Final Report Summary - CEED3 (Collaborative European effort to develop diabetes diagnostics)
Executive summary:
The 'Collaborative European effort to develop diabetes diagnostics' (CEED3) project brought together European leaders in beta-cell biology, genetics, and clinical diabetes. The main aim of CEED3 was to develop, test and validate genetic and non-genetic biomarkers as diagnostic tests in diabetes.
1. Diabetes diagnostics in monogenic diabetes
The major work was in identifying genetic subgroups particularly maturity-onset diabetes of the young who are frequently misdiagnosed as type 1 or 2 diabetes (T1D or T2D) and need completely different treatment once they are recognised and a genetic diagnosis made. Using shared resources from large patient groups, both novel and conventional biomarkers for the diagnosis of maturity onset diabetes of the young (MODY) were assessed. The most important novel biomarker was C-reactive protein (CRP) that identified HNF1A-MODY from both type 1 and T2D. This was tested in thousands of samples in 3 co-authored publications enabling the precision, sensitivity and specificity of the assay to be established. Proteoglycans was another novel marker that was developed on the basis of genomewide analysis results, this also showed excellent specificity and sensitivity for diagnosing HNF1A MODY but the difficulty in performing the assay limits its utility at present. CEED3 also established that conventional biomarkers, beta-cell antibodies and C-peptide measurement, were excellent at discriminating MODY from T1D.
An important innovation with wide clinical application was the development of an online MODY probability calculator. This allowed physicians or patients to enter basic clinical information and get a validated estimate of the likelihood that a patient has MODY. This allows appropriate patients to be chosen for genetic testing. A more refined model that integrates biomarker information improved the accuracy of the predictions particularly for insulin treated patients.
The CEED3 grouping have been select to write European guidelines on the diagnosis and management of monogenic diabetes enabling this important science to have a direct influence on guidelines.
2. Developing biomarkers to detect beta-cell dysfunction and Beta-cell loss
A variety of sophisticated approaches using human islets, animal models and cell lines have identified important lead compounds secreted by beta-cells either in health or after beta-cell destruction / dysfunction. The development of assays for these compounds was very challenging but initial studies have confirmed that some biomarkers measured in blood are detecting either beta-cell presence or beta-cell destruction. These will go forward for further validation studies. The testing of beta-cell markers will be assisted by the establishment of CEED3 shared resources from patients with type 1, T2D, different subtypes of MODY and at the extremes of distribution of beta-cell function. Genetic biomarkers utilising polymorphisms can predict deterioration of beta-cell function to diabetes but the effect sise is markedly less than the use of age and BMI and is not clinically useful.
3, Biomarkers to detect assess complications
Genetic and non-genetic biomarkers assessed for complications added little to the conventional markers cholesterol, glycaemia and blood pressure for vascular disease and micro-albuminuria measurement for nephropathy. Work in this area is continuing in the EU IMI SUMMIT project which includes many members of CEED3.
4. Education and dissemination
A major role of CEED3 was to disseminate to clinicians the novel and established work on identifying and treating monogenic diabetes. This was done by over 90 talks and lectures, the CEED3 clinician's course and seminar days, two educational forums, publishing over 100 papers, producing online videos and the CEED3 and http://diabetesgenes.org websites.
Project context and objectives:
CEED3 Project context
CEED3 responded to an important clinical need in diabetes care, in aiming to develop diagnostic tools to differentiate specific subgroups of diabetic patients. Diabetes is a major health problem that generates a significant burden, not only as the most common cause of renal replacement and blindness in young people but as a major risk for cardiovascular disease including myocardial infarction, stroke and amputation.
Effective clinical care for diabetic patients is often hindered by delayed or inadequate diagnosis and subsequent inappropriate treatment. Utilising the knowledge from leading European groups in the fields of beta-cell biology, biomarker discovery, diabetes genetics and clinical medicine the consortium aimed to address these challenges in diagnosis. Through the integration of the consortium's combined expertise the CEED3's core objective was the improvement of diagnostic tests able to differentiate patients into subsets or categories. Improved diagnosis would alter the treatment and management of diabetes, guiding the individualisation of therapy and having a direct benefit on patient care.
The timing of the project was highly relevant given the advances in basic and clinical science in the years before the project's inception. The human genetics of diabetes had progressed considerably allowing the ability to define both the rare monogenic causes of diabetes and the predisposing alleles for the common forms of diabetes and complications. The technical advances in polygenic analysis of common disease had also facilitated the use of cheap and effective large scale genome-wide analysis performed in large numbers of well characterised subjects.
