apoptotic Bcl-2 related protein) and Myt1 kinase (a negative regulator of the G2/M transition) [ 9 ]. TSAP6 has been shown to be interacting with TCTP (Translationally controlled tumor protein) and could be implicated in its secretion [ 10 ]. In this work we present evidence of remote homology of this family to other two families: the mainly eukaryotic Nox and the bacterial YedZ (Fig. 1 ), both involved in redox functions [ 11 - 17 ]. The Nox family is involved in the production of reactive oxygen species (ROS) [ 11 , 12 ]. The first member of the family (gp91phox) was discovered in phagocytes and contains an N-terminal transmembrane heme binding domain and two C-terminal domains with binding sites for both flavin adenine dinucleotide (FAD) and NADPH (Fig. 2 ) [ 11 - 13 ]. Originally, ROS were thought to be used just as a mechanism of host defence. The discovery of gp91phox homologues in several other tissues has suggested their implication in many other functions, such us signal transduction, cancer, mitogenic signalling, cellular growth, angiogenesis, and modification of extracellular matrix proteins [ 11 - 14 ]. Regarding the bacterial YedZ family, the only experimentally characterized member so far is the Escherichia coli YedZ protein, which binds a single heme and is involved in electron transfer to the molybdopterin cofactor in YedY, its operon neighbour gene [ 16 - 18 ]. The exact function of the operon YedZ/YedY remains unknown. Methods To do the sequence analysis of the new domain we took advantage of the possibility of connecting distant protein families via intermediate sequences [ 19 ] and methods of iterative profile sequence similarity search: HMMer [ 20 , 21 ] and PSI-BLAST [ 22 ] over the Uniprot 90% non redundant sequence database [ 23 ]. We used NAIL to view and analyse the HMMer results [ 24 ]. The alignment of transmembrane regions using standard substitution matrices might be inaccurate because of the different roles played by amino acids in globular proteins and in transmembrane media [ 25 ]. In the case of the ACRATA domain, the regions to be aligned mostly consist of amino acids located in transmembrane regions (Fig. 1 ). For this reason, we used a method based on a hidden Markov model (HMMer hmmalign), which does not rely on a general substitution matrix [ 20 , 21 ], using as a guide both the transmembrane predictions from TMHMM [ 26 , 27 ] and the results of multiple sequence alignment using T-Coffee [ 28 , 29 ]. The genomic neighborhood of the bacterial sequences (YedZ family) was analyzed to find potentially related genes in operons using STRING [ 17 , 18 ]. Results The global hidden Markov profile [ 20 , 21 ] generated for STEAP and related vertebrate proteins (STEAP family, henceforth) localized the first bacterial sequence (SpTrembl Q7VKI9 from Haemophilus ducreyi ) with an E-value of 0.63. This protein belongs to the large YedZ family of bacterial oxidoreductases. The corresponding YedZ global profile detected the STEAP family (most similar member: SpTrembl Q8IUE7, human STAMP1) with an E-value of 0.00087. The global profile of STEAP and YedZ detected the Nox family with an E-value of 0.032 and the corresponding Nox global profile localized the YedZ family with an E-value of 0.007 (Fig. 3 ). Only the regions from transmembranes 3 to 5 were considered to build the profiles because the transmembranes 1, 2 and 6 are highly variable among families. We have named this 6TM domain the ACRATA domain after A poptosis, C ancer and R edox A ssociated T ransmembr A ne domain. To investigate the consistency of our results we performed iterative database searches using the PSI-BLAST program [ 22 ]. We used as query the most conserved region of the ACRATA domain in E. coli YedZ protein (residues 71–166). These searches detected all of the ACRATA domain-containing families after 15 iterations (using a cut-off of E = 0.005 for the inclusion of retrieved sequences in the profile). None of these profile searches retrieved new unrelated sequences, and reciprocal searches produced convergent results. Therefore, we have concluded that ACRATA is a previously undetected, conserved domain that is commonly found in members of the STEAP, YedZ and Nox protein families. The similarity between these proteins was suspected before [ 30 ]. However, no alignment, domain definition or substantial statistical evidence was provided to demonstrate the evolutionary relationships of these proteins. The complete conservation of two histidines in all ACRATA domain containing proteins (Fig. 1 ) indicates that the STEAP protein family could bind at least an heme group, as was previously experimentally characterized for Nox and YedZ families [ 13 , 16 ], or for other analogous proteins such as cytochromes [ 31 ]. Discussion Experimental evidence show that Nox and YedZ families share heme binding capabilities and also involvement in electron transfer chains [ 11 - 13 , 16 ]. For the STEAP family, the electron transfer capability is consistent with the presence of an N-terminal cytoplasmic NADP oxidoreductase coenzyme F420 dependent domain, being the only exception the STEAP protein itself (Fig. 2 ). Therefore we conclude that ACRATA domain is a heme binding 6TM domain that originated before the onset of eukaryotes (ancestral, YedZ), transmitted by vertical descent in a conventional manner (Nox family), and further expanded in vertebrates (STEAP family) (Fig. 4 ). Although the mechanism of action of the ACRATA domain could be the same in all the proteins containing it, its variable cellular role is made conspicuous by the different effects produced by the modifications in the expression of the corresponding genes. For example, the knock-out of the whole YedYZ operon seems not to affect E. coli in a number of conditions tested (Brokx, S.J. and Weiner, J.H. personal communication). Very differently, changes in the expression patterns of human genes containing the ACRATA domain could be related with apoptosis or cancer [ 1 - 14 ]. The well known functional flexibility of oxidoreductases is exemplified in bacterial proteins such as cupredoxins and cytochromes, normally involved in electron transfer during respiration but that can enter in eukaryotic cells to induce apoptosis or inhibition of cell growth [ 32 ], or in the Nox family, with functions as different as host defence in phagocytes or extracellular matrix modification [ 11 , 12 ]. Conclusions We have described for the first time a 6TM domain present in three protein families (STEAP, YedZ and Nox): the ACRATA domain. The common functions of the proteins of those families suggest that this domain is involved in electron transfer, mediated by its heme binding capability. We hypothesize that STEAP, STAMP1, TSAP6, and TIARP have this function, and that they form part of electron transfer systems involved in cellular regulation, apoptosis, and cancer. Additional experimental approaches using different members of the ACRATA domain-containing families are required to confirm these hypotheses. Abbreviations STEAP, six-transmembrane epithelial antigen of the prostate; STAMP1, six transmembrane protein of prostate 1; TSAP6, tumor suppressor activated pathway-6; TIARP, Tumor necrosis factor-alpha-induced adipose-related protein; TCTP, Translationally controlled tumor protein; Nox, NADPH oxidase; ROS, reactive oxygen species; TM, Trans-membrane; FAD, flavin adenine dinucleotide; HMM, Hidden Markov Models; Competing interests The author(s) declare that they have no competing interests. Authors' contributions LSP, AMR, and MAA carried out the sequence analysis of the domain. LSP and MAA provided with the initial input of the research. LSP, AMR, AV, CMA, and MAA authored the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546010.xml
538257	The biological effects of diagnostic cardiac imaging on chronically exposed physicians: the importance of being non-ionizing	Ultrasounds and ionizing radiation are extensively used for diagnostic applications in the cardiology clinical practice. This paper reviewed the available information on occupational risk of the cardiologists who perform, every day, cardiac imaging procedures. At the moment, there are no consistent evidence that exposure to medical ultrasound is capable of inducing genetic effects, and representing a serious health hazard for clinical staff. In contrast, exposure to ionizing radiation may result in adverse health effect on clinical cardiologists. Although the current risk estimates are clouded by approximations and extrapolations, most data from cytogenetic studies have reported a detrimental effect on somatic DNA of professionally exposed personnel to chronic low doses of ionizing radiation. Since interventional cardiologists and electro-physiologists have the highest radiation exposure among health professionals, a major awareness is crucial for improving occupational protection. Furthermore, the use of a biological dosimeter could be a reliable tool for the risk quantification on an individual basis.	Introduction Over the last 30 years, medical cardiology imaging has rapidly grown, becoming an essential part of the cardiology clinical practice. Imaging procedures include conventional imaging tests such as echocardiography, radionuclide imaging, and angiography as well as a newer imaging techniques such as emission computed tomography and magnetic resonance imaging which promise to expand diagnostic capabilities [ 1 ]. These techniques widely differ not only for what concerns costs, availability and technical information, but they also differ in environmental and health hazards. Many cardiac procedures can deliver high radiation doses to the clinical staff [ 2 ]. This exposure may represent a significant health risk, resulting in deleterious clinical implications which can affect not only the personnel involved, but also their progeny [ 3 - 5 ]. Unfortunately, many physicians are unfamiliar with radiation biology or the quantitative nature of the risks and, frequently, ultrasound and ionizing radiation risks are misunderstood [ 6 - 9 ]. The purpose of this paper is to discuss the published evidence on health effects of cardiac imaging procedures employing ultrasound and ionizing radiation. Ultrasound imaging Ultrasound imaging, also called sonography, is a method of obtaining human body images through the use of high frequency sound waves. Ultrasounds are mechanical vibrations with frequencies above the human limit of audibility. The use of ultrasounds in order to obtain images for medical diagnostic purposes, typically employs frequencies ranging from 2 MHz to about 12 MHz [ 10 ]. Ultrasound does not use ionizing radiation, and it is the preferred image modality for monitoring both pregnant women and their embryos or fetus [ 10 ]. In contrast to ionizing radiation, which can damage biological materials by dislodging electrons from atoms and molecules, ultrasounds do not cause ionisation. They usually interact with human tissue primarily by generating heat, but also non-thermal effects which are ascribed to cavitation (i.e. micro-bubble) [ 11 ]. The process of cavitation includes ultrasounds mechanical effects which lead to hydrodynamic breaks of hydrogen bonds and oscillation of hydrogen ions, and chemical effects produced by the occurrence of free radicals in intercarionic space in the process of cavitation (Figure 1 ). Theoretically, these free radicals may interfere with DNA, causing chromosomal damage. Indeed, ultrasounds of diagnostic intensities induced detectable DNA damage in animal cells [ 12 , 13 ]. Currently, there is a body of studies on human DNA damage from exposure to therapeutic and diagnostic ultrasounds [ 14 - 20 ]. In particular, Stella et al. [ 15 ] reported that therapeutic ultrasound induce a significant increase in sister chromatid exchanges (SCEs) in human lymphocytes after treatment both in vitro and in vivo . In the same study, no increase in chromosomal aberrations was observed during and after ultrasound therapy [ 15 ]. Subsequently, some reports on human cells indicated that ultrasound was not able to induce SCEs or chromosomal damage (Table 1 ). Thus, there is at present no indication that exposure to medical ultrasound is capable of inducing genetic effects and representing a serious health hazard for clinical staff. However, very little information is available on the genetic effects of individuals occupationally exposed to chronic ultrasound. Medical staff can be exposed to hand-transmitted ultrasound waves in the work-place. Figure 1 At high acoustic pressure, ultrasound is capable of causing rapid bubble which grow and collapse among them (a) and cells (b). This mechanism results in the production of sufficient energy to disrupt chemical bonds and produce reactive free radicals, that may interfere with DNA. Table 1 Summary of studies on genetic effects of medical ultrasounds Author, Year (Ref) Assay System Endpoint Exposure Result Miller et al., 1983 (14) Human lymphocytes exposed in vitro SCE 2 MHz SPPA intensity 100 W/cm 2 Negative Stella et al., 1984 (15) Human lymphocytes exposed in vitro SCE CA 1 W/cm2; 0.860 MHz; for 40–160 sec Positive/ Negative Barnett et al., 1987 (16) Human lymphocytes exposed in vitro SCE 3.1 MHz SPPA intensities from 15 to 135 W/cm 2 . Negative Carrera P et al., 1990 (17) Chorionic villi exposed in vitro Chorionic villi from exposed pregnant women SCE 2 MHz at 1, 2, 3 h Diagnostic US for 20 min (in vivo exposure Negative Miller et al., 1991 (18) Human lymphocytes from exposed patients SCE 4 patients underwent therapeutic US 4 healthy persons underwent sham-therapeutic US Negative Martini et al., 1991 (19) Lymphocyte and lymphoblastoid cells exposed in vitro SCE 5 MHz for 20 sec, 1 min, 5 min, and 20 min Negative Sahin O et al., 2004 (20) Human lymphocytes from exposed patients MN 10 patients underwent 10 session of US therapy at 1 MHz for 10 min and 10 control subjects underwent sham-therapeutic US Negative Garaj-Vrhovac and Kopjar, 2000 (22) Human lymphocytes from cardiologists working with Doppler ultrasound CA SCE MN Unit working with colour Doppler US (transducer frequencies 2.5–7.5 MHz. SPPA intensity 60–110 W/cm 2 . Positive SCE: sister-chromatid exchange; MN: Micronuclei; CA: Chromosomal aberrations; SPPA: Spatial Peak Pulse Average Indeed, ultrasound sources do not transmit acoustic energy into air, and only low level ultrasound reaches medical personnel through handling of the probe [ 21 ]. Probably, occupational exposure to ultrasound occurs during training procedures [ 21 ]. In fact, medical personnel often apply diagnostic ultrasound to themselves during training or during technique demonstrations [ 21 ]. Consequently, ultrasound is not harmful like the other types and sources of radiation. However, a recent investigation indicated that medical personnel from a cardiology unit working with colour Doppler ultrasonic equipment had an increased genotoxic damage compared to the control subjects [ 22 ]. Therefore, this observation requires further studies in order to determine if chronic exposure to ultrasound might induce genotoxic effects. Ionizing radiation Ionizing radiation is known to cause harm. High radiation doses tend to kill cells, while low doses tend to damage or alter the genetic code (DNA) of irradiated cells. The biological effects of ionizing radiation are divided into two categories: deterministic and stochastic effects. Deterministic effects, such as erythema or cataract , have a threshold dose below which the biological response is not observed [ 23 - 25 ]. Some interventional procedures with long screening times and multiple image acquisition (e.g. percutaneous coronary intervention, radio-frequency ablation, etc) may give rise to deterministic effects in both staff and patients [ 26 , 27 ]. A stochastic effect is a probabilistic event and there is no known threshold dose. The likelihood of inducing the effect, but not the severity, increases in relation to dose and may differ among individuals. In fact, the effect of low doses of radiation -less than 50 mSv- do not cause an immediate problem to any body organ, but spread out over long periods of time after exposure. The biological effects are at DNA level and they may not be detected [ 23 - 25 ]. The cell has repair mechanisms against damage induced by radiation as well as by chemical carcinogens. Consequently, biological effects of low dose radiation on living cells may result in three outcomes: (1) injured or damaged cells repair themselves, resulting in no residual damage; (2) cells die; or (3) cells incorrectly repair themselves resulting in a biological change (Figure 2 ). Such biological changes include the development of cancer and genetic defects in the future children of exposed parents. At present, however, the effects of low-level exposure remain uncertain [ 28 ]. The associations between radiation exposure and the development of cancer are mostly based on populations exposed to relatively high levels of ionizing radiation (e.g., Japanese atomic bomb survivors). Since extraordinary large studies are required to quantify the risks of very low doses of radiation, it is unlikely that we will be able to precisely quantify cancer risk in human populations at doses below 10 mSv [ 28 ]. For instance, an epidemiological study of more than 5 million people would be needed to quantify the effect for a 10 -mSv dose or less [ 28 ]. Our inability to quantify risk does not, however, imply that this risk is negligible. Furthermore, the small (and often not so small) individual risk applied to a large number of individuals, and by protracted exposures, translates into a significant public health problem). As such, the international scientific community has adopted a prudent approach and acknowledged the fact that any level of exposure could potentially lead to biological effects. A linear, no-threshold dose response relationship is used by the IRCP in order to describe the relationship between radiation dose and the occurrence of cancer [ 29 ]. This dose-response model suggests that any increase in dose, no matter how small, results in an incremental increase in risk. Figure 2 Radiation damage of DNA. Damaged DNA is screened through the process of DNA repair and mismatch correction. DNA lesions that escape repair, has the ability to produce mutations, which lead to the development and the progression of both cancer and human diseases even decades after exposure. Genetic effects are the result of a mutation produced in the reproductive cells of an exposed individual that are passed on to their offspring. These effects may show up as birth defects or other conditions in the future children of the exposed individual and succeeding generation. Indeed, studies with laboratory animals have provided a large body of data on radiation-induced genetic effects [ 30 ]. Recently, these effects have been also observed in studies of people exposed to radiation from Chernobyl disaster, radiation workers and medical radiologists who have received doses of radiation [ 31 - 33 ]. However, no conclusive evidence exists yet [ 34 , 35 ]. Radiation exposure to cardiologists The use of radiation in medicine is the largest source of man-made radiation exposure. According to the latest estimation of the United Nations, an average of 2.4 mSv/year comes from natural sources [ 24 ]. In western countries, the exposure dose from medical radiation corresponds to 50 to 100% of the total natural radiation. In 1997, the German Federal Office for Radiation Protection reported 136 million x ray examinations and 4 million nuclear medicine diagnostic tests, resulting in a mean effective dose of 2.15 mSv per person per year [ 36 ]. Cardiac and interventional procedures account for a large percentage of nuclear and radiological examinations [ 36 ]. Of all radiological examinations, 28% are arteriographies and interventions. An additional 2% derive from chest X-rays and 37% from CT: many of them are cardiological referrals. Regarding nuclear medicine, 22% are cardiological scan. These percentages are likely higher now, since the use of cardiac and interventional procedures is increasing. Cardiac ionizing procedures expose both patients and medical staff to the highest radiation levels in diagnostic radiology, and recently, as the number of diagnostic and interventional cardiac catheterisation procedures has greatly increased, serious radiation induced skin injuries and an excess of cataract development have been reported in exposed staff [ 37 - 39 ]. Furthermore, it has been suggested that fluoroscopic procedures may be a health hazard and increase the risk for brain tumours in interventional cardiologists [ 40 ]. Today, interventional cardiologists represent, indeed, the most important group of exposed among professionally exposed physicians [ 41 , 42 ]. As known, the limit on effective dose for exposed workers should be 100 mSv in a consecutive five year period, subject to a maximum effective dose of 50 mSv in any single year. Radiation dose limits to adult occupational workers provided by the International Commission on Radiological Protection (ICRP) are shown in table 2 . Table 2 Recommended occupational dose limits by International Commission on Radiological Protection (ICRP). TISSUE INJURY OCCUPATIONAL DOSE LIMITS/YEAR whole body 20 mSv 2 rem Lens of the eye 150 mSv 15 rem Skin, hands, feet, and other organs 500 mSv 50 rem As a matter of fact, the head dose sustained by cardiologists may reach 60 mSv per year, and may in some cases exceed the occupational limit of 150 mSv per year recommended for the lens of the eye [ 41 ]. However, the correlation between occupational doses and staff radiological risks is not simple, and it is very dependent on equipment, the specialist, and protocols followed throughout the procedure [ 43 ]. Many factors can influence occupational doses for the same radiation dose imparted during cardiac procedure. One of the most important factors is that protection tools are available in catheterisation laboratories and are appropriately used [ 43 ]. In addition, another likely reason is a lack of knowledge, information and training in radiation protection [ 43 ]. Importantly, a recent survey showed that that most of cardiologists do not correctly evaluate the dose exposure, the medico-legal regulation, the environmental impact and individual bio-risks of the radiological investigations [ 9 ]. As shown in table 3 , this surprising lack of knowledge of both dose and clinical risk of commonly performed ionising test examinations, is not at all restricted to cardiologists, and seems to be democratically spread across all specialties – from surgeons to orthopaedics, to paediatricians [ 6 - 9 ]. Table 3 Doctors' knowledge of radiation dose and risk for medical ionising testing Author, year (Ref) Physicians Radiological Awareness Evaluation Results Shiralkar S et al., 2003 (6) British physicians Radiation doses for common radiological investigations. 97% of doctors underestimates dose. 5% believes that US use ionising radiation. 8% believes thatMRI use ionising radiation. Finestone A et al., 2003 (7) Istraeli orthopaedists Mortality risk of radiation-induced carcinoma from bone scan scintigraphy Mortality risk was identified correctly by less than 5% of respondents. Lee CI et al., 2004 (8) Emergency department (ED), physicians and radiologists Radiation dose and possible risks associated with CT scan Almost all doctors were unable to accurately estimate the dose. Only 9% ED physicians believed that there was increased risk. Correia MJ et al., 2005 (9) Adult and paediatric cardiologists Environmental impact, individual bio-risks, dose exposure and medico-legal regulation of medical ionising testing Only 11%, 5%, 29% and 42% of physicians correctly identified environmental impact, individual bio-risks, dose exposure and legal regulation, respectively. CT = computed tomography; MRI = magnetic resonance imaging; US = ultrasound Probably, this unawareness has its root in the difficult perception of a long-term risk associated to radiation exposure. In particular, the perception of cancer risk, which can have a latency period of many years after exposure, is often elusive. Furthermore, the exact risk at very low doses to a specific individual is further complicated by many factors, such as carcinogenic agents in our environment, cigarette smoke, diet and genetic background. However, a recent study has estimated that from 0.6% to 3% of all cancers are due to medical X-rays [ 44 ]. These figures are impressive but may largely underestimate the true risk, since they are referred to radiological data concerning the 1991–1996. Taking into account current radiological activities, medical radiation is likely to account for at least 20% of cancer in developed countries [ 45 ]. With regard to occupational exposure for radiologists and radiotherapists, available epidemiological studies have been recently reviewed by Yoshinaga et al [ 46 ]. An excess risk of leukaemia associated with occupational radiation was found among early workers employed before 1950, when radiation exposures were high. In addition, several studies provided evidence of a radiation effect for breast and skin cancer. To date, there is no clear evidence of an increased cancer risk in medical radiation workers exposed to current levels of radiation doses. However, given a relatively short period of time for which the most recent workers have been followed up and in view of the increasing uses of radiation in modern medical practices, it is important to continue to monitor the health status of medical radiation workers [ 46 ]. To the fatal cancer risk, one must add the risk of non-fatal cancer and major genetic damage transmitted to the offspring. It is relevant to underline that the long-term damage may not include only cancer but also other major degenerative diseases, including atherosclerosis [ 47 , 48 ]. However, it is important to realize that many difficulties are involved in designing epidemiological studies that can accurately measure the increases in health effects due to low exposures to radiation as compared to the normal rate of cancer. Studies with very large sample size are required in order to quantify the risks of very low doses of radiation. An alternative strategy could be based on the measure of biological effects by using biomarkers as predictors of delayed health outcomes [ 49 ]. Biomarkers in the assessment of radiation exposure Damage to deoxyribonucleic acid (DNA), which carries the genetic information in chromosomes in the cell nucleus, is considered to be the main initiating event by which radiation damage to cells results in the development of cancer and hereditary disease. Four biomarkers (Figure 3 ) -analysis of structural chromosome aberrations, micronucleus assay, sister chromatid exchange analysis and comet assay- in peripheral lymphocytes are currently employed in order to study human exposure to environmental carcinogens [ 50 ]. Among these, the test of chromosomal aberrations in peripheral blood lymphocytes has the most abundant literature validating that a high frequency of chromosomal breakage is a strong predictor of cancer risk in healthy subjects [ 51 , 52 ]. Figure 3 Biomarkers of DNA damage in human lymphocytes: a) Structural chromosomal aberrations (CA) are typical of cancer cells, probably as a manifestation of genetic instability. b) Micronuclei (MN) can originate from chromosome breaks or whole chromosomes that fail to engage with the mitotic spindle when the cell divides. Therefore, the micronucleus test can be considered just as a real "biological dosimeter" for evaluating both numerical and structural chromosome aberrations. c) Sister chromatid exchanges (SCEs) represent symmetrical exchanges between sister chromatids; generally they do not result in chromosomal alterations of the genetic information. c) The Comet assay is an especially sensitive method for detecting DNA single-strand breaks and oxidative DNA damage in individual cells. The entity of the DNA damage is proportional to the length of the comet. During the last years, the micronucleus assay has become popular since it is fast and inexpensive, and it is considered to be a "biological dosimeter" for exposure to ionizing radiation [ 53 ]. The importance of cytogenetic study of peripheral lymphocytes in subjects exposed to ionizing radiation has been reported for more than 20 years, especially in radiologists [ 54 - 68 ]. The available evidence suggests that chronic exposure to low dose radiation has a genotoxic effect on somatic DNA of professionally exposed workers (Table 4 ). This effect seems to be cumulative over time, although the majority of these studies failed to establish a dose-effect relationship for low doses. The absence of increase of somatic DNA damage in relation to the dose might be explained by various factors. Dosimetry records may underestimate the real dose exposure if the badges are not properly worn. The potential combined effect of other genotoxic exposures would also induce DNA damage, enhancing the effect of radiation exposure [ 63 ]. Moreover, genetic susceptibility may account for the inter-individual differences to radiation sensitivity. Such possible susceptibility may recognize sources of variability (genetic polymorphism) in people's DNA repair gene sequence [ 69 ]. However, it is interesting to underline that, in a group of radiologists, it has been documented an important parallelism between the decrease of the exposure to ionizing radiation in the hospitals and a reduction in the frequency of chromosome aberrations over the most recent decades [ 58 ] (Figure 4 ). This decrease was the result of an efficient protection policy among radiologists. Unfortunately, this is not the case for invasive cardiologists who need to know very well both the long-term risks and the doses involved in the large amount of examinations they prescribe and/or perform every day [ 40 , 41 ]. Table 4 Cytogenetic studies in hospital workers Author, year (ref) Exposed Subjects, n Non-exposed Subjects, n Endpoint Results Exposure Correlation with dose (Yes/No) Bigatti et al, 1988, (54) 63 (physicians, nurses and technicians) 30 (ward nurses and office personnel) CA Positive < legal limit. No Barquinero et al, 1993, (55) 26 (hospital workers) 10 (healthy individuals) CA Positive 1.6–42.71 mSv No Paz-y-Mino et al, 1995, (56) 10 (hospital workers) 10 (healthy individuals) CA Positive 1.84 mSv/year. No Vera et al, 1997, (57) 20 (medical staff working at an X-ray department) 20 (general population) CA MN Positive <25 mSv/year. No (Major DNA damage in subjects exposed to both ultrasound and X-ray) Bonassi et al., 1997, (58) 871 (hospital workers from 4 laboratories) 617 (healthy individuals) CA Positive Available only partially and variable. Yes/No Rozgaj et al, 1999, (59) 483 (radiologists, pneumologists, technicians) 160 (healthy individuals) CA Positive <20 mSv/year No Undeger et al., 1999, (60) 30 (technicians) 30 (nurses, technicians, office personnel) Comet Positive 50 mSv/ year. No Cardoso et al, 2001, (61) 8 (workers in X-rays, radiotherapy and nuclear medicine sectors) 8 (healthy individuals) CA MN SCE Positive 63.2 mSv/life No Maluf et al, 2001, (62) 22 (hospital workers) 22 (non-exposed workers) MN Comet Positive 0.2 – 121. mSv No Maffei et al, 2002, (63) 37 (physicians, technicians) 37 (non-exposed workers MN Negative/ Positive 35 mSv /life No Bozkurt et al, 2003, (64) 16 (nuclear medicine) 16 (non-exposed physicians) SCE Positive 3.39 mSv/year. Yes Garaj-Vrhovac and Kopjar, 2003, (65) 50 (physicians, 25 technicians, 10 nurses) 50 (healthy students and office employees) Comet Positive 0–8.5 mSv/year. No Maffei et al, 2004, (66) 34 (physicians, technicians) 35 (non-exposed workers) CA Positive 1.81–141.77 mSv/life. Yes Zakeri et al., 2004, (67) 71 (cardiologists, nurses and technicians) 36 (healthy individuals) CA MN Positive 3.0 mSv/year No Andreassi et al, 2004, (67) 31 interventional cardiologists 31 clinical cardiologists MN Positive 4 mSv/year No SCE: sister-chromatid exchange; MN: Micronuclei; CA: Chromosomal aberrations; Figure 4 a) Decrease in exposure to ionizing radiation in hospital radiologists over the most recent decades and b) a similar time-related reduction in the frequency of chromosome-type aberrations (redrawn from ref. 58) As matter of fact, our results and a recent monitoring of personnel working in angiocardiography laboratories in Iranian Hospitals showed a high frequency of chromosome aberrations in cardiologists s and technicians compared to unexposed subjects [ 68 , 69 ]. Taken together, these evidences highlight that the use of a biological dosimeter could complement the data obtained by physical dosimetry and reduce the uncertainties of low-dose radiation risk assessment [ 70 ]. The analysis of chromosome aberrations is the gold standard endpoint for radiation biological dosimetry. Limitations and strengths on biodosimetry have been fully discussed in the IAEA Report 405 [ 70 ]. A possible limitation is the response to high radiation dose (> 4 Sv) where cell death and delays in progression through the cycle represents a pitfall for estimation of acute irradiation particularly when non-uniform or partial body irradiation have occurred. Moreover, the method is laborious, time consuming and requires expert skills. Scoring of micronuclei has been proposed as an alternative to conventional chromosome aberrations analysis, being more sensitive and faster [ 71 ]. Although micronuclei method has been improved, inter-laboratories discrepancies have emphasized the need for better standardization [ 53 ]. However, in many countries the application of cytogenetic dosimetry has yet medical-legal recognition, and it is complementary to physical dosimetry. On the other hand, the usefulness of biomarkers as early biological effects, with special concern for the prediction of cancer, has been recently emphasized [ 72 ]. Therefore, the application of biodosimetry- that measures true cellular injury resulting from that radiation- could greatly enhance health risk, identifying susceptible individuals and enhancing the possibility of preventive measures, especially in occupational settings with a high volume of radiological activities (Figure 5 ). Figure 5 Illustration of potential use of biomarkers as early predictors of clinical disease. The evaluation of genetic effects such as chromosomal damage could be used to anticipate delayed health outcomes, providing a greater potential for preventive measures. Conclusion Occupational exposure can occur in cardiological procedures which employ ultrasound and ionizing radiation. Today, there are no consistent adverse biological effects on operators caused by exposures to ultrasound. However, it is clearly necessary to continually monitor both the potential risks and safety of ultrasound exposure. In contrast, exposure to ionizing radiation may result in adverse health effect on both cardiologists directly and on their progeny. Although the current risk estimates are clouded by approximations and extrapolations, most data from cytogenetic studies have reported an enhanced DNA damage in hospital workers exposed to chronic low doses of ionizing radiation. The occupational dose of interventional cardiologists, and electrophysiologists tend to be higher compared to other medical specialists as a result of the recent increasing use of interventional techniques. On the other hand, physicians are dramatically unaware of dose, long-term risks and populations health impact caused by the use of medical ionizing radiation. Thus, a major awareness appears to be crucial in order to improve both one's knowledge on the appropriateness of protective tools and also in trying to reduce the number of unnecessary procedures. The use of a biological dosimeter could be a reliable tool for risk quantification on an individual basis.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538257.xml
523844	Distribution of Major Health Risks: Findings from the Global Burden of Disease Study	ABSTRACT Background Most analyses of risks to health focus on the total burden of their aggregate effects. The distribution of risk-factor-attributable disease burden, for example by age or exposure level, can inform the selection and targeting of specific interventions and programs, and increase cost-effectiveness. Methods and Findings For 26 selected risk factors, expert working groups conducted comprehensive reviews of data on risk-factor exposure and hazard for 14 epidemiological subregions of the world, by age and sex. Age-sex-subregion-population attributable fractions were estimated and applied to the mortality and burden of disease estimates from the World Health Organization Global Burden of Disease database. Where possible, exposure levels were assessed as continuous measures, or as multiple categories. The proportion of risk-factor-attributable burden in different population subgroups, defined by age, sex, and exposure level, was estimated. For major cardiovascular risk factors (blood pressure, cholesterol, tobacco use, fruit and vegetable intake, body mass index, and physical inactivity) 43%–61% of attributable disease burden occurred between the ages of 15 and 59 y, and 87% of alcohol-attributable burden occurred in this age group. Most of the disease burden for continuous risks occurred in those with only moderately raised levels, not among those with levels above commonly used cut-points, such as those with hypertension or obesity. Of all disease burden attributable to being underweight during childhood, 55% occurred among children 1–3 standard deviations below the reference population median, and the remainder occurred among severely malnourished children, who were three or more standard deviations below median. Conclusions Many major global risks are widely spread in a population, rather than restricted to a minority. Population-based strategies that seek to shift the whole distribution of risk factors often have the potential to produce substantial reductions in disease burden.	Introduction Reliable and comparable analysis of risks to health is an important component of evidence-based policies and programs for disease prevention [ 1 , 2 ]. An important feature of risk assessment, with implications for specific interventions as well as broad prevention policies, is the distribution of disease burden among population subgroups. These subgroups may be defined by demographic factors, such as age and sex, or by socioeconomic status. Subgroups can also be defined by the level of exposure to a risk factor, if exposures are defined in multiple categories or continuously. Understanding the distribution of risk-factor burden is particularly important for targeting interventions. For example, the large number of road traffic accident injuries and deaths among young adult males may be largely associated with binge alcohol consumption by this group. Interventions that focus on this population subgroup and their specific drinking behaviors may be more effective or cost-effective than, for example, raising alcohol taxes, which would have a more diffuse impact on alcohol consumption. On the other hand, the majority of effects from risk factors such as blood pressure have been found to be among those at moderately elevated levels, motivating interventions beyond those intended for clinical hypertension [ 3 , 4 , 5 ]. The distributions of the health effects of risk-factor exposure by age and sex or by exposure level have been studied in specific cohorts and for specific risk factors [ 6 , 7 , 8 ]. Most notably Rose's seminal work stated that “a large number of people exposed to a small risk may generate many more cases than a small number exposed to high risk” [ 9 ]. Using the data from a global and regional assessment of multiple major risk factors, this paper reports the distribution by exposure levels, age, and sex of disease burden attributable to several major risk factors. The findings of this analysis confirm that Rose's observations have global relevance and also illustrate important new patterns on specific distributions of disease burden for multiple risks, in different age groups, and in populations at various stages of economic development. Methods The methods and data sources for individual risk factors and for estimating population attributable fractions (PAFs) and disease burden attributable to them have been fully described elsewhere [ 1 , 2 ] and are summarized below. The contribution of a risk factor to disease or mortality relative to some alternative exposure scenario (i.e., PAF, defined as the proportional reduction in population disease or mortality that would occur if exposure to the risk factor were reduced to an alternative exposure scenario [ 10 ]) is given by the generalized “potential impact fraction”: RR ( x ) is relative risk at exposure level x, P ( x ) is the population distribution of exposure, P ′( x ) is the alternative or counterfactual distribution of exposure, and m is the maximum exposure level. The alternative (counterfactual) scenario used in this work is the exposure distribution that would result in the lowest population risk, referred to as the theoretical minimum-risk exposure distribution ( Table 1 ) [ 1 , 2 ]. The theoretical minimum exposure distribution was zero in most cases since zero exposure reflected minimum risk (e.g., no smoking). For some risk factors, zero exposure was an inappropriate choice as the theoretical minimum, either because it is physiologically impossible (e.g., body mass index [BMI] and cholesterol) or because there are physical lower limits to exposure reduction (e.g., ambient particulate matter concentration and occupational noise). For these risk factors, the lowest levels observed in specific populations and epidemiological studies were used as the theoretical minimum. For factors with protective effects (i.e., fruit and vegetable intake and physical activity), a counterfactual exposure distribution was chosen based on levels in high-intake populations (e.g., fruit and vegetable intake in Greece) and the level to which the benefits may continue given current scientific evidence. Using theoretical minimum exposure distribution as the counterfactual has the advantage of providing a vision of potential gains in population health by risk reduction from all levels of suboptimal exposure in a consistent way across risk factors. Table 1 Leading Global Risk Factors, Exposure Variables, Theoretical Minima, and Attributable Deaths and Disease Burden (measured in DALYs) in 2000 See Table 1 in Ezzati et al. [ 2 ] for disease outcomes and data sources a The resulting hemoglobin levels vary across regions and age-sex groups (from 11.66 g/dl in children under five in a region of Southeast Asia to greater than 14.5 g/dl in adult males in developed countries) because the other risks for anemia (e.g., malaria) vary b Theoretical minimum for alcohol is zero, the global theoretical minimum. Specific subgroups may have a non-zero theoretical minimum c Theoretical minimum for lead is the blood lead levels expected at background exposure levels. Health effects were quantified for blood lead levels above 5 μg/dl where epidemiological studies have quantified hazards Table 1 Continued Table 1 Continued Estimates were made for eight age groups (0–4, 5–14, 15–24, 25–44, 45–59, 60–69, 70–79, and 80+ y), both sexes, and 14 Global Burden of Disease subregions ( Table 2 ). PAFs were estimated for mortality and incidence and were applied to regional cause-specific mortality and disease burden from the World Health Organization (WHO) Global Burden of Disease database ( Table 1 ). Burden of disease, reported annually in the annexes of the World Health Report, was expressed in disability-adjusted life-years (DALYs) with methods and assumptions described elsewhere [ 11 ]. Aggregate results (both mortality and disease burden) for all ages, sexes, and exposure levels have been reported elsewhere [ 1 , 2 ]. Many risks act simultaneously to cause disease, and joint effects have also been estimated [ 12 ], though the separate effects are presented in this paper. The aim of the analyses reported here was to obtain estimates of the distribution of disease burden by age, sex, and exposure level. To make separate estimates of disease burden by exposure level, the relationship in equation 1 was re-estimated with the entire exposure distribution divided into “narrow bands,” with each band corresponding to one level of exposure, and the estimation repeated for each such exposure band. Table 2 Global Burden of Disease 2000 Subregions High-mortality developing regions: AFR-D, AFR-E, AMR-D, EMR-D, and SEAR-D. Lower-mortality developing regions: AMR-B, EMR-B, SEAR-B, and WPR-B. Developed regions: AMR-A, EUR-A, EUR-B, EUR-C, and WPR-A a A, very low child mortality and very low adult mortality; B, low child mortality and low adult mortality; C, low child mortality and high adult mortality; D, high child mortality and high adult mortality; E, high child mortality and very high adult mortality Results The distribution of deaths and DALYs attributable to the risk factors by age and sex is shown in Table 3 . Disease burden attributable to being underweight and to micronutrient deficiencies in children was equally distributed among males and females. The total all-age disease burden from iron and vitamin A deficiencies was greater in females because of effects on maternal mortality and morbidity conditions. The disease burden of other diet-related risks, physical inactivity, environmental risks, and unsafe sex (sex with an infected partner without any measures to prevent infection) occurred almost equally in males and females. However, approximately 80% of disease burden from addictive substances and 60%–90% from occupational risks occured among men (bearing in mind that the assessment considered only formal employment). Women experienced an estimated two-thirds of the disease burden from childhood sexual abuse and all of the burden caused by a lack of contraception. Table 3 Distribution of Risk-Factor-Attributable Deaths and Disease Burden (DALYs) by Age and Sex See Table 1 , and Figure 1 in [ 2 ], for definition of each risk factor, data sources and methods, and total magnitude of mortality and DALYs The estimated disease burden from childhood and maternal undernutrition; unsafe water, sanitation, and hygiene; and global climate change (much of whose estimated effects are mediated through nutritional and water variables) was almost exclusively among children under 5 y of age. For these risks, more than 85% of total attributable burden occurred in this age group, with the exception of iron deficiency, for which 30% of burden was borne by women of childbearing age. Only a small fraction of disease burden from the risk factors considered was among the 5–14 y olds. This was because some of the leading diseases of this age group (e.g., injuries and depression) have complex causes that could not easily be included in the current risk-based framework [ 12 ]. For other leading diseases of this group (e.g., diarrhea and lower respiratory infections), most epidemiological studies have focused on children under 5 y of age and do not provide hazard estimates for older children. The disease burden from other diet-related risks, tobacco, and occupational risks (except injuries and back pain) was almost equally distributed among adults above and below the age of 60 y. For example, 43% of disease burden due to blood pressure and 61% of burden due to tobacco occurred in the 15–59-y age group. More than 90% of disease burden attributable to lack of contraception, illicit drugs, occupational ergonomic stressors, risk factors for injury, and childhood sexual abuse occurred in adults below the age of 60 y. Similarly, 67%–87% of the disease burden attributable to alcohol, unsafe sex, and unsafe health-care injections arose from events occurring between 15 and 59 y of age. Most of the risks whose hazards are concentrated in the younger adults are causes of injuries, neuropsychiatric diseases, maternal conditions, and HIV/AIDS. Assessment by the level of economic and demographic development illustrated that, with the exception of alcohol, which has global presence, the majority of disease burden from risks for mortality and disease among young adults is concentrated in developing countries (see Figure 1 in [ 2 ]). Stratification of economic and demographic development was also a determinant of the age distribution patterns for risk factors for chronic diseases, which occurred in younger age groups in developing countries than in developed regions. For example, in high-mortality developing regions, 69% of the tobacco burden occurred in people aged 15–59 y, whereas this share was 63% for lower-mortality developing countries and 55% for developed countries. Figure 1 Distribution by Exposure Level of Attributable Disease Burden Due to Selected Continuous Risk Factors Figure 1 shows the distribution of the estimated cardiovascular disease (CVD) burden of disease (in DALYs) attributable to four major continuous risk factors, by exposure levels. Half the attributable burden occurs to the left of the solid vertical line and half occurs to the right. The dashed vertical lines indicate commonly used thresholds—150 mm Hg for hypertension, 6.0 mmol/l for hypercholesterolemia, and 30 kg/m 2 for obesity. The blood pressure and cholesterol levels plotted are the estimated usual levels [ 22 ], which tend to have a smaller SD than levels based on one-off measurements commonly used in population surveys, because of normal day-to-day and week-to-week fluctuations. For example, the distribution of usual blood pressure is about half as wide as the distribution of one-off blood pressure measures, and so many fewer people would be classified as hypertensive (or hypotensive) if classifications were based on usual rather than one-off blood pressure. Thus, if a population mean SBP was 134 mm Hg, the SD of once-only measures might be 17 mm Hg (with about 18% of the population having one-off SBP over 150 mm Hg), and the SD of usual SBP based on many measures would be about 9 mm Hg (hence about 5% of the population would have usual SBP over 150 mm Hg). The distributions of attributable risk-factor burden by exposure levels are shown in Table 4 for those risks quantified using categorical variables and in Figure 1 for those with continuous variables. For most of these risks a substantial proportion of attributable disease burden occurred among those with modest elevations of risk. For example, while 35% of the large disease burden from being underweight in childhood, the leading risk factor for global loss of healthy life, occurred in severely underweight children who would be subject to clinical interventions (i.e., more than three standard deviations [SDs] from referent group median), the rest occurred in children only 1–3 SDs below the median. Even though the relative risks for the latter group are much lower, the number of children exposed to risk at this level is so great that the total attributable disease burden amounted to much more than that occurring in the severe category. Similarly, 52% of the attributable burden from physical inactivity occurred among those engaged in some but less than the recommended level of 2.5 h per week of moderate-intensity activity. For unsafe water, sanitation, and hygiene, almost all of the attributable disease burden was distributed among three of the five at-risk exposure categories, with approximately equal levels. This reflects the fact that the exposure categories were defined as the presence or absence of technology-based water and sanitation interventions. During decades of water and sanitation projects, many countries have “clustered” in a limited number of technology groups. There is likely to be large heterogeneity of exposure within each exposure category, however, because of factors such as quantity of water consumed and hygiene behavior [ 13 ]. Table 4 Distribution by Exposure Level of Attributable Burden Due to Selected Categorical Risk Factors Figure 1 shows that at the aggregated level a substantial proportion of the attributable disease burden for high blood pressure, cholesterol, and BMI and low fruit and vegetable intake occurred in the “mid-range” exposures. For example, the second and third quartiles (i.e., half of attributable disease burden) occurred at the following levels: systolic blood pressure (SBP) of 130–150 mm Hg, cholesterol of 5.0–6.1 mmol/l, BMI of 25–32 kg/m 2 , and fruit and vegetable intake of 150–300 g/d (2–4 servings/d). This was similar to or greater than the amount of disease burden occuring among individuals with risk-factor