CEED3 key objectives
In achieving its aims of individualising treatment across subgroups of diabetic patients, the project identified specific aims and objectives of its intended diagnostic tools:
(a) allow recognition of those who should be considered for molecular genetic testing, and identify patients within specific subgroups of monogenic diabetes;
(b) detect subgroups at high risk of beta cell dysfunction to help tailor treatment, follow-up or even identification of a novel treatment approach;
(c) improve identification of those patients at greatest risk of vascular complications. In all of these clinical areas the ability to differentiate subgroups of patients will alter and ultimately improve the treatment and management of diabetes.
The CEED3 project aimed to ensure progress went rapidly from scientific discovery to clinical advances in the form of validated diagnostic tools by bringing together experts in the laboratory with geneticists and clinical scientists.
The CEED3 objectives were achieved through a clear process of discovery of novel genetic and non-genetic biomarkers, validation within test and cohort data samples, development of clinical application and dissemination of this application.
In order to develop a clinically implemented diagnostic test, the core CEED3 pipeline consisted of:
- Discovery: Given the critical requirement of genetic and non-genetic biomarkers in order to establish effective diagnostic tests, the first steps for CEED3 were studies identifying new potential biomarkers as well as the integration of results from other existing scientific studies to identify all potential biomarkers.
- Validation: Potential markers were evaluated through validation in human samples (predominantly deoxyribonucleic acid (DNA) and plasma) from patients with known diagnostic status.
- Clinical application: Validated markers were then considered for clinical implementation and tested for diagnostic discrimination compared to existing clinical testing. A key consideration at this point was whether such tests were sufficiently robust and user friendly for widespread clinical use.
- Dissemination: Clinical tests that were considered clinically applicable and suitable for widespread use were then disseminated to clinicians internationally through the wide network generated by the leading experts brought together by CEED3.
These objectives provided the framework for CEED3 work packages, focusing the expertise of the consortium within the identified fields but feeding back to the broader aims of diagnostics development.
From this European-wide collaboration, CEED3 has successfully addressed its core objectives including:
(1) Development of an enhanced super national framework for research related to the development of diabetes diagnostic tools that assembles the best researchers, the best samples and data sets for primary discovery and subsequent validation, and the best development platforms for clinical tools to ensure widespread dissemination.
(2) The identification and validation of potential genetic and non-genetic biomarkers that can be used to differentiate between the discrete subtypes of diabetes using information from studies of humans and model systems.
(3) Development of a clinical tool to aid the diagnosis of monogenic diabetes integrating validated biomarkers with clinical phenotype assessment.
(4) The identification and validation of genetic and non-genetic biomarkers that can be used to identify individuals with early beta-cell dysfunction initial development of diagnostic tools based on these for potential use in the clinic.
(5) The identification of genetic and non-genetic biomarkers that can be used in early identification of patients with type 2 and monogenic diabetes at increased risk of developing chronic diabetic vascular complications.
(6) The dissemination of knowledge and provision of training on diabetes differential diagnosis and CEED3 results to all potentially interested patient, professional and societal groups.
Project results:
The main aim of CEED3 was to use a European collaboration to develop diabetes diagnostics. There have been three main areas where we have aimed to develop these diabetes diagnostics:
(a) differentiating subtypes of diabetes, particularly monogenic diabetes from the more common forms of diabetes;
(b) identifying markers of beta-cell dysfunction that would be useful in early diagnosis of diabetes and in monitoring the decline of diabetes; and
(c) markers for the development of complication.
The work carried out and main findings are described in detail below:
1. Differentiating subtypes of diabetes including monogenic diabetes
1.1 Differential diagnosis of diabetes: Basic science
1.1.1 Using genetic analysis of T1D and T2D
A massive international collaborative effort including Exeter, Oxford and Lund has enabled genetic data from large international cohorts, including those from the CEED3 partners to be integrated and analysed together. These have resulted in the identification of over 50 susceptibility variants for T2D. The search for genetic markers for T1D has resulted in over 70 variants being identified.
1.1.2 Using clinical resources and conventional and next generation sequencing technology to define prevalence of MODY and refine testing and develop potential diagnostic tests to T2D
Exeter has worked to identify how common MODY is in children with diabetes diagnosed before 18 years. They examined 391 patients from the 3 955 Swedish children from better diagnosis in diabetes study who did not have any pancreatic autoantibodies at diagnosis. Using Sanger sequencing they tested the common MODY genes and found 58 children with MODY (15 %). This is consistent with at least 1.5 % having MODY. This diagnosis was not known in over 60 % of cases and 45 % were on the wrong treatment with half of these being on unnecessary insulin.
To assess the prevalence in an older group and assess the role of novel sequencing methods. Lund and Paris have carried out a key project to sequence (Roche 454) 4 known MODY loci (HNF1A, HNF4A, HNF1B and glycokinase (GCK) in 541 GAD antibody-negative patients with age at onset of diabetes < 35 years from the 'All new diabetics in Scania' (ANDIS) project. This will allow detection of both high penetrance MODY causing mutations and also rare moderate penetrance variants predisposing to T2D. 32 variants are now being taken forward for validation using Sanger sequencing. This work (combined with the BDD) results for this region will allow the regional prevalence of MODY to be assessed in those diagnosed under 35 years
In addition, Lund and Paris have performed exome sequencing of 12 individuals with MODY-like diabetes (autosomal dominant diabetes with onset < 25 yrs of age). After filtering, a number of novel variants/mutations were found. We are now: (i) validating using Sanger sequencing;
(ii) sequencing in additional family members; and
(iii) exploring the functional consequences of the variants on the clinical phenotypes.
1.1.3 Generation of a collaborative cohort of DNA and serum / plasma samples from specific patient subgroups for early discovery and validation
Exeter, Oxford, Lund and Krakow have collaborated to produce large sample collections including samples from patients with monogenic diabetes, as well as Type 1 and T2D. These have been used extensively in biomarker testing and validation (Thanabalasingham et al Diabetologia 2011, McDonald et al Diabetes Care 2011, McDonald et al Diabetic Medicine 2011). Further details are in section '1.2 Assessment of diagnostic approaches in monogenic diabetes '.
1.1.4 Use of available samples for further rounds of proteomic and metabolomic biomarker discovery
Lund has shown, using state of the art metabolomics, that MODY2, due to mutations in the GCK gene, is metabolically a normal condition in contrast to all other types of hyperglycaemia (Spegel P et al. Diabetes 2012).
Helsinki pursued distinct T1D biomarkers by nuclear magnetic resonance (NMR) metabolomics. They initially observed dramatic differences between 800 T1D patients with no nephropathy form the FinnDiane Study and age and sex matched population controls from the Cardiovascular Risk in Young Finns Study. Unfortunately, these differences were later revealed to be strongly correlated with the difference in sample collection protocol between the two cohorts, and so were unable to replicate the biomarkers in a subset with identical sample procedures in cases and controls. Further work on the metabolic differences between T1D, T1D with nephropathy, and healthy controls are ongoing.
Oxford has also performed urine metabolic profiling in HNF1A-MODY, GCK-MODY and T2D, using both NMR and ultra-performance liquid chromatography - mass spectroscopy (UPLC-MS) (Gloyn et al, PLoS One 2012). This failed to detect any robust urinary markers for the MODY subtypes, but it was confirmed that those with HNF1A-MODY have raised urine glucose and relative aminoaciduria (which was no longer observed when correcting for glucose). Raised urine betaine was also observed in the HNF1A-MODY subjects, but a thorough examination of urinary 1-carbon metabolites did not replicate this finding.
1.1.5 Identification of candidate secreted proteins arising from mutations in HNF1A and HNF4A using genetic mouse models
Barcelona has profiled ribonucleic acid (RNA) from Hnf4a and Hnf1a deficient islets and liver from knockout mice to generate over 10 candidate secreted or shed biomarkers markers for testing in diagnostic assays and disease progression studies in HNF4A and HNF1A diabetes (Boj et al, Diabetes 2009, Servitja et al Mol Cell Biol 2009, Boj et al, Plos Genetics, 2010).
1.2 Assessment of diagnostic approaches in monogenic diabetes
1.2.1 Definition of key clinical data to discriminate between subtypes of monogenic diabetes from each other and T1D and T2D
Oxford has performed an evaluation of a set of approximately 250 clinical T1D and approximately 300 clinical early-onset T2D patients to estimate the extent of misdiagnosis of MODY, and to determine whether existing clinical pathways for differential diagnosis should be revised (Thanabalasingham et al, Diabetes Care 2012) Clinical features were defined that we hypothesised would be useful in differentiating HNF1A/4A- and GCK-MODY from T1D and T2D (as listed below). These features were then used as a pre-selection for genetic testing in our collections of young-onset T2D (n=300) and T1D (n=250).