levels above the commonly used, but arbitrary, thresholds for hypertension, hypercholesterolemia, and obesity indicated in Figure 1 . Despite the above finding on the important role of moderately elevated levels in total disease burden, the actual patterns of how disease burden is distributed among exposure levels varied across different regions and risk factors ( Figure 2 ). For example, Figure 2 shows that a comparatively larger fraction of the disease burden attributable to elevated blood pressure, cholesterol, and BMI occurred at lower levels in developing regions compared to developed regions, mainly because of lower age-specific exposure levels in those populations. Figure 2 also shows that the skewness of the distribution of disease burden was not substantially different across different age groups for BMI. This is because the comparatively larger relative risk per unit BMI at younger ages (which leads to more right-hand skew) is counterbalanced by the comparatively lower BMI at younger ages (which leads to left-hand skew). Therefore, the different distributions of BMI-attributable disease burden by region resulted not from the different age structures of populations across major world regions, but rather from the lower age-specific BMI levels in those countries [ 14 ]. This is in contrast to blood pressure, for which disease burden in younger age groups occurred at lower exposures because the age patterns of exposure and relative risk do not entirely compensate. Figure 2 Distribution of Attributable Cardiovascular Disease Burden Due to BMI, Blood Pressure, and Cholesterol by Exposure Level, Age, and Level of Development Conventions as for Figure 1 . Discussion The findings reported here should be considered within the context of limited available data and are subject to uncertainty, which varies across risk factors and is generally most marked in developing countries. Full discussion of uncertainty in the basic data sources and parameters is provided elsewhere and includes the uncertainties in estimates of disease incidence, duration, severity, and disability weighting [ 1 , 2 ]. Uncertainty in this risk assessment is by far dominated by absence or limitations of direct studies on exposure, hazard, and background disease burden, rather than parameter uncertainty, such as sampling error. This has motivated innovative assumptions and extrapolations even in the case of the best-studied risk factors in a limited number of countries [ 6 ]. While estimates of hazard size in individual studies were as much as possible adjusted for confounding effects, it is likely that residual confounding effects remain to some extent, hence leading to errors in estimation. Extrapolation of hazard from a limited number of studies to other populations is another source of potential error. While the robustness of relative measures of risk has been confirmed for more proximal factors in studies across populations [ 15 , 16 , 17 ], their extrapolation is an important source of uncertainty for more distal risks (e.g., childhood sexual abuse) or those whose effects are heterogeneous (e.g., alcohol and injuries versus alcohol and cancer). Direct exposure data for many risk factors are limited because of both difficulties in their measurement and underinvestment in risk-factor surveillance, especially in developing countries. Of particular relevance to the current analysis is the fact that, due to data limitations, some risks were measured using categorical variables (e.g., indoor smoke from solid fuels, underweight, and physical inactivity) even though the health effects occur along a continuum. Other risk factors were represented using indirect or aggregate indicators (e.g., smoking impact ratio for accumulated hazards of smoking, and motor vehicle accident registries for alcohol-caused accidents) that do not allow quantification of risks along continua of exposure. Two important sources of uncertainty with implications for interventions are correlations among multiple risk factors and the skewness of risk-factor distribution. Because risks are often correlated (e.g., undernutrition, poor water and sanitation, and the use of solid fuels are more common among poor households in developing countries, and smokers are more likely to have poor diets), the contributions of high-exposure groups are likely to be underestimated. In addition to being a source of underestimation at higher exposure levels, risk-factor correlation implies that the same individuals and groups are at the high end of multiple risk factors. Positive (rightward) skewness of exposure distribution, not modeled here, would also lead to an underestimation of events occurring in those who were hypertensive, hypercholesterolemic, or obese. The importance of skewness is emphasized by the observation that the recent increase in the proportion of people who are overweight and obese (e.g., in the United States) has involved not only a shift in the distribution, but also increasing skewness, which has extended the high-exposure tail. Coupled with risk-factor correlation, this should motivate more systematic data collection and reporting to determine mean exposure as well as the complete shape of distribution. The temporal nature of risk-factor exposure also has implications for the current cross-sectional estimates. There is an expectation that the size and rank order of risk-factor burden will alter in coming decades as a result of changes in exposure levels and delays between exposure and hazard. For example, it is predicted that by 2020 the leading risks to health will be (1) unsafe sex, principally because of HIV/AIDS and driven by increasing exposure, and (2) tobacco, because of market expansion of tobacco products in developing countries and delayed temporal effects on major health outcomes such as lung cancer [ 1 ]. This analysis in multiple age and exposure categories, or along a continuum of exposures, showed that, globally, a considerable proportion of the disease burden attributable to major risk factors occurred among those with only moderately raised levels, not the extremes such as those that define hypertension, obesity, or severe malnutrition. Further for many chronic-disease risk factors, such as tobacco and high blood pressure, as well as risk factors for injuries and sexual and reproductive health, significant proportions of disease burden occurred from events in middle ages, especially in developing countries. The concentration of disease burden from such a large number of risk factors in those below 60 years of age illustrates the large, and at times neglected, disease burden from risks that affect young adults in developing nations, with important consequences for economic development [ 18 ]. The distribution of risks and their determinants in a population have major implications for strategies of prevention. As stated by Rose, risk typically increases across the spectrum of a risk factor [ 8 ]. Rose's work led to one of the most fundamental axioms in disease prevention across risk factors: “A large number of people exposed to a small risk may generate many more cases than a small number exposed to high risk.” The analysis in this work showed that the risk factors for many of the major global diseases, such as lower respiratory infections, diarrhea, ischemic heart disease, and stroke exhibit such behavior, because they are caused by risks that occur along a continuum. Therefore, managing individual, high-risk crises, while appropriate for individuals, can only have a limited preventive effect at the population level and over long time periods because it relies on having adequate discriminative ability to predict future disease, and requires continued and expensive screening for new high-risk individuals. In contrast, population-based strategies that seek to shift the whole distribution of risk factors have the potential to substantially reduce total disease burden, and possibly over long time periods, if the interventions alter the underlying risk behaviors or their socioeconomic causes. This is particularly relevant in the context of risk factors such as those related to under- or overnutrition that affect societies in all stages of development. For example, a policy to reduce salt content in manufactured foods would result in a leftward shift in the population distribution of blood pressures and a surprisingly large reduction in cardiovascular disease [ 5 ]. Another example would be population-level measures that affected energy intake (such as the availability and price of energy-dense foods) and/or expenditure (such as the level of motorization and mechanization)—these can be expected to determine the distribution of a population's BMI levels, and hence largely determine its level of type 2 diabetes [ 1 ]. There were distinct patterns for the distribution of disease burden across risk factors, and across regions at various stages of development. At the extreme of these diverse patterns, for risk factors with acute exposure and acute outcomes, the underlying relationship is considerably more complex. For example while in many societies most alcohol-attributable injury (e.g., traffic accidents) involves people who on average drink moderate amounts [ 19 ], these people would be on the more extreme, high-risk spectrum in a different dimension: volume of drinking right before the injury. Therefore, if exposure to risk factors is clustered or the risk relationship does not follow a linear pattern, high-exposure groups may indeed play a disproportionately important role [ 20 , 21 ]. In summary, the analysis presented in this paper confirms and extends to a global level previous work indicating that the impact of many major risks is important across their exposure levels, not just among people with particularly high levels [ 8 ]. This analysis also illustrates that, beyond the general principle of population-wide shifts, there are important population-specific and risk-factor-specific patterns in how the burden of disease attributable to risk factors is distributed. Systematic assessment of multiple risks within any given population can provide the basis for selecting packages of interventions that include population-wide measures as well as highly targeted interventions provided to much smaller subsections of the population with constellations of major risks [ 1 , 5 , 18 ]. Patient Summary Background Health policy makers must understand existing health risks and which diseases cause the most health problems. The Global Burden of Disease database, maintained by the World Health Organization, collects information from around the world on risk factors such as malnutrition, childbirth, tobacco use, and cholesterol levels, as well as on diseases such as depression, blindness, and diarrhea. This information can be used to target health interventions. What Did the Researchers Do and Find? These researchers determined for 26 major risk factors the distribution of disease burden by age, sex, and exposure level. They found that many risk factors (such as high blood pressure and high cholesterol levels) occur across populations, and are not confined to one sex or age group. And for most risk factors, exposure to moderate risks (which is usually more common than exposure to severe risk) is responsible for causing most disease. What Do the Results Mean, and Who Will Use Them? Measures that reduce exposure to risk factors across whole populations, if only by a modest extent, can achieve large reductions in disease burden. This information is important for people who set global health policies and priorities. What Are the Problems with the Study? There are many challenges in estimating the impact of different major risks, each of which has distinct effects on a number of diseases. In addition, exposure to some risk factors today will cause disease only many years from now, so the picture will change over time. The major finding of the global distribution of many major risks is secure, but the exact extent of this remains uncertain due to the paucity of data in developing countries.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC523844.xml
406408	Defects in Ribosomal Protein Genes Cause Cancer in Zebrafish	xx	To investigate the genetic underpinnings of a particular biological process, geneticists screen large collections of mutant organisms to characterize their physical defects. By comparing the genetic makeup of nonmutant (called wild-type) organisms to mutants, it's possible to tease out the genes responsible for a defective appearance, or phenotype. In a classic study in the fruitfly, Christiane Nüsslein-Volhard and Eric Weischaus bred many lines of flies with mutations that were lethal: the fly embryos died, but not before displaying a wide range of developmental defects. Since it was known that the fruitfly needed only a single wild-type copy of these genes to survive, the mutations in these “embryonic lethals” had to be recessive, meaning that both copies, or alleles, of the gene had to be mutated for the lethal defect to appear. Nüsslein-Volhard and Weischaus's work revealed many such recessive genes crucial to early development and earned them a Nobel Prize. Zebrafish tumors caused by mutation of a ribosomal protein gene Among the model systems for studying development, the zebrafish has become prized because its transparent embryo develops outside the mother's body. The zebrafish has helped biologists identify many genes involved in embryogenesis and, because it's a vertebrate animal, has become a valuable resource for identifying genes involved in human disease. Now, a team led by Nancy Hopkins of the Massachusetts Institute of Technology, has created over 500 lines of zebrafish with lesions in key embryogenic genes and used them to identify a group of genes that predispose the fish to cancer, with some surprising results. All of the 500 lines created by the researchers carried a recessive embryonic lethal mutation; for about 400 of the lines, mutations in 300 distinct genes were identified as the cause of the embryonic phenotype. During the process of cultivating some of these mutant lines, the Hopkins team noticed that an abnormally large percentage of fish experienced early mortality (in some cases, over 50% compared to the 10%–15% seen in nonmutant fish), while the surviving fish in these lines developed large, highly invasive malignant tumors; both phenotypes persisted over successive generations. The tumors resembled malignant peripheral nerve sheath tumors (MPNSTs) that have been found in other fish species as well as in mammals. Suspecting that these mutant lines had elevated rates of cancer, the researchers investigated the genetic makeup of the fish and discovered to their surprise that each line was heterozygous for a mutation in a different ribosomal protein gene (rp) —that is, each line carried one healthy version and one defective version of a different rp gene. These proteins are components of ribosomes—the massive molecular complexes within cells that mediate protein synthesis—and are essential for embryonic development. All of the rp mutations, the researchers report, either reduced or eliminated expression of the corresponding rp gene. In the case of “classic” tumor suppressor genes, the wild-type allele must be lost for the defective allele to set the stage for cancer. Here, the wild-type allele appeared to remain intact in the tumor cells, implicating the proteins as “haploinsufficient” tumor suppressors—a reduction from two gene copies to one functional copy seems to be enough to increase the risk of cancer. Apart from the mutations in rp genes, the authors also found a loss-of-function mutation in a gene (called NF2 ) that acts as a tumor suppressor in mammals—establishing the soundness of this approach for identifying mammalian cancer genes. While these experiments do not explore how these mutations lead to cancer, the results suggest that some shared, ribosome-associated function allows these genes to act as tumor suppressors and that disrupting this function somehow leads to tumor formation. Though it's not clear what distinguishes the 11 rp genes whose mutations caused cancer from the five other rp genes whose mutations did not, the authors raise a number of possibilities for future study. And given the high degree of conservation of genes and pathways among vertebrates, it's likely that rp mutations also raise cancer risk in humans. Together, these results demonstrate that the tiny freshwater workhorse of developmental biology has a promising future as a model system for human cancer.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC406408.xml
550672	Stress-echocardiography in idiopathic dilated cardiomyopathy: instructions for use	A number of studies have suggested that stress-echocardiography may be used for prognostic stratification in patients with idiopathic dilated cardiomyopathy. There is no consensus on which protocol or which measurements of left ventricular contractile reserve to use. The most frequently used protocol is low-dose dobutamine stress-echocardiography, and most commonly used measures of left ventricular systolic performance are ejection fraction, wall motion score index and cardiac power output. Stress-echocardiography has been shown to predict improvement in cardiac function in patients with recently diagnosed dilated cardiomyopathy, as well as to predict which patients will benefit from the treatment with beta-blockers. Most importantly, stress-echocardiography can identify patients with worse prognosis in terms of cardiac death and need for transplantation. Additionally, contractile reserve is closely correlated with maximal oxygen consumption and can even be used for further stratification in patients with maximal oxygen consumption between 10 and 14 ml/kg/min. Future studies are needed for head-to-head comparison of various protocols in an attempt to make standardization in the assessment of patients with dilated cardiomyopathy.	Epidemiologic data from United States indicate that idiopathic dilated cardiomyopathy (DCM) is diagnosed in approximately 36/100.000 persons each year, and that it is responsible for more than 10.000 deaths per year [ 1 ]. Faced with the fact that the number of patients with DCM is constantly increasing [ 2 ], accurate assessment of patient's current status and prognosis is of the utmost importance for the implementation of optimal therapeutic algorithm as well as for the optimal utilization of resources. Why stress-echocardiography? There is a widespread belief that maximal oxygen consumption, assessed by cardiopulmonary testing, is one of most, if not the most important prognostic variables in DCM patients [ 3 ]. Maximal oxygen consumption is traditionally used for selection of patients for cardiac transplantation, with a values less than 12–14 ml/kg/min indicating poor prognosis and need for transplantation [ 4 , 5 ]. This approach is based upon assumption that maximal oxygen consumption during cardiopulmonary testing is determined exclusively by cardiovascular factors. However, it appears that other factors have considerable influence on maximal oxygen consumption, since it has been shown that regular physical exercise can augment oxygen consumption with little or no impact on other parameters of cardiovascular function [ 6 ]. Additionally, a normal blood flow through lower limbs was found in patients who stopped cardiopulmonary testing because of fatigue, indicating skeletal muscle dysfunction rather than pump failure [ 7 ] Patient's age and sex, as well as muscle mass are also shown to have strong influence on performance during cardiopulmonary testing [ 8 ]. Furthermore, it appears that patients with various degrees of cardiovascular impairment may yield similar maximal oxygen consumption, suggesting that there is a role for other procedures in risk stratification of patients with DCM [ 9 ]. A number of studies have shown that assessment of ventricular contractile reserve by means of stress-echocardiography may refine prognosis in patients with left ventricular systolic dysfunction [ 10 - 12 ]. Nevertheless, stress-echocardiography is widely underused in routine work-up in patients with heart failure, probably because of the unfamiliarity with the technique in this clinical setting. The aim of this review is to put stress-echocardiography in DCM patients in clinical context, to give practical tips how to perform it and what to measure, as well as to try to define its role in everyday clinical practice. How to perform stress-echocardiography in DCM? Unlike protocols for stress-echocardiography for coronary artery disease, there is no consensus about the protocol to be used in patients with left ventricular systolic dysfunction. The majority of authors have used either low- or high-dose dobutamine echocardiography [ 13 , 14 ]. Low-dose is usually defined as 10 mcg/kg/min of dobutamine [ 13 ], although some authors also consider 20 mcg/kg/min as a low-dose infusion [ 15 ] High-dose is uniformly defined as 40 mcg/kg/min [ 10 ]. There is no consensus either on duration of each stage of dobutamine infusion, since some authors use 3-minute stage [ 16 ] while the others use 5-minute stage [ 10 ] Most authors withdraw beta-blockers prior to stress-echocardiography, but some do not [ 15 ] Atropine is generally not used to achieve submaximal heart rate. Exercise testing has been used for the assessment of left ventricular contractile reserve but in conjuction with radionuclide angiography [ 17 ] and hemodynamic measurements [ 18 ]. The usual protocol is performed on supine bicycle in incremental stages of 25 W lasting three minutes each. There are no data about the value of exercise stress echocardiography in DCM patients. It can be assumed that exercise stress-echocardiography would have the same limitations as cardiopulmonary testing. Dipyridamole stress-echocardiography has been proposed for stratification of patients with DCM. Standard high-dose dipyridamole protocol (0.84 mg over 10 minutes) has been used for the assessment of contractile reserve [ 19 ]. What to measure? All studies on stress-echocardiography in DCM measured contractile reserve of the left ventricle. Contractile reserve is defined as the difference between values of an index of left ventricular contractility during peak stress and its baseline values. There is no consensus on what index to use. Ejection fraction This is the most frequently used index of left ventricular performance. However, it may not accurately reflect left ventricular contractility since it is heavily dependent on loading conditions [ 20 ] which is particularly important in patients with heart failure for the following reasons. First of all, mitral regurgitation is frequent in these patients, and can lead to overestimation of left ventricular contractility due to rise in ejection fraction caused by changes in loading conditions (higher preload, lower afterload) [ 21 ] Secondly, activation of neuroendocrine compensatory mechanisms may increase afterload, which in turn may subsequently decrease ejection fraction [ 22 ] Thirdly, left ventricular preload is dependant upon interventricular interaction which is exaggerated in cases of pulmonary hypertension [ 23 ] a frequent finding in DCM patients. Furthermore, dobutamine has variable influence on afterload, since it has been shown that it may decrease afterload by 10% in patients with mild heart failure, but may also increases afterload by 5% in patients with severe heart failure [ 24 ]. Despite all these potential drawbacks, the change in ejection fraction during stress has been shown to have crucial prognostic significance in patients with DCM. It is generally accepted that increase in ejection fraction by ≥ 5% or change from baseline ejection fraction by ≥ 20% during stress-echocardiography identifies patients with preserved left ventricular contractile reserve and better prognosis. Ejection fraction should be assessed by Simpson biplane formula. Wall motion score index Wall motion score index has been traditionally used in stress-echocardiography for the detection of coronary artery disease [ 25 ]. Only two reports used this index of left ventricular contractility to assess prognosis and functional recovery of DCM patients [ 10 , 15 ]. Wall motion score index was assessed in a standard manner, by using 16 segment model of the left ventricle according to the recommendations given by American Society of Echocardiography [ 26 ]. The major potential drawback for use of this index is semiquantitive assessment of wall motion, which is even more subjected to inter- and intraobserver variability in DCM patients due to preexisting wall motion abnormalities and substantial number of patients with left bundle branch. It has been suggested that dobutamine induced change in wall motion score index of ≥ 0.44 identifies patients who will do better during the follow-up. Cardiac power output This index is not sensitive to changes in afterload, and after optimization for preload accurately reflects contractile properties of the myocardium [ 27 ]. Noninvasive calculation of cardiac power output is relatively complex and requires special instrumentation [ 11 ]. The most practical formula to calculate cardiac power output in Watts was suggested by Cook and coworkers [ 28 ]: Cardiac power output = (cardiac output × mean arterial pressure) × 2.22 × 10 -3 where cardiac output is calculated by multiplying aortic velocity-time integral by aortic valve area, and mean arterial pressure is calculated in a standard manner. The major problem with this index is that its calculation is time consuming, requires skilled echocardiographer, and is subjected to numerous sources of error. Furthermore, very few cardiologist are familiar with this index which precludes its wider use. Suggested cut-off point between patients with respect to prognosis is dobutamine induced change in cardiac power output of ≥ 1 W. Prognostic significance There is no doubt that change in left ventricular contractility during stress has considerable prognostic significance and may have profound effect on therapeutic strategy. We will review available data according to the means how contractile response was elicited. Low-dose dobutamine Most authors prefer low-dose dobutamine stress-echocardiography, probably because it is considered safe and is not time consuming. A report by Paelinck and coworkers has suggested that low-dose dobutamine can identify patients with atrial fibrillation induced dilated cardiomyopathy who will improve following restoration of sinus rhythm [ 29 ]. These authors concluded that low-dose dobutamine may be used to identify patients with tachycardiomyopathy. It has been shown, in a small number of patients, that changes in left ventricular wall motion score index and ejection fraction during low-dose dobutamine echocardiography are predictive of improvement of left ventricular systolic performance during medium term follow-up (Figure 1 ) [ 15 ]. Since the degree of beta-receptor downregulation and desenzititation is a marker of progressive deterioration of left ventricular systolic function [ 30 ], the authors hypothetized that improvement in contractility during dobutamine infusion is greater in patients with preserved beta-receptor function who will subsequently show improvement in systolic performance. Figure 1 Change in ejection fraction during follow-up in patients wih preserved (group A) and diminished (group B) contractile reserve. Abreviations: LVEF, left ventricular ejction fraction. From: Kitaoka H, Takata T, Yabe N, Hitomi N, Furuno T, Doi YL: Low dose dobutamine stress echocardiography predicts the improvement of left ventricular systolic function in dilated cardiomyopathy. Heart 1999; 81 :523-27 . These findings are further extended, so it has been shown that the presence of myocardial contractile reserve identifies patients who will respond favorably to beta-blocker therapy [ 31 ]. Furthermore, Drozd and coworkers have demonstrated that the incidence of cardiac death or need for cardiac transplantation is lower in patients with preserved contractile reserve. In this paper, multivariate analysis identified left ventricular end-systolic volume of less than 150 ml after dobutamine infusion and no decrease of left ventricular end-diastolic volume after dobutamine infusion as significant predictors of combined end-point [ 32 ]. Contractile reserve has been shown to correlate well with peak oxygen consumption (Figure 2 ) [ 16 ]. At multivariate analysis in this report, only percentage change in end-systolic volume index was significantly associated with occurrence of cardiac death or hospitalization for worsening heart failure. The area under receiver-operating characteristic curve was similar for percentage change in end-systolic volume index and peak oxygen consumption (0.86 ± 0.04 vs. 0.80 ± 0.06). Additionally, a report by Paraskevidis and coworkers suggested that low-dose dobutamine may further refine prognosis in patients with maximal oxygen consumption between 10 and 14 ml/kg/min [ 13 ]. This finding may be used for prioritization of patients for cardiac transplantation. Figure 2 Linear correlation with 95% confidence interval between peak oxygen consumption and percent change in end-systolic volume index. Abreviations: % ESVI, percent change in end-systolic volume index; peak VO2, peak oxygen consumption. From: Scrutinio D, Napoli V, Passantino A, Ricci A, Lagioia R, Rizzon P: Low-dose dobutamine responsivness in idiopathic dilated cardiomiopathy: relation to exercise capacity and clinical outcome. Eur Heart J 2000; 21 :927-34. High-dose dobutamine Use of high-dose dobutamine is not associated with serious complications in DCM patients, and has an overall feasibility of 88.7%. The most common adverse event requiring discontinuation of dobutamine infusion is occurrence of complex ventricular arrhythmias, which was noted in 8% of patients (frequent multifocal ventricular extrasystoles in 6.4% and nonsustained ventricular tachycardia in 1.6%) [ 10 ]. Although there are no data on the association of complex ventricular arrhythmias and serum potassium concentrations, it may be postulated that complex arrhythmias are more frequent in potassium depleted patients. Therefore, it appears prudent to check serum potassium level prior to high-dose dobutamine stress-echocardiography. Hypotension, defined as decrease in systolic blood pressure by more than 30 mmHg, is very rare in the absence of complex ventricular arrhythmias and, in authors experience, occurs in less than 1% of patients with angiographically documented idiopathic DCM. Potential advantage of high-dose, as compared to low-dose, dobutamine echocardiography is that it may evoke more complete contractile response. Very intriguing finding is that early in the course of DCM, dobutamine induced change in left ventricular contractile response and geometry is able to predict late spontaneous recovery of left ventricular systolic performance [ 14 ]. It is interesting that this study confirmed previous findings that increased left ventricular mass is associated with better outcome in DCM [ 33 ], and suggested that the presence of left ventricular hypertrophy implies the presence of myocardial contractile reserve. The largest study that studied prognostic significance of high-dose dobutamine included 186 DCM patients. The major findings of this study are that dobutamine induced change in wall motion score index is able to identify patients at greater risk for cardiac death during the follow-up (Figure 3 ), and that change in wall motion score index carries superior prognostic information than change in ejection fraction [ 10 ] Additionally, it has been reported that dobutamine induced change in ejection fraction by ≥ 8%, assessed by radionuclide ventriculography, is prognostically superior to maximal oxygen consumption in patients with severe DCM [ 34 ]. Figure 3 Kaplan-Meir survival curves (only cardiac deaths were considered) in patients stratified according to the dobutamine induced change in wall motion score index. Abbreviations: ΔWMSi, change in wall motion score index. From: Pratali L, Picano E, Otašević P, Vigna C, Palinkas A, Cortigiani L, Dodi C, Bojić D, Varga A, Csanady M, Landi : Prognostic significance of the dobutamine echocardiography test in idiopathic dilated cardiomyopathy. Am J Cardiol. 200; 88 :1374-8. Recent data by our group demonstrate that contractile reserve indices assessed by high-dose dobutamine correlate with myocardial histomorphometric features, suggesting that contractile reserve is strongly related to the degree of hystological disruption in DCM patients. Myocyte diameter and interstitial fibrosis showed strongest correlation with change in wall motion score index (r = -0.667, p < 0.001, and r = -0.567, p = 0.004, respectively), followed by change in ejection fraction (r = -0.603, p = 0.002, and r = -0.467, p = 0.021, respectively) [ 35 ]. Dipyridamole It appears that dipyridamole may be used instead of dobutamine to evoke contractile response, since it is less arrythmogenic [ 36 ] better tolerated and yields similar prognostic information in patients with coronary artery disease [ 37 ]. Ability of dipyridamole to recruit contractile reserve is mediated through increase in coronary blood flow and accumulation of endogenous adenosine [ 38 , 39 ]. Potential advantage of dipyridamole over dobutamine stress-echocardiography is that the former is not affected by the use beta-blocking agents which are frequently used in DCM patients. Only one study recently examined ability of dipyridamole to predict prognosis in DCM patients. The authors concluded that increase in wall motion score index ≥ 0.15 during dipyridamole stress identifies patients who are more likely to survive during the mean follow-up of more than three years [ 19 ]. Reported overall feasibility of dipyridamole stress-echocardiography in this study was 99.2%, which is significantly higher than previously reported feasibility of dobutamine stress-echocardiography. Exercise As previously said, there are no echocardiographic studies on exercise induced contractile response in DCM patients. However, Nagaoka and colleagues used radionuclide ventriculography to measure increase in ejection fraction during exercise in DCM patients with mild symptoms, and concluded that change in ejection fraction <4% identifies patients with worse prognosis [ 17 ]. Additionally, it has been suggested that variables, such as cardiac power output, obtained by direct hemodynamic measurements during exercise may have important prognostic implications in patients with systolic dysfunction [ 12 ]. What about the right ventricle? Right ventricular contribution to global cardiac performance is minor in subjects with normal or mildly depressed left ventricular systolic function, but may become more important in patients with advanced left heart failure [ 40 ]. Previous studies have suggested that right ventricular enlargement is a strong marker for adverse prognosis in DCM patients [ 41 ], as well as that right ventricular long axis excursion is predictive of exercise tolerance [ 42 ]. However, there are only limited reports of the prognostic value of right ventricular contractile reserve in patients with DCM. DiSalvo and colleagues have demonstrated that an increase in RVEF to >35% during exercise is the only independent predictor of event-free survival in patients with advanced heart failure [ 43 ]. It has been also shown that preserved right ventricular contractile reserve (measured by pressure-area relations) induced by low-dose dobutamine infusion was associated with a good 30-day outcome in patients with NYHA class IV heart failure [ 44 ]. Data from our laboratory support prognostic significance of high-dose dobutamine induced change in right ventricular fractional area change [ 45 ]. It appears that fractional area change of >9% identifies patients with more favorable outcome. More importantly, these data suggest that patients in whom contractile reserve of both ventricles is preserved will most likely have good prognosis. What role for stress-echocardiography? Despite the wealth evidence that favor use of stress-echocardiography in patients with DCM, there is no clear-cut algorithm about its use in risk stratification and therapeutic strategy. The reasons for this are not clear, but probably reflect the lack of standardized protocol and measurements of left ventricular contractile reserve. We strongly believe that stress-echocardiography should be used as a standard procedure, at least in centers which do not have access to cardiopulmonary testing, since data obtained when patients are subjected to some form of stress have far greater prognostic significance than data obtained at rest. Furthermore, stress-echocardiography should be used in patients who are not able to exercise or fail to achieve expected work load. Stress-echocardiography may also play an important role for detailed risk stratification in patients with maximal oxygen consumption of 10–14 ml/kg/min. The choice of stress-protocol, at least for the time being, should be based upon local expertise and preferences of attending physician. Future directions There is an obvious lack of studies that will contribute to standardization of stress-echocardiographic protocol. Head-to-head comparison of stressors, including low- and high-dose dobutamine, dipyridamole, and exercise, has to be performed in order to rank their ability to predict prognosis. Similar comparisons have to be made for various indices of left ventricular contractility. Additionally, it is not clear should the patients be tested with or without beta-blocker therapy, and how this therapy may affect our choice of stressor. Last but not the least, novel echocardiographic techniques that can easily assess regional and global contractility, like tissue Doppler imaging and strain-rate imaging, have not yet been tested in a prospective manner. In conclusion, stress-echocardiography can be a valuable tool for the assessment of patients with DCM, but a lot of work has to be done before it becomes a part of a routine work-up.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC550672.xml
535940	SMAC is expressed de novo in a subset of cervical cancer tumors	Background Smac/Diablo is a recently identified protein that is released from mitochondria after apoptotic stimuli. It binds IAPs, allowing caspase activation and cell death. In view of its activity it might participate in carcinogenesis. In the present study, we analyzed Smac expression in a panel of cervical cancer patients. Methods We performed semi quantitative RT-PCR on 41 cervical tumor and 6 normal tissue samples. The study included 8 stage I cases; 16 stage II; 17 stage III; and a control group of 6 samples of normal cervical squamous epithelial tissue. Results Smac mRNA expression was below the detection limit in the normal cervical tissue samples. In contrast, 13 (31.7%) of the 41 cervical cancer biopsies showed detectable levels of this transcript. The samples expressing Smac were distributed equally among the stages (5 in stage I, 4 in stage II and 4 in stage III) with similar expression levels. We found no correlation between the presence of Smac mRNA and histology, menopause, WHO stage or disease status. Conclusions Smac is expressed de novo in a subset of cervical cancer patients, reflecting a possible heterogeneity in the pathways leading to cervical cancer. There was no correlation with any clinical variable.	Background Apoptosis is an evolutionarily conserved biological process that plays a fundamental role in development and tissue homeostasis in metazoans [ 1 ]. This type of cell death is executed by a family of proteases known as caspases [ 2 ]. There are two well-characterized apoptotic pathways that converge in caspase activation: the death receptor pathway and the mitochondrial pathway [ 3 ]. Inhibitors of Apoptosis Proteins (IAPs) are the most important regulators of caspases. These proteins inhibit caspase activation, thus preventing the induction of apoptosis [ 4 ]. In cells undergoing apoptosis, IAPs are inactivated by interaction with proteins containing the so-called IBM (IAP-binding motif) [ 4 , 5 ]. One IBM protein is the recently identified Smac/DIABLO [ 6 , 7 ]. Smac resides in the mitochondrial intermembrane space in healthy cells but is released into the cytosol during apoptosis, where it interacts with IAPs and disrupts their ability to bind caspases [ 8 ]. Smac is expressed ubiquitously, with high expression in adult testis, heart, liver, kidney, spleen, prostate and ovary and low expression in brain, lung, thymus, and peripheral blood leukocytes [ 9 ]. It is encoded in a nuclear gene and is post-translationally imported into the mitochondria via a targeting sequence in its amino terminus. Removal of this signal generates a mature polypeptide with the IBM at the amino terminal end [ 10 ]. Smac interacts with all mammalian IAPs examined so far: XIAP, cIAP-1, cIAP-2, survivin and ML-IAP [ 6 , 7 , 11 , 12 ]. The structure of the Smac-XIAP complex has been studied by X-ray crystallography [ 13 ] and high-resolution NMR [ 14 ]; it appears that the tetrapeptide AVPI is indispensable for the formation of this complex. IAPs are highly expressed in human tumor cells [ 15 - 17 ], contributing to the intrinsic resistance of these cells to endogenous death receptor-induced apoptosis and consequently to chemotherapy [ 18 ]. For this reason, peptides mimicking the action of Smac have been generated and analyzed. Four publications to date have shown promising effects of these Smac peptides in vitro and in vivo ; however, further studies are required prior to clinical testing [ 19 - 22 ]. Recently, Sekimura and colleagues found that Smac expression was significantly lower in primary lung cancers than in normal tissue [ 23 ]; patients with lower Smac mRNA levels had worse prognoses. These results indicate that Smac expression may play a role in the progression of primary lung cancer and may be useful for prognosis [ 23 ]. However, Smac expression has not been analyzed in other tumors. In view of the possible role of Smac in cervical carcinogenesis and its potential as a therapeutic target, we have investigated the expression of this apoptotic protein in cervical cancer patients. Methods Cell lines and tumor samples Cervical cancer cell lines (HeLa, SiHa, CaSki and CaLo) were obtained from ATCC and cultured as monolayers in Dulbecco Modified Eagle's Medium (DMEM) containing 10% (V/V) fetal bovine serum (GIBCO, Bethesda, MD, USA) at 37°C in a humidified atmosphere of 5% (V/V) CO 2 . Forty-one cervical cancer samples were obtained from the Instituto Nacional de Cancerologia of Mexico. Written consent was obtained from patients before the samples were collected. Tumors were staged according to the International Gynecology and Obstetric Federation (FIGO) system. The samples comprised 8 at stage IB, 16 at stage IIB and 17 at stage IIIB; and a control group comprising 6 samples of normal cervical squamous epithelial tissue (Table 1 ). The control samples were derived from hysterectomy specimens from patients with uterine myomatosis. Only samples with normal pathological reports were included. Histology Histopathological grading was done according to the WHO (World Health Organization) classification system (Table 1 ). RNA isolation and RT-PCR RNA extraction and RT-PCR analysis were performed as described previously [ 24 ]. Briefly, total RNA was extracted from cultured cells, tumors and non-neoplastic tissue samples with Trizol reagent (Invitrogen) following the manufacturer's protocol. RNA purity was confirmed by the 260/280 nm absorbance ratio and its integrity was established with agarose gels. Total RNA (2 μg) was reverse-transcribed in a final 20 μl reaction volume using 15 U ThermoScript reverse transcriptase, 2.5 × RT Buffer and random hexamers (ThermoScript RT-PCR, Invitrogen). The RT-PCR steps were 25°C for 10 min, 50°C for 50 min and 85°C for 5 min. Smac and GAPDH mRNA PCR reactions contained 0.25 μl Amplitaq gold polymerase (Applied Biosystems, ROCHE), 2.5 μl 10 × reaction buffer, 0.5 μl dNTP mix 10 mM, 1 μl sense primer 10 μM, 1 μl anti-sense primer 10 μM and 1 μl cDNA in 25 μl final volume. The Smac primers were: sense 5' GCGCGGATCCATGGCGGCTCTGAAGAGTTG 3' and anti-sense 5' AGCTCTCTAGACTCAGGCCCTCAATCCTCA 3'. The GAPDH primers were: sense 5' CCCCTTCATTGACCTCAACT 3' and antisense 5' TTGTCATGGATGACCTTGGC 3'. The PCR cycle parameters for Smac were: 10 min enzyme activation at 95°C followed by 3 cycles of 30 s at 95°C and 2 min at 72°C, then 30 cycles of 30 s at 95°C and 30 s at 68°C, and finally 5 min at 72°C. The corresponding parameters for GAPDH were: 10 min enzyme activation at 95°C followed by 25 cycles of 30 s at 95°C, 30 s at 60°C and 30 s at 72°C. The products were electrophoresed on 1% agarose gels and stained with ethidium bromide. Smac mRNA data were expressed as ratios between the densitometric values (Scion Image software) of Smac gene expression. The PCR products were normalized to the amplified GAPDH, the internal reference gene. Gene expression measurements were repeated at least twice. Statistical analysis To detect a correlation between pathological tumor parameters and normalized Smac expression we used ANOVA (stage, current disease and menopause status) and chi square tests (stage, histology of tumors, menopause and current status). Kaplan-Meier curves for status were generated and log rank was used to test for differences. The mean follow-up was 14.7 months. The statistical package Intercooled Stata 7.0 was used for analyses and statistical significance was accepted when the p value was less than 0.05. Results To ascertain whether Smac is expressed in cervical cancer we performed semiquantitative RT-PCR analyses on a panel of cervical cancer lines, including HeLa, SiHa, CasKi and CaLo cells. As shown in Figure 1 , the HeLa and CasKi lines contained Smac mRNA, but very low levels were observed in SiHa and CaLo cells. Next, we measured Smac mRNA levels using the same approach in 41 cervical tumor and 6 normal cervical samples. To ensure accurate determinations and to verify equal RNA input, GAPDH mRNA was amplified simultaneously. Figure 2 shows a representative panel of results, which are given in Tables 1 and 2 . Unexpectedly, Smac mRNA was below the detection limit in normal cervical samples. In contrast, as expected from the cell line data, 13 (31.7%) of the 41 cervical cancer biopsies contained detectable levels of this transcript. The samples expressing Smac were distributed equally among the stages (5 in stage I, 4 in stage II and 4 in stage III). We found no significant correlation between Smac mRNA level and histology, menopause, clinical stage or disease status (Table 2 ). When the Smac expression levels in the tumor samples were analyzed, there were no significant differences between clinical stages (Figure 3 ), menopause status (Figure 4 ) or disease status (Figure 5 ). Similarly, a survival analysis of the patients showed no statistical differences between patients expressing or not expressing Smac mRNA (Figure 6 ). Discussion Tumors proliferate beyond the constraints that limit growth in normal tissue. Therefore, the resistance of tumor cells to apoptosis is an essential feature of carcinogenesis. This has been confirmed by the finding that deregulated proliferation alone is not sufficient for tumor formation because there is concomitant induction of cell death [ 25 ]. Overexpression of growth-promoting oncogenes such as c-Myc sensitize cells to apoptosis [ 26 ]. Thus, tumor progression requires the expression of anti-apoptotic proteins or the inactivation of essential pro apoptotic proteins [ 27 , 28 ]. Indeed, it has been shown that survivin, a member of the Inhibitor of Apoptosis Protein (IAP) family, is upregulated in some tumors [ 29 ], correlating with prognosis [ 30 , 31 ]. Smac is a recently identified proapoptotic protein that interacts with and inhibits several IAPs, including survivin [ 6 , 11 ]. It has been shown that Smac mRNA levels in tumor tissues are significantly lower than in normal tissues [ 23 ]. Patients with lower Smac mRNA levels have worse prognoses. These results indicate that Smac expression may play a role in the progression of primary lung cancer, as expected by the known role of this protein in cell death induced by chemotherapeutic drugs. Unexpectedly, we found that during cancer progression, some cervical tumors express this protein de novo . Unfortunately, we found no correlation between Smac expression and any clinical variable. This could be attributed to differences in tissue expression of IAPs, which are reported to have different binding affinities for Smac. On the other hand, alternative IAPs such as the recently identified Omi/Htra2 [ 32 ] might play an important tissue- or tumor-specific role. This is supported by the recent report of a null phenotype in Smac-deficient mice, in which a role for other IAP inhibitory proteins is suspected [ 33 ]. Cancer treatment by chemotherapy and γ-irradiation kills cells primarily by the induction of apoptosis. However, few tumors are wholly sensitive to these therapies, and the development of resistance to therapy is an important clinical problem. Failure to activate the apoptotic programme represents an important mode of drug resistance in tumor cells [ 34 ]. Modulation of the key elements in apoptotic signaling should directly influence therapy-induced tumor-cell death. Indeed, it has recently been suggested that peptides mimicking the Smac amino-terminus could be a novel therapeutic weapon [ 19 ]. Tumors with low or null Smac expression, such as the ones reported in this study, could be more susceptible to this approach. Conclusions During cervical cancer progression, a subset of tumors express the apoptotic protein Smac de novo . This finding contrasts with a previous report for lung cancer [ 23 ], underlining the notion that downregulation or even expression of Smac could be dispensable for tumor progression, at least in cervical cancer. This could be because other mitochondrial molecules such as Omi might substitute for its known proapoptotic function. There was no correlation between Smac expression and any clinical variable. Competing interests The author(s) declare that they have no competing interests. Authors' contributions JMZ Conceived and coordinated the study. VAML: Conceived and coordinated the study. Statistical Analysis MEC: Performed RT-PCR assays DCL: Provided the clinical samples and coordinated patient study CMLG: Coordinated patient assessment, ethical guidelines. JGGS: Provided clinical assessment Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535940.xml
517505	Wireless local area network in a prehospital environment	Background Wireless local area networks (WLANs) are considered the next generation of clinical data network. They open the possibility for capturing clinical data in a prehospital setting (e.g., a patient's home) using various devices, such as personal digital assistants, laptops, digital electrocardiogram (EKG) machines, and even cellular phones, and transmitting the captured data to a physician or hospital. The transmission rate is crucial to the applicability of the technology in the prehospital setting. Methods We created two separate WLANs to simulate a virtual local are network environment such as in a patient's home or an emergency room (ER). The effects of different methods of data transmission, number of clients, and roaming among different access points on the file transfer rate were determined. Results The present results suggest that it is feasible to transfer small files such as patient demographics and EKG data from the patient's home to the ER at a reasonable speed. Encryption, user control, and access control were implemented and results discussed. Conclusions Implementing a WLAN in a centrally managed and multiple-layer-controlled access control server is the key to ensuring its security and accessibility. Future studies should focus on product capacity, speed, compatibility, interoperability, and security management.	Background The development of the Internet has encouraged doctors to use computers and hospitals to use wireless communications [ 1 ], since wireless technology offers many benefits over its wired counterpart, including ease of installation and access to network information [ 2 - 6 ], and higher productivity and convenience [ 2 ]. One study using personal digital assistants (PDAs) connected to a network showed that the device was of limited use in transmitting data in prehospital stroke management [ 7 ]. Another study showed that cellular phones, pagers, or other radio-based devices will remain an important communication mode in the near future [ 8 ]. The advancement of wireless local area network (WLAN) technology provides the potential to allow physicians to obtain a patient's information anywhere, even before the patient reaches the emergency room (ER) (Orthner, personal communication). Timely access to a patient's information may fundamentally improve patient care [ 9 ] in both pre- and in-hospital settings, due to earlier doctor interventions. At present, patient data such as electrocardiograms (EKG) and demographics are seldom sent from the prehospital environment to the ER before the ambulance arrives [ 10 ]. As a result, some preventive measures have to be given, regardless of need (e.g., aspirin [ 11 ] or thrombolysis [ 12 ] for presumed acute myocardial infarction. However, despite the potential benefits of wireless technology in prehospital settings, the application of this technology has been slow and few related studies have been carried out. The objective of this study was to assess the ability of wireless technology to facilitate data communication between a prehospital setting and an ER (Orthner, personal communication). The idea was for all the data collected by paramedical personnel to be transmitted to an ER server from the patient's home, on the way to the ER, or upon arrival at the ER. Thus the transmission rate is crucial to the usefulness and applicability of the technology. To test the feasibility of wireless data transmission under the various scenarios, two separate WLANs were created, one around our office and another in a house. In this report, we discuss our testing of the wireless technology, and its potentials and limitations in simulated prehospital settings. Methods The WLAN products used (Aironet 340 and 350 series wireless client adaptors and access points (APs); Cisco) offered 11-megabits-per-second (Mbps) transmission rates, built-in security features (including 40- to 128-bit encryption) and Web-based management. The transmission rates of files of different sizes were measured with different APs, patch antennae, clients, and transfer methods. The security of patient data was ensured using a centrally managed Access Control Server (ACS). Other issues such as standards, roaming, and cost are also discussed here. Within a WLAN, data are transmitted between a server and its wireless clients via an AP antenna. Both workstations and laptops were used here as servers, and file transfer rates were measured for both systems. We used Gateway Select series, Dell Inspiron series, and a Toshiba Satellite laptop computer as clients. The computers had CPU operating at 0.8–1.2 GHz, 256–1024 MB of RAM, and 10–40 GB hard drives, and all ran the Microsoft Windows 2000 Professional operating system. PDAs (Ipaq Pocket PCs, models 3550 and 3570, 200-MHz CPU, 32–64 MB RAM, Compaq) were also tested as wireless clients. The wireless coverage was tested using two APs (Aironet 340 and 350 series, Cisco) and a patch antenna (S2406P, Cushcraft Corporation). The feasibility of using cellular phones (StarTac 7868, Motorola) in data transmission in the area not covered by the WLAN was also tested. The various software used in the study for wireless client management, file transfer, and access control included the Aironet Client Utility (Cisco), Link Status Meter (LSM, Cisco), the ACS (V3.0, Cisco), Phone Tools (BVRP Software) for faxing, and file transfer protocol (FTP) for measuring the file transmission rate. The software LSM classifies the link status as the percentage of maximum signal strength and quality: "excellent" (>75%), "good" (40–75%), "fair" (20–40%), or "poor" (<20%); where signal strength and quality refer to the client adapter's radio signal at the time packets are being received, quantified as bytes transmitted and received and the errors that occur. Detailed descriptions of the mentioned software are available from the manuals provided by the vendors. The WLAN and its configuration The Aironet 340 and 350 series APs were tested by a two-step approach. In the first step, one AP was connected directly with the server that was not connected to the campus Ethernet backbone. In the second step, the AP was assigned a public Internet Protocol (IP) address and connected to the Ethernet backbone in the Susan Mott Webb Nutrition (Webb) Building at the University of Alabama at Birmingham. The IP address was assigned to an AP through either HyperTerminal or a Web console using a Web browser. An administrator ID and password were then created to enhance the Web console security. The client computer required a type II PCMCIA (Personal Computer Memory Card International Association) card slot. Every client needed a functional IP address to become associated with the AP. The Wired Equivalent Privacy (WEP) keys were enabled for both the AP and the clients to ensure two-way authentication. Comparison of different coverage of APs and patch antennas The coverage of the WLAN was found to highly structure dependent. The floor of the Webb building measures about 60 by 25 meters. A single Aironet 340 AP was unable to cover the entire floor with a "good" link status. This was achieved using two (more powerful) Aironet 350 APs. Achieving the "excellent" link status on the floor required the use of the S2406P (Cushcraft) patch antenna. The wireless clients associated with the AP had a "fair" or "good" status one floor up and one floor down from the floor where the AP was located. There was a small area outside the 5-floor building in which the clients could associate with the AP with a "poor" status. Results Link statuses around a simulated patient's home In order to test the feasibility of implementing an isolated LAN around a patient's home, we chose a two-story house with the layout shown in Figure 1 . Two simulated scenarios were tested: one with an ambulance parked on the street next to the house (Figure 1A and 1C ), and the other with the ambulance parked next to the house in its parking lot (Figure 1B and 1D ). Two connection modes under each scenario were also tested: one mode used the AP-client connection (Figure 1A and 1B ) and the other used a peer-to-peer connection without the AP (Figure 1C and 1D ). The link statuses both inside and outside the house were at least "good", and some of the area close to the AP in both scenarios had "excellent" coverage. This suggested that an ambulance with a patch antenna could communicate at a reasonable rate with handheld devices operating inside the patient's home, through either an AP or a direct peer-to-peer connection. File transfer rate with laptops To quantify the baseline signal strength, one Aironet 350 AP was tested in the open area: "excellent", "good", "fair", and "poor" link statuses were obtained within 10, 25, 50, and 100 feet (~4, 10, 20, and 39 meters), respectively. As shown in Table 1A , the file transfer rate for a 10-MB file was 224–2,000 kbps, depending on the distance. Interestingly, higher rates were reached for files of size 10–100 MB. The different methods and directions of file transfer might affect the rate. Methods such as FTP transfer, copying and pasting between folders, and Microsoft Briefcase synchronization were tested. Other factors that may affect transmission were also tested, including the initiation direction (pulling or pushing, in terms of the choice of client and server; see below) and relative physical motion. One Gateway workstation and two Inspiron 4000 laptops were used. To simplify the test, a single medium-sized file (50 MB) was chosen for the purpose. We chose a 50-MB file since it corresponds to a typical high-quality EKG image. As indicated in Table 1B , pushing refers to a transfer from the server to clients if initiated from the server side, and from a client to the server if initiated from the client side; whereas pulling refers to the transfer from the server to clients if initiated from the client side, and from a client to the server if initiated from the server side. Pulling a file led to a higher rate of transfer, in all link statuses (i.e., distances). The speed was lower when two clients transferred the 50-MB file at the same time, and it was generally lower for file transfer between two clients than between a server and a client (Table 1B , Sections III-IV). As expected, the transfer rate was slower while the client device was moving (as shown in Table 2B , section V). However, it was still found that in a WLAN environment, paramedics carrying data-collecting devices could move around and transmit data by different methods and directions at a reasonable rate. The simultaneous file transfer that involves multiple clients/users is a more likely scenario in real emergency settings. Under a "good" link status, the ability of five laptop computers to pull a single file from the FTP server was tested both individually (Table 2A ) and simultaneously (Table 2B ). The data rate reached 5.9 Mbps when a single client was transferring, and fell to below 2 Mbps when multiple clients were involved simultaneously (based on four independent tests). This lower rate, however, is still within a reasonable range, considering the file size. File transfer rate with Pocket PC and cellular phone We chose Ipaq Pocket PCs as PDAs due to their relatively large amount of RAM compared to other PDAs, and a cellular phone (Motorola) as alternative data transmission device when the PDA was outside the WLAN coverage (to simulate the scenario when a long-range antenna, such as a yagi antenna (Cisco), is not available). The transfer rates for single files of different sizes are summarized in Table 3A . The faxing speed through the phone did not appear to be correlated with the file size, since a 50-fold difference in file size resulted in a 30% difference in the time needed to complete the file transmission. The reason is unknown, and should be further investigated. Enhancing the WLAN security using an ACS A Cisco V3 ACS was used to improve the security of the network [ 13 ]. As shown in Figure 2 , the Active Directory facility of a Windows 2000 Server was used as a network access server to communicate with the ACS for authentication, authorization, and accounting (AAA) [ 14 ]. The ACS was able to control client access to the network through the AAA process. A private local area network When managing a WLAN with many clients, there are often insufficient public IP addresses. The solution is to use either the Dynamic Host Configuration Protocol (DHCP) or a private LAN with a Network Address Translation (NAT) server. DHCP services are not easily managed and sometimes create security challenges to network administrators in determining the user's identity. A private LAN seems to be a better solution and more applicable in most ER environments where many wireless clients may transmit at same times, thus maintaining relatively high-speed connections. Figure 3 shows a private LAN with a NAT server that was configured and tested. All wireless clients were centrally managed through an ACS server. The file transfer rates were similar to those achieved with public IP addresses under similar conditions (data not shown). Cost of a small-scale WLAN Health-care organizations have traditionally been slow in accepting WLAN technology in clinical practice. One of the major concerns has been its cost [ 15 , 16 ], followed by security [ 17 - 20 ], although the benefits of WLANs have been demonstrated in many fields including telemedicine [ 21 ]. The cost of a simple WLAN like the one tested here was calculated (Table 4 ). Discussion Synchronization After collecting all patient data at a patient's home, the data must ultimately be transferred to the ER. This involves two critical synchronization steps: (1) from the patient's home to a server on the ambulance, and (2) from the ambulance to the ER (while in transit or upon arrival). Here the WLAN was employed for both of these steps, using Microsoft Briefcase and Windows Workgroups. Automatic synchronization with the destination server and batch synchronization were desired. The ultimate objective is, however, to link the two synchronization steps using a long-range antenna that reaches up to 25 miles (e.g., a yagi antenna from Cisco). This would significantly shorten the time needed to transfer data from a patient's home to the ER, since the data will reach a ER WLAN earlier. We are currently performing the associated experiments. We also tested the use of a cellular phone and other types of PDA (e.g., a Palm Pilot) with network capabilities in transmitting a small (up to 50 MB) but critical file. The results suggested that cellular phones or PDAs with network cards can be effective alternatives to the use of a long-range antenna to transmit data from a patient's home to the ER. Wireless transfer of EKG data EKG data are considered very valuable in the early detection, early intervention, and possibly better outcome of heart attack patients [ 9 ]. The use of a wearable device with sensors to monitor specific physiological signals and communicate with a personal server has been reported [ 22 ]. Land-based telephone lines have also been employed to transmit EKG data and for monitoring by clinical personnel [ 23 ]. In our study, we showed that files up to 5–10 MB (the size of a typical high-quality digitalized EKG image) could be transferred using FTP or other file transfer methods within minutes. Handheld devices such as a Pocket PC and cellular phone may be useful in transmitting EKG files when the ambulance is still at the patient's home, as shown in Table 3 . The time required to transmit a file is proportional to its size in the case of a Pocket-PC-to-laptop transfer, but this was found not to be the case between a cellular phone and a fax machine. The reason for this discrepancy is unknown, and needs to be further investigated. The use of a long-range antenna may ultimately be needed to increase the transmission capacity and speed. Interferences The WLAN operates at 2.42 GHz with an output power of 100 mW, which may pose a risk of interference with medical devices using similar frequencies. Previous studies have shown that a WLAN may interfere with medical devices in close proximity [ 24 ] but is unlikely to be interfered with by such devices [ 25 ]. Further studies are needed to clearly address the possibility. In another study, infrared modems exhibited a similar performance to a wired system even in an electrically noisy environment [ 26 ], indicating that infrared wireless connectivity can be safely and effectively used in operating rooms. These studies suggest that a WLAN can be acceptable for use in prehospital settings if careful interference testing is conducted. Security and privacy The major concern over a WLAN is its security [ 17 - 20 ], especially when personal information is involved. It has been reported that the open-air clear-text transmission of WEP keys and MAC addresses increases network vulnerability [ 13 , 27 - 29 ]. One approach to minimizing the risk is to control the access of remote and/or wireless clients through the Remote Access Dial-in User Service and AP management using the Extensible Authentication Protocol. The regulation by the Healthcare Insurance Portability and Accountability Act may further delay an organization's decision to adopt WLAN technology, although both the Institute of Electrical and Electronic Engineering (IEEE) 802.3 and the OpenAir standard specifications offer security protection (these are the two major standards in the unlicensed commercial 2.4-GHz WLAN market). According to our experiences, the following steps are required to implement a secure WLAN. First, anonymous access should be disabled and the Service Set Identifier of an AP and data encryption key (WEP key) should be enabled. Secondly, a Web console should be used to designate an administrator and manage APs and clients. Thirdly, an ACS server such as Cisco Secure ACS should be implemented to work with Active Directory in order to offer both device- and user-dependent AAA services. Digital certificates should be applied whenever possible for mutual authentication to protect sensitive information through secure server access and secure Web access. In addition, the physical security of the APs, client, and server computers can never be overemphasized. Standards and interoperability The IEEE 802.11 specification addresses both the physical and MAC layers (Orthner, personal communication), and the OpenAir 2.4 interface standard is derived from the Wireless LAN Interoperability Forum [ 29 ] and needs to be interoperable with the IEEE 802.11 standards. The 5-GHz band WLAN standard (IEEE 802.11a) will become more popular once its cost decreases and the required components become more widely available. The use of standardized compliant devices facilitates communication and interoperability. Limitations of the study The present study was mainly based on the Windows operating system and Cisco wireless products. IEEE 802.11a products for the next generation of WLANs are emerging quickly from various vendors. Hence the stability, compatibility, and interoperability with other vendors require further evaluation. Although currently it is relatively expensive to implement a WLAN using this new protocol, the prices and capabilities are expected to improve within the near future. Conclusions Application of WLAN technology will help both paramedics and other health-care professionals in their daily acquisition of information in a localized area such as within a patient's home, an office, a small clinic, or an ER. Implementing a WLAN in a centrally managed and multiple-layer-controlled ACS is the key to ensuring its security and accessibility. Future studies should focus on product capacity, speed, compatibility, interoperability, and security management. Competing interests None declared. Authors' contributions D. Chen, the principal investigator, was most involved in conducting the experiments. H.F. Orthner participated in data collection in the simulated patient's home. H.F. Orthner was the sponsor and S.-J. Soong and G.J. Grimes were the advisors of the fellowship awarded to D. Chen from the National Library of Medicine, National Institute of Health, and they contributed significantly to the design, coordination, and performing of experiments. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517505.xml
340955	New Antibiotics—Resistance Is Futile	It is the certain fate of all antibacterials to be fought off eventually by the pathogens they target. We need new ways to defeat disease, and we will need them forever	By next summer, more than 40% of Streptococcus pneumoniae strains in the United States will resist both penicillin and erythromycin, according to a recent prediction from the Harvard School of Public Health. The forecast, based on mathematical modeling, was published in the spring of 2003. It's too early to tell whether that prediction is precisely on track, according to the senior author on that paper, Marc Lipsitch. But no one doubts that multidrug resistance in this common bug—responsible for diseases that range from sinus trouble and ear infections to meningitis and pneumonia—is speeding up. It is the certain fate of all antibacterials to be fought off eventually by the pathogens they target. The fact that the process is accelerating has been alarming public health officials for some time, especially in the United States. We need new ways to defeat disease, and we will need them forever. Tried and True—and Tired? Antibiotics have traditionally been plucked from nature's battleground. For billions of years, tiny organisms have engaged in an arms race, hurling toxic molecules at each other in the struggle to prosper. Nearly all of today's antibiotics are versions of weapons long wielded by microbes and fungi. Chemical synthesis of entirely human-created antibiotics has so far yielded only fluoroquinolones, a group of broad-spectrum antibiotics that includes Cipro, which became famously scarce during the 2001 anthrax scare, and linezolid (trade-named Zyvox), which is effective against some resistant strains of Staphylococcus , Streptococcus , and Enterococcus . The usual way to find a new antibiotic has been laborious screening of immense libraries of compounds, natural and otherwise. Some argue that screening chemical libraries is approaching a deadend. There may be diminishing returns from screening, but it's not quite dead yet: in October, researchers at the University of Wisconsin at Madison reported a new class of bacterial RNA polymerase inhibitors with antibiotic potential. They were found by screening for molecules that prevent Escherichia coli from transcribing RNA. Christopher T. Walsh of Harvard Medical School says screening's problem may be simply that libraries aren't good enough. Marine organisms have not been studied well, he points out, and 90% of organisms in the biosphere can't be cultured in standard ways. He says, “We're missing 90% of them every time we go and look in nature.” Walsh is doing his bit to create new libraries. He and his colleagues have recently employed combinatorial biosynthesis to learn how to use part of the machinery for assembling cyclic peptide antibiotics to control their architecture. The result was a small library of natural product analogs, some of which have improved antibiotic activity against common bacterial pathogens. “There are dozens of such enzymatic domains that in principle one could clone, express, and test with other substrates. I view that as the kind of thing we should do,” he says. For example, Walsh suggests, it is a reasonable approach to second-generation improvement of daptomycin, the antibiotic most recently approved for sale in the United States. Improving on Nature Walsh collaborates with Chaitan Khosla of Stanford University on finding ways to make existing antibiotics better. They are studying biosynthesis of rifamycin, an antibiotic that is increasingly less effective against its prime target, tuberculosis (TB) (see Figure 1 ). “In the course of learning about that pathway, we've learned a few interesting things lately about how that molecule is initiated, and we're trying to apply it in other contexts, especially in the context of erythromycin biosynthesis,” Khosla says. The idea would be to make a molecule that might be more effective against bacteria that are becoming resistant to rifamycin—and are already naturally resistant to molecules like erythromycin. Figure 1 TB Drug Resistance This 40-year-old Estonian truck driver's TB is resistant to drugs and his right lung was removed three days before this picture was taken. (Photograph by WHO/STB/Colors Magazine/J. Langvad.) “Basically, what we do is to try and figure out new ways to hijack the biosynthesis of antibiotics in nature so as to modify their structures with the goal of improving them,” Khosla explains. He works with an important class of natural antibiotics called polyketides that have generated dozens of drugs, including erythromycin. Polyketides are secondary metabolites (which give their producers a competitive advantage in their environment) produced mostly by bacteria and fungi and made by a complex and structurally diverse family of enzymes called polyketide synthases (see the primer by David Hopwood in this issue of PLoS Biology ). Among them are the anthracyclines, a group of anticancer drugs and antibacterials that includes tetracycline. In this issue of PloS Biology , Khosla and his colleagues report that they can make selective positional modifications in existing anthracycline antibiotics by starting in a different way with a different starting molecule. The molecule came from a natural anthracycline antibiotic, an estrogen receptor antagonist called R1128. R1128 is made via two modules of enzymes that work sequentially; the first module starts the process, and the second completes it. This division of labor permitted the researchers to tack the first R1128 module onto two other enzyme systems, thus engineering completely new anthracyclines. Some were more active in two types of assays than the natural parent molecule. “One setting was an assay on an estrogen-sensitive cancer cell line. Another setting was an assay to probe activity of an enzyme that's of particular interest in Type 2 diabetes, called glucose-6-phosphate translocase.” The work also revealed fundamental mechanistic features of the polyketide synthases, Khosla says. The researchers didn't study the new anthracyclines' effects on bacteria, but Khosla notes that the general principle should apply to other classes of compounds, although the details of how it's implemented will vary from system to system. He says, “The upshot of this paper is that it is now possible to modify a particular methyl group in just about any anthracycline antibiotic.” Finding New Targets Instead of searching for new antibiotics by modifying existing ones, some researchers are trying something completely different—first finding the most vulnerable targets in a bacterium and then designing something that hits one or more of them hard. “You have to understand a helluva lot more about how these little cells work. In fact, we think we understand a lot, but I think we can understand almost everything now that we have all the genomes,” says Lucy Shapiro of Stanford University School of Medicine. While having full genome sequences—more than 100 microbe sequences have been completed—is essential, Shapiro believes that knocking outs genes galore to find out which ones are necessary and going after them all is not a sensible strategy. She observes, “People have been doing that for a while with absolutely no success. That's really going after the problem with a Howitzer instead of with an intelligent approach.” So instead of screening libraries of existing compounds, Shapiro prefers using structural information about drug targets or their natural ligands to create new drugs, an approach known as rational drug design. And instead of looking at all essential genes in a bacterium and choosing one to target, she and her colleagues look at genetic circuitry that controls the cell cycle, the pathway that coordinates cell growth and differentiation. They have identified key control points, or nodes, in the circuitry for their favorite study subject, Caulobacter crescentus . Thus, they have found critical genes encoding proteins that control several critical functions in the cell. Their first candidate was an essential enzyme, a methyltransferase called CcrM, that prevents a particular piece of DNA from being expressed in a cell by tagging it with a methyl group. Antibiotic discovery is all chemistry, Shapiro says, which is why she joined with biochemist Stephen J. Benkovic of Pennsylvania State University. They didn't know the structure of CcrM, Benkovic explains, but the literature about other methyltransferases suggested that the adenine molecule, which is the substrate for CcrM within DNA, binds to a specific region of the enzyme. The researchers designed adenine-like molecules that would bind to CcrM and then developed inhibitors. Benkovic says, “We already knew what kind of structure we wanted, and we simply fine-tuned it.” They worked their way through 1,000 inhibitor candidates, ending up with a small subset—no more than about 20—that not only inhibited CcrM, but also killed Caulobacter very quickly. And not only inoffensive Caulobacter . The compounds knock out other gram-negative bacteria, such as the pathogens Brucella abortus and Francisella tularensis . Some even killed off anthrax, a big surprise because it is gram-positive and so has much thicker cell walls than gram-negative bacteria. The researchers undertook an exhaustive series of experiments to identify which gram-positive bacteria would be affected by which compounds. The list of sensitive pathogens now includes multidrug-resistant Streptococcus , Staphylococcus , and Mycobacterium tuberculosis . More recently, Shapiro reports, they have demonstrated efficacy against rats infected with anthrax or multidrug-resistant Staph , although the compounds save only about 60% of the rats at present. She notes, “So we have a long way to go. But this has proven that if you go after something using some rational approach instead of hit-and-miss, you'll probably have more success than by the other method.” Benkovic points out that theirs is an entirely new class of compounds, small molecular weight compounds that can be made in a few steps. He says, “They don't look like the normal antibiotic, so that's why I think they're fairly unique.” The basic research was done under a grant from the Defense Advanced Research Projects Agency (DARPA), the United States Department of Defense's (DOD) central research and development organization, and once the researchers realized they wanted to develop drugs against three agents that have been considered bioterrorism threats — Brucella , tularensis, and anthrax — they established a separate operation, Anacor Pharmaceuticals, which is developing them with DOD funding and without Shapiro. In her Stanford lab, she continues her fundamental research to define the complete genetic circuitry of Caulobacter , hoping to identify additional nodes in the circuit. She says, “I am not doing it to develop antibiotics; that's what comes out of the work. My goal is to understand how the cell works. I think a lot of studies in pathogenesis should not be just to understand pathogenic organisms, but to understand the complete network of regulatory mechanisms that controls the bacterial cell.” Phage Therapy The most radical approach to new antibiotics may be the resurrection of an old idea: bacteriophage therapy (see Figure 2 ). Late in the 19th century, a researcher noticed that water from some of India's sacred rivers combated cholera. Some years later, the active agents were identified as viruses that infected bacteria. Such viruses are called bacteriophage, or phage for short. There were reports of phage success against dysentery, typhoid, and plague, and bacteriophage therapy had a brief heyday, especially in the 1920s. Results on other diseases were mixed, and with the appearance of antibiotics, phage therapy became unfashionable in the United States, although it has continued in Russia and Eastern Europe. Figure 2 Phage Negative stain electron micrograph of the gamma phage from which the PlyG lytic enzyme was cloned for use to control B. anthracis . (Photograph courtesy of Vincent Fischetti and Raymond Schuch, The Rockefeller University.) Phage were the model organisms of choice for genetics research in the 1930s and 1940s, but became less fashionable as research tools when investigators moved on to eukaryotes. A few held on, like Ry Young of Texas A&M University, who has made phage-induced cell lysis his life's work. “The cell is basically genetically dead as soon as the phage goes in there, but it will keep living as sort of an infected zombie for as long as the phage wants it to, with virus particles accumulating inside the cell,” he explains. “Only when the phage is ready and has decided that it's the right time will it pull the trigger. And the cell blows up.” The freed phage then spew forth to infect new cells. Antibiotic resistance has led to new interest in phage therapy by several small biotech companies. Young continues basic research at Texas A&M, but has also joined one of them, GangaGen, providing bacteriophage expertise to its labs. Phage do kill pathogenic bacteria effectively, and they do it without penetrating human cells, which they can't even recognize. So what is keeping phage therapy out of the clinic? Problems that some doubt can be overcome. Because bacteria develop resistance to phage rapidly, phage therapy companies will need to direct cocktails against a single pathogen, according to Vincent Fischetti at The Rockefeller University. Phage are also antigenic, and the antibodies they stimulate will neutralize their effects during subsequent treatment, he says. But the chief problem appears to be regulatory—regulatory in the political, rather than the genetic, sense. When bacteriophage package their DNA, they occasionally include varying amounts of their hosts' DNA, too. This miscellany, Fischetti points out, is likely to make the Food and Drug Administration unhappy. “Phage normally are very fragile, their tails break, so lot-to-lot homogeneity could be a problem too,” he adds. “So even though it will work, I think they'll have an uphill battle.” Phage may well enter agricultural or veterinary use, he predicts, but are probably not going to be available to patients in the United States any time soon. Fischetti chose a different approach to phage therapy. It does not rely on phage themselves, but on enzymes that phage produce to smash their way out of their host bacteria so they can infect new hosts. He and his colleagues employ these enzymes externally to kill bacteria. He reports, “We now have enzymes that will kill Strep pyogenes , pneumococci, Strep pneumoniae , Bacillus anthracis , Enterococcus faecalis , and group B Strep . The beauty of these enzymes is that they are targeted killing. You only kill the organism you intend to kill, without destroying or affecting the surrounding organisms that are necessary for health.” The enzymes can be loaded into a nasal spray that wipes out pathogens such as Pneumococcus , Staphylococcus , and group A Strep on contact with mucous membranes. The strategy might prevent bacterial infections from spreading in close quarters like hospitals, nursing homes, and daycare centers. Fischetti says, “Clinical trials would tell us how often we had to treat, but more important, we'd have a reagent that could treat people who walk out the door of the hospital to eliminate or reduce the transmission of resistant organisms into the community. We don't have that capability right now.” Fischetti and his colleagues have moved on to using the enzymes systemically to wipe out Bacillus anthracis spores, preventing them from germinating and seething through the bloodstream, producing deadly toxins. An IV drip would be started after exposure to the spores. The method, Fischetti reports, is already successful in mice; clinical trials will determine how long treatment must be continued, perhaps a week or so. They have also eliminated septicemia from pneumocci with the same intravenous method. Up to now the enzymes must make contact with bacteria to kill, but Fischetti is hoping that a new generation of engineered enzymes will be able to kill pathogens inside cells too. A second disadvantage is that they are effective only against gram-positive bacteria, although that group includes many vicious pathogens. But phage enzymes seem to offer one very big advantage: resistance to them has yet to develop. Fischetti says, “We've tried very hard to identify resistant bacteria, but so far we haven't found resistant organisms in all three of the enzymes we're working with. It appears to be a very rare event, much rarer than resistance to antibiotics.” Fischetti cautions against expecting that gladsome state to last forever, but he points out that even if widespread resistance takes the same 40 or 50 years that antibiotics required to become significantly resistant, phage enzymes could buy researchers decades for inventing other approaches. Antibiotics in the 21st Century There is no shortage of ideas for unearthing new antibiotic candidates. Why are they so slow to enter medical practice? The bottleneck, researchers agree, lies in the development process of turning them into effective therapies. Several researchers blame the big pharmaceutical companies that got so big by leading the way to new drugs for battling infectious disease, but in recent years have dropped out. Fischetti complains, “These are the big companies that have the money to develop antiinfectives, but they leave it to small biotech companies, and it's not going to happen as rapidly as it should. I think it's really unconscionable for these big companies to drop the ball because it's not going to be a billion-dollar market for them and that's what they're looking for.” Half a billion at least, says Francis Tally, a big pharmaceuticals veteran who is now chief scientific officer at Cubist Pharmaceuticals, a biotech company located in Lexington, Massachusetts. According to Tally, Cubist produced daptomycin, approved in September 2003, by licensing it from Eli Lilly, which shelved the new compound after concluding its potential market was only $250 million. But, Tally argues, the size of the market is not the only barrier to new antibiotics. Combinatorial chemistry and the genomics revolution have simply not delivered on their early promise. “The pipeline is very dry,” he says. “There's been a real lag at the basic research level.” “Antibiotic discovery is hard,” Shapiro says. “It's a huge long process to get a decent antibiotic.” Walsh agrees. “It's easier to find inhibitors of particular enzymes for particular processes—and a very long road to convert that into something for development.” In the meantime, there is a rising clamor to slow down the rate at which bacteria develop resistance. Doctors are exhorted to cut back on prescribing antibiotics and decline to prescribe for viral diseases, which antibiotics can't combat, even when their patients badger them. But even if antibiotic consumption slowed, we will still need new antibiotics. “I always say it's not a matter of if, it's only a matter of when,” says Walsh. “There will always be a need for new antibiotics because the clock starts ticking on the useful lifetime of any antibiotic once you start to use it. That cannot be argued.”	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC340955.xml
538280	Characterization of the chicken inward rectifier K+ channel IRK1/Kir2.1 gene	Background Inward rectifier potassium channels (IRK) contribute to the normal function of skeletal and cardiac muscle cells. The chick inward rectifier K + channel cIRK1/Kir2.1 is expressed in skeletal muscle, heart, brain, but not in liver; a distribution similar but not identical to that of mouse Kir2.1. We set out to explore regulatory domains of the cIRK1 promoter that enhance or inhibit expression of the gene in different cell types. Results We cloned and characterized the 5'-flanking region of cIRK1 . cIRK1 contains two exons with splice sites in the 5'-untranslated region, a structure similar to mouse and human orthologs. cIRK1 has multiple transcription initiation sites, a feature also seen in mouse. However, while the chicken and mouse promoter regions share many regulatory motifs, cIRK1 possesses a GC-richer promoter and a putative TATA box, which appears to positively regulate gene expression. We report here the identification of several candidate cell/tissue specific cIRK1 regulatory domains by comparing promoter activities in expressing (Qm7) and non-expressing (DF1) cells using in vitro transcription assays. Conclusion While multiple transcription initiation sites and the combinatorial function of several domains in activating cIRK1 expression are similar to those seen in mKir2.1 , the cIRK1 promoter differs by the presence of a putative TATA box. In addition, several domains that regulate the gene's expression differentially in muscle (Qm7) and fibroblast cells (DF1) were identified. These results provide fundamental data to analyze cIRK1 transcriptional mechanisms. The control elements identified here may provide clues to the tissue-specific expression of this K + channel.	Background The inward rectifier potassium channel IRK1/Kir2.1 helps controls cell excitability through setting the resting membrane potential [ 1 ]. Its dominant role of inward rectification for the normal function of skeletal and cardiac muscles is shown by the complete loss of inward rectifying current and K + -induced dilations in arterial myocytes from Kir2.1 knockout mice [ 2 ] and periodic paralysis, and by cardiac arrhythmias in Anderson's syndrome caused by point mutation of human Kir2.1 [ 3 ]. Kir2.1 expression is detected in excitable cells in brain, heart, and skeletal muscle in both mouse and chick [ 4 - 8 ]. In addition, chicken IRK1 (cIRK1) is expressed in the cochlea [ 8 ], a feature not observed in mammals [ 9 , 10 ]. In this report, we first analyzed cIRK1 genomic DNA to identify transcriptional initiation sites and distinct motifs that are important for the expression of this potassium channel gene. Using in vitro promoter assays with fragments of the cloned cIRK1 locus, we also identified several candidate control domains that may participate in regulating the channel's exquisite tissue-specific transcription. Results Structure of the chick IRK1 genomic locus We began this study by isolating cIRK1 genomic clones from a chicken genomic DNA phage library. A series of overlapping clones were isolated by screening the library using full length cIRK1 cDNA as a specific probe. Approximately 6.5 kilobase pairs (kb) of the cIRK1 5'-flanking region were sequenced (Genbank AF375660), including exon 1 and a portion of the intron. cIRK1 contains two exons: exon 1 includes only upstream non-coding sequence (5'-untranslated region, 5'UTR) while exon 2 includes 5'UTR (216 bp), the full open reading frame (1,284 bp), and the 3'UTR (520 bp) (Figure 1A ). The single intron is estimated to be approximately 4.9 kb in length. A comparison of the cDNA and genomic sequences shows the splice site to be located between positions 103 and 104 of cIRK1 cDNA (GenBank U20216). Sequence data also showed that the intron had consensus donor (GU) and acceptor (AG) sequences (Fig. 1B ). This genomic structure of the cIRK1 locus resembles that of mKir2.1 [ 11 ], which possesses two exons separated by a 5.5 kb intron. In a previous report [ 8 ], an approximately 5.5 kb cIRK1 transcript was detected, in addition to one 2.5 kb in length, in brain, cerebellum, heart, skeletal muscle, and cochlea. Since we have identified polyadenylation signals at bp positions 1,645–1,650 and 1,865–1,870 of the cDNA, we conclude that cIRK1 has 2 exons and no additional exon in the 3'UTR. This is also supported by the fact that multiple attempts to extend the cDNA by library screening or 3'RACE were not successful. Figure 1 Genomic structure of chick IRK1 and determination of transcription initiation sites (A) Restriction map of chick IRK1/Kir2.1 genomic locus. The four overlapping genomic clones are shown above the restriction map. Solid boxes indicate exons. E; EcoRI site, X; XbaI site. The probe used for northern blots is underlined. (B) Genomic sequence surrounding splice sites. Exon 1 ends at +103 and exon 2 begins at +104 (based on cIRK1 cDNA, U20216), as indicated. Intronic sequences are in italics; bold letters indicate donor site GT and acceptor site AG. (C) Primer extension analysis. Approximate size, in bases, is indicated on the left. Products of about 80 bases were present in brain, but not in DF1 fibroblasts or in controls (tRNA). (D) Determination of transcription initiation sites using 5'RACE. Shown are the results of the secondary PCR using the 5'-nested primer and a primer designed at 64-41 of cIRK1 cDNA (N-R2), and controls using the 5'-nested primer only (N only) or the reverse primer only (R2 only). Identification of transcription initiation sites Before studying the motifs and elements of the promoter region, we sought to determine the transcription initiation sites of cIRK1 using primer extension and 5'RACE. When the specific reverse complementary sequence to the cIRK1 5'UTR at positions 61–41 was utilized as a primer for reverse transcription, a product of approximately 80 bases was generated from brain mRNA (Fig. 1C ). Specificity of cDNA synthesis was determined by the absence of the band from yeast tRNA and DF1 cells – a chicken embryonic fibroblast cell line which does not express detectable levels of cIRK1 mRNA (Fig. 3 ). To confirm the precise transcriptional initiation sites, oligo-capped RNA based 5'RACE [ 12 ] was carried out using brain total RNA as template. This method results in RNA of which only transcripts with intact 5'-ends are capped by an RNA-oligo. We then reverse-transcribed this RNA using a specific primer R1 designed at positions 146–122 of cIRK1 cDNA. PCR was first performed with the supplied (GeneRacer) 5'-primer and the reverse primer R1. Subsequent nested PCR, using the supplied 5'-nested primer and a second reverse primer R2 designed at positions 64–41 of cIRK1 cDNA, produced multiple fragments. Those of approximately 110 bp were determined to be specific products of the primer set (N-R2) because they were not detected when the PCR was carried out using either the 5'-nested primer alone (N only) or the R2 primer alone (R2 only, Fig. 1D ). The PCR fragments were cloned and multiple clones were sequenced. All of them contained 30 bp of 5'-capped oligo plus several lengths of cIRK1 transcript (80, 81, 82, and 100 bp), which occurred with similar frequencies. Therefore, the actual transcribed 5'-flanking regions were 16, 17, 18, or 36 nucleotides longer than the 5'-end of the clone originally isolated from a chick cochlea cDNA library [ 8 ]. The confirmed initiation sites are shown in bold in Fig. 2 . The 3'-most initiation site, located 16 nucleotides upstream of the 5'-end of the previously reported cIRK1 cDNA, was numbered +1. Our identification of multiple transcription initiation sites indicates that exon 1 of cIRK1 can be 119, 120, 121, or 139 bases in length. Figure 2 Motifs in 5'-flanking region of cIRK1 DNA sequence of the 5'-flanking region (from -417 to +33) is shown. Highlighted are the positions of two E boxes (asterisks), a NF-κB site (heavy underline), a putative TATA box (plus signs), four Sp1 sites (dashed double underlines), and an NF-Y binding site (dotted underline). Transcription initiation sites are shown in bold, with the most downstream site numbered +1. The 5'-ends of the fragments used in the promoter assays are shown with slashes and the construct names (see also Figure 4). Figure 3 IRK1 is expressed in Qm7 cells, but not in DF1 cells. Northern blot analyses were performed using 5 μg of poly A(+) RNA from each cultured cell type. Two bands whose sizes (5.5 kb and 2.5 kb) are similar to those previously reported in chick tissues [8] were detected in Qm7 cells but not in DF1 cells. g3pdh is included as a loading control. Comparison of the 5'-flanking regions of cIRK1 and mKir2.1 We then analyzed the motifs contained in the cIRK1 5'-flanking region, based on the presumption that there might be significant divergence between chicken and mammal IRK1/Kir2.1 promoter regions that could account for their different gene expression patterns. The 5'-flanking region of cIRK1 , shown in Fig. 2 , shares no evident homology with the comparable region of the mKir2.1 promoter [ 11 ]. The corresponding region of the putative human Kir2.1 promoter shows high similarity to the mKir2.1 promoter (62.6% over1,017 bp), however, the hKir2.1 promoter also failed to show significant homology with the cIRK1 5'-flanking region. While both the cIRK1 5'-flanking region and the mKir2.1 promoter have high GC contents, the chick promoter is substantially more GC-rich (71.0% in chick and 62.8% in mouse from position -390 to -1) and contains 38 CpG dinucleotides as compared to 17 CpGs in mouse. The cIRK1 promoter region has four consensus Sp1 binding elements (at positions -216, -194, -182, and -70; Fig. 2 ), while the mKir2.1 promoter contains three Sp1 sites [ 11 ]. Consensus binding sites for several other factors were also identified. E boxes (CANNTG) were present at -316 and at -46, and an inverted CCAAT (NF-Y element) [ 13 ] was found at -98, while the mKir2.1 promoter region contains 3 E boxes and one NF-Y element. A striking difference from mKir2.1 is that the cIRK1 promoter was found to contain a putative TATA box (TATTAA), absent in mKir2.1 , at -56. Overall, while GC-rich domains and motifs are present in both the chick and mouse promoter regions, the extent and number of the motifs are different between the cIRK1 and mKir2.1 promoters. In addition, a TATA box-like sequence is present only in the chick promoter. In vitro promoter analysis and cell-specific regulatory regions We next sought to determine which of these avian-specific upstream elements, including the putative TATA-box, might play a role in cIRK1 transcription. First, we tested whether the 5'-flanking region up to approximately 2 kb contained sufficient elements to activate cIRK1 transcription, and then we investigated the minimal promoter region by constructing a series of 5'-flanking region deletion fragments, including a portion of exon 1, inserted into plasmids upstream of the luciferase reporter (Fig. 4A ). Promoter activities were measured following transient transfection into DF1 or quail myoblast Qm7 cells, which do not and do express endogenous avian IRK1, respectively (Fig. 3 ). We hypothesized that identification of tissue specific regulatory regions of cIRK1 could be achieved by a comparison of reporter gene expression data obtained from these two cell lines. Figure 4B shows that cIRK1 transcription was activated in many of the deletion constructs in both DF1 and Qm7 cells, indicating that both contain sufficient factors to activate cIRK1 expression, whether or not they express the gene endogenously. However, promoter activities were differentially regulated in the two cell types when the relatively long 5'-flanking regions were included in the constructs upstream of the reporter luciferase gene (Fig. 4B ). First, in Qm7 cells, deletion of a domain from position -2142 to -976 resulted in an approximately 60% loss of promoter activity (3B-2256 versus 3B-1089). In addition, deletion of the segment from -975 to -727 resulted in a recovery of activity (3B-840). This indicates that the region between -975 and -727 can function to repress transcription, but that this repression can be overcome by upstream elements. These effects were generally not seen in DF1 cells, except that 3B-840 exhibited somewhat higher activity than did 3B-1229. The region between -1115 and -976 showed a weak inhibitory effect on reporter expression in DF1 cells, while it had a positive regulatory function in Qm7 cells. Finally, a dramatic loss (59%) of promoter activity was observed upon deletion of the domain from -417 to -286 in Qm7 cells (3B-531 versus 3B-399). While this effect was substantial in Qm7 cells, it was absent in DF1 cells. These data indicate that the element(s) in this region activate IRK1 gene expression only in the myoblast cell line Qm7. Figure 4 Identification of domains that affect cIRK1 expression (A) Constructs used in in vitro promoter assays. The putative TATA box, an NF-Y binding site, and an E box are shown in black, hatched and lined boxes, respectively. Numbers in parentheses indicate the length in bp of the promoter region/exon 1 included in each test plasmid. The transcription initiation site is marked with an asterisk. Luc, luciferase reporter gene. (B) In vitro transcription analysis. Plasmid constructs were transiently transfected into DF1 (left) or Qm7 cells (right). Results are shown as fold increases compared with the basal luciferase activity from cells transfected with control pGL-3B vector. Standard errors are shown as bars. ; p value < 0.05, *; p < 0.005. The relative promoter activities exhibited by 3B-399 and shorter constructs were similar in the two cell lines. Activities decreased significantly between 3B-399 and 3B-226, but not between 3B-226 and 3B-206. Candidate active regulatory elements in this region included three Sp1 consensus sites arranged in tandem at -216, -194, and -182. The construct 3B-206, which lacks the NF-Y element activated transcription as well or slightly better than 3B-226. Point mutation of this element in the mKir2.1 minimal promoter resulted in a significant increase of promoter activity, suggesting a significant role for the NF-Y element in mouse compared with chicken [ 11 ]. Specific mutation of the putative TATA box led to a 60.3 % loss of activity in DF1 cells, and an 80.6 % loss in Qm7 cells, but not to a complete loss of promoter activity. The construct 3B-109 showed no significant activity compared with the mock vector, indicating that the 5'-flanking region up to -92 upstream is necessary for promoter activity. Taken together, two distal regions, -1115 to -727 and -417 to -286, were found to be candidate cell/tissue specific regulatory domains in the cIRK1 promoter, while the 5' flanking region proximal to -285 is predicted to activate ubiquitously. The GC rich regions, including Sp1 motifs, and the putative TATA box regulate the promoter positively. Discussion This is the first report describing a promoter analysis of an avian potassium channel. Our data show that cIRK1 and mKir2.1 have similar genomic organizations comprised of a single intron dividing the 5'UTR. In addition, in the 5'-flanking region, the cIRK1 promoter appears to contain a unique putative TATA box. Our results demonstrating positive regulatory activity of the domain that includes the TATTAA sequence strongly suggest that this element does indeed recruit and bind TATA-binding protein [ 14 ], thereby contributing significantly to gene expression. No TATA boxes are found in mammalian K + channel promoters such as those for Kir2.1 [ 9 ], Kir3.1 [ 15 ], Kir3.4 [ 16 ], Kv1.4 [ 17 ], Kv1.5 [ 18 ] and Kv3.1 [ 19 , 20 ]. Further studies may determine whether chick orthologous K + channel genes share their genomic organizations with those of mammals. It should also be emphasized that elements other than the putative TATA box also participate in, although to a lesser extent, transcription initiation of cIRK1 . Just upstream of the transcription initiation site lie several binding sites for transcriptional regulation, including Sp1 sites and an E box; these are also present in mKir2.1 and have been shown to activate that gene's expression [ 11 ]. While the significant role shown for the NF-Y site in mKir2.1 [ 9 ] was not apparent in the case of cIRK1 promoter activation, Sp1 and E box-binding proteins may work cooperatively in promoter activation, as this mechanism has been reported in the study of telomerase reverse transcriptase gene regulation [ 21 ]. Overall, the difference in transcriptional regulation of IRK1/Kir2.1 between chicken and mouse is that regulation of cIRK1 expression depends not only on the Sp1 sites and E box, but also on the putative TATA box. It is known that IRK1/Kir2.1 is expressed in the sensory epithelium of the chick cochlea [ 8 ], but that it is not expressed in the mature cochlea of the mouse [ 9 ] or rat [ 10 ]. Whether the putative TATA box leads to the expression of cIRK1 in the cochlea awaits further analysis. While the results of our in vitro transcription analyses indicate that cIRK1 transcription can be activated both in expressing Qm7 and in non-expressing DF1 cell lines, and therefore likely in a variety of tissues, they also showed the existence of several cell specific regulatory domains. Examination of transcriptional regulation by the region -417 to -286 indicated that element(s) in this region increase promoter activity only in the myoblast cell line Qm7. Candidate elements within this region that might have cell-specific activity are an E box (CAGGTG) at -316 and an activating component of transcription activator NF-κB (GGGRNNYYCC; where R is A or G, Y is C or T, and N is any base) at -296 [ 22 , 23 ]. Various mouse helix-loop helix transcription factors such as atonal homolog Math1 [ 24 ] and MyoD [ 25 ] are reported to associate with E boxes and are required for tissue-specific cell type determination. The chicken Math1 ortholog Cath1 is expressed in the developing brain regions that express cIRK1 [ 26 ]. However, our co-expression of Math1 with the cIRK1 promoter construct 3B-531 did not enhance promoter activity (data not shown). We also identified a region between -1115 and -727 with weak repressor activity that exhibited different properties in the two cell lines. Potential binding motifs present in this region are: a site for interferon regulatory factor-1 at -1030, an AP-4 site at -978, a thyroid hormone beta receptor binding element (TRE) at -962, another E box at -954, a binding site for the potent repressor CTF/NF-I at -859 [ 29 ], and a Drosophila homeotic gene Antennapedia binding motif [ 30 ] at -786. Among these, TRE has been shown to regulate a fast-activating K + conductance in inner hair cells [ 27 ], and an E box has been shown to function as a repressive element when bound by Mad1/Max in differentiated HL60 cells [ 28 ]. Our observation of the transcriptional activation of cIRK1 in DF1 cells which do not express the gene endogenously clearly suggests the existence of additional mechanisms to repress cIRK1 in DF1 cells. Since domains acting to strongly silence cIRK1 expression in DF1 cells were missing in the 5'-flanking region characterized in this study, additional distant control regions may lie further upstream, downstream, or in the intronic sequences. Although at this point clearly speculative, the existence of many CpG dinucleotides in cIRK1 suggests the possibility of gene repression through DNA methylation of CpG dinucleotides leading to heterochromatin condensation of the adjacent genomic locus [ 31 ]. Tissue-specific expression of cIRK1 might consequently be explained by a combination of regulation via the specific motifs studied here, regulation via additional domains outside this region, and the potential modification of the cIRK1 genomic locus through DNA methylation. Conclusions We have identified multiple transcription initiation sites and several candidate regulatory elements in the chicken potassium channel gene cIRK1 . These results provide fundamental data to further analyze cIRK1 transcriptional mechanisms. While the use of multiple transcription initiation sites and the combinatorial participation of multiple domains in activating cIRK1 expression are similar to those seen for mouse Kir2.1 , the cIRK1 promoter is distinct in that it exhibits a higher GC-content than does the mKir2.1 promoter, and by the presence of a functional putative TATA box that is not observed in the mKir2 .1 promoter. Transcriptional control domains identified here form the foundation of an in-depth analysis of tissue-specific expression of this K + channel as well as the species-specific expression of cIRK1 in the chicken cochlea. Methods cIRK1 genomic cloning and upstream sequence analysis Chick IRK1 genomic clones were isolated by screening a chick genomic library in the lambda FIX II vector (Stratagene) using cIRK1 cDNA (GenBank U20216) as a screening probe. Briefly, the probe was random primer-labeled with 32 P-α-dCTP and hybridized with phage plaques blotted onto HybondN nylon membranes (Amersham Biosciences) overnight at 65°C in 6x SSC, 250 μg/ml salmon sperm DNA, 5x Denhardt's solution, and 0.1% SDS. The membranes were washed in 1x SSC and 0.1% SDS, and exposed to x-ray film. Four positive clones were further analyzed following digestion with XbaI and/or EcoRI, and subcloning into pBlueScript-KS. The promoter region (AF375660) was sequenced from both ends. Comparisons between the putative human Kir2.1 promoter region (AC005242), the mouse Kir2.1 promoter region (AF072673), and the coding regions of mKir2.1 cDNA (AF021136) and hKir2.1 cDNA (U24055) were performed using the Wilbur-Lipman DNA alignment method. Transcription factor binding elements were predicted based on the TRANSFAC algorithm [ 32 ] and the transcription element search system . Primer extension, 5'RACE, and northern blot analysis Total RNA was extracted from tissues and cells using TRIZOL reagent (Invitrogen) and treated with DNaseI. Three micrograms of total RNA (brain, DF1, or yeast tRNA) were reverse transcribed with 32 P-end-labeled primer in 10 mM DTT, 50 μg/ml actinomycin D and 0.5 mM dNTP using Superscript II (Invitrogen) at 47°C for 60 min. The primer was designed at 61–41 of cIRK1 cDNA (U20216) [5'-TGT TAA GAT CCG CGG GGA CAC-3']. Reaction mixtures were fractionated on 7% polyacrylamide (PAA)/7 M Urea gels. The 3'-most transcription initiation site was numbered +1. RNA ligase-mediated rapid amplification of 5' cDNA ends (5'RACE) was carried out using the GeneRacer kit (Invitrogen). In brief, 3 μg of total RNA were dephosphorylated, decapped and ligated with a 44-base RNA-oligo, according to manufacturer protocols. Next, the RNA was reverse transcribed using R1 primer designed at 146-122 of cIRK1 cDNA [5'-GCA GAG TTA GCT TAA CAA GTA ACC G-3'] at 42°C for 1 hr. The PCR was performed in a reaction mix containing 1x PCR buffer, 200 nM each of 5'-forward primer [5'-CGA CTG GAG CAC GAG GAC ACT GA-3'] and R1, 100 μM dNTPs, 5 μCi 32 P-α-dCTP, 5% DMSO, and 1.25 U of AmpliTaq (Perkin-Elmer). Reaction conditions were; 94°C for 3 min, 5 cycles at 94°C for 1 min, 57°C for 5 min, 72°C for 2 min, and then 30 cycles at 94°C for 1 min, 57°C for 1 min, 72°C for 1 min, followed by 72°C for 5 min. Nested PCR was performed using the same conditions, with first round PCR products as template, and using the 5'-nested forward primer [5'-GGA CAC TGA CAT GGA CTG AAG GAG TA-3'] and R2 primer designed at 64-41 of cIRK1 cDNA [5'-GGG TGT TAA GAT CCG CGG GGA CAC-3']. Reaction products were eluted from the 7% non-denaturing PAA gel, re-amplified, then ligated into the pCR4-TOPO-TA cloning vector (Invitrogen) and sequenced. Northern blotting was performed using poly A(+) RNA isolated using the FastTrak 2.0 kit (Invitrogen). Five micrograms of poly A(+) RNA were applied to each lane, fractionated on a 0.8% agarose gel, and transferred to a HybondN membrane. The membrane was hybridized with a random-primed 32 P-labeled cIRK1 cDNA fragment (522 bp; 1,384–1,915) overnight at 42°C, washed, and exposed to x-ray film. A quail glyceraldehyde-3-phosphate dehydrogenase (g3pdh) gene fragment (696–968 bp, 97.4% identical to chick) was used as an internal control. Message sizes were estimated by comparisons with RNA molecular weight markers (Invitrogen) run in the adjacent lane. In vitro transcription assay Primers were designed to amplify each promoter region of interest. Constructs 3B-2256 (4201–6456 bp of cIRK1 5'-flanking region, AF375660), 3B-1229 (5228–6456 bp), 3B-531 (5926–6456 bp), and 3B-226 (6231–6456 bp) were generated by PCR, cloned into pBS-KS, and then subcloned into the multiple cloning site (between KpnI and XhoI) of pGL-3B (Promega). Constructs 3B-1089 (5368–6456 bp) and 3B-840 (5617–6456 bp) were obtained from a PCR fragment (1229 bp) with the 5'-region digested by SpeI and PvuII or XbaI, respectively. 3B-399 (6058–6456 bp) was derived from 3B-531 by removing the 5'-region by SacI and SmaI digestion, and 3B-206 was generated by exonuclease III digestion of 3B-226 followed by recircularization. 3B-109 (6348–6456 bp) was generated by SpeI and NheI digestion of 3B-206. Construction of 3B-206 with a mutated putative TATA box (3B-206dT) was accomplished by replacing the putative TATA box (TATTAA) with a StuI site (AGGCCT) by amplifying the original construct using primer sets designed to be complementary to sequences just outside the elements of interest, yet including StuI sites at their 5'-ends. The chick fibroblast cell line DF1 was purchased from ATCC and cultured at 5% CO 2 and 39°C in Dulbecco's minimum essential medium (DMEM) with 1.5 g/L sodium bicarbonate, 10% fetal calf serum, and 100 U/ml penicillin/streptomycin. The quail myoblast cell line Qm7 [ 25 ] was cultured in medium 199 with 10% tryptose phosphate, 10% fetal calf serum and 100 U/ml penicillin/streptomycin in 5% CO 2 at 37°C. The ability of Qm7 cells to differentiate into myotube-like morphology was confirmed periodically as previously described [ 33 ]. One microgram of each construct with 0.1 μg of control vector pRL-TK were mixed in 50 μl of calcium phosphate buffer (140 mM NaCl, 5 mM KCl, 750 μM Na 2 HPO 4 , 6 mM dextrose, 25 mM HEPES at pH = 7.15, and 120 mM of CaCl 2 ), and were incubated with 2 × 10 4 cells in 24-well dishes for 8 hrs in 500 μl of DMEM. The transfection mixes were replaced with fresh growth media and incubation was continued for 48 hrs. In vitro promoter activities were determined using the Dual-Luciferase Reporter Assay System (Promega). Briefly, transfected cells in each well were lysed in 100 μl of 1x Passive Lysis Buffer for 15 min at room temperature. Lysates were centrifuged briefly, and 20 μl of the supernatant was mixed with 100 μl of Luciferase Assay Reagent II, followed by 100 μl of SG reagent. Luciferase activities reflecting cIRK1 promoter activities and control thymidine kinase promoter activities were measured for 10 seconds after premeasurement periods of 2 seconds. All experiments were performed in duplicate and repeated at least 3 times (n= 3–6). The data were analyzed using a t-test assuming equal variance between two samples. Results are shown as fold increases ± standard error when compared with the basal luciferase activities from cells transfected with mock vector pGL-3B. Authors' contributions HM obtained the sequence of cIRK1 genomic DNA and did all the database searches to identify motifs and control regions in the 5'-flanking region of cIRK1 . He also conducted 5'RACE, Northern blot, cell culture, construction of deletion mutant, in vitro transcription analysis, and drafted the manuscript. LCK mapped much of the genomic structure and determined some of the sequence. EL and NK participated in maintenance of the cell lines and in vitro transcription analysis. JCO participated in design and coordination of the study and finalized the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538280.xml
546004	A minor alternative transcript of the fumarylacetoacetate hydrolase gene produces a protein despite being likely subjected to nonsense-mediated mRNA decay	Background Coupling of alternative splicing with nonsense-mediated mRNA decay (NMD) may regulate gene expression. We report here the identification of a nonsense alternative transcript of the fumarylacetoacetate hydrolase ( fah ) gene, which produces a protein despite the fact that it is subject to NMD. Results During the characterization of the effects of the W262X nonsense mutation on FAH mRNA metabolism, two alternative transcripts (del100 and del231) of the fah gene were identified. Del100 lacks exon 8 and as a consequence, the reading frame is shifted and a premature termination codon appears at the 3'end of exon 10. Exons 8 and 9 are skipped in del231, without any disruption of the reading frame. Specific amplification of these transcripts demonstrate that they are produced through minor alternative splicing pathways, and that they are not caused by the W262X mutation per se . As shown with an antiserum raised against the C-terminal part of the putative DEL100 protein, the del100 transcript produces a protein, expressed at different levels in various human tissues. Interestingly, the del100 transcript seems to be subjected to nonsense-mediated mRNA decay, as its level was stabilized following a cycloheximide treatment. Conclusions The del100 and del231 transcripts arise due to minor alternative splicing pathways and del100 is likely subjected to nonsense-mediated mRNA decay. However the remaining amount of transcript seems sufficient to produce a protein in different human tissues. This suggests that NMD has a broader role than simply eliminating aberrant transcripts and when coupled to alternative splicing, may act to modulate gene expression, by allowing the production of low amounts of protein.	Background Cells have evolved surveillance mechanisms to ensure the fidelity of gene expression. One such mechanism, nonsense-mediated mRNA decay (NMD), was discovered about twenty years ago in yeast [ 1 ] and then described in human inherited diseases caused by nonsense or frameshift mutations [ 2 - 4 ], which introduce premature termination codons (PTCs). Contrary to what would be predicted based on the nature of the mutation (a premature translational arrest), the resulting nonsense mRNAs rarely code for truncated protein products and are rather rapidly degraded [ 5 ]. Hence, NMD was first envisaged as a mean to protect cells against the effects of deleterious truncated proteins, with potential dominant-negative effects or a gain of function. Moreover, it seems that NMD has not solely evolved under the pressure of nonsense mRNAs originating from mutations but it also monitors PTC-containing transcripts arising from abnormalities in gene expression [ 6 ]. NMD plays a role in normal cellular development as examplified by the production of functional TcR and Ig genes. During lymphocyte maturation, these genes are subjected to extensive rearrangements and somatic mutation events. Approximately two-thirds of the rearranged genes are not in the proper translational reading frame and the resulting transcripts are down-regulated by NMD, ensuring that only functional TcR and Ig genes are expressed [ 7 , 8 ]. More recently, it was suggested that NMD could play a role in the regulation of gene expression. This was first suspected following the identification of unproductive splicing products of the SRp20 and SRp30b proteins or the ribosomal proteins L3, L7a, L10a and L12 in Caenorhabditis elegans [ 9 , 10 ]. The coupling of alternative splicing, which generates transcripts containing PTCs and NMD, which degrades these transcripts, enables the negative regulation of gene expression [ 9 , 10 ]. This system, termed RUST (for Regulated Unproductive Splicing and Translation, [ 11 ]) is also used in humans [ 12 - 14 ]. It was suggested that alternative splicing leading to NMD might prove to be a common mechanism of autoregulation of many splicing factors [ 12 , 15 , 16 ]. For example, the polypyrimidine tract binding protein (PTB), which generally acts as a splicing repressor, downregulates its own expression by repressing exon 11 inclusion in the mature mRNA [ 12 ]. The resulting alternative PTB mRNA lacking exon 11 contains a PTC and is subjected to NMD [ 12 ]. This negative autoregulation prevents the accumulation of PTB, and therefore the inappropriate processing of its targets [ 12 ]. This coupling of alternative splicing and NMD seems to be a rather common mechanism, as in silico analyses show that 35% of EST-suggested alternative transcripts contain PTCs [ 17 ]. Alternative splicing is thought to occur in 30–60% of human genes [ 18 ], and in addition to expanding proteome diversity, it may play a role in gene expression regulation by generating PTC-containing alternative isoforms. Interestingly, 10 to 30% of nonsense transcripts can escape NMD and are further immune to degradation [ 19 ]. Whether such transcripts code for proteins with a physiological function is unknown. Fumarylacetoacetate hydrolase (FAH, E.C. 3.7.1.2) is the last enzyme of the tyrosine catabolic pathway. A deficiency in FAH causes hereditary tyrosinemia type I (HTI; OMIM 276700), the most severe disease of the pathway [ 20 ]. This inherited metabolic disorder is characterized by severe hepatic and renal dysfunctions often resulting in death in the first years of life if untreated. HTI displays phenotypic heterogeneity with both chronic and acute forms [ 20 , 21 ]. The fah coding gene located on chromosome 15 in the q23-q25 region [ 22 ] spans over 35 kb and contains 14 exons [ 23 ]. Forty-seven mutations have been identified so far in the fah gene, including 7 nonsense mutations [ 24 - 26 ]. While characterizing the effects of the W262X nonsense mutation on FAH mRNA metabolism, we identified two alternative transcripts, del100 and del231 in a HTI patient homozygous for W262X. These transcripts are found in normal cells and thus are not due to the nonsense mutation per se . Interestingly, del100 has skipped exon 8 and as a consequence, the reading frame is shifted, with the appearance of several new PTCs. This transcript is therefore likely subjected to NMD, as suggested by a block of translation by cycloheximide. However, the amount of nonsense transcript which escapes NMD seems to be sufficient to produce a protein of 31-kDa, detected in several human tissues. This report suggests that NMD may allow for the production of low amounts of protein. Results Identification of two alternative transcripts of the fah gene The W262X mutation is a G->A transversion located in exon 9 of the fah gene at nucleotide position 786 and is frequent in the Finnish population [ 27 ]. No protein was detected in the liver or the fibroblasts of homozygous patients [ 28 ]. Consistent with this type of mutation, we demonstrated that W262X mRNAs are degraded by NMD in the cytoplasm [ 29 ]. While studying the decay of W262X nonsense FAH transcripts in lymphoblastoid cell lines, we repeatedly observed two additional RT-PCR products in homozygous W262X/W262X cells (Figure 1A ). Purification and sequence analysis of these two products revealed that they were alternative transcripts of the fah gene. The first one, del100, lacks exon 8 (Figure 1A ) and the second one, del231, which is less abundant, lacks both exons 8 and 9 (Figure 1A ). In del100, the skipping of exon 8 from the mature transcript causes a shift in the reading frame and as a consequence, several new PTCs appear, the first one being located at the 3' end of exon 10, at the new amino acid position 270 (Figure 1A ). The G786A mutation does not code for a stop codon in del100 but causes the replacement of a glycine by a glutamate residue (Figure 1A ). Del231 does not show further disruption of the open reading frame downstream of the deletion and is predicted to code for a shorter FAH-like protein (about 34-kDa) missing the region encoded by exons 8 and 9 (Figure 1A ). Del100 and del231 were first identified in the liver of a patient harboring another mutation (Q279R), which weakens the donor splice site of exon 9 [ 30 ]. Because the W262X mutation is located in the same exon and that in some cases a nonsense codon can affect splicing [ 31 ], we wondered whether these two transcripts were due to an effect of the W262X mutation on a cis-acting splicing element in exon 9. To test this hypothesis, RT-PCRs were performed using primers spanning the exon7-exon9 junction or the exon7-exon10 junction to specifically amplify del100 and del231 respectively. As shown in Figure 1B , both del100 and del231 transcripts were detected by this method, in homozygous mutant cells (W262X/W262X; Figure 1B , middle panel), as well as in normal cells (wt/wt; Figure 1B , middle panel). The identity of these amplification products was verified by sequencing (data not shown). A similar result was found in various human cell lines (Figure 1B , right panel). Indeed, del100 and del231 were amplified in fibroblasts, HeLa cells (Figure 1B , right panel) and in human liver (Figure 1B , right panel), the tissue where FAH is the most expressed [ 32 ]. Moreover, del100 was amplified in two HTI cell lines (Figure 1B ), which harbor either a splice mutation in intron 12 (IVS12/IVS12; [ 33 ]) or two nonsense mutations in exon 13 (E357X/E364X; [ 33 ]). Del231 was not detected in these two cell lines as expected, since these mutations introduce PTCs either following exon 12 skipping (IVS12/IVS12) or due to the two nonsense mutations themselves (E357X and E364X), that likely target the nonsense transcripts to the NMD pathway. Altogether, these data strongly argue in favor of del100 and del231 being minor alternative transcripts of the fah gene, rather than resulting from the presence of the W262X mutation. The del100 transcript is translated into a protein The identification of two minor alternative transcripts of the fah gene raised the question whether they resulted from errors of the splicing apparatus and were unproductive alternative transcripts or whether they could produce protein products with potential physiological roles. There is presently no reported indication for the existence of additional FAH isoforms. The DEL231 open reading frame is identical to FAH, except for the missing region encoded by the skipped exons 8 and 9 and corresponding to amino acids 203 to 280 (Figure 2A ). The open reading frame of the del100 transcript is identical to FAH and del231 from the ATG start codon to amino acid 202. However, the last 67 amino acids of the putative DEL100 protein encoded by exons 9 and 10 are completely different, due to the shift in the reading frame following exon 8 skipping (Figure 2A ). To find out if del100 was translated into a protein, we raised an antiserum against the last 67 amino acids of the putative DEL100 protein and used it to search for the presence of this protein in different adult human tissues. These tissues were obtained after an autopsy and only one sample per tissue was tested, due to the difficulty to obtain them. As shown in Figure 2B (middle panel), a cross-reacting band was present in heart, liver, kidney, spleen, suprarenals and bladder. The DEL100 protein has an apparent molecular weight of 31-kDa, consistent with the value of 29.7-kDa calculated from its sequence. Interestingly, the expression level of the DEL100 protein varied between the different tissues and it differed from that of FAH (Figure 2B , top panel). For example, FAH was barely detected in the spleen, whereas the expression level of the DEL100 protein was the highest in this tissue (Figure 2B ). A monoclonal antibody directed against the N-terminal part of the FAH protein was used to detect both FAH and DEL100 in the tissues where FAH is the less expressed (Figure 2C ). DEL100 was barely detected in the spleen, suggesting that the protein is synthezised in very low amounts. The specificity of the signal was verified by adsorbing the antiserum on the purified C-terminus of DEL100, used for the mouse immunization. As shown in Figure 3A (top panel), the affinity-purified antiserum still recognized the 31-kDa protein, whereas the non-adsorbed fraction did not show any cross-reactivity (Figure 3A , bottom panel). The protein of low molecular weight, which is recognized by the non-adsorbed antiserum in the spleen, is unspecific background (Figure 2B , middle panel and Figure 3A , bottom panel; indicated by a star). A DEL100 protein with the Myc tag was synthesized in an in vitro transcription-translation assay and immunoprecipitated using an anti-Myc antibody. The anti-DEL100 antiserum recognizes the immunoprecipitated DEL100-Myc protein, further demonstrating its specificity (Figure 3B ). Altogether these results suggest that the del100 alternative transcript is translated into a protein of 31-kDa whose expression in different tissues differs from that of FAH. The del100 transcript seems to be subjected to NMD The skipping of exon 8 in del100 causes a change in the reading frame and as a consequence, several new stop codons appear (different from the W262X mutation). To verify if the nonsense del100 transcript was subjected to NMD, lymphoblastoid cells were treated with cycloheximide (Figure 4 ) an inhibitor of translation. Stabilization of nonsense transcripts following such a treatment suggests that they are degraded through the NMD pathway [ 34 ]. The effectiveness of the treatment was previously verified on the full-length W262X containing transcript [ 29 ] and an example is given in Figure 4A (upper panel). The full-length transcript was up-regulated in the homozygous cell line (Figure 4A ; W262X/W262X) but remains unaffected in wild-type cells (Figure 4A ; wt/wt). The same treatment was used to determine the fate of the del100 transcript (Figure 4A and 4B ). A stabilization of this alternative transcript was observed when the FAH transcripts were amplified from exons 6 to 14 in wild-type and homozygous cells (indicated by a star in Figure 4A ). The same result was obtained using a specific amplification of the del100 transcript (Figure 4A , second panel and Figure 4B ). The amount of del100 following cycloheximide treatment increased about 5-fold the level observed in untreated cells (Figure 4B ). These results suggested that del100 is indeed subjected to NMD in homozygous cells and in normal cells as well. In contrast del231, which does not contain any PTC as a result of the skipping of both exons 8 and 9, seemed relatively unaffected by the cycloheximide treatment as expected (Figure 4C ) and does not seem to be subjected to NMD. This result confirmed those obtained in E357X/E364X or IVS12/IVS12 HTI fibroblasts (Figure 1B ), where del231 was undetectable probably because of the introduction of PTCs in these transcripts and their targeting to the NMD pathway. Discussion Del100 and del231 were originally identified while studying the impact of NMD on hereditary tyrosinemia type I. During the characterization of the effects of the W262X mutation on FAH mRNA metabolism [ 29 ], we detected two minor alternative transcripts. If due to the W262X mutation, they should only be produced when the nonsense mutation is present, i.e. in heterozygous (W262X/wt) and homozygous (W262X/W262X) cell lines. However, by using specific primers for each transcript, we found that both del100 and del231 are produced in normal lymphoblastoid cells (Figure 1B ) and in normal human liver. Del100 was also present in two different HTI cell lines harboring different nonsense mutations (E357X/E364X) or a splice mutation (IVS12+5g->a) in exon 13 or intron 12 respectively. These data argue in favor of del100 and del231 resulting from alternative splicing pathways rather than from a W262X-associated altered splicing mechanism. We suggest that both transcripts result from a weak definition of exon 8 (Figure 5 ). Indeed, exon 8 is subjected to many alterations as a result of splice mutations in the region encompassing exons 6 to 9. For example, due to a splice mutation in intron 6 (IVS6-1g->t; [ 30 , 35 ]), a cryptic acceptor site in exon 8 is activated or exon 8 is skipped [ 30 , 35 ]. Del100 and del231 were also identified in the case of the Q279R mutation, a splicing mutation that weakens the donor splice site of exon 9 [ 30 ]. Figure 5 presents with a model that could explain these observations: we suggest that in the major splicing pathway, intron 8 is removed before introns 7 and 9, leading to a splice intermediate that contains the merged exons 8 and 9. Both exons are subsequently defined as a single exon. The order of intron removal is an important determinant of the outcome of splice-site mutations and could explain some unusual alterations, like the skipping of contiguous exons, as strongly suggested by studies of the COL1A1 and COL5A1 splice mutations [ 36 , 37 ]. Altogether, these data suggest that at least del231 may arise through a minor splicing pathway due to an error-prone splicing apparatus because of the weak definition of exon 8 and the order of intron removal. Del100 could originate from a second minor splicing pathway, in which exon 8 is skipped alone because of its weak definition (Figure 5 ). Del100 and del231 are thus the first cases of alternative splicing for the fah gene. However, it remained to see whether they were unproductive splice isoforms or whether they could code for protein isoforms. The del231 transcript retains an unchanged open reading frame when compared to FAH. The putative DEL231 protein would be similar to FAH except for the lower molecular weight (about 35-kDa), due to the missing region encoded by exons 8 and 9. We have been unable to detect a protein species of the size that could correspond to DEL231 using an antibody against full-length FAH. Whether this reflects the absence of such a protein or its presence in a very low amount undetectable with the presently available antibodies remains unknown. The latter explanation seems plausible since the del231 transcript, although not subjected to NMD, is much less abundant that the full-length FAH transcript or del100, as it is barely detected in W262X cells with the RT76 and RT025 primers (Figure 1 ) and the number of PCR cycles needed for its visualization is higher than for del100 when using the specific primers. The structure of the putative DEL100 protein in the N-terminal part is identical to that of FAH. But due to exon 8 skipping, the reading frame is very different in the last 67 amino acids. DEL100, a 31-kDa protein, was detected in different human tissues using an antiserum raised against the specific C-terminal part of the putative protein. The antiserum is specific for the DEL100 protein and does not cross-react with FAH. In addition, the cross-reacting 31-kDa band was lost after adsorbing the antiserum against the purified peptide used for the immunization and the antiserum recognizes an in vitro translated DEL100-Myc protein, demonstrating its specificity. Thus the del100 transcript seems to direct the synthesis of a protein. This result is surprising because this transcript contains PTCs and seems to be subjected to NMD, as shown by a block of translation following a cycloheximide treatment. Interestingly, FAH and DEL100 have converse expression patterns in the human tissues examined. This suggests a post-transcriptional regulation of the expression of the two proteins, since the two transcripts originate from the same pre-mRNA. Alternative splicing, a highly regulated process, which can be developmental-stage or cell-specific, could be responsible for this difference of expression. For example, exon 8 may be more prone to skipping in the spleen given the concentration of specific trans -acting splicing regulators. This could be a way to downregulate the level of FAH transcript, by producing an alternative transcript, which is further eliminated by NMD. Indeed recent in silico analyses and observations on splicing factors have suggested that NMD, when coupled to alternative splicing, could regulate gene expression [ 10 - 12 , 17 ]. Del100 could be another example of such a coupling of alternative splicing and NMD. In such a case, DEL100 would not be expected to play any function in the cell. Very low levels of proteins can sometimes have enormous effects. Interestingly, 10–30% of nonsense transcripts escape NMD and when associated with polysomes are stable [ 19 ]. Is the coupling of alternative splicing with NMD in order to degrade the unproductive isoform the only option? An alternative, as proposed by Neu-Yilik et al . [ 38 ], may be that NMD could function in quantitatively controlling the expression of low amounts of protein. In this view, the PTC-containing del100 transcript may produce a protein with a physiological, although still unknown, function in the cell. The FAH structure contains a C-terminal part of 300 residues, which presents a novel arrangement of β-strands and plays a functional role in Ca 2+ binding, dimerization and catalysis of its substrate, fumarylacetoacetate [ 39 ]. Many of the residues encoded by exon 8 are part of the β-strands and residue 233 serves to bind the Ca 2+ [ 39 ]. The DEL100 protein, which lacks these residues, is thus very unlikely to function in catalyzing the hydrolytic cleavage of carbon-carbon bonds. While the function of DEL100 is unknown at this time, it may have a function in tyrosinemia. Indeed, not all mutations affecting the fah gene will similarly affect the DEL100 protein. For example, mutations affecting exons downstream of exon 10 will affect FAH production but not that of the DEL100 protein. This might be reflected in the phenotypic heterogeneity observed in HTI patients [ 21 ]. Preliminary computer analyses of DEL100 motifs using Proscan at PBIL suggest that it contains a putative DNA-binding motif (RVFLQNLLSvSQARLR with 89% similarity found to the consensus sequence). Whether DEL100 can function as a regulating factor remains unknown. Conclusions NMD was first envisaged as a mechanism to prevent the accumulation of faulty transcripts, arising from mutations or processing abnormalities. Recent in vivo observations and in silico analyses have suggested a new role of NMD in gene expression regulation, when coupled to alternative splicing. We report here the identification of an alternative nonsense transcript of the fah gene, which despite being subjected to NMD, produces a protein in different human tissues. This provides an interesting starting point for the analysis of the role of NMD in the regulated productive splicing and translation. Methods Cell culture The lymphoblastoid cell lines were established from lymphocytes of a HTI patient and his parents as described in Tremblay and Khandjian [ 40 ]. Cells were maintained in RPMI-1640 supplemented with 15% fetal bovine serum. The other human cell lines, HeLa (cervix) and normal fibroblasts, were cultured in DMEM 10%. Fibroblasts of other HTI patients (WG1647, mutation IVS12/IVS12; and WG1922, mutation E357X/E364X [ 33 ]) obtained from the Montreal Children's Hospital (C.R. Scriver) were maintained in DMEM 10% FBS. In translation inhibition experiments, lymphoblastoid cells were treated with 100 μg/ml cycloheximide 3 hours prior to RNA extraction (see below). RT-PCR analysis Total RNA was extracted from 5·10 6 cells with Trizol reagent (Gibco-BRL). RNA from human normal liver was extracted using the RNAqueous kit (Ambion). 1 μg RNA was reverse transcribed using an oligo(dT) and Stratascript (Stratagene). FAH cDNA was amplified from exons 6 to 14 using the following primers: RT76 (5'-CGT GCC TCC TCT GTC GTG-3') and RT025 (5'-GGG AAT TCT GTC ACT GAA TGG CGG AC-3'). Sense primers were designed to specifically amplify del100 and del231. RT84 (5'-TGG AGC TGG AAA TGC ACG-3') spans the exon 7 to exon 9 junction, whereas RT85 (5'-TGG AGC TGG AAA TGG ACC-3') spans the exon 7 to exon 10 junction. Amplification of the alternative transcripts with RT84 or RT85 was performed using HotStart (Qiagen) and PCR conditions were optimized in order to minimize nonspecific hybridization of the primers. Moreover, for each amplification (the FAH transcripts and RAR), the kinetic of the reactions were performed and the number of cycles used for each PCR was in the exponential phase. Analysis of the cycloheximide treatment was done as previously described [ 29 ]. Production of an antiserum against the DEL100 protein An antiserum against the C-terminal part of the DEL100 protein was raised in mouse. The antigen is the C-terminal part (the last 67 amino acids) of the DEL100 protein and is different from FAH or the DEL231 protein. FAH cDNA was amplified from exon 9 to exon 14 using the primers hFAHsstermdel100 (5'-CGG GAT CCC TGC AGC ACG AGA CAT TCA GAA GTG G-3') and RT025 with Expand High Fidelity. The PCR product was inserted into pET30a (Novagen) at the BamHI and EcoRI sites, in order to express the reading frame of the C-terminal part of the DEL100 protein. The His-Tag fusion protein used for immunization was purified by affinity chromatography on a Ni-NTA column (Qiagen) in denaturing conditions with 6 M urea. Preparation of the anti-hFAH monoclonal antibody The anti-hFAH monoclonal antibody was raised against the N-terminus of the protein. The 161 residue peptide was obtained by cutting the pET30a-FAH vector [ 41 ] with the NcoI and XhoI restriction enzymes (New England Biolabs), overhangs were then filled using T4 DNA polymerase (New England Biolabs) and the vector was ligated with the T4 DNA ligase (New England Biolabs). The peptide was expressed in the GJ1158 strain of Escherichia coli as previously described [ 41 ] and purified by affinity chromatography on a Ni-NTA column under denaturing conditions (6 M urea). The purified peptide was injected into BALB/c mice and hybridomas were prepared according to the procedures described in [ 42 ]. Western blot analysis Cells were harvested and lysed in 1 × SDS sample buffer (62.5 mM Tris-Hcl, pH 6.8, 2% SDS, 2.5% 2-β mercaptoethanol, 75 mM DTT, 10% glycerol and 0.005% Bromophenol blue). Human tissues obtained at autopsy and stored at -70°C until used [ 32 ] were homogenized in 10% (w/v) 0.01 M K-phosphate buffer (pH 7.3) and centrifuged for 20 min at 15,000 g. The supernatant was used for immunoblot assay. Samples were electrophoresed on SDS-15% polyacrylamide gels and proteins transferred to a nitrocellulose membrane. Antiserum against the DEL100 protein was used at a dilution of 1:20,000 against purified proteins or 1:1,000 against human tissues. FAH was detected using the polyclonal antibody #488 (1:25,000) as described previously [ 32 ] or using a monoclonal antibody directed against the N-terminal part of the FAH protein (dilution 1/500). Protein loading was verified by using a monoclonal antibody against β-actin (dilution 1/400; Neomarker). Tests of the specificity of the anti-DEL100 antiserum In some experiments, the mouse antiserum against the DEL100 protein was adsorbed on the recombinant His-tag C-terminal protein blotted on nitrocellulose. After an overnight incubation at 4°C, the non-adsorbed fraction was removed and conserved for further characterization. The adsorbed antibody fraction (affinity purified) was eluted using 1 ml of glycine-HCl 0.1 M, pH 2.8 and the pH immediately neutralized by adding 100 μl of 1 M K 2 HPO 4 , pH 8.2. The del100 cDNA was obtained by RT-PCR on total RNA extracted from W262X/W262X cells using ND1 (5' CCC AAG CTT CAG CAT GTC CTT CAT CCC GGT GG 3') and ND2 (5' TGC TCT AGA TTT ATT TGT CAC TGA ATG GCG G 3'). The amplification products were cloned into pDrive (Qiagen) and different clones were sequenced. One clone containing the del100 cDNA was used for further cloning. It was amplified using 5'del100-Eco (5' GGA ATT CCA GCA TGT CCT TCA TCC 3') and ND2. The amplified fragment was digested with EcoRI and XbaI and ligated into EcoRI-XbaI-digested pcDNA3-myc RANGAP, replacing the insert coding for RANGAP. pcDNA3-mycRANGAP was kindly provided by Dr M. J. Matunis (Johns Hopkins University, Baltimore, MD) The construct was used for coupled in vitro transcription-translation using the TNT coupled reticulocyte lysate system (Promega) according to the manufacturer's recommendations. 25 μl of the reaction were used for immunoprecipitation using the anti-Myc antibody as follows: the anti-Myc (1/100) was incubated 1 hour with protein A-sepharose beads (Sigma). The antibody was next immobilized on the beads using 20 mM dimethyl pimedilate (Sigma) in 0.2 M borate sodium (pH 9.0) for 30 min at room temperature. The reaction was stopped by washing the beads twice in 0.2 M ethanolamine and incubation in this solution for 2 hours at room temperature. The antigen (25 μl of the in vitro translated DEL100-Myc protein) was incubated with the beads for two hours at 4°C in a dilution buffer containing 10 mM Tris-HCl pH 8.0, 1 mM EDTA and 10% glycerol. The immunoprecipitated protein was eluted by adding 25 μl of SDS loading buffer (62.5 mM Tris-HCl, pH 6.8; 2% SDS; 2.5% β-mercaptoethanol; 75 mM DTT; 10% glycerol and 0.01% bromophenol blue). Samples were electrophoresed on SDS-15% polyacrylamide gels and proteins transferred to a nitrocellulose membrane. The anti-Myc was used at a dilution of 1/2,000 and the anti-DEL100 antiserum at a dilution of 1/1,000. List of abbreviations CHX, cycloheximide; ESE, exonic splicing enhancer; FAH, fumarylacetoacetate hydrolase; HTI, hereditary tyrosinemia type I; Ig, immunoglobulin; NMD, nonsense-mediated mRNA decay; PCR, polymerase chain reaction; PTB, polypyrimidine tract binding protein; PTC, premature termination codon; RAR, retinoic acid receptor; RT, reverse transcription; RUST, regulated unproductive splicing and translation; TcR, T-cell receptor. Author's contributions ND carried out the experiments and wrote the manuscript. AM participated in the design of the study. JFBL participated in the cloning of Del100 into pcDNA3. AB raised the mAb directed against the N-terminal part of the FAH protein. RMT participated in the design, coordination of the study, and in the writing of the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546004.xml
550666	Role of religion and spirituality in medical patients: Confirmatory results with the SpREUK questionnaire	Background Spirituality has become a subject of interest in health care as it is was recognized to have the potential to prevent, heal or cope with illness. There is less doubt that values and goals are important contributors to life satisfaction, physical and psychological health, and that goals are what gives meaning and purpose to people's lives. However, there is as yet but limited understanding of how patients themselves view the impact of spirituality on their health and well-being, and whether they are convinced that their illness may have "meaning" to them. To raise these questions and to more precisely survey the basic attitudes of patients with severe diseases towards spirituality/religiosity (SpR) and their adjustment to their illness, we developed the SpREUK questionnaire. Methods In order to re-validate our previously described SpREUK instrument, reliability and factor analysis of the new inventory (Version 1.1) were performed according to the standard procedures. The test sample contained 257 German subjects (53.3 ± 13.4 years) with cancer (51%), multiple sclerosis (24%), other chronic diseases (16%) and patients with acute diseases (7%). Results As some items of the SpREUK construct require a positive attitude towards SpR, these items (item pool 2) were separated from the others (item pool 1). The reliability of the 15-item the construct derived from the item pool 1 respectively the 14-item construct which refers to the item pool 2 both had a good quality (Cronbach's alpha = 0.9065 resp. 0.9525). Factor analysis of item pool 1 resulted in a 3-factor solution (i.e. the 6-item sub-scale 1: "Search for meaningful support"; the 6-item sub-scale 2: "Positive interpretation of disease"; and the 3-item sub-scale 3: "Trust in external guidance") which explains 53.8% of variance. Factor analysis of item pool 2 pointed to a 2-factor solution (i.e. the 10-item sub-scale 4: "Support in relations with the External life through SpR" and the 4-item sub-scale 5: "Support of the Internality through SpR") which explains 58.8% of variance. Generally, women had significantly higher SpREUK scores than male patients. Univariate variance analyses revealed significant associations between the sub-scales and SpR attitude and the educational level. Conclusions The current re-evaluation of the SpREUK 1.1 questionnaire indicates that it is a reliable, valid measure of distinct topics of SpR that may be especially useful of assessing the role of SpR in health related research. The instrument appears to be a good choice for assessing a patients interest in spiritual concerns which is not biased for or against a particular religious commitment. Moreover it addresses the topic of "positive reinterpretation of disease" which seems to be of outstanding importance for patients with life-changing diseases.	Background Spirituality has become a subject of interest in health care, and an increasing number of studies, commentaries and reviews examine the connection between religiosity/spirituality and health, its potential to prevent, heal or cope with diseases [ 1 - 10 ]. Moreover, research has confirmed that spiritual well-being is positively associated with quality of life, fighting-spirit, but also fatalism, yet negatively correlated with helplessness/hopelessness, anxious preoccupation, and cognitive avoidance [ 11 ]. Indeed, there is evidence that spirituality is important in coping with illness, as spiritual well-being offers some protection against hopelessness and despair in terminally ill patients [ 12 - 16 ]. However, although religiosity and spirituality were interchangeable words, these constructs may not be identical. It is well established to divide Religiosity into three sub-constructs: Intrinsic, Extrinsic, and Quest Religiosity [ 17 - 20 ], while the construct Spirituality was divided into the following sub-constructs: Cognitive Orientation Towards Spirituality, Experiential/Phenomenological Dimension of Spirituality, Existential Well-Being, Paranormal Beliefs, and Religiousness [ 21 ]. The measurability and operability of spirituality and religiosity remains a problem and thus several questionnaires address this topic. Most of them measure beliefs of specific religious groups, and ask about the relationship with God (i.e. the Spiritual Well-Being Scale [ 22 ], the Daily Spiritual Experience Scale [ 23 ], or the Santa Clara Strength of Religious Faith Questionnaire [ 24 ], while only a few took into account that several patients are offended by institutional religion, but may have an interest in distinct forms of spirituality, respectively in a more personal search for spiritual fulfilment [ 25 , 26 ]. The Functional Assessment of Chronic Illness Therapy – Spiritual Well-Being (FACIT-Sp) scale has a much more open design [ 27 ], but, however, the 12 items of this instrument which made up 2 main factors (labelled "Meaning/Peace" and "Faith") may not really meet the situation of patients with severe and life-threatening diseases. In the post-treatment orientation phase of cancer patients, more existentialistic issues in the patients' attempt to manage the implications of their disease in daily life are of outstanding importance [ 28 , 29 ] The same is true for hospitalised cancer patients [ 30 ]. There is less doubt that values and goals are important contributors to life satisfaction, physical and psychological health, and that goals are what gives meaning and purpose to people's lives [ 31 - 33 ]. Moreover, health can be conceptualized as a competence to gain control for the design of the biography [ 34 ]. But in face of a life-threatening diseases, do patients find meaning and purpose in their life? Many of them rely on religious beliefs to relieve stress, retain a sense of control, maintain hope and their sense of meaning and purpose in life [ 35 ], while others may lose faith in their religious beliefs, and seek for alternatives [ 28 , 29 ]. There is as yet but limited understanding of how patients themselves view the impact of spirituality on their health and well-being, and whether they are convinced that spirituality may offer some beneficial effects. To raise these questions and to more precisely survey the basic attitudes of those patients towards spirituality/religiosity (SpR) and their adjustment to their illness, we developed the SpREUK questionnaire [ 28 , 29 , 36 - 39 ]. We defined the multi-dimensional construct "Spirituality" as an "individual and open approach in the search for meaning and purpose in life, as a search for transcendental truth which may include a sense of connectedness with others, nature, and/or the divine" [ 28 ]. The main sub-scales of our instrument may thus correspond to MacDonald's spirituality constructs of an "Existential Well-Being" [ 21 ] which describes a meaning and purpose for existence, and the perception of self as being competent and able to cope with the difficulties of life and limitations of human existence, and to the construct of an "Cognitive Orientation Towards Spirituality" which is identified by beliefs, attitudes, and perceptions regarding the nature and significance of spirituality, as well as having relevance and importance for personal functioning. In this article we report the re-validation of the SpREUK 1.1 questionnaire (SpREUK is an acronym of the German translation of "Spiritual and Religious Attitudes in Dealing with Illness"), an instrument designed to examine attitudes of patients with life-threatening and chronic diseases towards spirituality/religiosity. Methods Procedure and subjects All individuals were informed of the purpose of the study, were assured of confidentiality, and gave informed consent to participate. The patients were recruited consecutively in the cancer service, the multiple sclerosis service, and two internal medical units of the Communal Hospital in Herdecke (Germany). All subjects completed the questionnaire by themselves. Demographic information is provided in Table 1 . Table 1 Demographic data and SpREUK scores of 257 subjects % Search Meaning (50.6 ± 25.9) Message Disease (70.4 ± 20.9) Trust Guidance (70.0 ± 26.5) Support External (59.0 ± 23.9) Support Internal (62.3 ± 24.3) sex * ** * female 70 54.7 ± 26.0 73.4 ± 20.3 72.9 ± 24.4 61.8 ± 23.1 64.9 ± 22.6 male 30 40.8 ± 22.9 63.4 ± 20.8 63.0 ± 30.0 52.4 ± 24.5 55.6 ± 27.1 age () * * * < 30 years 3 31.8 ± 19.8 60.9 ± 24.6 40.6 ± 20.5 38.0 ± 18.1 43.0 ± 20.7 30–49 years 38 49.7 ± 26.2 70.4 ± 20.2 63.4 ± 26.9 55.3 ± 23.4 58.8 ± 23.9 50–69 years 45 54.2 ± 26.2 72.2 ± 19.6 74.7 ± 24.9 62.5 ± 22.4 65.8 ± 23.2 > 70 years 12 46.2 ± 23.6 67.7 ± 25.7 79.8 ± 23.5 63.4 ± 28.2 64.5 ± 28.0 marital status * * * married 65 47.2 ± 26.0 69.3 ± 21.3 70.1 ± 27.0 56.1 ± 24.6 60.5 ± 26.0 living with partner 11 52.5 ± 20.1 73.3 ± 21.3 57.1 ± 24.7 58.3 ± 16.8 63.2 ± 13.9 divorced 9 65.0 ± 26.9 75.4 ± 19.3 74.1 ± 25.8 69.9 ± 19.7 68.5 ± 24.7 alone 10 55.9 ± 29.6 71.3 ± 20.4 72.3 ± 25.8 61.6 ± 27.2 65.9 ± 23.7 widowed 5 51.4 ± 17.8 68.2 ± 19.6 84.3 ± 17.8 73.5 ± 18.2 62.5 ± 22.9 education 1 level 1 25 36.6 ± 24.9 58.6 ± 20.4 69.0 ± 26.1 55.0 ± 24.1 58.9 ± 24.6 level 2 29 44.4 ± 27.8 67.8 ± 19.5 68.9 ± 29.6 50.6 ± 25.7 53.5 ± 29.2 level 3 37 65.5 ± 24.7 76.9 ± 16.7 75.8 ± 19.8 68.2 ± 19.2 69.1 ± 15.3 other 9 65.5 ± 24.94 74.6 ± 21.6 74.2 ± 21.2 67.9 ± 19.8 75.0 ± 19.0 disease Cancer 51 55.4 ± 24.6 73.8 ± 19.7 74.2 ± 23.4 62.3 ± 22.5 64.3 ± 22.3 Multiple Sclerosis 24 35.8 ± 22.5 59.0 ± 19.1 56.8 ± 28.0 48,0 ± 23.3 52.9 ± 25.8 Chronic diseases 16 56.5 ± 25.7 75.0 ± 19.3 69.8 ± 27.5 63.1 ± 26.1 68.6 ± 26.1 Acute diseases 7 44.7 ± 26.4 74.8 ± 27.1 76.4 ± 30.6 60.6 ± 23.9 67.6 ± 23.9 duration of disease < 0.5 years 19 48.2 ± 26.1 73.3 ± 18.6 68.8 ± 27.7 58.4 ± 21.7 63.3 ± 20.2 0.5–1 years 12 53.1 ± 24.9 79.0 ± 19.7 67.9 ± 28.6 58.1 ± 20.6 58.0 ± 26.0 1–3 years 26 54.1 ± 24.4 72.5 ± 20.2 71.1 ± 23.1 60.6 ± 22.9 63.7 ± 22.9 3–5 years 12 46.8 ± 27.6 61.9 ± 19.4 66.4 ± 28.2 56.7 ± 26.1 56.7 ± 28.3 > 5 years 31 47.3 ± 26.4 70.8 ± 22.1 68.0 ± 28.1 56.7 ± 27.0 61.9 ± 26.0 confession Christian 80 52.9 ± 25.4 71.2 ± 20.8 76.0 ± 21.3 62.2 ± 22.8 63.8 ± 24.3 Others 3 58.3 ± 18.0 70.8 ± 17.1 85.4 ± 15.7 65.7 ± 23.5 83.0 ± 16.8 None 17 38.7 ± 26.9 67.0 ± 22.2 38.2 ± 28.3 41.7 ± 22.8 50.9 ± 22.4 Spiritual attitude R+S+ 32 71.1 ± 20.2 77.9 ± 18.5 85.4 ± 15.0 75.7 ± 14.0 74.6 ± 19.9 R+S- 36 42.1 ± 21.0 68.2 ± 19.8 81.3 ± 16.6 59.1 ± 20.3 61.8 ± 22.2 R-S+ 9 66.8 ± 14.7 76.7 ± 22.0 50.2 ± 15.6 63.8 ± 20.6 67.9 ± 16.7 R-S- 23 29.0 ± 17.6 61.1 ± 21.3 37.8 ± 22.6 33.3 ± 19.0 42.9 ± 23.6 1 Increasing educational level (based on German school system): 1 = secondary education (Hauptschule), 2 = secondary education (junior high; Realschule), 3 = high school education (Gymnasium). Scores are significantly different (** p < 0.01; * p < 0.05; () 0.05 < p < 0.10; Kruskal-Wallis-Test for asymptomatic significance). Deviations of >15% from the mean were highlighted. The sample contained 257 subjects of whom 70% were women. The mean age was 53.3 ± 13.4 years. The majority had a Christian nomination (80%), 17% had no religious orientation, and 3% other nominations. Cancer was diagnosed in 51%, multiple sclerosis in 24%, and other chronic diseases in 16% (i.e. Hepatitis C, liver cirrhosis, inflammatory bowel disease, severe hypertension etc.); 7% of the individuals were patients with acute diseases (i.e. prolapsed intervertebral disc, stomach ulcer, heart arrhythmia etc). Patients in final stages of their disease were not enrolled. Measures The items of the SpREUK 1.0 were developed with the patients' input (cancer service of the Herdecke Community Hospital) and experts' statements (physicians, priest and chaplains working with patients) [ 28 , 36 ], rather than from theoretical concepts. Nevertheless, the original SpREUK 1.0 questionnaire heeded the concept of "internal resp. external locus of control" by Rotter [ 40 ] and Levenson [ 41 ], "passive, active or collaborative religious coping" by Pargament [ 41 ], and the search for "meaning in life" described by Emmons [ 32 , 33 ]. In the final step of the questionnaire design, the items were improved with respect to already existing questionnaires dealing with the topics of religion and spirituality in patients care [ 36 ]. According to a previously conducted reliability and factor analysis [ 36 , 37 ] the SpREUK 1.0 version had the following scales: (1) Search for meaningful support, (2) Guidance, control and message of disease, (A) Support in relations with the external through spirituality/religiosity, and (B) Stabilizing the inner condition through spirituality/religiosity. In order to more precisely differentiate the three topics guidance, control and message of disease in scale 2 of the version 1.0, for the current version of the questionnaire, six new items were added (i.e. F3.5: "My illness is a chance for my own development."; F3.7: "Because of my illness, I reflect on what is essential in my life"). All items were scored on a 5-point scale from disagreement to agreement (0 – does not apply at all; 1 – does not truly apply; 2 – don't know; 3 – applies quite a bit; 4 – applies very much). The SpREUK scores are referred to a 100% level (4 "applied very much" = 100%). Statistical analysis Reliability and factor analysis of the new inventory were performed according to the standard procedures. Next, to combine several items with similar content, we relied on the technique of factor analysis which examines the correlations among a set of variables, and to achieve a set of more general "factors." Factor analyses were repeated rotating different numbers of items in order to arrive at the solution which demonstrated both the best simple structure and the most coherence. Differences in the SpREUK scores were tested using the Kruskal-Wallis-Test for asymptomatic significance. We judged p < 0.05 significant, and 0,05 < p < 0.10 as a trend. To tested the impact of several variables on the SpREUK sub-scales, we performed analysis of univariate variance (ANOVA). As in several cases Levene's test for equality of variances was significant, and we judged p < 0.01 as significant. All statistical analyses were performed with SPSS for Windows 10.0. Results Reliability In order to eliminate items from the item pool that were not contributing to the questionnaire reliability, the reliability of the scale and distinct sub-scales was evaluated with internal consistency coefficients, which reflect the degree to which all items on a particular scale measure a single (unidimensional) concept. Our item pool consisted of the previously established set of items [ 28 , 36 , 37 ] and 6 new items which were added to differentiate the topic "Guidance, control and message of disease" of the version 1.0. As some of the questions require a positive attitude towards SpR, these items (item pool 2) were separated from the others (item pool 1). Reliability analysis revealed that 6 items from the new item pool 1 had a poor corrected item-total correlation and thus were eliminated (however, several from the previous "Locus of Control" topic): F1.2 ("I do not need spiritual advice, I know by myself what should be done"; 0.033), F1.3 ("Spiritual/religious ideas are out-of-date"; 0.091), F2.1 ("I have no influence on my life, it is fixed by fate"; 0,025), F2.2 ("I accept my illness and bear it calmly"; 0.073). F2.3 ("My doctor or therapist helps me to keep my illness at bay"; 0.037), and F3.1 ("Whatever happens, I have trust in my inner strength"; 0.173). One item (F3.6 "The "true being" ("inner core") can not be affected by illness") was omitted because of a weak reliability (0.2997) and – even more important – it points to a distinct "field of meaning" that would need more items in the questionnaire, and thus will be used as marker item until the construct will be revised for this topic. As shown in Table 2 , the 15-item construct derived from the item pool 1 had a good quality (Cronbach's alpha = 0.9065). The 14-item construct which refers to the item pool 2 (which is identical to the old item pool 2 as described in [ 36 ]) had a very good quality (Cronbach's alpha = 0.9525). Table 2 Mean values of the items from SpREUK 1.1 and reliability analysis Factors and Items Mean value (Score 0–4) Standard deviation loading corrected Item-Total correlation Alpha if Item deleted (α = 0.9065) 1: Search for meaningful support 1.5 finding access to a spiritual source can have a positive influence on illness 2.21 1.30 .776 .7597 .8940 1.1 Spiritual attitude 1.96 1.37 .733 .6231 .8994 1.6 searching for an access to SpR 1.81 1.38 .730 .7641 .8935 1.9 urged to spiritual/religious insight 1.99 1.32 .721 .7593 .8940 1.7 others might teach and help to develop spirituality 2.13 1.31 .704 .6881 .8970 1.4 illness has brought renewed interest in SpR questions 1.97 1.40 .636 .6054 .9002 2: Positive interpretation of disease 3.5 illness as a chance for development 2.55 1.30 .813 .7135 .8959 3.4 illness has meaning 2.40 1.33 .703 .6901 .8968 3.2 illness as a hint to change life 2.86 1.05 .604 .6061 .9006 3.7 reflect on what is essential in life because of the illness 3.35 0.77 .570 .2975 .9085 2.4 able to affect the course of illness by themselves 2.58 1.18 .526 .4067 .9066 3.3 illness encourages me to get to know myself better 2.98 1.05 .457 .5026 .9037 3: Trust in external guidance 2.6 Religious attitude 2.71 1.24 .846 .5087 .9033 2.5 trust in a higher power. 2.71 1.24 .810 .5327 .9028 1.8 looking for purpose and meaning in life 3.02 1.21 .295 .4005 .9072 Factors and Items Mean value (Score 0–4) Standard deviation loading corrected Item-Total correlation Alpha if Item deleted (α = 0.9525) 4: Support in relations with the External life through SpR 4.1 plays a major role in life 2.23 1.38 .819 .7934 .9479 4.3 helps to manage life more consciously 2.64 1.18 .814 .9187 .9452 4.2 provides deeper connection with the world around 2.52 1.20 .807 .82112 .9475 4.4 helps to cope better with illness 2.49 1.23 .806 .8718 .9463 4.7 helps to restore mental and physical health 2.30 1.17 .720 .8530 .9465 4.8 practicing with others deepens SpR 1.80 1.34 .663 .6189 .9523 4.6 helps to view disease as a beneficial challenge for own development 2.06 1.24 .657 .8130 .9474 4.9 practicing alone and in silence deepens SpR 2.56 1.21 .594 .6139 .9523 4.10 distinct places stimulate SpR 2.64 1.32 .584 .5709 .9537 4.5 People who share SpR attitudes are important 2.47 1.17 .385 .7956 .9479 5: Support of the Internal life through SpR 5.4 refers to an inner power 2.13 1.27 .765 .4374 .9572 5.1 provides feeling of contentment and inner peace 2.64 1.19 .732 .8671 .9465 5.2 promotes inner strength. 2.46 1.21 .713 .8757 .9462 5.3 refers to a higher (external) power 2.74 1.30 .574 .7470 .9491 Thus, the internal consistency of the 29-item SpREUK 1.1 construct was sufficiently high. The level of difficulty (LoD = 2.482 [mean value] / 4) is 0.6205 for item pool 1 resp. (LoD = 2.406 [mean value] / 4) 0.6014 for item pool 2. With the exception of item F3.7 ("I reflect on what is essential in my life because of the illness"; LoD = 0.838), all values are in the acceptable range from 0.2 to 0.8. Factor analysis To combine several items with similar content, we relied on the technique of factor analysis which examines the correlations among a set of variables, and to achieve a set of more general "factors." Factor analyses were repeated rotating different numbers of items in order to arrive at the solution which demonstrated both the best simple structure and the most coherence. With a Kaiser-Mayer-Olkin value of 0.850 (item pool 1) resp. 0.939 (item pool 2), which measures the degree of common variance, the 15 resp. 14-item-pool seems to be suitable. Barlett's test for non-sphericity was highly significant (p < 0,001). Primary factor analysis of item pool 1 pointed to a 5-factor solution. However, due to a low item number in the tentative subscales 2–5 (with 2 or 3 items each), we favoured the more appropriate 3-factor solution which explains 53.8% of variance (Table 2 ). Sub-scale 1 ("Search for meaningful support") with its 6 items had a Cronbach's alpha of 0.8549, sub-scale 2 ("positive interpretation of disease ") with its 6 items had an alpha of 0.8000, while the 3-item sub-scale 3 ("trust in external guidance") had an alpha of 0.6625. An spiritual attitude loads to sub-scale 1, while a religious orientation loads to sub-scale 3. Factor analysis of item pool 2 (Table 2 ) pointed to a 2-factor solution which explains 58.8% of variance. The 10-item subs-scale 4 ("Support in relations with the External life through SpR") had a Cronbach's alpha of 0.9400, while sub-scale 5 ("Support of the Internality through SpR") with its 4 items had an alpha of 0.7828. Thus, this the internal consistency of the item pool 1 was sufficiently high. However, there are several inter-correlations between the sub-scales (Table 3 ). "Search for meaningful support" correlated negatively with "trust in external guidance" and slightly with the "positive interpretation of disease". Moreover, "trust in external guidance" negatively correlated with "positive interpretation of disease". Table 3 Component Transformation Matrix Scale Search Meaning Message Disease Trust Guidance Support External Support Internal 1 Search Meaning .745 .566 .352 2 Message Disease -.304 .758 -.577 3 Trust Guidance -.594 .323 .737 4 Support External .861 -.509 5 Support Internal -.509 .861 Components 1, 2 and 3 explain 53.8% of variance, while components 4 and 5 explain 58.8% of variance. Analysis of the "side-loadings" of item pool 1 (only values > 0.35 were take into account) reveal that items F1.8 ("looking for purpose and meaning in life") load good on sub-scale 2 (0.414). Analysis of the side-loadings of item pool 2 revealed that several items load also on the other sub-scale. Moreover, sub-scale 4 showed a strong but negative inter-correlation with sub-scale 5 (Table 3 ). Relation between SpREUK scores and demographic variables The highest scores were found for the sub-scales 2 and 3 ("positive interpretation of disease" resp. "trust in external guidance"), the lowest for sub-scale 1 ("Search for meaningful support"). Means and standard deviations for study variables are provided in Table 1 . Women had significantly higher SpREUK scores than male patients. With respect to age, the lowest SpREUK scores were found in the group of < 30 years of age. With increasing age, the trust in a higher supporting presence (sub-scale 3) and the beneficial effects of resp. support through SpR increased. With respect to the marriage status, widowed patients obviously has to rely on external guidance (sub-scale 3) but not the patients living with a partner not married with. Widowed and divorced patients find support in external relations through their SpR engagement (sub-scale 4), while – in contrast to married patients which may find hold in their partnership – especially divorced patients are in search for meaningful support (sub-scale 1). Search for meaningful support and positive interpretation of diseases were depending on the educational level, as patients with lower educational level had significantly lower scores than those with a higher level. A higher educational level was associated with higher scores in the sub-scales 4 and 5 which deals with the beneficial effects of SpR. Illness itself (but not the duration of disease) has a significant impact on the SpREUK scores, as MS patients had the lowest scores in all 5 sub-scales. The SpREUK scores of cancer patients revealed slight differences when compared to patients with other chronic diseases. With the exception of sub-scale 2, patients without confessional affiliations had the lowest scores for all sub-scales, indicating that the "message of disease" was not depending on a denomination. Surprisingly, the few patients with other than a Christian orientation had the highest scores for sub-scales 1, 3, 4, 5. Since nominational affiliation is not necessarily identical with religiosity or spirituality, we asked whether the patients would describe themselves as religious or spiritual [ 28 , 34 , 35 ]. Thirty-two % reported themselves as both religious and spiritual (R+S+) ; 35% as religious, but not spiritual (R+S-) ; 23% as neither religious nor spiritual (R-S-) ; 10% claimed that they were spiritual, but not religious (R-S+) . Thus, the numbers of patients with denominational affiliation and self-reported spiritual/religious attitudes is somewhat similar. A spiritual attitude (R+S+ and R-S+) was associated with "search for meaningful support" and "positive interpretation of disease", while a religious attitude (R+S+ and R+S-) was associated with the highest scores for the "trust in external guidance" sub-scale 3. The living area and the duration of diseases had no significant impact on the SpREUK scores. Correlation with SpR practice As shown in Table 4 , there were moderate to strong correlations between the SpREUK sub-scales and the engagement in a SpR practice as measured by the SpREUK-P manual [ 28 , 29 ]. The new version of SpREUK-P [ 29 ] measures (1) conventional religious practice (praying, church attendance etc.), (2) nature-oriented practice (healing effect on environment etc.), (3) existentialistic practice (self-realization, spiritual development, higher level of consciousness etc.), (4) unconventional spiritual practice (meditation, rituals, body-mind discipline etc.), and (5) humanistic practice (make an effort for other people etc.). The "nature-oriented practice", "humanistic practice" and the "unconventional spiritual practice" were only weakly associated with "Trust in external guidance". In contrast, a "conventíonal religious practice" was strongly correlated with "Search for meaningful support", "Trust in external guidance" and both "Support through SpR" scales 4 and 5. An "unconventional spiritual practice" was associated more with "Search for meaningful support" and " Support in relations with the External life through SpR", while an "esistentialistic practice" was associated stronger with "positive interpretation of disease" and "Support in relations with the External life through SpR. Table 4 Pearson correlation between SpREUK sub-scales and SpR practice 1 Search Meaning Message Disease Trust Guidance Support External Support Internal SpREUK-P engagement scores conventional religious practice .577 * .424 .642 .691 .624 nature-oriented practice .247 .246 .266 * .358 .334 existentialistic practice .437 .530 .411 .479 .439 unconventional spiritual practice .498 .459 ** .223 * .506 .431 humanistic practice .362 .306 .241 .381 .327 Selected SpREUK-P items praying .432 .348 .670 .528 .516 church attendance .290 .145 .473 .425 .324 meditation .452 .378 .137 .421 .374 make an effort for others .160 .167 -.077 .065 -.006 1 engagement in SpR practice was measured with an additional manual of the SpREUK questionnaire, the SpREUK-P manual (Büssing et al ., 2005). Bivariate correlations are statistically significant with ** p < 0.01; * p < 0.05 (2-tailed significance) In detail (Table 4 ), praying and church attendance were strongly correlated with "Trust in external guidance" and both "Support through SpR" scales 4 and 5, while church attendance did not correlate at all with "Message of disease" and only weakly with "Search for meaningful support". In agreement with the results of our study, meditation did not correlate with "Trust in external guidance". However, an attitude of "making and effort for others" did not correlate at all with our SpREUK sub-scales. Analyses of variance Next we tested the impact of several variables on the SpREUK sub-scales, such as sex and marital status, educational level and confession, age and SpR attitude, and disease and duration of disease. Using the method of univariate analyses of variance we identified several sources of variability (Table 5 ): Table 5 Univariate variance analyses Variables F-value significance (1) Search for meaningful support SpR attitude age 46.429 0.784 0.000 n.s. Confession educational level 0.205 4.205 n.s. 0.007 disease duration of disease 3.784 0.991 0.011 n.s. (2) Positive interpretation of disease SpR attitude age 7.710 1.128 0.000 n.s (3) Trust in external guidance SpR attitude age 82.148 1.717 0.000 0.007 Confession educational level Confession * education 9.117 3.630 3.822 0.000 0.015 0.006 (4) Support in relations with the External life through SpR SpR attitude age 51.319 1.120 0.000 n.s Confession educational level 0.511 2.697 n.s 0.049 (5) Support of the Internal life through SpR SpR attitude age 51.319 1.120 0.000 n.s In this table, only significant results were given. Levene's test for equality of variances was significant and thus the level of significance should be p < 0.01. • The SpR attitude is an important covariate for the "Search for meaningful support", "Positive interpretation of disease", "Trust in external guidance", and both "Support through SpR" sub-scales. • The educational level is an important covariate for "Search for meaningful support", "Trust in external guidance" and to a minor content for "Support in relations with the External life through SpR" – but not for the "Positive interpretation of disease". • Age is an important covariate only for "Trust in external guidance". • Confession is an important covariate for "Trust in external guidance". • Disease itself has an impact on the "Search for meaningful support Discussion Data from the current analysis demonstrate the reliability and validity of the SpREUK construct. Moreover, the sub-scales 1 and A (= 4) and B (= 5) of the preliminary version 1.0 were confirmed in the new version 1.1. In order to more precisely differentiate the three topics guidance, control and message of disease from the SpREUK version 1.0, six new items were added. Due to this fact, some items from the original item pool decreased the reliability of the construct and thus, two items from the sub-scale 1 had to be deleted ("I do not need spiritual advice" and "Spiritual/religious ideas are out-of-date", and four items which deal with the "internal/external locus of control" topic as described by Rotter [ 40 ] and Levenson [ 41 ] ("I know by myself what should be done"; "Whatever happens, I have trust in my inner strength"; "I have no influence on my life, it is fixed by fate"; "I accept my illness and bear it calmly"; "My doctor or therapist helps me to keep my illness at bay"). However, the current item pool 1 made up the new sub-scale 2 which highlights the positive interpretation of disease ("message of disease") and the new sub-scale 3 which asks for the trust in an external guidance ("God"). To improve the quality of this 3-item-scale, we have added two additional items. The search for "meaning in life" as described by Emmons [ 32 , 33 ] respectively the concept of "meaning-based coping" are important topics of our questionnaire. However, the item "looking for purpose and meaning in life" loads to the sub-scale 3 which is obviously not identical with the "Search for meaningful support" through spirituality as measured in sub-scale 1. The items of sub-scale 3 fit well to the concept of "external locus of control" and share several topic with Belschner's scale "Transpersonal Trust" [ 43 , 44 ], while the items which made up the new sub-scale 2 (which addresses the "message of disease" and how the patients actively respond to their illness) may fit to the concept of "internal locus of control". This topic of "meaning of disease" is of outstanding importance for cancer patients [ 45 - 49 ], in as much as health can be conceptualized as a competence to gain control for the design of the biography [ 34 ]. In consequence, loss of control due to a life-threatening illness might be interpreted by patients as "punishment", "weakness" or "irreparable loss" – illness has no positive meaning, no "signal" to change aspects of life. As reported by Degner et al . [ 46 ], women who ascribed a negative meaning of illness had significantly higher levels of depression and anxiety and poorer quality of life than women who indicated a more positive meaning. As Spiritual Well-Being can be described as a 2-factor construct, i.e. Religious Well-Being and Existential Well-Being [ 48 ], addressing existentialistic concerns and the possibility to find some kind of sense and meaning even in illness are thus functions of spiritual well-being. In breast cancer patients, Levine and Tarq [ 8 ] found significant correlations of spirituality and spiritual well-being with functional well-being, while items pertaining to meaning and peace tended to correlate significantly with physical well-being. Moreover, the spirituality scales accounted for 40% of the variance in functional well-being, thus confirming the importance of spirituality and spiritual well-being in both physical and functional well-being of cancer patients. Conclusions The SpREUK questionnaire may have important strengths. First, it appears to be a good choice for assessing a patients interest in spiritual concerns which is not biased for or against a particular religious commitment. Moreover, as several patients may be offended by institutional religion, even terms such as God, Jesus, praying, church etc. were avoided. Moreover, the subscale "Search for meaningful support" thus had a good correlation with both, an engagement in conventional religious practice and unconventional spiritual practice. A second strength is that the subscale "positive reinterpretation ("message") of disease" has a good correlation with an existentialistic practice, which seems to be of outstanding importance for patients with life-changing diseases. It may be desirable to use such a measure that allows to assess attitudes which are independent of any religion or specific belief. A third strength is that the validation was performed in a sample with at least two different types of life-changing diseases (cancer and MS, and other chronic diseases) and a healthy control group. Beyond conceptual boundaries, our instrument differentiates the self-addressed "religious" and "spiritual" attitudes of the patients with life-threatening diseases and heeds their search for support and meaning, and integrates the topic of "meaning in illness". We cannot exclude the possibility that these topics are not relevant for healthy individuals. In future studies we have to correlate our scales with other relevant instruments which measure aspects of SpR. Nevertheless, evaluation of the SpREUK questionnaire indicates that it is a reliable, valid measure of distinct topics of SpR that may be especially useful of assessing the role of non-religious spirituality in health related research. The focus of a larger study is to enrol patients from the highly secular Eastern Europe, and to run longitudinal studies with cancer, multiple sclerosis patients, but also cardiac failure and spinal cord damage. The SpREUK with its additional SpREUK-P manual to measure a patient's engagement in distinct forms of SpR practice is currently available in English and German language. Authors' contributions AB conceived the study, designed and developed the questionnaire, performed statistical analysis and drafted the manuscript. TO participated to conceive and design the study, performed additional statistical analysis and helped to draft the manuscript. PFM conceived the study and participated in the design and development of the questionnaire. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC550666.xml
515306	Case-control study on uveal melanoma (RIFA): rational and design	Background Although a rare disease, uveal melanoma is the most common primary intraocular malignancy in adults, with an incidence rate of up to 1.0 per 100,000 persons per year in Europe. Only a few consistent risk factors have been identified for this disease. We present the study design of an ongoing incident case-control study on uveal melanoma (acronym: RIFA study) that focuses on radiofrequency radiation as transmitted by radio sets and wireless telephones, occupational risk factors, phenotypical characteristics, and UV radiation. Methods/Design We conduct a case-control study to identify the role of different exposures in the development of uveal melanoma. The cases of uveal melanoma were identified at the Division of Ophthalmology, University of Essen, a referral centre for tumours of the eye. We recruit three control groups: population controls, controls sampled from those ophthalmologists who referred cases to the Division of Ophthalmology, University of Duisburg-Essen, and sibling controls. For each case the controls are matched on sex and age (five year groups), except for sibling controls. The data are collected from the study participants by short self-administered questionnaire and by telephone interview. During and at the end of the field phase, the data are quality-checked. To estimate the effect of exposures on uveal melanoma risk, we will use conditional logistic regression that accounts for the matching factors and allows to control for potential confounding.	Background Although a rare disease, uveal melanoma of the eye is the most common primary intraocular malignancy in adults, with an incidence rate of up to 1.0 per 100,000 person years (age-standardized, world standard) in Europe [ 1 ]. Only a few consistent risk factors have been identified for this disease. One set of uncommon risk factors include predisposing diseases like the dysplastic nevus syndrome, atypical ocular nevi, ocular and oculardermal melanocytosis [ 2 , 3 ]. Another set of host risk factors are ancestry, light skin and iris pigmentation [ 4 - 6 ]. In addition, a number of environmental factors including UV radiation [ 7 , 8 ] are weakly or inconsistently associated with uveal melanoma. Some uveal melanoma are associated with neurofibromatosis. However, the vast majority of familial cases reported are non-syndromic [ 9 ]. Some recent studies suggests that mutations in the breast cancer susceptibility locus, BRCA2 on chromosome 13, may be involved in the development of uveal melanoma [ 9 , 10 ]. Occupation may be also relevant, and may include chemical work [ 11 , 12 ], arc welding [ 8 , 12 ] and agriculture and farming work [ 13 , 14 ]. Two recent studies found an increased risk of uveal melanoma among cooks [ 8 , 15 ]. Electromagnetic waves with frequencies of 300 kilohertz (kHz) to 300 gigahertz (GHz) are called radio-frequency radiation. Typical occupational sources transmitting radio-frequency radiation in Germany include walkie-talkies in the military and security services, in plants, radio sets on ships, transporters, freight trains, police cars and wireless phones including cellular phones (C-net: 450–465 MHz, since the 1990 ies D-net: 890–960 MHz and E-net: 1710–1800 MHz) and cordless phones (800–1900 MHz) with different modulation types. The population-wide introduction of analog and digital mobile phone techniques in the recent years, which has been coined as the mobile revolution [ 16 ], has resulted in an increasing number of people who fear that radio-frequency radiation (RFR) may have adverse health effects [ 17 ]. There is currently much uncertainty about the role, if any, of radio frequency transmitted by radio sets or mobile phones in human carcinogenesis. The assessment of the potential association of radio-frequency radiation and cancer risk is hampered by uncertainties about effective electromagnetic frequency ranges, the lack of a clear biological mechanism, as well as by difficulties of exposure assessment. Until now, the majority of epidemiological cancer studies focussed on brain cancer because the brain may be exposed to RFR [ 18 - 20 ]. With the exception of one study by Hardell et al. [ 21 ], all brain cancer studies showed no association between RFR as emitted by mobile phones and brain tumour risk until now. In contrast, the pooled analysis of two recent German case-control studies on the aetiology of uveal melanoma showed that frequent use of radiofrequency radiation devices including radio sets and mobile phones at the work place is associated with an about 4.2-fold elevated risk for uveal melanoma [ 22 ]. However, several methodological limitations including a small study size and a crude exposure assessment complicated the interpretation of these findings. Here we present the study design of an ongoing incident case-control study on uveal melanoma (acronym: RIFA study) that focuses on radiofrequency radiation as transmitted by radio sets and wireless telephones. We expect to publish the results of the study in summer 2005. Methods/Design Study questions The RIFA study is planned to answer several etiologic questions with a special focus on electromagnetic radiation especially radio-frequency radiation as emitted by mobile phones and radio sets. First, is the finding of an increased risk of uveal melanoma among subjects with frequent use of RFR devices reproducible? Second, if there is an association, is this association site-specific in terms of laterality of the uveal melanoma and major site of mobile phone use? Third, if there is an association, is there a dose-response relationship between RFR and uveal melanoma risk? Another set of etiologic question relates to pigmentation characteristics including iris colour, hair colour, tendency of the skin to burn and to tan, freckling, number of cutaneous nevi. A further study questions relates to exposure to work and leisure time related ambient ultraviolet radiation and uveal melanoma risk. In addition, our study focuses on several occupational exposures or jobs that are suspected to be associated with an increased risk of uveal melanoma [ 23 , 15 ] including working in the chemical industry, farming, coal mining, welding, cooking, working in the health service sector etc. Finally, we focus on the association between cancer history of the index persons and their relatives (especially breast cancer) and risk of uveal melanoma. Case recruitment The case recruitment is hospital-based and takes place in the Division of Ophthalmology, University of Essen which is a the referral centre for eye cancer in Germany, currently treating about 400–500 eye cancer patients per year. Eligible uveal melanoma cases have to fulfil several criteria. Patients with newly diagnosed first uveal melanoma located in the choroid, iris, and/or ciliary body [ 24 ] during the recruitment period from September 25 th , 2002 to September 24 th , 2004 are eligible, if they are referred to the Division of Ophthalmology, University of Duisburg-Essen during the recruitment period, are aged 20–74 years at diagnosis, are living in Germany, and are capable to complete the interview in German. The majority (about 70–80%) of uveal melanoma treated at the University Hospital Essen receive episcleral plaque therapy without histological verification. For this reason we did not include a reference pathologist who reviews the diagnostic certainty of the cases. Experiences from our previous case-control studies showed that there is a nearly perfect agreement between the local eye doctors in the reference centre and the international reference pathologist (Dr. Ian Cree, London) [ 22 ]. We considered a diagnosis of uveal melanoma as definite if the results of the clinical examination of the eye (ophthalmoscopy) and ultrasound (sometimes supplemented by fluorescence angiography, computertomography or magnetic resonance imaging) were unambiguous. The inclusion and exclusion criteria of the cases are listed in table 1 . Control recruitment Interim analyses showed that the majority of cases comes from the territory of former West Germany. Figure 1 displays the geographic distribution of cases treated for uveal melanoma at the Division of Ophthalmology, University of Duisburg-Essen, from September 24 th , 2002 through March 31 th , 2004. Assuming comparable incidences of uveal melanoma in the federal states of Germany, the crude rate (referred cases divided by population at risk aged 20–74 years) may be considered as an indicator of the referral effect. Obviously, the referral effect varies by federal states. However, it is difficult to judge whether the case referral to the Division of Ophthalmology, University of Duisburg-Essen is a random sample of all newly diagnosed uveal melanoma cases in Germany. We therefore decided to recruit three different control groups. First, if we assume that cases treated in Essen are a random sample of all cases in Germany, a population-based control group would be the most appropriate control group. For this approach, we randomly select controls from mandatory lists of residence that cover the total population of the city or local district. These lists are regarded as the most complete sampling frame for population-based studies in Germany. Second, if the referred cases are not a random sample of all newly diagnosed cases of uveal melanoma in Germany, a control group sampled from those ophthalmologists who referred cases to the Division of Ophthalmology, University of Duisburg-Essen, would be most appropriate [ 25 , 26 ]. To increase the statistical power, we decided to include two controls per case. In addition, we recruit sibling controls of cases (matching ratio 1:1) in order to assess whether genetic factors may confound the effect of exposure. The sibling controls are matched in genetic background. The inclusion criteria of the three control groups are presented in table 2 . Power calculations Based on our former uveal melanoma case-control studies [ 15 , 22 ] we estimated to identify 480 eligible cases within a recruitment period of 24 month. With an anticipated response proportion of about 80%, we expect to interview 380 cases overall. Population-based prevalence estimates of mobile phone use in the general population are scant. A recent telephone survey from 2001 showed that 82% of male and 74% of female participants aged 14–44 years use mobile phones; within the age group 45–59 years, 63% of male and 58% of female participants use mobile phones. The oldest age group (> = 60 years) shows considerable lower prevalences of mobile phone use (men: 49%, women: 25%) [ 27 ]. To determine the statistical power of the case-control study, we assumed several mobile phone prevalence estimates in the control group. We conducted all power calculations two-sided according to formulas of Woodward [ 28 ]. We chose α to be 5% and 1-β to be 90%. Detectable increases of odds ratio estimates by varying prevalences of mobile phone use in the control group are presented in table 3 . A case-control interview ratio of 380 to 760 would enable us to detect increased odds ratios in the range of 1.5 to 2.2 depending on the exposure prevalence in the control group (table 3 ). Exposure assessment Table 4 presents a list of exposures that are assessed in the RIFA study. The questionnaire on mobile phone use is the same instrument which has been used by the international case-control study on brain cancer and mobile phone use sponsored by the International Agency for Research on Cancer, called Interphone study [ 29 ]. In contrast to the Interphone study, we do not perform personal interviews but telephone interviews. For this reason, we cannot show photographs of different types of mobile phones in order to assess the detailed type of mobile phone used. Compared to other studies on the aetiology of uveal melanoma, the RIFA study uses a detailed assessment of pigmentary characteristics. The self-administered questionnaire contains a eye and hair colour card that allows the participants to choose the most appropriate colour of their eyes and hairs at age 20 years. During the telephone interview, the skin reaction to sun exposure (tanning ability, burning tendency) is asked according the concept of Fitzpatrick [ 30 ]. Fitzpatrick's original question contains both items (tanning ability, burning tendency) within a single question which is of methodological concern because the categorical answers given by the Fitzpatrick question may not presents all types of skin reactions as has been demonstrated by Rampen et al. [ 31 ]. To reduce this potential misclassification, we separated the Fitzpatrick items into two questions that separately ask about burning tendency and tanning ability. For the assessment of nevi of the upper arms and the dorsum of the feet with a diameter of at least 3 mm, participants receive a template with a 3 mm hole that enables them to count all nevi of this minimum size. In addition, the CATÍ (computer assisted telephone interview) includes a detailed history of sunburns in the recent 15 years before interview and tendency to freckle as a child. The CATI contains a question on eye colour with the typical categorical answers as has been used by several others (blue, grey, green, hazel, brown, black, [ 22 ] that will enable us to study the agreement between eye colour assessment by colour cards and colour categories. The course of the exposure assessment is displayed in figure 2 . The exposure assessment starts with a self-administered questionnaire among subjects who agreed to participate and gave written informed consent. Subjects are chronologically asked about each job held for at least six months and included questions on the job tasks and industries as has been done in previous studies on the aetiology of uveal melanoma [ 22 ]. In addition, subjects are asked questions related to eye colour, hair colour, ever use of mobile phone, wireless telephone and radio set, and number of nevi. Subjects who have sent back the self-administered questionnaire undergo a computer-assisted telephone interview which takes about 30–40 minutes. For subjects who reported selected work tasks (e.g. cooking and food processing, welding and others), we use 16 job-specific supplementary questionnaires to obtain details of the job tasks and materials used. Quality assurance The quality control program includes several procedures. The study is designed to fulfill the recommendations of the German Good Epidemiologic Practises [ 32 ]. Eight month before the main study started, a manual of standard operating procedures (MOP) was written and a pilot study of four weeks was conducted to test the field work and exposure questionnaires. After some minor revisions of the MOP, report forms, and questionnaire instruments, the principal investigator and participating epidemiologists had to sign that they fully agree with the final version of the MOP. Interviewers of the study were introduced into the field work and were blinded against our study hypotheses. After an initial interviewer training course, interviewers are regularly monitored and receive regular training courses. The recruitment progress, given as number of registered cases and controls, distribution of inclusion and exclusion criteria, response proportions, is monitored monthly. The analysis of nonresponse reasons is supplemented by an short questionnaire for subjects not willing to participate. This questionnaire includes few demographic and exposure items that help us to assess potential selection effects due to nonresponse. A plausibility control of the interview data is done quarterly and is the basis for the regular training courses of the interviewers. The completeness of case registration is checked by regular comparison of the list of registered cases with lists of admissions to the referral centre. In addition we compare our list of cases with data of the hospital information system that includes information on diagnoses. The self-administered short questionnaires are visually edited by the study personnel before the telephone interview starts. The visual editing includes a completeness check and coding of the life-long job history. For each job period, the occupation and branch of industry is coded according to ISCO-68 [ 33 ] and NACE 1993 [ 34 ]. These classifications have been repeatedly used in occupational case-control studies. Self-administered questionnaires with incomplete information or missing data are marked and questions are prepared for the telephone interviewer who is responsible to ask these questions before the main telephone interview starts. The CATI contains internal quality checks that prevent data entry errors. For example, interviewers are not able to fill in the detailed questions on mobile phones, if the entry question on ever having used mobile phone has been answered with no. Planned analyses At the end of the field phase, the data are quality-checked. To estimate the effect of exposures on uveal melanoma risk, we will use conditional logistic regression that accounts for the matching factors and allows to control for potential confounding. We will classify people exposed to an occupational category if they ever worked within this category for at least six months. The quantification of mobile phone use will be based on average number of phone calls and average duration of phone calls per time unit. The association between pigmentary characteristics and uveal melanoma risk will be assessed by detailed matrix containing information on hair colour at age 20 years, eye colour, freckling tendency, and skin colour. Final results of these analyses are scheduled to be published in summer 2005. Competing interests None declared. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC515306.xml
517511	Durable cytotoxic immune responses against gp120 elicited by recombinant SV40 vectors encoding HIV-1 gp120 ± IL-15	Background A vaccine that elicits durable, powerful anti-HIV immunity remains an elusive goal. In these studies we tested whether multiple treatments with viral vector-delivered HIV envelope antigen (gp120), with and without IL-15, could help to approach that goal. For this purpose, we used recombinant Tag -deleted SV40-derived vectors (rSV40s), since they do not elicit neutralizing antibody responses, and so can be given multiply without loss of transduction efficiency. Methods SV(gp120) carried the coding sequences for HIV-1NL4-3 Env, and SV(mIL-15) carried the cDNA for mouse IL-15. Singly, and in combination, these two vectors were given monthly to BALB/cJ mice. Cytotoxic immunity and cytotoxic memory were tested in direct cytotoxicity assays using unselected effector cells. Antibody vs. gp120 was measured in a binding assay. In both cases, targets were P815 cells that were stably transfected with gp120. Results Multiple injections of SV(gp120) elicited powerful anti-gp120 cytolytic activity (>70% specific lysis) by unselected spleen cells. Cells from multiply-immunized mice that were rested 1 year after their last injections still showed >60% gp120-specific lysis. Anti-gp120 antibody was first detected after 2 monthly injections of SV(gp120) and remained elevated thereafter. Adding SV(mIL-15) to the immunization regimen dramatically accelerated the development of memory cytolytic responses, with ≥ 50% specific lysis seen 1 month after two treatments. IL-15 did not alter the development of antibody responses. Conclusions Thus, rSV40s encoding antigens and immunostimulatory cytokines may be useful tools for priming and/or boosting immune responses against HIV.	Introduction The development of an effective vaccine against HIV has been hindered by a variety of problems. The high mutation rate of the virus itself is such that it represents a moving antigenic target during the course of an infection [ 1 - 4 ]. Furthermore, HLA-A and -B expression is directly downregulated by HIV (via intracellular blocking of class I MHC-export to the cell surface by HIV-1 Nef and Vpu), so that efficient antigen presentation is compromised [ 1 , 5 ]. Compared to administration of protein antigen or naked DNA, an infectious vector could be more effective at enhancing antibody and cytotoxic responses against a transgene product. Application of such a strategy, however, has been often complicated by the development of neutralizing immune responses, principally antibodies, against vector coat antigens [ 6 - 10 ]. These neutralizing antibodies arise because the viral vectors enter cells largely through endocytic pathways. Their capsids, like most other particulate antigens, are processed at the time of infection and presented to the immune system. Resulting immune responses neutralize subsequent injections of the vector, and so limit the ability of that vector to be used repeatedly to boost immune responses. This limitation can be circumvented by repeatedly changing the serotype of the antigen-carrying vector, or by using recombinant Tag -deleted SV40-derived gene delivery vectors (rSV40s) for immunization. Several studies have shown, both directly and indirectly, that rSV40 vectors do not elicit detectable neutralizing antibodies [ 11 - 13 ]. Even repeated administration of single [ 11 , 12 ] or different [ 13 ] rSV40 vectors in normal, immunocompetent hosts does not generate antibodies against the vector capsid proteins sufficiently to impair the ability of these vectors to deliver their genes efficiently in vivo . The explanation for this unusual state of affairs may lie in the fact that SV40 enters cells via caveolae and thence travels directly to the nucleus, bypassing cellular antigen processing [ 14 - 16 ]. Thus, only proteins expressed by virus can elicit immune responses. Since, for Tag -deleted rSV40 vectors (unlike wild type SV40), capsid proteins are not expressed, immune responses can only be generated by transgene products [ 11 , 12 ]. Whether for this or for other reasons, rSV40 vectors can be used multiple times to prime and/or boost immune responses against antigens encoded by the transgenes they carry [ 13 , 14 ]. We have previously shown that powerful transgene-specific cytolytic and serum antibody responses can be detected in mice inoculated with rSV40 carrying the cDNA for SIVmac239 envelope glycoprotein gp130 [ 12 ]. Four to five monthly immunizations were adequate to produce >50% specific lysis of envelope-expressing target cells, even with effector:target ratios of 10:1 [ 12 ]. Other investigators have reported that co-administration of vectors carrying immunostimulatory cytokines was useful in augmenting anti-lentiviral immune responses [ 17 - 19 ]. IL-15 has various immunostimulatory and immunomodulatory effects, among which is the ability to upregulate activated T cell proliferation and induce cytotoxic T cell activity [ 20 ]. It also promotes cytotoxic T cell memory [ 21 , 22 ]. Both antibody and cell-mediated immune responses may be useful to protect from HIV infection and progression to AIDS [ 23 - 26 ]. However, there is a particularly good correlation between long-term non-progression to AIDS and strong CTL responses in HIV-positive individuals [ 22 , 27 - 31 ]. Weak CTL responses are generally seen in those who progress rapidly to disease, and in children. Because of the importance of a virus-specific cytotoxic T cell (CTL) response, one of the major aims of any vaccine should be to elicit strong HIV-specific CTL responses [ 32 , 33 ]. We used rSV40s to study the generation and longevity of both humoral and cell-mediated responses in an effort to generate immune responses against the HIV-1 envelope glycoprotein, gp120. We also tested whether co-immunization regimens involving rSV40 delivery of both IL-15 and gp120 augmented and/or accelerated SV40-mediated immune responses further. Methods Cell Lines The murine mastocytoma cell line P815 (ATCC, Bethesda, MD, USA) was used, and maintained in culture with Dulbecco's Modified Eagle's medium (DMEM), supplemented with 10% newborn calf serum (NCS) (Gibco BRL/Life Technologies, Grand Island, NY, USA). COS-7 cells (ATCC, Bethesda, MD, USA), were used to expand stocks of recombinant SV(gp120) and SV(mIL-15) viruses. Cytotoxic lymphocytes were obtained from spleens of immunized mice, and cultured in RPMI-1640 (Gibco BRL/Life Technologies, Grand Island, NY, USA) supplemented with 10% NCS (RPMI-10). Rabbit kidney fibroblasts (RK13 cells) and CV-1 cells (African green monkey kidney cells) were obtained from ATCC (Bethesda, MD, USA). RK13 cells were used to propagate stocks of VCB41, a vaccinia virus vector carrying HIV-1NL4-3 envelope gp120 sequence. Mice BALB/cJ mice aged 6–8 weeks were purchased from Jackson Laboratories, Bar Harbor, ME, USA. They were fed and housed in accordance with American Association for Accreditation of Laboratory Animal Care standards. Use of mice in the laboratory protocols described was approved by the Thomas Jefferson University Institutional Animal Care and Use Committee. Generation of SV(gp120) A 1.6 kb DNA fragment encoding gp120 from HIV-1NL4-3 was made by PCR using primers with engineered restriction sites. This PCR product was cloned into pT7A5 (a plasmid containing an SV40 genome, in which large T antigen gene was replaced by cytomegalovirus (CMV) immediate early promoter and downstream polylinker), giving pT7A5-gp120. To make SV(gp120), the SV(gp120) genome was released from the carrier plasmid by restriction digestion, and used to make recombinant virus in COS-7 cells as described previously [ 34 ]. Virus stocks were purified and titered, as described elsewhere [ 34 ]. SV(HBS), a control virus for these studies, carries hepatitis B surface antigen (HBsAg), and has been reported previously [ 11 ]. Generation of SV(mIL-15) To generate a recombinant SV40 virus with the murine IL-15 transgene (SV(mIL-15)), mIL-15 cDNA was cloned into pSL-4p, which contains a Tag -deleted SV40 genome [[ 35 ], Vera, et al ., in preparation], to yield prSVmIL-15. Virus was made from this plasmid in COS-7 cells as previously reported [ 32 ]. Immunization of mice Mice were given monthly 1 × 10 9 infectious units (IU) of SV(gp120) ± SV(mIL-15) intraperitoneally (IP). In some studies, final administrations included both IP and subcutaneous (SQ) inoculations. SV(HBS) was used as a control antigen-carrying vector. Specific immunization schedules are described in the Results section, below. Stably-transfected P815 target cells for cytotoxicity assays Production of HIV Env-expressing stably transfected targets is similar to the procedure used for generating SIV Env-expressing targets [ 12 ]. Briefly, gp120 cDNA was cloned into pCDNA3. The resulting plasmid, pcgp120, was co-transfected into P815 cells together with the neomycin resistance-carrying plasmid, pSV2Neo. Transfected cells were selected in G418-supplemented DMEM-10, then cloned by limiting-dilution. Viable clones were expanded, assayed for gp120 expression by flow cytometry, and maintained thereafter in G418-supplemented medium. Flow cytometric detection of cell surface gp120 expression Flow cytometry was used to verify gp120 expression on the surface of P815 cells. A recombinant vaccinia virus carrying HIV-1NL4-3 gp120 (VCB41, NIH AIDS Reference Reagent Repository Program (NIH-ARRRP)) was used both as a positive control for gp120 expression and also to generate gp120-expressing target cells in some experiments. Cells that had been stably-transfected with plasmid gp120, or infected with VCB41 both expressed gp120, as assayed by flow cytometry (Coulter-Epic, Kimmel Cancer Center, TJU) (data not shown). The gp120-expressing P815 population was then cloned by limiting dilution. Clonal outgrowths were then reanalyzed by cytofluorimetry (FACS, data not shown) and the single clone expressing the highest levels of gp120, clone 24, was used in subsequent studies as a target for cytotoxicity assays. Anti-gp120 binding-antibody detection using a CELISA An ELISA method was used to assay the activity of anti-gp120 antibodies elicited by immunization of the mice with SV(gp120) ± SV(mIL-15). The strategy for our appraoch to testing for antibodies vs. HIV Env is similar to one we have used to measure binding activity vs. SIV Env [ 12 ]. Briefly, a cell-based assay was developed using VCB41-infected P815 cells (cells were infected with virus for 48 hours prior to being used in assay) as control targets. Sera were taken from mice at 2- and 4-week intervals after immunization(s). Antibody reactivity vs. cell membrane-expressed gp120 was tested by measuring A 405nm of test sera vs. VCB41-infected P815 cells, subtracting A 405nm due to binding to wild type (wt) VV-infected P815 cells, and also subtracting A 405nm of control sera from mice immunized in parallel with a control rSV40, SV(HBS) Measurement of cytotoxic lymphocyte activity by specific lysis of 51 Cr-labeled target cells Wild type (wt) P815 cells, or clone 24 P815 cells expressing gp120 were the target cells for unselected lymphocytes from spleens of mice immunized with SV(gp120) (± SV(mIL-15)), or SV(HBS) as control. Where SV(gp120) was used alone, P815 cells infected with wt vaccinia virus or VCB41 were target cells for spleen cells of immunized mice. Mice were boosted simultaneously with 1 × 10 9 IU intraperitoneally (IP) and 1 × 10 8 IU subcutaneously (hind footpads) usually 4 d before assay, but up to 1 month prior to assay, to test cytolytic lymphocyte memory. Spleen cell concentrations adjusted to 2 × 10 6 / ml with RPMI-10. In some assays of cytotoxic lymphocyte memory, effector cells were harvested 1 month after the final injection. Effector cells from immunized mice were prepared as described, but in addition, were incubated with 5 μg/ml Concanavalin A (Con A, Sigma Chemical Co., St. Louis, MO) overnight, prior to assay. Con A-stimulated cells were then harvested, washed once in RPMI-10, and then used with target cells in the assay. P815 target cells were washed, then labeled with 51 Cr (ICN Biomedicals, Inc., Irvine, CA, USA) (100 μCi per 1 × 10 6 cells) at 37°C, 5% CO 2 for 4 h as described previously [ 12 ]. Afterwards, target cells were washed, then plated in triplicate with effector cells (splenocytes) at effector:target (E:T) ratios of 20:1 and 10:1, and incubated at 37°C, 5% CO 2 for 4 hours. Supernatant 51 Cr was counted (1282 Compugamma CS, LKB) [ 12 ]. Mean specific lysis was calculated as: Mean c.p.m. for gp120-immunized effector cells mixed with gp120-expressing targets, minus the mean c.p.m. control (SV(HBS))-immunized effector cells vs gp120-expressing targets, and expressed as a percentage of the maximal target cell lysis (target cells incubated with 1% Triton-X). Background release of 51 Cr from wild type target cells was subtracted. Thus: % specific 51 Cr release = {[c.p.m. 51 Cr released by SV(gp120)-immune populations from gp120-expressing P815 cells] minus [c.p.m. 51 Cr released by SV(HBS)-immune lymphocytes from gp120-expressing P815 cells]}, divided by [c.p.m. 51 Cr released by Triton X-100 from gp120-expressing P815 cells]. The same calculations were done for lysis of wild type P815 cells by gp120-immune and control-immune effector populations. These numbers were then subtracted from the calculated 51 Cr release above to determine the gp120-specific lysis of target cells by SV(gp120)-immunized effector cells. Western analysis of IL-15 expression in SV(mIL-15)-transduced P815 cells P815 cells were transduced with SV(mIL-15) ×1 at m.o.i. = 100. Culture supernatants were harvested at several times post-transduction, and stored at -80°C. At day 6 post-transduction, a well of cells was harvested and lysed (2% NP40, 50 mM Tris pH7.4,150 mM NaCl, 1 mM EDTA, 10% Glycerol + protease inhibitor cocktail (25× stock Complete™ EDTA-free protease inhibitor cocktail, Roche Diagnostics GmbH, Mannhein, Germany)). Remaining wells were activated non-specifically with 5 mg/ml Con A, and supernatants harvested at various times thereafter. 3 days after con A stimulation, cells were lysed as described above. 50 μg of each culture supernatant or lysate were loaded on a 4–20% Tris-HCl gradient gels (Ready Gel, Bio-Rad, Hercules, CA, USA). 50 ng recombinant human IL-15 was used as a positive control. Samples were electrophoresed, and blotted to PVDF membranes (Immobilon™-P, Millipore Corporation, Bedford, MA). Blots were blocked overnight at 4°C with 5% milk in PBS + Tween-20 (0.05%). Rabbit anti-mouse IL-15 (Abcam, Cambridge, UK) was used as primary antibody, (diluted 1:500 with PBS-Tween), for 2 h at 37°C. Horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG (Jackson ImmunoResearch, West Grove, PA) was used at 1:10,000 dilution in PBS-Tween, for 1 h at room temperature. Signal was detected with chemiluminescence reagent (ECL Plus, Amersham Pharmacia Biotech UK Ltd., Little Chalfont, UK,) Assaying for IFNγ production stimulated by IL-15 COS-1 cells were infected with SV(mIL15) or SVLUC (carrying luciferase) as a negative control. 24 h later, the media were changed and cells incubated 48 h in 500 μl of RPMI 10% serum/well. Fresh mouse spleen cells (5000/well) then cultured 48 h with 100 μl of cell supernatant + 100 μl of RPMI 10% serum. IFNγ ELISA (Pharmingen) was performed on the supernatant from these cultures. Results Stably transfected HIV-1 gp120-expressing P815 cells P815 cells stably transfected to express HIVNL4-3 gp120 were selected and cloned by limiting dilution (see Methods ). We used flow cytometry to identify the clone most strongly positive for cell membrane gp120. Compared to other stably-transfected clones, "clone 24" expressed gp120 at the cell membrane best (data not shown). VCB41-infected and SV(gp120)-transduced P815 cells also expressed substantial cell membrane gp120. Control wtP815 cells, or P815 cells infected with wt VV did not (data not shown). Therefore, clone 24 cells were used to assay gp120-specific immune responses. In both antibody and cytotoxicity assays, two different types of background were subtracted from the responses of gp120-immunized mice: serum binding or cellular reactivity from gp120-immunized animals vs. wt P815 cells and reactivity from control (i.e., SV(HBS))-immunized rSV40-immunized mice vs. clone 24 cells. Thus, data presented below reflect gp120-specific responses against clone 24. Immunization with SV(gp120) Normal BALB/c mice were inoculated with SV(gp120), and their sera were assayed for reactivity vs. gp120 by CELISA. The details of this cell-based ELISA, or CELISA, as described in Methods . Specific binding antibody activity was first statistically significant, compared to prebleed sera 2 weeks after the second inoculation of SV(gp120) (P = 0.000332, using two-tailed Student's t-test) and reached a plateau after the third inoculation (P = 0.000000316 by the same analysis) (Figure 1 ). Additional immunizations beyond the third did not further increase detectable antibody levels (data not shown). Figure 1 Serum antibody against HIV-1NL4-3 envelope glycoprotein gp120 in mice receiving multiple inoculations of SV(gp120) BALB/cJ mice were immunized at monthly intervals with 1 × 10 9 infectious units (IU) SV(gp120), IP. They were bled biweekly. Gp120-specific antibody reactivity was assayed by CELISA, as described in Methods and in reference #12. Specific binding of HIV-1 Env is shown here as specific A405nm, ± S.E.M. Cytolytic responses against gp120: testing for cytotoxic lymphocyte memory An effective anti-lentiviral immunization regimen should generate cytotoxic memory cells. To see if SV(gp120) treatment could do this, mice were immunized once with SV(gp120) IP, then sacrificed 1 month later, without further treatment. In order to lyse target cells, committed cytotoxic cells require activation. However, to avoid antigen-specific selection and specific stimulation only of gp120-reactive cytotoxic cells, splenic lymphocytes were non-specifically stimulated by overnight incubation with Con A. A single immunization with SV(gp120) alone elicited only weak memory lytic responses (≤ 10% specific lysis) against gp120-expressing target cells (Figure 2 ). Figure 2 Specific cytolytic activity against HIV-1NL4-3 gp120 in mice immunized with SV(gp120) and assayed one month after final injection. BALB/cJ mice were immunized twice with SV(gp120) IP at monthly intervals. Splenocytes were harvested one month after final inoculation. Unselected effector cells were added to 51 Cr-labelled target cells and specific lysis of gp120-expressing cells was calculated as described in Methods , ± S.E.M. Results shown here represent ≥ 3 independent determinations per data set. A second group of animals received a second inoculation with SV(gp120) one month after the first, then were assayed the same way one month later for anti-gp120 cytolytic activity. These mice made stronger specific memory responses (15–20% specific lysis) than did animals given only a single inoculation (P ≤ 0.04, by Student's t-test, comparing 2 injections with just one) (Figure 2 ). To test whether SV(gp120) could elicit very long term cytotoxic lymphocyte memory, mice were immunized monthly ×8 with SV(gp120) IP. A final IP inoculation with SV(gp120) was given 1 year after their eighth immunization . They were sacrificed 4 days later, and direct gp120-specific splenocyte cytotoxicity was measured. Unselected spleen cells from all animals made very strong (≥ 50% specific lysis) gp120-specific cytolytic responses (mean specific lysis of 61% ± 4.2). IL-15 expression and secretion in SV(mIL-15)-transduced P815 cells Because higher levels of durable memory cytotoxic responses could be achieved with repeated injections of SV(gp120), and lower levels were seen with 2 injections, we tried to accelerate development of such responses using IL-15, delivered by transduction. To determine if IL-15 could be expressed by transduction, P815 cells were transduced with SV(mIL-15) at m.o.i. = 100. Culture supernatants were harvested 36, 72 and 144 hours later, at which point the cultured cells were activated with Con A. Culture supernatants were collected at 24 and 72 hours post-activation. Supernatants were assayed for IL-15 secretion by Western analysis as described in Methods , using rabbit antibody vs. murine IL-15 (Figure 3 ). The positive control, recombinant human IL-15, has approximately 60% sequence homology with murine IL-15). IL-15 secretion was detectable, but just barely so, in unstimulated culture supernatants. It was abundant by 72 hrs post-stimulation. These data were used in planning co-transduction experiments. Figure 3 Western analysis of IL-15 expression in culture supernatants from SV(mIL-15)-transduced P815 cells P815 cells were transduced with SV(mIL-15) at a m.o.i. = 100. 1.5, 3 and 6 days post-transduction, culture supernatants were collected. Cultures were then stimulated with Con A and additional supernatants harvested 24 and 72 hours later. IL-15 secretion into supernatants was visualized by Western analysis. Recombinant human IL-15 (rhIL-15) was the positive control (non-adjacent lane). Supernatants from unstimulated cultures were tested in parallel (non-adjacent lane). The functionality of the IL-15 produced in this fashion was tested by exploiting the ability of IL-15 to elicit production of IFN-γ by lymphocytes. Thus, CV-1 and COS-7 cells (African green monkey kidney cells) were transduced by SV(mIL-15), then cultured for 72 hrs. Control cultures of the same cells were transduced with SVLUC (carrying luciferase as a transgene). Normal mouse spleen cells were cultured for 42 hrs in 200 μl of the resulting culture supernatants. Production of IFN-γ by the spleen cells was measured by ELISA. Supernatants from COS-7 and CV-1 cells elicited respectively 2056 ± 363 pg/ml and 880 ± 196 pg/ml IFN-γ. Supernatants from SVLUC-transduced cells did not elicit detectable interferon secretion by spleen cells (<20 pg/ml). Effects of IL-15 in cytotoxic lymphocyte responses against gp120 To determine whether coordinate administration of SV(mIL-15) plus SV(gp120) improved cytolytic responses against gp120, mice were given two sets of injections IP, one month apart. Normal BALB/c mice received IP with 10 9 IU of rSV40: one group was given SV(mIL-15) alone, and one group SV(gp120) alone. Three other groups received both SV(mIL-15) and SV(gp120): either SV(mIL-15) followed 3 days later by SV(gp120), or SV(gp120) first, followed by SV(mIL-15). The final group was given both SV(mIL-15) and SV(gp120) simultaneously. The 3 day separation between the two vectors was used because of the strength of the signal for secreted IL-15 by Western blotting at 72 hours post-stimulation (see above). One month later, unselected spleen cells were assayed as described above for cytolytic activity against gp120-expressing clone 24 cells. Adding IL-15 to the immunization regimen greatly increased gp120-specific cytolytic responses (Figure 4 ). Also, the timing of cytokine administration relative to SV(gp120) inoculation significantly affected the responses seen. Mice given SV(mIL-15) 3 d after SV(gp120) did not make detectable gp120-specific cytotoxic responses. Simultaneous inoculation with SV(mIL-15) and SV(gp120), however, increased specific cytolysis to ≥ 20%, which was significant at E:T = 20:1 (P ≤ 0.05, using Student's t-test) compared to SV(gp120) alone. The most dramatic results were observed when SV(mIL-15) was administered 3 d before SV(gp120). Those mice demonstrated highly significant augmentation by SV(IL-15) of gp120-specific lysis, which was ≥ 60% at both 20:1 and 10:1 E:T ratios (P ≤ 0.02 using Students' t-test, compared with immunization of SV(gp120) alone). Mice injected ×2 with SV(mIL-15) alone made no significant gp120-specific cytolytic responses at either 20:1 or 10:1 effector:target ratios. Mice given only SV(gp120) demonstrated ≈ 10% specific lysis at E:T = 20:1. Figure 4 Specific cell-mediated responses against gp120-expressing target cells by splenic effectors from co-immunized mice Mice were given two monthly injections with either SV(mIL-15), SV(gp120) or both cytokine and antigen sequentially or simultaneously IP. One month after the final inoculation(s), unselected spleen cells were assayed for specific cytolytic activity against gp120-expressing clone 24 cells labeled with 51 Cr, as described in Methods . Results shown here represent ≥ 3 independent determinations per data set. Effect of IL-15 co-administration on anti-gp120 antibody responses Mice receiving SV(mIL-15) and/or SV(gp120) (or the control vector, SV(HBS)) according to the schedules outlined above were tested to determine the effect, if any, of such co-administration in anti-gp120 serum antibody responses. CELISA and calculation of gp120-specific antibody binding were performed as described in Methods . Slight binding antibody activity was detected 2 weeks after the first inoculation(s), in all SV(gp120) recipient groups. Levels of SV(gp120)-induced groups made detectable antibody responses were not appreciably affected by coadministration of SV(mIL-15) (data not shown). Discussion In this study, we used rSV40 vectors to elicit HIV-1NL4-3 gp120-specific cytotoxic lymphocyte and antibody responses. We have observed that these vectors may be administered repeatedly to boost those responses. Further studies also suggested that such responses are durable in vivo . Our results here demonstrate several important strengths of using rSV40 vectors to immunize against lentiviral antigens: Among these are the ability of the vector to be administered multiple times without eliciting neutralizing responses [ 11 - 13 ], and the magnitude of the cytotoxic responses to the vector-encoded lentiviral target antigen. When SV(gp120) was given alone, i.e. without added SV(mIL-15), levels of specific cytotoxicity increased with additional SV(gp120) injections: After 2 injections, ≈ 20% specific lysis was seen, which increased to >70% specific lysis after 7 injections. The potency of rSV40 immunization to elicit cytotoxic immune responses is underscored by the fact that these responses were measured in direct 51 Cr-release assays: unselected lymphoid organ populations were added directly to labeled target cells at low E:T ratios, and specific 51 Cr release was measured. Analysis to confirm CD8 expression, or expression of other CTL markers was not performed on the effector cells. However, it is unlikely that these data reflect the cytotoxic activity of NK cells. NK cytolytic activity is non-specific and does not increase with repeated immunization. The patterns of 51 Cr release observed in the current studies were extensively controlled to ascertain the antigen-specificity of the cytolysis observed: background lysis of wild type P815 cells was subtracted, as was lysis by lymphocytes from mice immunized with an irrelevant rSV40 vector. We also found that cytolysis increased with increasing numbers of immunizations, which is not a characteristic of NK cell-mediated lysis. Since a key goal for a vaccine against HIV is to generate immune responses that are durable in vivo , we tested whether cytotoxic lymphocyte activity elicited by SV(gp120) immunization, was detectable one month after inoculation. Thus, cytolytic responses, assayed one month after a second injection, were ≈ 20%, which is comparable to those of splenic cytotoxic cells assayed four days following a third inoculation (data not shown). Further, mice given multiple injections of SV(gp120), then rested for one year, gave ≈ 70% specific lysis when challenged with SV(gp120). Therefore, SV(gp120) administration may thus favor development of cytotoxic lymphocyte memory. In an attempt to accelerate and to improve upon these specific cytotoxic and particularly cytotoxic memory responses, we co-immunized mice with SV(gp120) and a rSV40 carrying mouse IL-15. IL-15 promotes cytotoxic lymphocyte responses, and in particular, cytotoxic memory responses [ 23 , 24 ]. The biological effects of IL-15 are less well understood than are those of some of the other immunostimulatory cytokines that have been applied to these types of immunization protocols, such as IL-2, IL-12 and IFN-γ. IL-15 is not a T cell-derived product, but rather appears to be produced by a variety of cells, such as epithelial cells, stromal cells and muscle. It acts on activated T cells, sometimes similarly to IL-2, but it has activities distinct from those of IL-2. IL-15 may play a role in T cell activation in the CNS. It also promotes cytotoxic responses, cytotoxic T cell memory, and natural killer (NK) cell maturation [ 33 , 34 ]. Accordingly, our analysis of the contribution by IL-15 to cytotoxic responses, focused mainly on the ability of SV(mIL-15) to augment specific cytotoxic responses of spleen cells from animals rested 1 month following immunization. Because quiescent cytotoxic T cells are not strong effectors, we non-specifically activated the splenocytes prior to assay with Con A. Non-specific activation was used to avoid specifically enriching effector cell population for gp120-specific cells in vitro . Furthermore, low effector:target ratios (20:1 and 10:1) were used in these assays. Our immunization protocols tested both simultaneous and staggered administration of rSV40s carrying HIV-1NL4-3 gp120 and murine IL-15. IL-15 co-immunization dramatically accelerated cytotoxic responses, depending on the immunization regimen used: Animals given SV(mIL-15) alone made no gp120-specific cytolytic responses. Mice receiving 2 treatments with a mixture of SV(gp120) and SV(mIL-15) gave much higher specific lysis, depending on the coadministration regimen, as compared to those receiving SV(gp120) alone (≈ 10% specific lysis). Thus, among mice given staggered injections of SV(mIL-15) and SV(gp120), the order of cytokine administration greatly affected the response: if SV(gp120)was given first, no detectable gp120-specific cytolysis was observed. However, if the cytokine was given first, followed 3 days later by SV(gp120), ≥ 60% specific lysis was seen at both 20:1 and 10:1 effector:target ratios. Why the order of cytokine administration should affect antigen-specific responses so dramatically is not yet clear. Cytokine given after, or together with antigen, may have insufficient time to augment cytotoxic responses. In addition, Western analysis of IL-15 production showed that IL-15 secretion was not detectable in supernatants beyond 36 hours, but could be stimulated subsequently. Thus, a specific, possibly brief, window for IL-15 expression and secretion may need to be attained, in order for its effects on gp120-specific responses to be detectable. We observed very high levels of specific lysis by these unselected effector populations following just two tandem injections of SV(mIL-15) followed by SV(gp120). The strong anti-lentiviral cytolytic responses we report were observed in a strain of mouse, BALB/cJ, that generally mounts relatively weak type 1 T cell responses. The finding of >60% cytolysis with two administrations of SV(gp120) + SV(mIL-15), suggests that a strategy similar to that described herein may be helpful in individuals who would generate relatively low cytolytic responses. Serum antibody levels assayed by CELISA where SV(gp120) was administered alone, multiple times, were detectable after two immunizations, and continued to increase up to week 4 following the third immunization. These responses were not further enhanced by subsequent boosting immunizations. While specific antibody responses against gp120 were detected in all experimental groups following SV(gp120) and SV(mIL-15) co-immunization, IL-15 co-administration did not augment anti-gp120 antibody levels, compared to gp120 alone. This was to be expected, since IL-15 reportedly acts primarily on T cell and NK cell functions, rather than on humoral immune responses. Our data argue in favor of using IL-15 as an adjuvant for antigen-specific immune responses, particularly cytotoxic lymphocyte responses. We also demonstrate that a single transgene, administered multiple times (>3), may be very effective at eliciting both humoral and cell-mediated responses. These results thus both corroborate and extend our previous observations [ 13 , 14 , 32 ], and suggest that combining rSV40s encoding antigens and immunostimulatory cytokines sequentially in multi-administration regimens may provide high levels of long-lasting immunity against the target antigen. Conclusions Recombinant SV40-derived gene-delivery vectors, being transparent to the immune system, can be given multiple times to prime and boost immune responses against the delivered antigens. Anti-vector immunity does not overwhelm responses against the target antigens. As well, these vectors elicit very high levels of antibody, and especially cell-meditated immunity. Finally, combining the delivery of rSV40s bearing antigens with those bearing cytokines such as IL-15 can enhance levels of immunity, particularly long-term immunity. Clearly, much work remains. However, this approach offers promise as a strategy to immunize against pathogens for which classical approaches have not been adequately effective. List of Non-Standard Abbreviations Used Table 1 Abbreviation Meaning CELISA cell-based ELISA E:T effector cell:target cell ratio gp120 major HIV envelope glycoprotein, 120 kDa HBS hepatitis B surface antigen IFNγ interferon-gamma mIL-15 mouse interleukin-15 NIH-ARRRP National Institutes of Health, AIDS Research Reference Reagent Program pNPP p-nitrophenyl phosphate VCB41 strain of vaccinia virus carrying gp120 coding sequences wt wild type Competing Interests None declared. Authors' Contributions HJM devised all the assay systems for cell- and antibody-mediated immunity against lentiviral antigens, performed all the immunization studies and assays. HM also wrote this manuscript. PYT generated the SV(gp120) construct. MV and PF generated the SV(mIL-15) and SVLUC constructs described here and performed the ELISA for IFNγ stimulated by SV(mIL-15). DSS is the Principal Investigator for this work, oversaw and planned the experimental strategies, worked with HJM in interpreting the experimental data and writing the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517511.xml
518971	Screening for Parkinson's disease with response time batteries: A pilot study	Background Although significant response time deficits (both reaction time and movement time) have been identified in numerous studies of patients with Parkinson's disease (PD), few attempts have been made to evaluate the use of these measures in screening for PD. Methods Receiver operator characteristic curves were used to identify cutoff scores for a unit-weighted composite of two choice response tasks in a sample of 40 patients and 40 healthy participants. These scores were then cross-validated in an independent sample of 20 patients and 20 healthy participants. Results The unit-weighted movement time composite demonstrated high sensitivity (90%) and specificity (90%) in the identification of PD. Movement time was also significantly correlated (r = 0.59, p < 0.025) with the motor score of the Unified Parkinson's Disease Rating Scale (UPDRS). Conclusions Measures of chronometric speed, assessed without the use of biomechanically complex movements, have a potential role in screening for PD. Furthermore, the significant correlation between movement time and UPDRS motor score suggests that movement time may be useful in the quantification of PD severity.	Background The success of dopaminergic interventions in the treatment of Parkinson's disease (PD) symptoms has been significant. Nevertheless, a misdiagnosis of PD can cause psychological trauma and unnecessarily expose patients to PD drugs. Additionally, as new, and possibly neuroprotective, drugs become available for the treatment of PD, early and accurate diagnosis will become increasingly important. As the diagnosis of PD is usually based on subjective clinical assessment of overt symptomatology [ 1 ], the need for an objective and reproducible battery of diagnostic tests is great. Medical imagery offers some hope for the objective diagnosis of PD (e.g. 18 F-dopa positron emission tomography [ 2 ]), but these techniques tend to be expensive, and inaccessible to patients living in remote areas. What is truly needed is a low-cost objective test battery that might be used in situations where (a) movement disorder specialists are unavailable to render expert diagnoses, and (b) medical imaging is inaccessible. Montgomery et al. [ 3 , 4 ] have published one of the better known objective PD batteries, incorporating measures of motor performance, olfaction, and mood. The aggregate of all subtests of this battery has demonstrated good diagnostic properties, with a sensitivity of approximately 70%, and a specificity of approximately 90%. Given that the primary symptoms of PD are motoric [ 1 ], however, it is interesting to note that the sensitivity of the motor task in this PD battery is approximately 50% [ 3 , 4 ], indicating that diagnoses based solely on this subtest are not much better than chance. As the predictive power of a battery increases with the addition of each valid and independent subtest, it is important to evaluate motor performance paradigms that may produce better predictive validity. The global slowing that is consistently demonstrated by PD patients suggests that measures of cognitive or motor speed are logical methods for obtaining quantitative measurements of PD severity. As reaction time (RT) and movement time (MT) have repeatedly been demonstrated to show substantive and significant deficits in PD populations (for a review of this literature, see Gauntlett-Gilbert et al. [ 5 ]), these indicators are (by definition) capable of distinguishing individuals with PD from healthy participants. Despite this fact, however, the motor subtest of the PD battery described by Montgomery et al. [ 3 , 4 ] remains the only significant attempt at evaluating diagnostic accuracy with these chronometric indicators. Although this subtest measures both RT and MT, the task is performed in a biomechanically complicated fashion that requires the participant to move his/her hand in an arc (i.e. wrist flexion and extension) to aim at LED targets. This test assesses both rigidity and bradykinesia within the same task – and while this is a conceptually defensible measurement decision, the resulting inter-subject variability may overwhelm group differences, and confound diagnostic accuracy. It is, therefore, worth examining the extent to which a simpler paradigm might be used to distinguish PD patients from healthy participants. The goal of this study, therefore, is to evaluate the diagnostic properties of a choice reaction time task that uses a simple external response console (i.e. a "button box"), similar to other similar response time tasks extant within the PD literature. Methods Two independent samples were drawn for this study, the first consisting of 40 PD patients (Age: M = 62.13, SD = 9.59) and 40 healthy participants (Age: M = 65.02, SD = 8.84), and the second consisting of 20 PD patients (Age: M = 64.50, SD = 10.88) and 20 healthy participants (Age: M = 62.65, SD = 12.02). To ensure that no baseline ability differences existed between groups, Wechsler Adult Intelligence Scale (WAIS) full scale IQ estimates were computed for all participants, using the National Adult Reading Test (NART) [ 6 ]. No significant age or IQ differences were identified between patients and controls, in either sample. During the course of testing, patients were also assessed by an experienced clinician (using the motor subscale of the Unified Parkinson's Disease Rating Scale; UPDRS), to determine the severity of their motor symptoms [ 7 ]. Patients demonstrated mild to moderate motor symptoms in both the first ( M = 24.49, SD = 9.79), and the second sample ( M = 22.73, SD = 7.66), with no significant severity differences demonstrated between samples. The spectrum of motor severity within the clinical group is graphically depicted with an area graph in Figures 1 and 2 , corresponding to the norming sample and the cross-validation sample, respectively. Finally, all participants were demonstrated to have a Mini-Mental Status Examination (MMSE) score of at least 27 at the time of testing. The response time tasks used in this study started with an instruction to watch a fixation point (asterisk) in the centre of the computer screen, while depressing the home key (measuring 1.905 cm × 1.905 cm) in the centre of the response console. For the 'uncued' task, participants were not given any advance information concerning the location of the upcoming stimulus. For the 'cued' task, an arrow appeared in place of the fixation point (i.e. in the center of the screen) for a period of 2 seconds, immediately following the disappearance of the fixation point, and correctly cued the location of the upcoming stimulus on all trials. The visual stimulus to which the subject responded was presented on the right or left side of the monitor, at a random interval (between 500 and 1500 ms) following the fixation point ('uncued') or the arrow ('cued'). Participants responded to the stimulus by moving the index finger of their dominant hand from the home key to a response key (measuring 1.905 cm × 1.905 cm) placed 3.175 cm to the left or right of the home key, as directed by the stimulus placement on the screen. The time measured between the onset of the visual stimulus and a participant's movement from the home key was defined as reaction time (RT), and the time measured between a participant's lift from the home key and depression of the response key was defined as movement time (MT). Each task consisted of 10 practice trials, and 40 experimental trials. A participant's RT and MT was computed as the unit-weighted average of scores on the 'cued' and 'uncued' choice response time tasks. This testing apparatus is described in further detail by Johnson et al. [ 8 ]. All patients involved in the study were asked to remain drug-free overnight, and to delay taking their morning anti-Parkinsonian medications until after the testing. To avoid any confounding effects resulting from different levels of caffeine intake among participants, all participants were asked to have a normal caffeine-free breakfast prior to testing. None of the participants reported any acute physiological conditions that may have precluded them from putting forth their best effort during the testing session. All procedures and materials were approved by the Health Sciences Research Ethics Board at the University of Western Ontario. Results In the first sample (40 patients and 40 healthy participants), separate receiver operator characteristic (ROC) curves were generated for the composite RT and MT scores. The best prediction (i.e. the largest area under the ROC curve) was achieved using the composite MT score. The cutoff score was identified as the point on the curve that maximized sensitivity, with a specificity of at least 70%. This cutoff score was determined to be 230 ms (i.e. individuals with a MT of at least 230 ms were identified as having PD). To control for the possibility that classification success in the first sample was the result of a capitalization on sample-specific variability [ 9 ], this classification strategy (i.e. the cutoff score identified from the ROC within the first sample) was cross-validated in the second sample (20 patients and 20 healthy participants). Classification results and diagnostic efficacy variables for both samples are presented in Table 1 . To identify the extent to which response time predicted disease progression, correlations were computed between the UPDRS and the aggregate RT and MT scores. As the samples demonstrated no significant differences on the UPDRS, correlations were computed across all data collected in both samples. Both RT (r = 0.23, n.s.) and MT (r = 0.59, p < 0.025) were positively correlated with the UPDRS, suggesting that these response time tasks are good predictors of the severity of Parkinsonian symptoms, particularly when considering the MT component of response time. Discussion This study confirms previous research that has shown significant movement time (MT) differences between PD patients and healthy participants [ 5 ]. The results of the present study also suggest that MT composites on biomechanically simple response time tasks demonstrate high cross-validated sensitivity and specificity for 'unmedicated' patients (i.e. patients that have been temporarily withdrawn from their dopaminergic medications) – and that these values may be higher than the demonstrated sensitivity and specificity of the motor subtest employed by Montgomery et al. [ 3 , 4 ]. Standardized objective test batteries will be diagnostically useful in two general scenarios: (a) as an adjunct to the physical examination performed by a specialist (to improve diagnostic accuracy), and (b) as a standardized preliminary screening tool, for situations in which a movement disorders expert is unavailable for the physical examination. The latter situation is more important than the former, as primary care physicians are often the first point of contact for these patients. Given the waiting times to see a movement disorders specialist, patients that are considered likely to have PD (based on a screening measure) might be assigned a higher priority in their wait for an initial appointment. This assumes, of course, that primary care physicians will have access to the appropriate motor performance testing devices – and while this is currently prohibitive from a logistical standpoint, it is technologically feasible for the simple tasks described herein to be packaged in smaller (perhaps handheld), less expensive devices. A MT battery may also allow for the communication of results in a "common metric" – without relying on subjective clinical judgments, thereby complementing other clinical tools. Aside from its diagnostic utility, MT batteries may also be useful in tracking a patient's progress as he/she undergoes treatment. At present, motor evaluations conducted during the clinical exam are the only method for tracking change, and this is considerably more qualitative than the MT measures described herein. Along similar lines, MT batteries could make useful adjuncts to clinical drug trial protocols, as they provide good quantitative measures of motor skill that may be used to gauge the effectiveness of the medication under study – in the present study, MT was able to explain 34.81% of the variability in UPDRS motor scores. It should, of course, be noted that the present research was only used to separate PD patients from healthy participants, and so it has not been demonstrated to have any differential diagnostic capabilities (e.g. distinguishing PD from progressive supranuclear palsy). Future research in this area should, therefore. investigate the differential diagnostic power of response time batteries – it may be that the MT battery administered in this study are detecting a general 'impairment' factor, and is not useful as a standalone instrument for the diagnosis of Parkinson's disease. Extending the study base to include patients with disorders such as progressive supranuclear palsy would provide important information concerning appropriate norming that may be done to maximize diagnostic utility. At the very least, however, these results suggest that the use of simple response time batteries may serve as a useful adjunct to other clinical assessment batteries, and may also open interesting avenues of exploration into the consideration of the biological underpinnings of reaction time, and its relationship to movement disorders in general. Competing interests None declared. Authors' contributions AJ conceived and designed the study, collected all data in the norming sample, supervised data collection in the cross-validation sample, supervised data analysis, and contributed to the writing of the paper. PV supervised data collection in the norming sample, assisted in the development of the response time tasks, contributed to the data analysis, and to the writing of the paper. QA assisted with data collection in the cross-validation sample, and contributed to the writing of the paper. LG did all clinical testing on patients in the norming sample. RS assisted with data collection in the cross-validation sample. MJ provided patient diagnoses for participants in the norming sample and the cross-validation sample, and contributed to the writing of the paper. All authors have read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC518971.xml
515299	A genome-wide screen identifies a single β-defensin gene cluster in the chicken: implications for the origin and evolution of mammalian defensins	Background Defensins comprise a large family of cationic antimicrobial peptides that are characterized by the presence of a conserved cysteine-rich defensin motif. Based on the spacing pattern of cysteines, these defensins are broadly divided into five groups, namely plant, invertebrate, α-, β-, and θ-defensins, with the last three groups being mostly found in mammalian species. However, the evolutionary relationships among these five groups of defensins remain controversial. Results Following a comprehensive screen, here we report that the chicken genome encodes a total of 13 different β-defensins but with no other groups of defensins being discovered. These chicken β-defensin genes, designated as Gallinacin 1–13 , are clustered densely within a 86-Kb distance on the chromosome 3q3.5-q3.7. The deduced peptides vary from 63 to 104 amino acid residues in length sharing the characteristic defensin motif. Based on the tissue expression pattern, 13 β-defensin genes can be divided into two subgroups with Gallinacin 1–7 being predominantly expressed in bone marrow and the respiratory tract and the remaining genes being restricted to liver and the urogenital tract. Comparative analysis of the defensin clusters among chicken, mouse, and human suggested that vertebrate defensins have evolved from a single β-defensin-like gene, which has undergone rapid duplication, diversification, and translocation in various vertebrate lineages during evolution. Conclusions We conclude that the chicken genome encodes only β-defensin sequences and that all mammalian defensins are evolved from a common β-defensin-like ancestor. The α-defensins arose from β-defensins by gene duplication, which may have occurred after the divergence of mammals from other vertebrates, and θ-defensins have arisen from α-defensins specific to the primate lineage. Further analysis of these defensins in different vertebrate lineages will shed light on the mechanisms of host defense and evolution of innate immunity.	Background Defensins constitute a large family of small, cysteine-rich, cationic peptides that are capable of killing a broad spectrum of pathogens, including various bacteria, fungi, and certain enveloped viruses [ 1 - 5 ]. These peptides play a critical role in host defense and disease resistance by protecting the hosts against infections. Transgenic mice expressing human enteric defensin HD5 are fully protected against the doses of Salmonella typhimurium that are otherwise lethal to the wide-type mice [ 6 ]. Conversely, mice deficient in the matrilysin gene, which is responsible for activating enteric defensins, become more susceptible to oral infection with S. typhimurium [ 7 ]. Defensins have been identified in species ranging from plants, insects to animals and humans [ 1 - 5 ]. Characterized by the presence of 6–8 cysteine residues in relatively defined positions, all defensins are structurally related in that they form 3–4 intramolecular disulfide bonds and 2–3 antiparallel β-sheets with or without an α-helix. Based on the spacing pattern of cysteines, these peptides are broadly divided into five groups; namely plant, invertebrate, α-, β-, and θ-defensins [ 1 - 5 ]. Alignment of all known defensin sequences revealed the consensus defensin motif of each group as follows: plant defensin: C-X 8–11 -C-X 3–5 -C-X 3 -C-X 9–12 -C-X 4–11 -C-X 1 -C-X 3 -C; invertebrate defensin: C-X 5–16 -C-X 3 -C-X 9–10 -C-X 4–7 -C-X 1 -C; α-defensin: C-X 1 -C-X 3–4 -C-X 9 -C-X 6–10 -C-C; and β-defensin: C-X 4–8 -C-X 3–5 -C-X 9–13 -C-X 4–7 -C-C. The α- and β-defensins are unique to vertebrate animals with α-defensins only being found in rodents and primates, while β-defensins are present in all mammalian species investigated [ 1 - 3 ]. On the other hand, θ-defensins have only been found in certain primates as a result of posttranslational ligation of two α-defensin-like sequences [ 8 - 10 ]. A pseudogene for θ-defensin is also present in humans [ 11 ]. Analysis of human and mouse genomes indicated that β-defensins form 4–5 distinct clusters on different chromosomes with each cluster consisting of multiple defensin genes [ 12 ]. Interestingly, the single mammalian α-defensin locus is located on a β-defensin cluster with θ-defensins residing in the center of α-defensins [ 12 ]. Studies with mammalian defensins suggested a rapid duplication followed by positive selection and diversification within each group [ 13 - 18 ]. However, the evolutionary relationships among three groups of mammalian defensins and among plant, invertebrate, and mammalian defensins remain controversial. Similarity in spatial structure and biological functions favors the notion that all mammalian defensins are evolutionarily related [ 19 ], although a phylogenetic analysis suggested a closer relationship between β- and insect defensins than between α- and β-defensins [ 16 ]. Existence of a large number of expressed sequence tag (EST) sequences and recent completion of chicken genome sequencing at a 6.6× coverage [ 20 ] provided a timely opportunity to discover a complete repertoire of defensin-related sequences in birds for studying the evolutionary relationship between invertebrate and mammalian defensins. Here we report identification of a single β-defensin cluster that is composed of 13 genes located on the chicken chromosome 3q3.5-q3.7. Evolutionary and comparative analyses of these chicken β-defensins with mammalian homologues strongly suggested that all mammalian defensins have evolved from a common β-defensin-like ancestor, which has undergone rapid duplication, positive diversifying selection, and chromosomal translocations, thereby giving rising to multiple gene clusters on different chromosomal regions. Results and Discussion Discovery of novel chicken defensins To identify novel defensin genes in the chicken, all five groups of known defensin-like peptide sequences from plants, invertebrates, and vertebrates were first queried individually against the translated chicken nonredundant (NR), EST, high throughput genomic sequence (HTGS), and whole-genome shortgun sequence (WGS) databases in the GenBank by using the TBLASTN program[ 21 ]. All potential hits were then examined manually for the presence of the characteristic cysteine motifs. For every novel defensin identified, additional iterative BLAST searches were performed until no more novel sequences could be found. In addition to three known chicken β-defensins ( Gal 1–3 ) [ 22 , 23 ], nine novel putative sequences, namely Gal 4–12 , have been found in the EST database with at least two hits for each, and such sequences have also been confirmed in genomic sequences (Table 1 ). Because of the fact that mammalian defensins tend to form clusters [ 12 , 14 , 15 , 18 ], all chicken HTGS and WGS sequences containing defensin sequences were also retrieved from GenBank, translated into six open reading frames, and manually curated. As a result, an additional putative β-defensin, Gal13 , was identified in several genomic clones (Table 1 ). The open reading frame of Gal13 was predicted by GENSCAN [ 24 ] and confirmed by directly sequencing of RT-PCR product amplified from chicken kidney. Table 1 Identification of chicken β-defensins Gene EST 1,2 HTGS WGS Gene Size (bp) 3 E 1 I 1 E 2 I 2 E 3 I 3 E 4 Gal1 BX260462 AC110874 AADN01058097 70 972 88 482 127 496 217 Gal2 BX540940 AC110874 AADN01058097 66 1113 143 183 121 674 204 Gal3 AC110874 AADN01058097 AADN01058096 53 980 109 1180 215 Gal4 BU451960 AADN01058096 136 461 127 117 141 Gal5 BU389548 AADN01058096 290 445 127 355 187 Gal6 CF251501 AC110874 AADN01058097 AADN01058098 52 704 86 705 130 249 234 Gal7 CF251115 AC110874 AADN01058098 50 656 86 201 130 234 248 Gal8 BU242665 AC110874 AADN01058098 71 915 91 259 134 706 494 Gal9 BX270804 AC110874 AADN01058098 220 1592 130 781 343 Gal10 AW198592 AC110874 AADN01058099 118 268 133 1719 381 Gal11 BM440069 AC110874 AADN01058101 63 966 129 1001 460 Gal12 BX257296 AC110874 AADN01058102 84 396 420 Gal13 AC110874 AADN01058102 61 1016 118 3322 91 1 Abbreviations: EST , expressed sequence tag; HTGS , high throughput genomic sequence; WGS , whole-genome shortgun sequence; E , exon; I , intron. 2 One EST sequence entry is given only for the exemplary purpose. In each case, more than two independent EST sequences have been found, except for Gal3 and Gal13 , both of which have no EST sequences. Gal3 was found through homology cloning [23], and Gal13 was predicted by us from the genomic sequence. 3 All Gal genes are predicted to consist of four exons separated by three introns, except for Gal12 , whose last two exons are fused together. The absence of additional sequence information at the 5'-untranslated regions of the cDNA sequences prevented prediction of the sizes of first exon and intron for Gal 3–5 and Gal 9–13 genes. No other sequence containing β-defensin-like six-cysteine motif has been found in NR, EST or genomic databases, suggesting that 13 Gal genes constitute the entire repertoire of the β-defensin family encoded in the chicken genome. Although it is highly unlikely, we could not rule out the possibility that additional defensin-related genes with distant homology might be uncovered in the chicken by different computational search methods such as the use of Hidden Markov models [ 12 , 15 ]. It is noted that none of other groups of defensins have been discovered in the chicken, indicating that plant, invertebrate, α-, and θ-defensins are absent in the chicken lineage. Similar to Gal 1–3, 10 novel β-defensins, deduced from either EST or genomic sequences, vary from 63 to 104 amino acid residues in length. Alignment of these peptides revealed a conservation of the signal sequence at the N-terminus and the characteristic six-cysteine defensin motif at the C-terminus (Figure 1 ). Consistent with the fact that all β-defensins are a group of secreted molecules in response to infections, the signal sequences of all chicken defensins are hydrophobic and rich in leucines. In addition, the mature C-terminal sequences are all positively charged due to the presence of excess arginines and lysines. Interestingly, Gal11 contains two tandem, but highly divergent, copies of the six-cysteine motif at the C-terminus, and is the only defensin having such sequences. Functional significance for existence of such two defensin motifs remains to be studied. Figure 1 Multiple sequence alignment of chicken β-defensins. The intervening region between signal and mature peptide sequence is the short propiece. The conserved residues are shaded. Also shown is the length of each peptide. Notice the six-cysteine defensin motif is highly conserved. The six cysteines in the second tandem copy of the defensin motif in Gal11 are boxed. Evolutionary analysis of vertebrate β-defensins Phylogenetic analysis of vertebrate β-defensins showed that chicken defensins clustered with various different groups of mammalian β-defensins (Figure 2 ). However, the bootstrap support for these patterns was very weak (less than 50% in all cases). The clustering of certain chicken β-defensins with mammalian homologues suggests that major subfamilies of β-defensins arose before the last common ancestor of birds and mammals, estimated to have occurred about 310 million years ago [ 25 ]. This in turn implies that some duplication of β-defensin genes must have taken place before the divergence of birds and mammals. The apparent lack of α-defensins in the chicken and other non-mammalian species (G. Zhang, unpublished data) suggests that α-defensins may have evolved after mammals diverged from other vertebrates. Figure 2 Phylogenetic relationship of vertebrate β-defensins. The tree was constructed by the neighbor-joining method and the reliability of each branch was assessed by using 1000 bootstrap replications. Numbers on the branches indicate the percentage of 1000 bootstrap samples supporting the branch. Only branches supported by a bootstrap value of at least 50% are indicated. Chicken β-defensins are highlighted in yellow. Abbreviations: BNBD , bovine neutrophil β-defensin; LAP , lingual antimicrobial peptide; EBD , enteric β-defensin; TAP , tracheal antimicrobial peptide; PBD , porcine β-defensin; DEFB/Defb , β-defensin; Gal , Gallinacin; GAPDH , glyceraldehyde-3-phosphate dehydrogenase. Comparison of the numbers of synonymous and nonsynonymous nucleotide substitutions provides a powerful test of the hypothesis that positive Darwinian selection has acted to favor changes at the amino acid level [ 26 ]. This approach has previously been applied to both α- and β-defensins of mammals and has revealed positive selection acting on the mature defensin but not on other regions of the gene [ 16 , 17 ]. In the comparison of the chicken β-defensin sequences, synonymous sites were saturated with changes or nearly so, making it impossible to test the hypothesis of positive selection in every case. In pairwise comparisons among all sequences, mean p S in the propeptide region was 0.551 ± 0.036 (S.E.), while mean p N was 0.369 ± 0.040. In the mature defensin region, mean p S was 0.673 ± 0.027, while mean p N was 0.534 ± 0.051. Mean p N in the mature defensin was significantly greater than that in the propeptide (z-test; P < 0.05), indicating lesser functional constraint on the amino acid sequence of the former. The high mean p S shows that chicken β-defensin genes have not duplicated recently, unlike β-defensin genes of the bovine [ 16 ]. In the comparison between the most closely related pair of sequences ( Gal6 and Gal7 ), mean p S in the mature defensin was 0.221 ± 0.082, while mean p N was 0.331 ± 0.076. While these values are not significantly different at the 5% level, the fact that p N was higher than p S suggested that positive selection may have acted to diversify the mature defensin region between these two genes. Genomic organization and chromosomal localization of the chicken β-defensin gene cluster Searching through HTGS database led to identification of two overlapping bacterial artificial chromosome (BAC) sequences, TAM31-54I5 (accession no. AC110874) and CH261-162O9 (accession no. AC146292), both of which were sequenced and deposited earlier by one of us (J.F. Chen). Alignment of these two sequences allowed to re-order three DNA fragments in AC110874 and to construct a continuous, gap-free genomic contig that includes 11 Gal genes except for Gal4 and Gal5 . Later search of chicken WGS sequences released on February 29, 2004 confirmed the order of the genomic contig that we assembled and also revealed the locations of two remaining genes, Gal4 and Gal5 , both of which reside on a WGS (accession no. AADN01058096) that overlaps with AC110874 (Figure 3 ). The position and orientation of each Gal gene were obtained by comparing its cDNA with the assembled DNA sequence. As shown in Figure 3 , all 13 Gal genes were clustered densely within a distance of 86.0 Kb on the genome. It was also confirmed by aligning such a contig with the chicken genome assembly, in which 13 Gal genes are located on six WGS contigs (Table 1 ) of chromosome 3 that are only ~3.3 Mb from the distal end. Consistent with this, the Gal gene cluster was physically mapped to the tip of chicken chromosome 3 at the region of q3.5-q3.7 by fluorescence in situ hybridization (FISH) using the TAM31-54I5 BAC DNA as probe (Figure 4 ). Figure 3 Genomic organization of the chicken β-defensin gene cluster. The horizontal lines at the bottom represent the three overlapping genomic clones that were used to assemble the continuous, gap-free contig. The position of each gene is represented by a solid vertical bar and the width of each bar is proportional to the size of each gene. The direction of transcription is indicated by the triangle above each gene. The genes with solid triangles are transcribed in the direction opposite to the ones with open triangles. Slanted lines refer to the sequences omitted. Note that the three fragments of AC110874 sequence have been re-ordered and the gaps have been filled following alignment with AC146292. Figure 4 Chromosomal localization of the chicken β-defensin gene cluster by fluorescence in situ hybridization. The BAC clone TAM31-54I4, which harbors 11 Gal genes, was mapped to chicken chromosome 3q3.5-q3.7. Arrows indicate the hybridization signals. Comparing the cDNA with genomic sequences also revealed the structure of each Gal gene. Unlike most mammalian β-defensin genes, which primarily consist of two exons and one intron, the Gal genes were found to be composed of four short exons separated by three introns with variable lengths ranging from 117 bp to 3,322 bp (Table 1 ). Gal12 is an exception, in which the last two exons have been fused together. While the first exon of the Gal genes encodes 5'-untranslated region (UTR) and the majority of the last exon encodes 3'-UTR as well as a few C-terminal amino acids, two internal exons resemble mammalian β-defensin genes in that one exon encodes the signal and pro-sequence and the other encodes the mature sequence with six-cysteine motif [ 19 , 27 - 29 ]. Apparently, the first two and the last two exons of the Gal genes have joined together during the evolution as a result of exon shuffling, which occurred in many other evolutionarily conserved gene families [ 30 ], including invertebrate defensins [ 5 ]. The fusion of defensin exons in mammals is presumably adaptive because it allows a faster mobilization of such host defense molecules to better cope with invading microbes. Tissue expression patterns of chicken β-defensins It has been shown that Gal1 and Gal2 are expressed in bone marrow and lung, while Gal3 is more preferentially expressed in bone marrow, tongue, trachea, and bursa of Fabricius [ 23 ]. To study the tissue expression patterns of novel Gal genes that we identified, RT-PCR was performed with a panel of 32 different chicken tissues. Similar to Gal 1–3 , Gal 4–7 are highly restricted to bone marrow cells with Gal5 also expressed in tongue, trachea, lung, and brain at lower levels (Figure 5 ). By contrast, the six remaining genes, Gal 8–13 , were not found in bone marrow, but instead in liver, kidney, testicle, ovary, and male and female reproductive tracts (Figure 5 ). These results clearly suggested that all chicken β-defensin genes can be divided into two subgroups. Seven genes ( Gal 1–7 ) are predominantly expressed in bone marrow and the respiratory tract, whereas the other six genes ( Gal 8–13 ) are more restricted to liver and the urogenital tract. However, the functional significance and transcriptional regulatory mechanisms of these genes during inflammation and infection remain to be investigated. Figure 5 Tissue expression patterns of 10 novel chicken β-defensins by RT-PCR. See Materials and Methods for details. The number of PCR cycles was optimized for each gene, and the specificity of each PCR product was confirmed by sequencing. The house-keeping gene, GAPDH, was used for normalization of the template input. Comparative analysis of chicken and mammalian β-defensin gene clusters To study the origin and evolution of mammalian defensins, a comparative analysis of β-defensin gene clusters in the chicken, mouse, and human was performed by employing additional, more phylogenetically conserved gene markers surrounding the defensin clusters. As shown in Figure 6 , two genes, CTSB (Cathepsin B, accession no. NP_680093) and a human EST sequence (accession no. BE072524) immediately located centromeric to chicken defensins, were also found to be conserved in the defensin gene clusters on human chromosome 8p22 and mouse chromosome 14C3. Similarly, another gene, HARL2754 (accession no. XP_372011) that is 6-Kb telemetric to Gal4 is also conserved in another defensin cluster in human (8p23) or mouse (8A1.3) (Figure 6 ). Figure 6 Comparative analysis of defensin clusters among the chicken, mouse, and human. The gene clusters were drawn proportionally according to their sizes. Each vertical line/bar represents the position of a gene, and the width of each line/bar is proportional to the size of each gene. Three highly conserved genes ( CTSB , BE072524, and HARL2754 ) surrounding the defensin clusters in the chicken, mouse, and human were connected by solid lines. The position of the α-defensin locus ( DEFA ) was indicated as an open square. Note that the human θ-defensin pseudogene resides in the DEFA locus. The positions and orders of defensin genes in human and mouse were drawn based on the genome assemblies released in July 2003 and October 2003, respectively. Abbreviations: GGA , chicken chromosome; MMA , mouse chromosome; HSA , human chromosome; Tel , telomere; Cen , centromere. These results strongly suggested that all vertebrate β-defensins are evolved from a single gene. This conclusion is further supported by the fact that there are three highly similar β-defensin-like sequences present in the largely finished zebrafish genome (G. Zhang, unpublished data). In addition, a group of homologous β-defensin-like sequences, namely crotamine and myotoxins, have been found in several Crotalus snakes [ 31 ], which are presumably derived from a single ancestral gene. The appearance of multiple β-defensin gene clusters on different chromosomal regions in mammalian species [ 12 ] is apparently a result of rapid gene duplication, positive diversifying selection, and chromosomal translocation following divergence of mammals from other vertebrate lineages. In addition to the structural conservation between β-defensin-like sequences in the rattlesnake and mammals [ 32 ], a growing body of evidence suggests that their functions appear to be largely conserved in that both are capable of interacting negatively-charged lipid membranes followed by formation of ion channels or pores [ 32 - 34 ]. It is noteworthy that the conservation of Cathepsin B (CTSB) adjacent to β-defensins is perhaps not surprising, given the recent finding that cathepsins are involved in the cleavage and inactivation of β-defensins [ 35 ]. Conclusions We have showed that chicken genome encodes a total of 13 different β-defensin genes clustered densely within a 86-Kb distance on the chromosome 3q3.5-q3.7, but with no α-defensin genes. These peptides exhibit homology to different subgroups of mammalian β-defensins-, consistent with the hypothesis that α-defensins and β-defensins arose by gene duplication after the divergence of birds and mammals. The θ-defensins are specific to primates; and thus appear to have arisen from α-defensins by gene duplication specific to the primate lineage. Apparently, the evolution of defensins is rapid and driven by duplication and positive diversifying selection. Collectively, this study represents the first large-scale detailed investigation of defensins in non-mammalian vertebrates. There is no doubt that further analysis of these defensin genes will lead to a better understanding of host defense mechanisms and evolution of innate immunity. Methods Computational search for novel chicken defensins To identify novel defensins in the chicken, all known cysteine-containing defensin-like peptide sequences discovered in plants, invertebrates, birds, and mammals were individually queried against the translated chicken NR, EST, HTGS, and WGS databases in the GenBank by using the TBLASTN program [ 21 ] with default settings on the NCBI web site [ 36 ]. All potential hits were then examined for the presence of the characteristic defensin motif. For every novel defensin identified, additional iterative BLAST searches were performed until no more novel sequences could be revealed. Because mammalian defensins tend to form clusters [ 12 , 14 , 15 , 18 ], all chicken genomic sequences containing defensin sequences were also retrieved from the GenBank and translated into six open reading frames and curated manually for the presence of the defensin motif in order to discover potential sequences with distant homology. Alignment and phylogenetic analysis of chicken β-defensins Multiple sequence alignment was constructed by using the ClustalW program (version 1.82) [ 37 ]. A phylogenetic tree of amino acid sequences of mature β-defensins was constructed by the neighbor-joining method [ 38 ]. So that a comparable data set would be used for all pairwise comparisons, any site at which the alignment postulated a gap in any sequence was excluded from the analysis. To maximize the number of sites available for analysis, certain sequences with large deletions were excluded from the analysis. Because the sequences were very short (25 aligned sites), no correction for multiple hits was applied. The reliability of clustering patterns within the tree was assessed by bootstrapping; 1000 bootstrap pseudo-samples were used. The proportion of synonymous nucleotide differences per synonymous site (p S ) and the proportion of nonsynonymous nucleotide differences per nonsynonymous site (p N ) were estimated by the method of Nei and Gojobori [ 26 ]. Again, no correction for multiple hits was applied because a small number of sites were examined. Assembly of the chicken β-defensin gene cluster To generate a continuous defensin gene cluster, the HTGS and WGS sequences containing the putative defensin genes were retrieved from the GenBank, aligned to generate a longer contig, which was confirmed later by searching through the assembled chicken genome released on February 29, 2004, by using the BLAT program [ 39 ] under the UCSC Genome Browser web site [ 40 ]. The relative positions, orientations, and structural organizations of individual genes were determined by comparing its cDNA sequence to the continuous genomic contig that we assembled. Chromosome localization of the chicken β-defensin gene cluster Fluorescence in situ hybridization (FISH) was used for chromosomal assignment of the chicken β-defensin gene cluster by using the BAC clone TAM31-54I4 as probe, which harbors 11 Gal genes. Metaphase chromosome speads were prepared from mitogen-stimulated chicken splenocyte culture as we described [ 41 , 42 ]. The BAC clone was labeled by nick translation with biotin 16-dUTP (Roche Diagnostics), hybridized to metaphase chromosome DNA, followed by detection with FITC-labeled avidin (Roche Diagnostics) and staining with propidium iodide to simultaneously induce the R-banding. RT-PCR analysis of the tissue expression patterns of chicken β-defensins Total RNA was extracted with Trizol (Invitrogen) from a total of 32 different tissues from healthy, 2-month-old chickens (see Figure 5 ). A total of 4 μg RNA from each tissue were reverse transcribed with random hexamers and Superscript II reverse transcriptase by using a first-strand cDNA synthesis kit (Invitrogen) according to the instructions. The subsequent PCR was carried out with 1/40 of the first-strand cDNA and gene-specific primers for each β-defensin and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as described [ 28 , 43 ]. Every pair of primers were designed to locate on different exons to aid in distinguishing PCR products amplified from cDNA vs. genomic DNA (Table 2 ). The PCR program used was: 94°C denaturation for 2 min, followed by different cycles of 94°C denaturation for 20 sec, 55°C annealing for 20 sec, and 72°C extension for 40 sec, followed by a final extension at 72°C for 5 min. The number of PCR cycle was optimized for each gene to ensure linear amplification (Table 2 ). A half of the PCR products were analyzed by electrophoresis on 1.2% agarose gels containing 0.5 μg/ml ethidium bromide. The specificity of each PCR product was confirmed by cloning of the PCR product into T/A cloning vector, followed by sequencing of the recombinant plasmid. Table 2 Primer sequences used for RT-PCR analysis of novel chicken β-defensins Gene Primer Sequence Product Size (bp) Cycles Used Sense Antisense cDNA Genomic Gal4 CATCTCAGTGTCGTTTCTCTGC ACAATGGTTCCCCAAATCCAAC 321 899 36 Gal5 CTGCCAGCAAGAAAGGAACCTG TGAACGTGAAGGGACATCAGAG 300 1100 36 Gal6 AGGATTTCACATCCCAGCCGTG CAGGAGAAGCCAGTGAGTCATC 249 1203 36 Gal7 CTGCTGTCTGTCCTCTTTGTGG CATTTGGTAGATGCAGGAAGGA 230 665 35 Gal8 ACAGTGTGAGCAGGCAGGAGGGA CTCTTCTGTTCAGCCTTTGGTG 261 967 35 Gal9 GCAAAGGCTATTCCACAGCAG AGCATTTCAGCTTCCCACCAC 211 1802 33 Gal10 TGGGGCACGCAGTCCACAAC ATCAGCTCCTCAAGGCAGTG 298 2285 33 Gal11 ACTGCATCCGTTCCAAAGTCTG TCGGGCAGCTTCTCTACAAC 301 1299 33 Gal12 CCCAGCAGGACCAAAGCAATG GTGAATCCACAGCCAATGAGAG 335 731 36 Gal13 CATCGTTGTCATTCTCCTCCTC ACTTGCAGCGTGTGGGAGTTG 175 4514 50 GAPDH GCACGCCATCACTATCTTCC CATCCACCGTCTTCTGTGTG 356 876 30 Note added in proof Following submission of this manuscript, Lynn et al. reported independently discovery of seven novel chicken β-defensins in the chicken EST database by using homology search strategies [ 44 ]. Consistent with our conclusion, they also revealed occurrence of positive selection particularly in the mature region of chicken β-defensins following evolutionary analysis. Moreover, albeit the use of a different nomenclature, they confirmed that the expressions of Gal 4–7 are primarily in bone marrow, while other genes are more restricted to liver and the genitourinary tract. List of abbreviations Abbreviations: Gal, Gallinacin; NR, nonredundant; EST, expressed sequence tag; HTGS, high throughput genomic sequence; WGS, whole-genome shortgun sequence; BAC, bacterial artificial chromosome; FISH, fluorescence in situ hybridization; UTR, untranslated region; GAPDH, glyceraldehyde-3-phosphate dehydrogenase. Authors' contributions YX carried out the tissue collection, RT-PCR analysis of tissue expression patterns, and drafted the manuscript. ALH carried out the phylogenetic and molecular evolutionary analyses. JA and YM carried out the fluorescence in situ hybridization. JFC carried out the sequencing of two chicken defensin-containing BAC clones. DSN participated in tissue collection and preparation. GZ conceived of the study, carried out all computational analyses and annotation, drafted the manuscript, and participated in its design and coordination. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC515299.xml
512290	Elevated expression of MMP-13 and TIMP-1 in head and neck squamous cell carcinomas may reflect increased tumor invasiveness	Background Matrix metalloproteinases [MMPs], which degrade the extracellular matrix, play an important role in the invasion and metastasis of squamous cell carcinomas. One MMP, MMP-13, is thought to play a central role in MMP activation. The purpose of this study was to investigate MMP-13 and TIMP-1 expression in squamous cell carcinomas of the head and neck and to relate these levels of expression to histologic patterns of invasion. Methods This study included T1 lesions obtained via biopsy from the larynx, tongue, and skin/mucosa of 78 patients with head and neck squamous cell carcinomas. The relationship between expression of MMP-13 and TIMP-1 and the mode of tumor invasion [MI] was evaluated immunohistochemically, using breast carcinoma tissue as a positive control. Results Increased expression was observed in highly invasive tumors, as reflected by the significant correlation between the degree of staining for MMP-13 or TIMP-1 and MI grade [p < 0.05]. There was no significant relationship between the degree of staining for MMP-13 or TIMP-1 and patient age, sex, tumor site, or tumor histologic grade. In addition, levels of staining for MMP-13 did not correlate with levels of staining for TIMP-1. Conclusion The expression of MMP-13 and TIMP-1 appears to play an important role in determining the invasive capacity of squamous cell carcinomas of the head and neck. Whereas additional studies are needed to confirm these findings, evaluating expression of these MMPs in small biopsy samples may be useful in determining the invasive capacity of these tumors at an earlier stage.	Background The invasion of surrounding tissues by neoplastic cells is one of the most important steps in tumor progression. Proteolytic enzymes such as matrix metalloproteinases [MMPs] contribute to tumor expansion by degrading components of the extracellular matrix [ECM]. MMPs are a 21-member family of zinc-dependent endopeptidases, which are capable of degrading most ECM components including collagen, elastin, fibronectin, and gelatin. Whereas this facilitates processes such as wound healing, enhanced MMP activity has been observed in a variety of pathologic conditions including osteoarthritis, rheumatoid arthritis, cardiovascular disease, and neoplasia [ 1 - 6 ]. MMPs can be divided into subgroups, which include collagenases, stromelysins, stromelysins-like MMPs, gelatinases, membrane-type MMPs, and others [ 5 ]. MMP-13 [collagenase 3] is a member of the collagenase family. It was first identified in human breast cancer and is active against a wide variety of ECM components [ 7 ]. MMP-13 also plays a central role in the MMP activation cascade, both activating and being activated by several MMPs [ 8 ]. Elevated MMP-13 expression has been found in a number of different malignancies, and expression has been related to tumor behavior and patient prognosis [ 7 , 9 , 10 ]. MMPs can be inactivated by specific tissue inhibitors of matrix metalloproteinases [TIMPs]. Thus far, four different TIMPs [TIMP-1, -2, -3, -4 ] have been identified [ 11 ] and implicated in connective tissue turn-over and remodeling. TIMPs inhibit MMPs by binding to them and forming non-covalent complexes [ 5 ]. Recent studies have linked increased MMP expression and decreased TIMP expression with tumor aggressiveness; however, other studies have shown overexpression of TIMPs in some patients with advanced tumors [ 12 , 13 ]. Squamous cell carcinoma [SCC] of the head and neck region is a major problem. Many studies have shown that MMPs are expressed in SCCs [ 5 , 14 , 15 ], with particular involvement by MMP-2 and MMP -9 [ 5 , 13 , 16 ]. MMP-13, which plays a central role in MMP activation, has also been shown to be highly expressed in head and neck SCCs [HNSCCs] [ 5 ]. As MMPs appear to be essential for tumor invasion and metastatic spread, characterizing their expression might help to determine a patient's treatment or prognosis. Therefore, we investigated the expression of MMP-13 and TIMP1 in biopsy specimens of HNSCCs to determine the relationship between the expression of these proteins and the mode of tumor invasion. Methods Patient and tumor characteristics Between 1998 and 2003, we obtained 78 incisional and excisional biopsy samples from the primary tumors of 78 patients with invasive SCC who were admitted to the Head and Neck Surgery and Plastic Surgery Departments. Fifty-four men and 24 women were included in the study, and the mean patient age was 64 years [range: 26 to 88 years]. T1 stage tumors from the larynx, tongue, and skin or mucosa of the face, cheek, lip, nose, and ear were studied. Table 1 summarizes clinical characteristics of the patients. None had undergone prior chemotherapy or radiotherapy. All patients provided informed consent to participate, and the study protocol was approved by our institution's ethics committee. Tumors were histologically graded as well [G1], moderately [G2], or poorly [G3] differentiated [Table 1 ]. The mode of tumor invasion [MI] was graded as follows: grade 1, well-defined border; grade 2, less-marked border; grade 3, groups of cells with no distinct border; grade 4, diffuse invasion [a, cordlike; b, widespread] [Table 1 ]. Immunohistochemical procedures The most representative block of tumor tissue was chosen in each case, and 5-μm sections were obtained and mounted on poly-L-lysine-coated slides for immunohistochemical staining. A standard streptavidin-biotin immunoperoxidase method was used for immunostaining with MMP-13 [7 ml, MS-825-R7, Ready-to-use, Neomarkers, USA] and TIMP-1 [7 ml, MS-608-R7, Ready-to-use, Neomarkers, USA] antibodies. The tissue sections were deparaffinized in xylene, rehydrated in an alcohol series, and immersed in distilled water. The sections were then boiled in a citrate buffer solution [10 mmol/L, pH= 6.0] in a microwave oven X3 for 10 minutes for antigen retrieval of both the MMP-13 and TIMP-1 antibodies. Endogenous peroxidase activity was blocked by exposing sections to a 0.3% solution of hydrogen peroxidase in phosphate-buffered saline [PBS] for 10 minutes at room temperature. After the sections had been rinsed with TRIS buffer, primary antibodies were applied for 60 minutes at room temperature followed by TRIS buffer. Linking antibody and streptavidin peroxidase complex were then added consecutively for 10 minutes at room temperature, and sections were washed again in TRIS buffer. After applying AEC chromogen, the sections were washed in deionized water, counterstained and mounted. Breast carcinoma tissue (which showed positive staining) was used as positive control during the evaluation of MMP-13 and TIMP-1 immunostaining. Evaluation/scoring of the staining One pathologist [NC] evaluated the stained slides associated with each case. The degree of staining for MMP-13 and TIMP-1 was scored as follows: 0, no staining of the tumor or stromal cells; 1+, weak [< 50%] positive staining of the tumor cells and/or weak staining of stromal cells; 2+, moderate [> 50%] positive staining of the tumor cells and/or moderate staining of stromal cells; 3+, extensive staining of the tumor cells and/or strong staining of stromal cells [Figs. 1 , 2 ] [ 12 ]. No normal epithelial cells were stained [Fig. 3 ]. Follow-up If necessary, surgery or radiotherapy was performed after the diagnosis. All patients were followed-up postoperatively for a mean interval of 20 months [range: 7 to 36 months]. During this period, 2 of 78 [2.6%] patients died from their tumors [one in larynx, one in tongue], and 1 patient died of an unrelated cause. Seventy-five of 78 [96.2%] patients were free of disease at the end of the follow-up period. Statistical analysis The chi-square test was used for statistical analysis. Data were analyzed using SPSS for Windows 10,0. A p level <0.05 was considered to be statistically significant. Results Although expression of cytoplasmic MMP-13 was primarily detected in tumor cells, it also was seen in stromal cells. MMP-13 expression was evident in tissue from 44 [56.4%] of the 78 patients. High levels of MMP-13 were noted in highly invasive tumors. There was no difference in the amount of staining between tumor centers or margins. Significant TIMP-1 expression was detected in tissue from 42 [53.8%] patients. Prominent labeling was confined to the stroma surrounding the tumor cells. Table 2 shows the relationship between MMP-13 expression and MI in the 78 biopsy specimens. Increased expression of MMP-13 was observed in highly invasive tumors [MI grades 4a and b]. The relationship between MMP-13 expression and MI grade was statistically significant [p < 0.05]. Table 3 summarizes the relationship between TIMP-1 expression and MI in the 78 biopsy specimens. Increased expression of TIMP-1 was observed in highly invasive tumors. The relationship between TIMP-1 expression and MI grade was statistically significant [p < 0.05]. There was no statistically significant relationship between the degree of staining for MMP-13 or TIMP-1 and patient age, sex, tumor site, or histologic grade. Also, the degree of staining for MMP-13 did not significantly correlate with the degree of staining for TIMP-1. Discussion As tumor invasion and metastasis affect a patient's prognosis, it is important to predict the invasive potential of tumors such as SCCs at an early stage. Several steps are involved in the invasion and metastasis of malignant cells including the attachment of cells to the ECM, the breakdown of matrix components, cell detachment, and migration of cells through the degraded matrix. This complex process requires several proteases, and the local balance between these proteases and protease inhibitors appears to be crucial. The MMPs represent one family of degrading proteases, which are expressed in various tumors. Many studies have shown that MMPs, especially MMP-2, -3, -9, are expressed in SCCs [ 5 , 12 , 16 - 18 ]. In the present study, we characterized the expression of one MMP–MMP-13, which plays a key role in the MMP activation cascade–and one inhibitor of MMPs, TIMP-1, in HNSCCs. Our finding that expression of both proteins was upregulated in markedly invasive tumors is consistent with prior reports and may have important therapeutic and prognostic implications. The expression of several MMPs has been investigated in SCCs including MMP-9 and MMP-13. MMP-9 expression was found to correlate the most strongly with advanced pathological stages, and patients with HNSCCs also may have elevated serum concentration of MMP-9 [ 17 , 19 ]. Furthermore, expression of some MMPs such as MMP-9 appears to vary not only between the primary tumor and sites of lymph node metastasis, but also between the early and late stages of lymph node metastasis [ 16 ]. MMP-13 is a member of the collagenase family, which degrades fibrillar collagens of types I, II, III, IV, X, and XIV, tenascin, fibronectin, aggrecan, versican, and fibrillin-I. It is now accepted that MMP-13 plays a key role in the MMP activation cascade, both activating and being activated by several MMPs [MT1-MMP, MMP-2, MMP-3] [ 8 , 11 ]. MMP-13 expression is also susceptible to stimulation by several cytokines and growth factors such as TNF-α and TGF-β [ 11 , 15 ]. Elevated MMP-13 expression has been documented in numerous malignancies [e.g., breast carcinoma, colorectal cancer, vulvar SCC, skin SCC, HNSC, BCC (in focal areas of keratinized cells)] and associated with tumor behavior and patient prognosis [ 1 , 7 - 10 , 20 ]. Both small and large tumors appear to express MMP-13, with expression being the most prevalent in undifferentiated tumors [ 11 ]. MMP-13 expression has also been observed in transformed, but not primary, human epidermal keratinocytes [ 11 , 15 ]. Although MMP-1, -2, -3 have been detected in actinic keratosis [ 12 , 21 ], premalignant and benign tumors were mostly negative for MMP-13 in one study [ 20 ]. It is thought that MMP-13, alone, can markedly enhance the invasive capacity of malignant cells [ 22 ]. In transformed keratinocytes, p 53 plays an important role in suppressing MMP-13 expression [ 23 ]. Whereas some investigators have not found a significant relationship between MMP-13 expression and HNSCC behavior [ 17 ], we found significant MMP-13 expression in highly invasive tumors; this finding suggests that MMP-13 likely plays a role in regulating tumor invasion. In regard to location, cellular events at the tumor-stromal interface are thought to be more closely related to metastatic potential than events at the tumor's center [ 17 , 24 ]. Others have reported that MMP-13 is not only expressed by tumor cells in the invading periphery of most SCCs but also by stromal cells in a subset of tumors [ 15 ]. In the present study, we found no difference in levels of MMP-13 expression between tumor centers or margins. TIMP-1 is another protein that has been implicated in tumor growth. TIMP-1 mRNA has been detected in well-differentiated cancer cells, proliferated keratinocytes, and endothelial cells [ 2 , 14 ], and TIMP-1 overexpression has been observed in almost every case of HNSCC [ 24 ]. Whereas some investigators have reported that increased expression of TIMP-1 and TIMP-2 correlates with less aggressive tumors, others have reported the opposite finding [ 12 , 13 ]. In the present study, TIMP-1 expression was significantly increased in markedly invasive tumors, suggesting that TIMP-1 also plays a role in regulating tumor invasiveness. In some studies, no association between MMP-13 and TIMP-1 expression and any clinicopathological variables was found, or TIMP expression did not correlate with tumor growth [ 13 , 17 ]. Our results are not compatible with these studies. However, we did fail to identify any statistically significant relationship between the degree of staining for MMP-13 or TIMP-1 and the patient's age, sex, tumor site, or tumor histologic grade. Also, there was no statistically significant correlation between the degree of staining for MMP-13 and TIMP-1. This suggests that the balance between MMP and TIMP expression may not be as important to tumor invasion as their overexpression; that is, MMP and TIMP may play important separate roles in tumor invasion in which they act via different mechanisms. Because of the relatively short follow-up interval, we are unable to demonstrate whether MMP-13 and TIMP-1 expression is related to long-term survival. However, 75 of our 78 patients were free of disease after a mean follow-up interval of 20 months. As poor outcomes are the result of local recurrences and distant metastasis, factors that reflect the invasive and metastatic potential of SCCs, such as MMPs, could be helpful in predicting patient prognosis. Thus, further investigation of MMPs may not only clarify their role in tumor invasion but also may facilitate the development of new therapeutic approaches. Conclusion The results of this study suggest that MMP-13, which plays a central role in MMP activation, and the MMP inhibitor, TIMP-1, help regulate the invasiveness of HNSCCs. Although other methods [e.g., Western blots] are needed to confirm these findings, examination of MMP-13 and TIMP-1 expression in small biopsy samples might be useful in determining the invasive capacity of these tumors at an earlier stage. Competing interests None declared. Authors' contributions NC drafted and wrote the manuscript. KM and EC performed the surgery and follow-up of the patients. ED participated in the design and coordination of the study. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC512290.xml
516790	Immature rats show ovulatory defects similar to those in adult rats lacking prostaglandin and progesterone actions	Gonadotropin-primed immature rats (GPIR) constitute a widely used model for the study of ovulation. Although the equivalence between the ovulatory process in immature and adult rats is generally assumed, the morphological and functional characteristics of ovulation in immature rats have been scarcely considered. We describe herein the morphological aspects of the ovulatory process in GPIR and their response to classical ovulation inhibitors, such as the inhibitor of prostaglandin (PG) synthesis indomethacin (INDO) and a progesterone (P) receptor (PR) antagonist (RU486). Immature Wistar rats were primed with equine chorionic gonadotropin (eCG) at 21, 23 or 25 days of age, injected with human chorionic gonadotropin (hCG) 48 h later, and sacrificed 16 h after hCG treatment, to assess follicle rupture and ovulation. Surprisingly, GPIR showed age-related ovulatory defects close similar to those in adult rats lacking P and PG actions. Rats primed with eCG at 21 or 23 days of age showed abnormally ruptured corpora lutea in which the cumulus-oocyte complex (COC) was trapped or had been released to the ovarian interstitum, invading the ovarian stroma and blood and lymphatic vessels. Supplementation of immature rats with exogenous P and/or PG of the E series did not significantly inhibit abnormal follicle rupture. Otherwise, ovulatory defects were practically absent in rats primed with eCG at 25 days of age. GPIR treated with INDO showed the same ovulatory alterations than vehicle-treated ones, although affecting to a higher proportion of follicles. Blocking P actions with RU486 increased the number of COC trapped inside corpora lutea and decreased ovulation. The presence of ovulatory defects in GPIR, suggests that the capacity of the immature ovary to undergo the coordinate changes leading to effective ovulation is not fully established in Wistar rats primed with eCG before 25 days of age.	Introduction Ovulation, the release of mature oocytes from the ovary, requires proteolytic degradation of the follicle wall, as well as the overlying ovarian tissues. This happens through the expression of a series of critical genes, triggered in a precise temporal and spatial pattern by the preovulatory LH surge [ 1 , 2 ]. It is worthy to note that, for successful ovulation, follicle rupture has to occur just at the site of the follicle wall facing the ovarian surface, thus allowing release of the cumulus-oocyte complex (COC) to the periovarian space, while preventing proteolytic damage of the perifollicular tissues at the basolateral follicle sides. A large amount of information on the ovulatory process was accumulated during the last century (reviewed in [ 1 - 5 ]), and the involvement of crucial genes such as those encoding cyclooxygenase-2 (COX-2), and progesterone receptor (PR) has been clearly established. However, the mechanisms underlying the spatial targeting of the follicle rupture remain poorly understood. Although mechanical factors are likely involved in stigma formation and rupture [ 6 ], the mechanisms responsible for the specific location of proteolytic breakdown of the theca layers and perifollicular connective tissue at the apex of the follicle are not known. In recent studies [ 7 - 9 ] we have proposed that both prostaglandins (PG) and progesterone (P), classically recognized as essential ovulatory factors [ 1 , 2 ], play complementary roles in the spatial targeting of follicle rupture. This was supported by detailed morphological studies in cycling rats treated with indomethacin (INDO), a strong inhibitor of PG synthesis, and RU486 (a PR antagonist), showing antiovulatory effects [ 1 , 2 , 10 - 12 ]. Gonadotropin-primed immature rats (GPIR) constitute a useful model for the study of ovulation. The administration of a single dose of equine chorionic gonadotropin (eCG) to immature animals induces the growth of abundant follicles, that reach preovulatory size in two days. Ovulation is then triggered by a single dose of human chorionic gonadotropin (hCG), thus providing a large number of synchronized ovulatory follicles [ 13 - 25 ]. An additional advantage of this model is the absence of regressing corpora lutea of previous cycles. This is relevant because structural luteolysis, that is temporally coincident with ovulation in cycling rats, also involves tissue remodeling and proteolytic degradation of the extracellular matrix [ 5 ]. For these reasons, GPIR (ranging from 21 to 28 days of age, at the time of eCG treatment [ 13 - 25 ]), have been widely used in studies focused on the ovulatory process, and a large amount of the information in this topic is derived from studies in immature rats. However, it should be kept in mind that GPIR constitute a non-physiological model and the possible immaturity of the pathways leading to ovulation cannot be ruled out. In order to examine further the role of PG and P in follicle rupture and ovulation, we performed detailed morphological analysis of the ovulatory process in inmature rats primed with gonadotropins at different ages. Surprisingly, these animals showed age-related alterations of the follicle rupture similar to those of adult rats lacking PG and P actions. We report herein the morphological alterations of follicle rupture and ovulation in GPIR and the effects of treatment with P and/or PG of the E series, as well as the response of GPIR to PG synthesis inhibition with INDO and to a PR antagonist (RU486), whose inhibitory effects in ovulation are clearly established [ 1 , 2 , 8 , 13 , 14 ]. Materials and methods Animals and drugs Wistar female rats bred in the vivarium of the University of Cordoba were used. The day the litters were born was considered as day 0. Litter size was adjusted to 8 pups. The animals were maintained under controlled conditions of light (14L:10D; lights on 0500-1900) and temperature (22°C). The animals had free access to pelleted food and tap water. Experimental procedures were approved by the Cordoba University Ethical Committee for animal experimentation and were conducted in accordance with the European Union guidelines for care and use of experimental animals. Indomethacin (INDO), Progesterone (P), equine chorionic gonadotropin (eCG), human chorionic gonadotropin (hCG) and prostaglandins were purchased from Sigma.(St Louis, MO). The progesterone antagonist RU486 was obtained from Exlegin (Paris, France). Gonadotropin-priming at different ages Wistar immature rats were injected sc with 10 IU of eCG at 1700 h at 21, 23 or 25 days of age, and 48 h later were injected sc with 10 IU of hCG, a frequently used schedule [ 23 , 26 , 27 ]. Body weights were in the recommended range [ 23 , 26 ] (46.0 ± 0.34, 57.0 ± 1.30 and 61.8 ± 1.4 g, mean ± SEM for n = 5). The animals (5 per group) were sacrificed at 0900 h on the day following hCG injection (i.e. at 24, 26 and 28 days of age). The ovaries, including the ovarian bursa, oviducts and periovarian fat pad, were fixed in Bouin-Hollande's fluid for 24 h and processed for paraffin embedding. Gonadotropin-primed rats treated with prostaglandins and/or progesterone Immature rats were primed with eCG at 21 days of age and with hCG 48 h later as described above. In addition, these animals were injected with 100 μg of PGE1 or PGE2, 1 mg of P or 1 mg of P plus 100 μg of PG E1 or PGE2 or vehicles (70% ethanol in saline) at 1730 h (30 min after hCG treatment). These dosages were effective, in the same body weight basis, in previous studies in adult rats [ 9 ]. The animals (5 per group) were killed at 0900 h on the day after hCG treatment (i.e. at 24 days of age) and the ovaries were processed as described above. Gonadotropin.primed rats treated with indomethacin or RU486 Immature rats were primed with eCG at 23 days of age and with hCG 48 h later as described in previous experiments. Indomethacin-treated rats received a sc injection of 0.5 mg of INDO or vehicle (olive oil) at 1200 h at 25 days of age. RU486-treated rats were injected sc with 0.5 mg of RU486 or vehicle (olive oil) at 0900 h at 24 and 25 days of age.The animals (5 per group) were killed at 0900 h on the day following hCG injection (i.e. at 26 days of age) and the ovaries were processed as described above. Histological analysis of follicle rupture and ovulation The right ovaries were serially sectioned (6 μm) and stained with hematoxylin and eosin. All sections were examined under the microscope. The total number of cumulus-oocyte complexes (COCs) per ovary were counted. The number of COCs trapped inside the corpus luteum, released to the ovarian interstitium, retained in the bursal cavity or found in the oviducts, was recorded. COCs released to the ovarian interstitium were clearly recognizable by the presence of the oocyte in metaphase II and the first polar body, dispersed cumulus and follicular fluid. The total number of COCs per ovary matches the total number of corpora lutea (unruptured or not). In addition to absolute values, the number of COCs in each location was expressed as the percentage with respect to the total number of COCs (or corpora lutea) per ovary, to assess ovulatory efficiency, avoiding variability in the total numbers of corpora lutea among the different groups. Unruptured follicles measuring less than 575 μm in diameter, showing signs of atresia, such as irregularities of the granulosa cell layer, presence of apoptotic granulosa cells, as well as lack of dispersion of the cumulus or resumption of meiosis, were not considered. Statistical analysis was performed by ANOVA followed by the Student-Newman-Keuls method for multiple comparison among means. Significance was considered at the 0.05 level. Results Ovulatory process in immature rats primed at different ages Absolute numbers of COCs, that match the numbers of corpora lutea per ovary, are presented in Table 1 , whereas relative data (the proportion of COCs found in each specific location) are shown in Fig. 6 . On the day following hCG treatment, the total number of COCs per ovary was equivalent for the different age-groups (Table 1 ). Immature rats primed with eCG at 21 or 23 days of age showed frequent ovulatory defects (see Figs 1 , 2 , 3 , 4 , 5 for representative micrographs and Fig. 6 . for quantitative data). The COC was trapped in about 60% and 40% of the luteinized follicles in rats primed with eCG at 21 and 23 days of age respectively (Figs. 1A , 2 , 6 ). However, many luteinized follicles containing the COC showed rupture of the theca layers and release of follicular fluid to the ovarian interstitium (Fig. 1A ). Furthermore, 6–10% of the COCs were released to the ovarian interstitum (Figs. 1B , 2 , 3 , 6 ). In these cases, degradation of the ovarian stroma with invasion of the blood and lymphatic vessels and formation of emboli containing the COC and follicular fluid (Figs 1C,1D,1E,1F , 2 , 3 ) were observed. In some animals, massive embolism of the ovarian vein at the ovarian hilus with follicular fluid and COCs (Fig. 3 ) caused ovarian hyperemia and vascular congestion. In addition, proteolytic degradation of the ovarian bursa by follicular fluid and granulosa cells, with invasion of the periovarian fat pad (Fig. 1G,1H ) and even release of the COC to the peritoneal cavity (Fig. 4 ), was also observed. A characteristic feature of some corpora lutea ruptured at the apex, was the presence of an unruptured ovarian surface epithelium, and the COC was trapped in lacunae of follicular fluid and blood in the tunica albuginea (Fig. 5 ). Variable numbers of COCs (from 0 to 25%) were located in the bursal cavity (Fig. 6 ). From 33% to 60% of COCs were found in the oviducts, in rats primed with eCG at 21 and 23 days of age respectively. Otherwise, in immature rats primed with eCG at 25 days of age, the vast majority of COCs were found in the oviducts and abnormally ruptured corpora lutea (with trapped COC) were only occasionally found. Table 1 Number of COCs per ovary in control rats primed with eCG at different ages. Age (days) at eCG treatment Total COCs COCs located in Corpus luteum Interstitium Ovarian Bursa Oviduct 21 42.5 ± 8.72 22.5 ± 4.32 3.17 ± 1.47 0.67 ± 0.42 16.2 ± 6.93 23 58.4 ± 4.74 18.8 ± 0.70 3.20 ± 1.74 2.40 ± 0.51 34.0 ± 6.54 25 38.6 ± 3.76 0.6 ± 0.60 a,b -- -- 37.4 ± 3.20 Values are the mean ± SEM for n = 5. ANOVA and Student-Newman-Keuls multiple range test. a p < 0.05 vs 21 days. b p < 0.05 vs 23 days Figure 1 Representative micrographs from the ovary of immature rats primed with eCG at 21 ( A-D , G, H ) or 23 ( E, F ) days of age, stained with hematoxylin and eosin. A , luteinized follicle showing trapped COC and release of follicular fluid ( FF ) to the ovarian interstitium. The rupture of the theca layers are indicated by arrows . B , COC released to the ovarian interstitium, in a lacunae of follicular fluid ( FF ). Clusters of granulosa cells are indicated by open arrows . C, D , COCs in the lymphatic ( A ) or blood ( D ) vessels at the ovarian hilus ( OH ). E, F , COC inside a blood vessel located in the periovarian fat pad ( PFP ) near the ovarian hilus ( OH ). The framed area is shown at higher magnification in F showing rupture of the blood vessel and escape of red blood cells ( arrows ). G, H , Non-consecutive serial sections showing a COC released to the bursal cavity ( BC ), adhered to the ovarian bursa. Degradation of the ovarian bursa by follicular fluid and granulosa cells ( open arrows ) and invasion ( arrows ) of the periovarian fat pad ( PFP ) can be observed. Figure 2 Micrograph of the ovary of a rat primed with eCG at 21 days of age, stained with hematoxylin and eosin. Several COCs trapped inside corpora lutea, showing dispersion of the cumulus, and interstitial COCs inside a blood vessel. Figure 3 Non-consecutive serial sections of the ovary of a rat primed with eCG at 21 days of age and stained with hematoxylin and eosin. The ovarian vein is embolized with folicular fluid and two COCs (in A and B ). A COC released to the ovarian hilus can be also observed in A . The ovarian surface is indicated by empty arrows in the inset . Figure 4 Micrographs from non-consecutive serial sections from the ovary of an immature rat primed with eCG at 21 days of age. The ovarian bursa ( OB ) has been degraded ( A ) allowing release of the COC to the peritoneal cavity ( B ) Clusters of granulosa cells ( arrows ) can be observed free or attached to the ovarian bursa. Hematoxylin and eosin. Figure 5 Representative micrographs from the ovary of immature rats primed with eCG at 21 ( A ) or 23 ( B ) days of age, showing luteinized follicles ruptured at the apex. The ovarian surface epithelium ( OSE, arrows in A and B ), and its basement membrane ( arrowheads in B ) remain intact, whereas the underlying ovarian tissue has been degraded ( asterisk in A ). The COC is retained under the ovarian surface. OB , ovarian bursa. Hematoxylin and eosin. Figure 6 Percentage of cumulus-oocyte complexes (COCs) trapped inside the corpus luteum, released to the interstitium, located in the bursal cavity, or in the oviducts, in immature rats primed with eCG at 21, 23 or 25 days of age and ovulating at 24, 26 or 28 days of age respectively. See Table 1 for absolute counts. Different superscripts mean significant (p < 0.05) differences. Effects of P and/or PGE treatment The total number of COCs per ovary was equivalent in rats primed with eCG at 21 days of age and treated with vehicles, P, PGE1, PGE2, P plus PGE1 or P plus PGE2 (Table 2 ). Neither P, PG or combined P plus PG treatment, prevented ovulatory defects and these animals showed the same morphological alterations as gonadotropin and vehicle-treated rats. However, degradation of the interstitial tissue and embolism of blood vessels with follicular fluid and granulosa cells were very scarce in prostaglandin-treated animals. Although the number of effectively ovulated oocytes were apparently increased in prostaglandin-treated rats, differences were not large enough to be statistically significant (Fig. 7 ). Table 2 Number of COCs per ovary in rats primed with eCG at 21 days of ages. Treatment Total COCs COCs located in Corpus luteum Interstitium Ovarian Bursa Oviduct Vehicles 35.4 ± 3.52 17.2 ± 1.47 2.6 ± 0.85 0.6 ± 0.24 15.0 ± 2.72 PGE 1 33.2 ± 4.20 11.0 ± 2.42 2.3 ± 1.02 1.2 ± 0.58 19.0 ± 3.40 PGE 2 33.2 ± 1.62 10.2 ± 2.80 0.6 ± 0.39 3.6 ± 2.59 18.8 ± 3.38 P 4 45.6 ± 7.77 17.2 ± 2.97 3.4 ± 1.06 5.8 ± 3.10 19.2 ± 5.45 P 4 + PGE 1 34.6 ± 5.17 14.6 ± 3.37 0.6 ± 0.39 0.8 ± 0.37 18.6 ± 2.82 P 4 + PGE 2 35.2 ± 6.85 15.6 ± 3.10 0.6 ± 0.39 5.7 ± 2.85 14.2 ± 1.63 Values are the mean ± SEM for n= 5. Figure 7 Percentage of cumulus-oocyte complexes (COCs) trapped inside the corpus luteum, released to the interstitium, located in the bursal cavity, or in the oviducts, in rats primed with e CG at 21 days of age and injected two days later with vehicles (Veh), prostaglandin E1 (PGE1), prostaglandin E2 (PGE2), progesterone (P) or progesterone plus PGE1 or PGE2. See Table 2 for absolute counts. No significant differences were found. Effects of INDO or RU486 treatments The total number of COCs per ovary was equivalent in rats treated with vehicle, INDO or RU486 (Table 3 ). Immature rats treated with INDO during the preovulatory period showed the same morphological alterations of the ovulatory process as gonadotropin and vehicle-treated rats, although a higher proportion of follicles were affected (Fig. 8 ). A significantly (p < 0.05) higher number of COCs were released to the ovarian interstitium, whereas the number of COCs found in the oviducts was significantly (p < 0.05) decreased. Rats treated with RU486 showed increased numbers of COCs retained inside the follicle (Fig. 8 ), whereas the number of COCs released to the interstitium or effectively ovulated were significantly decreased. Most of the luteinized follicles in which the COC was trapped, did not show evident rupture of the theca layers. The numbers of COC trapped in the bursal cavity was significantly increased (Table 3 and Fig. 8 ). Table 3 Number of COCs per ovary in rats primed with eCG at 23 days of ages. Treatment Total COCs COCs located in Corpus luteum Interstitium Ovarian Bursa Oviduct Vehicle 39.6 ± 1.60 12.8 ± 1.0 2.6 ± 1.73 1.0 ± 0.62 23.2 ± 1.53 INDO 34.4 ± 3.76 19.2 ± 2.55 10.0 ± 1.35 a 0.4 ± 0.39 5.0 ± 0.05 a RU486 41.8 ± 3.86 22.8 ± 4.80 0.6 ± 0.39 b 3.2 ± 0.57 a,b 15.2 ± 3.10 a,b Values are the mean ± SEM for n = 5. ANOVA and Student-Newman-Keuls multiple range test. a p < 0.05 vs vehicle, b p < 0.05 vs INDO. Figure 8 Percentage of cumulus-oocyte complexes (COCs) trapped inside the corpus luteum, released to the interstitium, located in the bursal cavity, or in the oviducts, in rats primed with eCG at 23 days of age and treated with indomethacin (INDO) or RU486. see Table 3 for absolute counts. Different superscripts mean significant (p < 0.05) differences. Discussion Gonadotropin-primed immature rats (GPIR) is a popular model for the study of ovulation. In this model, eCG stimulates the growth of a large cohort of follicles, that reach preovulatory size in about 48 h and are then induced to ovulate by hCG [ 1 , 2 , 28 ]. Early data indicate that the ovulation rate in immature rats was age-dependent, and that the number of ova found in the oviducts increases, although not linearly, with age [ 29 , 30 ]. However, these early studies do not provide information on the number of ova that were not effectively ovulated, and the marked variability in the rate of ovulation at different ages could be due to differences in the number of recruited growing follicles, in the responsiveness of preovulatory follicles to the ovulatory stimulus or both. In the present study, we evaluated the ovulatory process in immature rats primed with gonadotropins during the 21–25 day-old period, commonly used in studies focused on ovulation [ 13 , 17 , 18 , 20 - 23 , 25 , 26 ]. In agreement with previous data [ 29 , 30 ], the proportion of COCs found in the oviducts (i.e. effectively ovulated) increases in parallel to age. Detailed morphological evaluation indicated that ovulation was unsuccessful in 30–60% of preovulatory follicles in rats primed with eCG before 25 days of age. The nearly normal ovulatory process in immature rats primed with eCG at 25 days of age indicates that ovulatory defects were not due to gonadotropin treatment itself, but to ovarian immaturity at earlier ages. The ovulation failure described herein (in rats primed with eCG at 21 or 23 days of age) was not due to decreased recruitment of growing follicles, as indicated by the equivalent numbers of follicles that reach preovulatory size at all ages tested. The data of this study strongly suggest that the relative ovulatory incompetence of young immature rats was due to a defective capacity of preovulatory follicles to undergo the coordinate network of interactions that leads, in reponse to hCG, to COC release to the periovarian space. Defective ovulation in GPIR was related to abnormal follicle rupture. Whereas some LH-driven morphological changes in preovulatory follicles, such as cumulus expansion, resumption of the meiotic process and initial luteinization were present and showed normal features in GPIR, aberrant follicle ruptures were frequently observed. This was indicated by the presence of COCs released to the ovarian interstitium, and of corpora lutea showing rupture of the theca layers at any site of the follicle wall with release of follicular fluid to the ovarian stroma. The presence of follicle ruptures at the basolateral follicle sides suggests that the mechanisms underlying the spatial targeting of follicle rupture at the apex are not fully established in immature rats primed with eCG before 25 days of age. Almost identical ovulatory defects have been previously reported in INDO-treated adult cycling rats that also show abnormally ruptured follicles and proteolytic degradation of perifollicular tissues [ 7 - 9 ]. Furthermore, INDO-treated GPIR in the present study showed similar morphological alterations of the ovulatory process as vehicle-treated GPIR, even though the drug increased the number of affected follicles. The similarity of the ovulatory alterations found in GPIR and INDO-treated adult rats raises the question of whether the COX-2-prostaglandin pathway is fully established or not in GPIR. In this context, treatment of GPIR with PG of the E series was carried out in order to analyse whether ovulatory defects in GPIR were due to defective PG synthesis. Treatment with prostaglandin E resulted in a decrease in the invasion of the ovarian stroma and blood vessels by follicular fluid and granulosa cells, but ovulation was not improved significantly. In contrast, supplementation with exogenous PG of the E series inhibits abnormal follilce rupture and restores ovulation in INDO-treated adult rats [ 2 , 8 , 31 ]. This suggests that the possible immaturity of the COX-2-prostaglandin pathway in GPIR would be beyond prostaglandin synthesis, in agreement with previous studies reporting normal PG generation in GPIR in response to hCG treatment [ 28 ]. Otherwise, recent studies in COX-2 defective mice [ 32 ] indicate that the COX-2-PG pathway is not fully established in immature rodents. It has been clearly established that activation of PR plays a crucial role in ovulation [ 10 , 11 , 33 - 37 ]. Accordingly, adult cycling rats treated with PR antagonists [ 11 ] or inhibitors of the P synthesis [ 34 ] during the preovulatory period showed almost complete inhibition of follicular rupture [ 9 , 11 , 34 ]. Noteworthy, the response of GPIR to PR blockage (showing incomplete inhibition of follicle rupture) was similar to that of adult rats lacking both P and PG actions. In this sense, combined treatment of adult rats with RU486 and INDO resulted in follicle rupture in about 25% of preovulatory follicles [ 9 ]. Furthermore, some morphological features of the ovulatory process in GPIR, such as the persistence of the ovarian surface epithelium in spite of the degradation of the underlying tissues, was also a characteristic feature of adult rats treated with both INDO and RU486 [ 9 ]. However, exogenous P supplementation did not restore ovulation in GPIR, that suggest a defective response of preovulatory follicles to the ovulatory LH-dependent P surge. RU486-treated GPIR showed a significant increase in the number of COCs retained into the bursal cavity. Alterations in the transport of COCs to the oviduct has been previously described in adult rats treated with a PR antagonist [ 11 ], although the existence of a role for P in the COC transport to the oviduct cannot be ascertained from the present data. The end-point of the complex interaction network leading to follicle rupture and COC release, is the expression/activation of proteolytic enzymes, reponsible for extracellular matrix breakdown [ 24 , 38 , 39 ]. Proteolytic activity has to be closely modulated by protease inhibitors to prevent damage to perifollicular tissues. However, the characterization and regulation of ovarian proteolytic inhibitors has not been completely elucidated. The presence in GPIR of degradation of the ovarian stroma, invasion of blood and lymphatic vessels in abnormally ruptured follicles, strongly suggest that defective ovulation in GPIR was not due to decreased proteolytic activity but due to untargeted proteolysis. This was particularly suggested by the degradation of the ovarian bursa and invasion of the periovarian fat pad by cumulus cells and follicular fluid released at the follicle apex (see Figs. 1G,1H , 3 and 4 ), permitting, in some cases, escape of the COC to the peritoneal cavity. In this sense, the use of very young immature rats to characterize proteolytic activity regulation during ovulation could provide conflicting data. In this study, we used Wistar rats. The possible existence of slight differences in the age of ovarian maturation among different rat strains cannot be ruled out. In conclusion, the presence of aberrant ovulations in very young GPIR, closely resembling those in adult rats lacking P and PG actions, strongly suggests that the capacity of the immature ovary to undergo changes leading to effective ovulation is not fully established in Wistar rats primed with eCG before 25 days of age. This should be taken into account when using the GPIR model for studies focused on ovulation.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC516790.xml
512284	Discovery of induced point mutations in maize genes by TILLING	Background Going from a gene sequence to its function in the context of a whole organism requires a strategy for targeting mutations, referred to as reverse genetics. Reverse genetics is highly desirable in the modern genomics era; however, the most powerful methods are generally restricted to a few model organisms. Previously, we introduced a reverse-genetic strategy with the potential for general applicability to organisms that lack well-developed genetic tools. Our TILLING (Targeting Induced Local Lesions IN Genomes) method uses chemical mutagenesis followed by screening for single-base changes to discover induced mutations that alter protein function. TILLING was shown to be an effective reverse genetic strategy by the establishment of a high-throughput TILLING facility and the delivery of thousands of point mutations in hundreds of Arabidopsis genes to members of the plant biology community. Results We demonstrate that high-throughput TILLING is applicable to maize, an important crop plant with a large genome but with limited reverse-genetic resources currently available. We screened pools of DNA samples for mutations in 1-kb segments from 11 different genes, obtaining 17 independent induced mutations from a population of 750 pollen-mutagenized maize plants. One of the genes targeted was the DMT102 chromomethylase gene, for which we obtained an allelic series of three missense mutations that are predicted to be strongly deleterious. Conclusions Our findings indicate that TILLING is a broadly applicable and efficient reverse-genetic strategy. We are establishing a public TILLING service for maize modeled on the existing Arabidopsis TILLING Project.	Background Rapid progress being made in genome sequencing projects provides raw material for the potential understanding of gene function, so effective reverse genetic strategies are increasingly in demand [ 1 ]. Sequence information alone may be sufficient to consider a gene to be of interest, because sequence comparison tools that detect protein sequence similarity to previously studied genes often allow a related function to be inferred. Hypotheses concerning gene function that are generated in this way must be confirmed empirically. Experimental determination of gene function is desirable in other situations as well, for example, when a genetic interval has been associated with a phenotype of interest. In such cases, the functions of genes in an interval can be deduced from the phenotypes of induced mutations. Furthermore, the dissection of gene interactions often requires the availability of a range of allele types. However, most available methods for inferring function rely on techniques that produce a limited range of mutations, are labor-intensive or unreliable, or are limited to species in which special genetic tools have been developed [ 2 ]. Just as the discovery of induced mutations led to forward genetics, the introduction of rapid reverse genetic methods can have great impact. Several general strategies have been used to obtain reduction-of-function or knockout mutations in model organisms, including insertional mutagenesis [ 3 ] and RNA suppression [ 4 ], which have been widely used in plants. Insertional mutagenesis is now largely an in silico procedure for Arabidopsis researchers, as searchable databases of flanking sequences from T-DNA and transposon insertions are available on-line [ 5 ]. RNA suppression currently requires considerable manual effort, but it has the potential of reducing expression of repeated genes, which are especially common in plants, including Arabidopsis. However, because these techniques rely either on Agrobacterium T-DNA vectors for transmission or on endogenous tagging systems, their usefulness as general reverse genetics methods is limited to very few plant species. In maize, the majority of reverse genetic resources have exploited the endogenous Mutator (Mu) and Activator (Ac) transposon families. Mutator transposable elements are frequently present in high copy number and tend to insert in or near genes [ 6 ] thus providing mostly gene knock-out and potential loss-of-function alleles due to insertions in regulatory elements. Resources to identify Mu insertions in target genes include the Trait Utility System for Corn [ 7 ], the Maize Targeted Mutagenesis project [ 8 ], and the Photosynthetic Mutant Hunt [ 9 ]. Additionally, a modified Mutator element has been engineered to allow plasmid rescue of the element and flanking genomic DNA into E. coli [ 10 , 11 ] facilitating the sequencing of tagged genes and allowing the generation of an insertional database analogous to the T-DNA database currently available for Arabidopsis. Activator transposons are low copy number elements that, like Mu, preferentially insert in or near genes [ 12 ]. Large-scale isolation and sequencing of Ac elements for a reverse genetics database is advantageous, because single to low copy number insertions can be obtained. Reverse genetic strategies based on induced mutations have the potential for general applicability. Two such methods have been described for plants. One is deletional mutagenesis using fast neutron bombardment, which appears to be an effective means of knocking out tandemly repeated genes [ 13 ]. Another is TILLING (Targeting Induced Local Lesions IN Genomes), in which treatment using traditional chemical mutagens causes point mutations that are then discovered in genes of interest using a sensitive method for single-nucleotide mutation detection [ 14 ]. TILLING can provide an allelic series that includes missense and knockout mutations. The utility of allelic series that has been demonstrated in traditional forward genetic studies makes TILLING an especially desirable reverse-genetic strategy as genomic sequences become increasingly available. We have introduced a high-throughput screening method for TILLING based on the use of a mismatch-cleavage endonuclease, followed by fluorescent display of cleaved products on polyacrylamide electrophoretic gels using the LI-COR analyzer system [ 15 ]. We systematized this method and established a public TILLING facility for the general Arabidopsis community [ 16 ]. Our Arabidopsis TILLING Project (ATP) screened on average ~3000 ethylmethanesulfonate (EMS)-mutagenized M2 Arabidopsis plants per 1-kb gene fragment, which resulted in the delivery of >4000 mutations in >400 genes. Analysis of the TILLING data revealed that EMS is a nearly ideal mutagen, producing G/C-to-A/T transitions >99% of the time with only minor local sequence biases [ 17 ]. Our analysis also indicated that the high-throughput cleavage-based detection method is highly efficient: at least 3/4 of all mutations present were detected and essentially all mutations detected were confirmed by sequencing. ATP demonstrated the practicality of high-throughput TILLING in a production setting. However, the small genome size of Arabidopsis and the ease with which it can be cultured might have made Arabidopsis easier to TILL than a field-grown crop plant with a large genome. To determine whether the procedures that were developed for ATP can be generalized, we chose to TILL maize, a crop plant with a genome that is ~20 times larger than Arabidopsis. We find that essentially the same procedures that provided efficient TILLING of Arabidopsis can be applied to maize, yielding comparable results. Despite the relatively small size of our mutagenized maize TILLING population, we obtained useful mutations. These include a promising allelic series for a chromomethylase gene that had been previously implicated in non-CpG DNA methylation, whose counterpart in Arabidopsis is responsible for epigenetic gene silencing and genome surveillance. Results TILLING consists of a series of steps, beginning with chemical mutagenesis of reference individuals and culminating in the determination of mutant base pairs by DNA sequencing [ 14 ]. High-throughput TILLING utilizes the CEL I mismatch-cleavage enzyme on heteroduplexes with detection of end-labeled cleavage products on electrophoretic gels [ 15 , 18 ]. The procedure for maize TILLING is identical to that for Arabidopsis, except that pollen rather than seed was treated with EMS (Figure 1 ). For this study, two separate mutagenized B73 maize populations, designated UI (M2 families from 384 mutagenized lines) and NS (366 individual M1 plants), were screened in 4-fold pools (4 UI families or 4 NS plants) in a 96-well format. To compensate for the larger genome size of maize relative to Arabidopsis, we increased the amount of genomic template DNA amplified 20-fold; otherwise, all protocols and default parameters were the same as used for Arabidopsis TILLING [ 16 ]. We found that 11 of the 14 primers gave high quality products, compared to ~90% success that is obtained for Arabidopsis. We proceeded to screen for mutations within these 11 gene fragments in the 750 DNA samples and discovered 21 point lesions (Table 1 ). All 21 were verified by sequencing. Of the 21 lesions, 17 appeared to be EMS-induced. These 17 mutations were G/C-to-A/T mutations, as expected for EMS [ 17 ]. The other four lesions were found in a single plant (Table 1 ) and only one was a G/C-to-A/T transition, suggesting that this plant is a non-B73 contaminant, likely due to cross-pollination. Contaminations seen as an excessively high frequency of polymorphisms in single plants have been occasionally observed by ATP [ 17 , 19 ]. The presumed EMS-induced mutations were detected in single plants, except for DMT102 G878A, where the exact same mutation was found in two different plants; this circumstance is expected and observed to occur 4% of the time in Arabidopsis based on random distribution of induced mutations, GC content of the genome and distribution of location of mutations discovered in fragments [ 17 ]. Finding one coincidence among 17 mutations is not significantly different from expectation. Furthermore, finding that all 17 mutations are G/C-to-A/T transitions effectively rules out the possibility that they are naturally occurring polymorphisms: with four possible single-base changes, the chance probability of observing that all 17 conform to the expectation for EMS mutagenesis is only (1/4) 17 or ~1/10 10 . Importantly, each population yielded 8–9 confirmed new mutations, for an overall mutation density of approximately two mutations/megabase. Taking into account the fact that pollen treatment mutagenizes only one of two genomes, whereas seed treatment mutagenizes both (although by screening M2s in the latter case, 1/4 of the mutations are lost as +/+ segregants), the estimated mutation density for both maize populations is ~3/4 as high as our average for Arabidopsis per mutagenized genome. We have previously demonstrated reliable detection of mutations in 8-fold pools based on analysis of ATP-generated data [ 17 ]. For example, we obtained almost precisely the expected 2:1 heterozygote:homozygote ratio for ~1900 mutations in 8-fold pools, indicating that detection of 1/16 is no different from detection of 1/8 by TILLING. To confirm this detection efficiency in maize, we screened the primer sets for DMT101 and DMT103 with 8-fold pools from the 750 B73 DNA samples that had already been screened in pools of four. In this test, we detected only three of the five mutations that we had discovered by 4-fold pooling. Inadequate data quality does not account for missing these two base changes (UI20291 and NS3471.9) in 8-fold pools, because the gel images were typical of what is seen in our ATP operation. Therefore, we considered the possibility that failure to detect two of five mutations in this limited test was caused by variation of DNA amounts in the pools. One potential source of variability in DNA amount is that degraded DNA is difficult to measure accurately when visualized by agarose gel analysis. We noticed that some of the genomic DNA samples, including NS3471.9, were partially degraded. Inaccuracies in measuring the amount of DNA in a sample will compromise normalization in pools: any variation in the amount of DNA contributed by each plant in the pool will lead to reduced representation of one or more plants. As the amount of DNA in a pool from a particular plant decreases, mutation detection becomes limiting. Recognizing that the quality of genomic DNA could potentially hinder the throughput of mutation detection, we sought an alternative method of sample quantification and normalization. We have found that running samples on 3% Metaphor ® (Cambrex) agarose gels reduced "smearing" of fragments, thus facilitating quantification. While lower DNA quality and inaccurate normalization could account for missing the base change in NS3471.9, this is not a likely explanation for missing the mutation in UI20291, which was from an apparently undegraded sample. Therefore, we considered that another source of sample-to-sample variation arose from the sampling of leaves from M2 sibling plants in the UI series rather than using M1 plants directly. Individual DNA samples from the UI population were generated by pooling approximately 10 individuals from an M2 family. M1 plants that are heterozygous for a new mutation will yield M2 families segregating 1:2:1 for the new mutation. A sample over a large M2 family should yield DNA that is equal parts wild type and mutant allele, assuming good viability and fertility of the mutant allele. However, a small M2 sample may be biased and contain too large a proportion of wild-type DNA which would then prevent a given target sequence containing a mutation from being detected among 8-fold pooled DNAs, depending on the limit for robust mutation detection. Direct evidence that sampling from M2 families was a problem came from our finding that two of the mutations found in plants sampled as leaves from multiple M2 plants were scored as homozygous by our usual criteria. Pollen mutagenesis can only produce M1 heterozygotes, and we interpret the homogeneity of the mutant as resulting from limited sampling of heterogeneous M2 plants. For example, if only one or two of the planted M2 seed germinated for these families, and the limited sample favored homozygotes, then they would sometimes appear to be purely homozygous in the sequence trace. Similarly, an equal or greater number of mutations will be underrepresented because wild type will be in excess, leading us to miss detecting the mutation in 8-fold, but not in 4-fold pools. This underscores the importance of using DNA from M1 pollen-mutagenized individuals and taking care to avoid collection procedures that could exacerbate degradation during DNA isolation. By assaying DNA concentrations on Metaphor gels and sampling only the M1 generation, we should be able to pool 8-fold without reducting detection efficiency. To test this, we screened a population of 768 M1 W22 maize DNAs (pollen mutagenesis) with similar degradation patterns as the B73 samples described above and normalized using 3% Metaphor agarose gels. We then made 8-fold and 4-fold pools from these 768 samples and screened with four of the 11 primer sets in the original screen. This screening of 6-Mb of total sequence led to the independent detection of the same 3 mutations in both the 4-fold and the 8-pools (data not shown). Therefore, we conclude that even partially degraded DNA from M1 pollen-mutagenized samples, such as might be extracted from material collected in the field, can be used for TILLING. The 17 induced mutations discovered in the screen were distributed as expected, consisting of 10 missense, 7 silent and no truncation mutations, compared with 51% missense, 44% silent and 5% truncation mutations based on ~4000 TILLed Arabidopsis mutations. Considering that about half of missense mutations are expected to be damaging to a typical protein [ 20 ], we expect that even a small allelic series will be useful for phenotypic analysis. Indeed, we discovered that all three different DMT102 missense mutations are likely to be deleterious to the protein, based on SIFT and PSSM Difference scores (Figure 2 ). The SIFT algorithm predicts deleterious missense mutations with ~75% overall accuracy based on analysis of experimental mutagenesis data [ 21 , 22 ] and comprehensive human polymorphism and disease data [ 23 ]. Therefore, the DMT102 allelic series appears to be essentially complete after screening a 1-kb region within only 750 maize plants. Discussion We have shown that TILLING is an efficient method for reverse genetics in a crop plant. The density of mutations that we discovered appears to be only slightly lower than what is obtained for Arabidopsis using the same methodology. For Arabidopsis, we currently screen ~2300 M2 plants to obtain a suitable allelic series, which averages ~12 mutations per 1.5-kb segment screened. Based on this work, we estimate that screening ~4000 maize plants will provide a comparable series ([4000] [17 mutations] [1.5-kb/1-kb]÷([11-kb] [750 plants]) = ~12 mutations). Since this study was completed, we have expanded the size of our mutagenized population to ~2500 M1 individuals in the B73 genetic background and ~2000 M1 plants in the W22 genetic background that are ready for screening, and populations of ~10,000 M1 plants are being prepared. For effective pooling, each individual in a pool must be represented at a concentration that is equivalent to the other members in a pool. Failure to accomplish this could result in a mutation represented in the pool below the level of detection. The goal is to maximize throughput by increasing pooling while still detecting all possible mutations. We have shown that in Arabidopsis, heterozygous mutations can be as efficiently discovered as homozygous mutations in 8-fold pools, thus providing a minimum estimate of robust discovery in a production setting of 1 in 16 [ 16 ]. We have described here two possible sources of error that might hinder the ability to pool samples effectively: inaccurate DNA quantitation, and sampling error in tissue collection. DNA quantitation using gel electrophoresis is difficult when samples are degraded, as the standard band of DNA appears as a smear, although the problem can be minimized using high percentage agarose gels. Sampling bias in the collection of individuals descended from a single mutagenized parent can also lead to non-equivalent representation of a mutation in a pool. The present study revealed normalization inaccuracy or sampling bias by detecting homozygous mutations in lines that should have yielded only heterozygous mutations. To minimize sampling bias, only DNA from M1 individuals will be used to create our library for a maize TILLING service. At least one of the maize genes that we screened yielded an excellent allelic series. We discovered three missense mutations in DMT102, all of which are predicted to damage the protein based on sequence conservation. This analysis used two different programs: SIFT ( S orting I ntolerant F rom T olerant [ 23 ], which uses PSI-BLAST alignments, and PARSESNP P roject A ligned R elated S equences and E valuate SNPs [ 24 ], which provides a PSSM ( P osition- S pecific S coring M atrix) Difference score based on alignment blocks (Figure 2 ). DMT102 is a member of the chromodomain-containing "chromomethylase" subfamily of cytosine-5-DNA methyltransferases. The Arabidopsis CMT3 chromomethylase is the first example of a gene to be TILLed [ 14 ], and a nonsense mutation was responsible for sharply reducing CpNpG methylation [ 25 ]. This had confirmed a study in which a Mutator insertional mutation into maize DMT102 was shown to reduce CpNpG methylation [ 26 ]. Plant chromomethylases have received considerable recent attention. For example, studies of other mutations affecting CpNpG methylation reveal the first links between DNA methylation, histone methylation [ 27 ] and the small interfering RNA (siRNA) machinery [ 28 ] in a higher eukaryote. A methylation profiling study has revealed that transposons are in vivo targets of CMT3-dependent methylation [ 29 ]. Together with the Mutator insertional mutation [ 26 ], our TILLed DMT102 allelic series may now be applied to understanding the relationship between DNA methylation, chromatin structure, siRNAs and transposon biology in maize. Conclusions TILLING has several advantages as a general reverse-genetic tool, especially for organisms for which other options are limited. The high density of mutations resulting from chemical mutagenesis means that, relative to insertional or deletional mutagenesis, far fewer plants are required for screening and much smaller genes can be effectively targeted. EMS is a stable and reliable mutagen, whereas the stability, penetrance and accuracy of RNAi-based silencing is uncertain [ 30 , 31 ], and insertional mutagenesis can cause chromosomal rearrangements that complicate subsequent phenotypic analysis [ 32 ]. TILLING provides an allelic series of mutations, and is the only method that can focus the search for missense mutations to just part of a protein, such as in a single domain of a multidomain protein. TILLING lines can be produced in a homogeneous wild-type genetic background, which avoids problems of heterogeneity often required for insertional mutagenesis, especially in maize. Finally, given the high regulatory and intellectual property costs associated with transgenics and the current concerns about genetically modified crop plants, there is likely to be agricultural interest in producing phenotypic variants without introducing foreign DNA of any type into a plant's genome. We are currently establishing the reference population necessary to provide TILLING as a service to the maize community. Methods Maize mutagenesis, culture and DNA preparation B73 pollen was mutagenized with EMS and applied to the silks of B73 ears [ 33 ]. Ears were harvested at 5–6 weeks post pollination. For the NS population, M1 seed were planted and a whole young leaf harvested from each plant and lyophilized for DNA sampling. For the UI population, the M1 were selfed to make M2 seed, then families of twenty M2 siblings were planted and a total of 60 leaf discs were punched from the youngest leaves using all members of the family, and the pooled sample for an entire family lyophilized. Samples were prepared from lyophilized leaf tissue essentially as described [ 34 ], except that dried tissue was homogenized into a powder in a FastPrep homogenizer before adding buffer, and 20 mg of this powder was used to prepare DNA. High-throughput TILLING The same procedure used for TILLING Arabidopsis [ 16 ] was adapted for maize with only minor modifications. Primers were designed to amplify ~1-kb segments using the CODDLE program [ 35 ] based on either known maize genomic sequence or from maize cDNAs that are orthologous to intronless rice sequence. Genes were chosen from the NSF Plant Chromatin Project web site [ 36 ] based on the availability of genomic sequence or on the prediction of an exonic region at least 1-kb in size. Because rice and maize typically have identical placement of introns, by aligning the predicted maize coding sequence with the rice gene model, we could choose maize-specific primers that would amplify only exonic DNA. To find such regions, cDNA sequences were searched against Arabidopsis and rice genomic sequences using a version of BLAST that was modified to identify large exonic regions in maize based on the corresponding regions being exonic in rice and/or Arabidopsis. In all, we were able to design 14 primer pairs from plant chromatin genes for screening, and primer sets were ordered. In other cases, maize genomic sequence was available from ChromDb. Amplification of pools and individual DNA samples in 96-well plates, annealing, cleavage by CEL I, electrophoresis, image analysis, rescreening and DNA sequencing were performed as described [ 16 ]. Screening of mutations using LI-COR gel analyzers was performed as previously described [ 34 ]. Sequence trace information was analyzed using the Sequencher program as described [ 17 ] Authors' contributions SHR, KY and EB prepared the samples, CB, CAC, LCE and ARO performed the high-throughput screening and data processing, CW and NS provided the mutagenized plant material, EAG performed the data analysis and BJT, LC and SH participated in the design and execution of the study and wrote the paper.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC512284.xml
544954	Immunity & Ageing: a new journal looking at ageing from an immunological point of view	In the elderly, many alterations of both innate and clonotypic immunity have been described. Alterations to the immune system in the elderly are generally viewed as a deterioration of immunity, leading to the use of the term immunosenescence. However, although many immunological parameters are often notably reduced in the elderly, retained function of both innate and clonotypic immunity in the elderly is tightly correlated to health status. Recognising the important role of the immune system in ageing, over the last few years, journals oriented towards gerontology and geriatric sciences have increasingly published articles dealing with the immunology of ageing, but a specialised journal in this area does not exist. Immunity & Ageing is a new Open Access, peer reviewed journal that aims to cover all the topics dealing with innate and clonotypic immunity which are relevant to ageing. The journal will provide an opportunity to focus on this topic, which is emerging as one of the critical mechanisms of ageing. Furthermore, as an online, Open Access journal, Immunity & Ageing will promote immediate accessibility to research, which is generally not possible for articles published in printed journals. We hope this forum, concentrating on the themes of ageing and immunology with a strong focus on human studies, will create a new perspective for viewing a world that is inevitably becoming older.	Immunity & Ageing is a new Open Access, peer reviewed journal that aims to provide a forum for articles examining ageing from an immunological point of view. During the past century, humans have gained more years of average life expectancy than in the last 10,000 years; we are now living in a rapidly ageing world. The sharp rise in life expectancy, coupled to a steady decline in birth rates in all developed countries, has led to an unprecedented demographic revolution characterized by an explosive growth in the number and proportion of older people. The number of people aged 60 years or older exceeded 635 million in 2002, and is expected to grow to nearly 2 billion by 2050. The proportion of people aged 60 and over stands about 1 in 4 in many Western European countries as well as in Japan. Should the present trend continues, this ratio is expected to reach 1 in 3 by 2050 [ 1 ]. Among the aged, the oldest old (>85) make up the fastest growing category. As access to medical care improves worldwide, the rate of population ageing will accelerate. If global communications is making the world "young and fast", then global ageing is surely "maturing and slowing" it. In any case, these epidemiological facts underscore the importance of studies on successful and unsuccessful ageing and necessitate the prompt spread of knowledge about ageing in order to satisfactorily decrease the medical, economic and social problems associated with advancing years. Ageing Ageing is a post-maturational process that, due to a diminished homeostatic capacity and increased vulnerability, reduces responsiveness to environmental stimuli and is generally associated with an increased predisposition to illness and death. At the beginning of the 19 th century, mortality was described as increasing exponentially with respect to progression through the lifespan [ 2 ]. This trend, also described in invertebrates, persists: in Western countries the mortality rate increases 25 times more rapidly in individuals over 60 years old compared to people aged 25–44. Causes of death in aged people are increased compared with individuals between 25 and 44 years old: cancer 43-fold, pneumonia and influenza 89-fold, heart disease 92-fold and stroke and chronic lung disease greater than 100-fold [ 3 ]. Thus far, to understand ageing mechanisms, much attention has been paid to gene mutations in invertebrates and caloric restriction in rodents. However, these data suggest a key role for immunity in the survival of the elderly because susceptibility to these diseases depends at least in part on optimal immune function [ 4 , 5 ]. So, a better understanding of the ageing immune system may provide the most important clues for slowing the inevitable decline associated with the passage of time. Immunity in ageing In the elderly, many alterations in innate and clonotypic immunity have been described and viewed as deleterious, hence the term immunosenescence. In 1969, Roy Walford published his landmark book, "The Immunologic Theory of Aging", and first coined the term immunosenescence [ 6 ]. Significantly, most of the areas that he pioneered during his illustrious research career remain the "hot" areas of current gerontological research. On the other hand, immunosenescence is a complex process involving multiple reorganizational and developmentally regulated changes, rather than simple unidirectional decline of the whole function [ 7 , 8 ]. However, some immunological parameters are commonly notably reduced in the elderly and, reciprocally good function is tightly correlated to health status [ 4 , 5 ]. Innate immunity in ageing The process of maintaining life for the individual is a constant struggle to preserve his/her integrity. This can come at a price when immunity is involved, namely systemic inflammation [ 9 ]. Inflammation is not a negative phenomenon per se: it is the response of the immune system to the invasion of viruses or bacteria and other pathogens. The problem is that in the course of evolution the human organism was set to live 40 or 50 years. Today, however, the immune system must remain active for longer. This very long activity leads to a chronic inflammation that slowly but inexorably damages all the organs: this is the typical phenomenon linked to ageing and it is considered the major risk factor for age-related chronic diseases, such as osteoporosis, sarcopenia, type 2 diabetes, Alzheimer's disease and atherosclerosis, though progression seems also dependent on the genetic background of individuals [ 4 , 5 , 8 , 10 , 11 ]. Emerging evidence suggests that pro-inflammatory genotypes are related to unsuccessful ageing, and, reciprocally, controlling inflammatory status may allow a better chance of successful ageing [ 4 , 5 , 8 , 12 ]. In other words, age-related diseases are "the price we pay" for an active immune system that defends us thorough out life, but also has the capacity to harm us later, as its fine tuning becomes compromised [ 13 ]. In fact, our immune system has evolved to control pathogens, so pro-inflammatory responses are likely to be evolutionarily programmed to resist fatal infections with an increased resistance to pathogens. Thus, inflammatory genotypes are an important and necessary part of the normal host responses to pathogens in early life, but the overproduction of inflammatory molecules might also cause immune-related inflammatory diseases and eventually death later. Therefore, low responder genotypes might better control inflammatory responses and age-related disease development, resulting in an increased chance of long life survival in a facilitory environment with reduced pathogen load and medical care, such as might be present in Western societies [ 12 , 14 - 16 ]. Clonotypic immunity in ageing Recently, longitudinal studies have shown that a cluster of immunological parameters can be used to evaluate the expectation and quality of life, i.e. the immunological risk phenotype [ 17 , 18 ]. Senescence of clonotypic immunity is most likely principally a result of alterations to T cells. Lifelong and chronic antigen load seems to be the major driving force of immunosenescence, which impacts on human lifespan by reducing the number of virgin antigen-non experienced T cells, and, simultaneously, fills the immunological space with expanded clones of memory and effector, antigen-experienced T cells [ 18 - 20 ]. Gradually, the T cell population shifts to a lower ratio of naïve cells to memory cells, the thymus pumps out fewer naïve T cells with age and those T cells remaining, especially the CD8+ subset, also show increased oligloclonality with age [ 4 ]. Thus, the repertoire of cells available to respond to antigenic challenge from previously unencountered pathogens is reduced. In addition, older organisms are often overrun by memory cells that carry a single type of T cell repertoire, i.e. clonal expansion. Thus, the memory cells from old individuals might recognize a limited set of antigens despite being plentiful in number. Many of the clonal expansions crowding an elderly person's immune system result from previous infections by persisting viruses [ 18 , 19 ]. In contrast to T cells, no evidence for a loss of B cell function has been found as neither the total number of B cells or immunoglobulin secreting cells have been shown to be profoundly decreased with age. However, the B-cell repertoire is influenced by ageing during an actual immune response, where the spectrum of expressed immunoglobulin genes, as well as the frequency of somatic mutations, affects the quality, though not necessarily the quantity, of the antibody response, which is highly relevant in clinical practice [ 7 , 21 ]. Immunity & Ageing . Why do we need an Open Access journal? Considering the paramount function of the immune system during ageing, journals oriented towards gerontology and geriatric sciences are now publishing an increasing number of articles dealing with immunology and ageing, but a specialised journal in this area does not exist. Immunity & Ageing was conceived to cover all topics relevant to immunity and ageing in an interdisciplinary manner. It is clear that immunological mechanisms are involved in process and manifestation of ageing in many systems. An example is displayed by the increase in serum and tissue levels of circulation pro-inflammatory cytokines which is accompanied by typical cellular ageing phenomena such as telomeric loss, oxidative damage, DNA defects, accumulation of advanced glycosylation products, cellular loss and others. These defects may be stimulators of cytokine secretion, and subsequently, cytokines are released into the systemic circulation. This results in a slowly progressive endo-crinosenescence and neurosenescence. The different factors may influence each other in form of a vicious spiral [ 22 ]. The journal will provide an opportunity to focus the topic of immunology of ageing, which is emerging as one of the critical mechanisms of ageing which aspects should get to all scientists physicians and other professions who are involved in the different inter-related disciplines, and the bridge that is emerging between these fields. Furthermore, as an online, Open Access journal, Immunity & Ageing will promote "immediate" accessibility to this fast moving area of research, which is generally not possible for articles published in printed journals. From a modern perspective, electronic publishing is obviously attractive for its speed, easy global access and low cost, which all present considerable advantages over print and are seen as attractive factors by authors, readers and publishers. Open Access appears to provide exactly what people would like: rapid publication for authors, free access to information for readers, and inexpensive, global availability to readers for publishers [ 23 ]. We believe knowledge should belong to those who want it and access should not be unjustly and unacceptably expensive or difficult. Immunity & Ageing will take advantage of the Open Access policy to make peer-reviewed information widely and almost immediately available. We hope this forum, focussing on the themes of ageing and immunology, will create a new perspective for looking at a world that is inevitably becoming older. Creating and pushing forward a research knowledge base in Immunity & Ageing provides a unique opportunity to dissect out some parameters which might make health span equate with increasing lifespan.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544954.xml
544940	Emerging Themes in Epidemiology: Form and function	Emerging Themes in Epidemiology is a new, online Open Access journal. This editorial – which coincides with the Journal's launch – describes its unique review and publication model. The editorial board and review process of ETE will be managed by research degree students and will therefore be a training ground for students (though final editorial control rests with senior faculty Associate Editors). With our mix of Open Access publishing and the strong involvement of students in editing and reviewing, we believe that Emerging Themes in Epidemiology will be a progressive medium for promoting new ideas in epidemiology.	The idea that has given birth to Emerging Themes in Epidemiology (ETE) was conceived late on a Friday afternoon. As the rest of the academic and professional work week was coming to an end, the creators of this journal were just beginning to think straight. Clearly the work of students. However, it was not just clandestine meetings at unsociable hours that brought this diverse group of epidemiology PhD students together to launch and direct a new journal. Through training and various research projects PhD students become versed in the technical, methodological and practical aspects of epidemiology. While some will stay in academia and some will use their skills in other sectors of public health, our time as PhD students is seen as a time to learn and a time to be a part of the academic community. However, aside from the occasional correspondence with editorial boards, research students receive little training or involvement in the process of scientific publication. We think that this is an important process for us to learn about since, as researchers, our careers are dependent on publication. So we came together to form ETE , which will be a training ground for PhD students. This Open Access journal will also be our contribution to the field of epidemiology. Students will manage the editorial board and they will be a part of the review committee for each paper. Through this process we hope to learn. But, we also hope to facilitate the publication of new ideas and to raise issues that concern us as students of epidemiology. In the other editorial in this launch 'issue' of Emerging Themes in Epidemiology , the philosophical underpinnings of the Journal have been illustrated. Now, some explanation is required of how we will achieve our vision of ETE . The Editorial Board The editorial board of ETE is made up of doctoral students and faculty. The research students are from various academic and public health institutes in London (though we are open to the prospect of overseas expansion). The student members are responsible for the running of the Journal, for managing the review process and communicating with authors and readers. The board is headed by our Editor-in-Chief, Professor Peter Smith, and is overseen by an international panel of Associate Editors. The role of these senior faculty is to consider the major decisions of the Board and to preside over final publication judgements. And, because PhD studentship is a transient state (of varying length), these Associate Editors will maintain the continuity of ETE by recruiting new student editors as the current board graduates. The review process and student reviewers At present, there is a lively debate as to whether peer-review should be 'open' or 'closed'. Recognizing the forceful arguments on both sides, we believe that the scientific community is best served by a system that encourages a constructive dialogue between colleagues. Authors and reviewers will know each others' name and affiliation. However, we recognize that some highly specialized fields of research are small and competitive. In reality, this means that there are often repercussions for giving both criticism and praise to a colleague's work. So at the reviewers' request, we will withhold their names from the authors. When a paper is submitted to ETE , it is assigned to one of our two Deputy Editors. A committee of five research students then screens the manuscript to decide whether it fits within the scope of the Journal. If they decide it does, the paper is sent for external peer review. All papers are seen by a minimum of two reviewers. As part of our dedication to involving research students in the editorial process and facilitating the publication of new ideas, at least one PhD student and one expert in the appropriate field will be asked to review each paper. How will PhD students perform as reviewers? There is evidence that young reviewers (under 40 years old) trained in epidemiology or statistics, and those who can dedicate about 3 hours provide better quality reviews [ 1 ]. Clearly, these are characteristics of most of our student reviewers, so we are optimistic about our review process. Furthermore, in accord with our objective to make the Journal a training resource, all student reviewers will be informed of the Editorial Board's decision. They will see how their review influenced that decision and how it complemented or contradicted with the reviews of more experienced reviewers (the 'experts'). Final decisions to accept or reject will always be done with the endorsement of one of the Associate Editors, who are senior researchers in various epidemiologic fields. The involvement of students in the review process is admittedly novel, but we believe there are sufficient checks and balances in our review system to ensure quality. Testament to this is the decision by PubMed and PubMed Central to index ETE . Authors can therefore be confident that their papers will have the highest visibility and will be available in PubMed's archives. Conflict of interest in a journal dedicated to discourse The issue of conflict of interest is anathema to many scientists. We are trained to pursue objectivity; the suggestion that one's methods, results or conclusions are in any way affected by funding sources or other affiliations can be seen as a direct challenge to credibility. It shouldn't be. In the complex funding and political environments in which we work, we can no longer deny that these issues must be faced. Evidence has mounted that funding is associated with research outcomes (in tobacco and food) and, recently, high-profile cases related to vaccine safety and nutritional supplements have furthered the case for explicit statements of potential conflicts [ 2 - 5 ]. Once again, ETE will encourage openness from authors as well as reviewers. An individual's financial or professional associations do not necessarily constitute conflicts of interest, but failing to declare them can be highly misleading to the reader and the scientific community [ 6 ]. ETE is a journal dedicated to active discourse in epidemiology. We will ask contributors to disclose their source(s) of funding – as it could be seen to influence opinions or research agenda. We will also ask authors to disclose links to other associations that are not explicit from their institutional affiliation. Open Access and financing of ETE Perhaps there is no discipline more appropriate for open-access publication than epidemiologic research and the allied population health sciences. Our work has the aim of revealing the determinants of disease in human populations, with the ultimate goal, through knowledge empowerment, intervention and policy, to reduce suffering from ill health. With such an ethos, how can we justify keeping research findings locked away in expensive medical journals, inaccessible to those who may need it most? The open-access publishing model has the potential to revolutionize the way authors use scientific literature and the manner in which it is received by other scientists and the community at large. The advent of open-access publishing means "it is now possible to make our treasury of scientific information available to a much wider audience, including millions of students, teachers, physicians, scientists, and other potential readers, who do not have access to a research library that can afford to pay for journal subscriptions" [ 7 ]. Currently, the most common economic model for sustaining Open Access journals is one where the author pays for the costs incurred in the editorial and publishing process. BioMed Central, our publisher, charges authors per accepted publication (currently $525), but fees are waived for authors from institutions with membership to BioMed Central or authors with financial hardship. Our hope is that a long-term solution will in future be developed by the scientific community as a whole, perhaps through the provision of grant schemes for Open Access publication. Nonetheless, recognizing the difficulties with the current 'author pays' model, we aim to cover the article processing fees for at least the next few years for all submissions not already covered under institutional memberships or other grants or subsidies. Special issues and online publishing From time to time ETE will call for papers on a particular subject pertinent to the theory and practice of epidemiology. The series of articles will be built around a framework, designed by the editorial board and a guest editor. The series of articles will be published in tandem, and will be highlighted on a special page on the website. The flexibility of online publishing will allow for articles submitted after the initial publication of the issue and for commentary on the original articles to also be included. Guest editors for special issues will work with the Editorial Board to develop an issue, commission appropriate articles and write a leading editorial. Anyone interested in collaborating with ETE to curate a special issue is encouraged to contact the Editorial Board. The birth of ETE Epidemiology is 'young' science, rapidly evolving in its theory, methods, and subject matter. With our mix of Open Access publishing and the strong involvement of students in editing and reviewing, we believe that Emerging Themes in Epidemiology will be a progressive medium for promoting new ideas that will contribute to the development of the field.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544940.xml

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.