Differentiating HNF1A-MODY from T1D: Subjects with significant C-peptide secretion post-honeymoon period were selected (C-peptide > 0.02 nmol/l or stimulated C-peptide increment > 0.02 nmol/l) and sequenced for HNF1A/4A and 3 HNF1A mutations, 8.5 % (n=21) of the apparent T1D fitted these criteria. Mutations were found in 10 % of those sequenced (1 % of the T1DM cohort).
Differentiating HNF1A-MODY from T2D: Subjects with either (a) age of onset up to 30 or (b) age of onset <45 yrs + absence of metabolic syndrome (IDF criteria) were selected. 28 % (n=82) of the apparent T2D fitted at least one of these criteria and were sequenced for HNF1A/4A. 10 HNF1A and 2 HNF4A mutations were found which was 4% of the overall T2D group (15% of those sequenced).
Differentiating GCK-MODY from T1D or T2D: Subjects with characteristics consistent with the specific features of GCK-MODY (FPG 5.5 - 8.5 mmol / l, HbA1c = 8.5 %, OGTT 2 hour increment < 4.5 mmol / l) and C-peptide positive were selected. 7.6 % (n=42) from the combined T1D and T2D groups were sequenced for GCK-MODY. 1 mutation was found in the T2D group (< 0.5 % of overall T2D, 2 % of those sequenced). Using the extended criteria more than doubled the cases of HNF1A/4A-MODY that would have been identified based on traditional guidelines (Diagnosis age<25, with parental history and evidence of endogenous insulin). Our strategy for finding GCK-MODY was less successful and this may have reflected that the HbA1c of 8.5% was too high studies from Exeter suggest 7.5 not 8.5 %.
In addition, Exeter investigated clinical characteristics (age at diagnosis, parent affected with diabetes, HbA1c, age at referral/recruitment, treatment, time to insulin (if applicable), gender and BMI) to discriminate MODY from type 1 and T2D and developed a clinical prediction model. This model utilises a weighted combination of clinical features to provide a probability on a continuous scale of whether a particular patient has MODY. The model performs well, with an ROC AUC of 0.95 for discriminating MODY from T1D and an AUC=0.98 for discriminating MODY from T2D, with similar AUCs when tested on an external dataset. This algorithm is now available as an online calculator at http://www.diabetesgenes.org/content/mody-probability-calculator with a link from the CEED3 website and has been published (Shields et al. 2012, Diabetologia).
Exeter has also produced a model based on the same clinical features to help identify HNF1A/4A MODY from GCK MODY (abstract accepted for Diabetes UK annual professional conference 2013). This model shows good discrimination (ROC AUC=0.88) and may help determine the order of testing until next generation sequencing is established.
1.2.2 Integration of data from ongoing efforts to identify phenotypic biomarkers with the potential to discriminate between diabetes subtypes
Exeter, Oxford, Lund and Krakow have collaborated to share data and samples from in excess of 3n 000 patients including 801 with MODY (400 HNF1A, 325 GCK, 55 HNF4A, 21 HNF1B), 730 with type 1 and approximately 1 521 with T2D which has allowed the assessment of the diagnostic potential of a number of biomarkers.
1.2.3 Testing of the ability of known and potential novel genetic and non-genetic biomarkers to differentiate subtypes of monogenic diabetes from each other and from T1D and T2D
Known biomarkers
(a) Urinary C-peptide/Creatinine Ratio (UCPCR): Exeter has shown a UCPCR > 0.2 nmol / mmol is an extremely sensitive (97%) and specific (96%) marker for differentiating MODY from T1D in adults outside the honeymoon period (Besser et al., Diabetes Care, 2011). Work in pediatric cases has shown UCPCR>=0.7 nmol / mmol is sensitive (97%) and specific in discriminating MODY and T2D from T1D >2y post-diagnosis (Besser et al., Pediatric Diabetes, 2013)
(b) Islet autoantibodies: Islet autoantibodies (glutamate decarboxylase (GAD) and IA2) were found in 82 % of type 1 patients but in < 1 % of MODY patients (Exeter), suggesting a positive result is very good at ruling out MODY in young-onset patients (McDonald et al., 2011 Diabetic Medicine).
(c) Lipoproteins: Samples from Exeter, Oxford and Krakow have been used to investigate lipoproteins in discriminating MODY from T2D. The most promising finding was HDL>1.12 mmol / L which had 75 % sensitive and 64 % specific for discriminating MODY from T2D (McDonald et al., 2012, Clin Chim Acta).
Novel biomarkers
(d) Highly sensitive CRP: Findings from genome-wide association study (GWAS) that common variation near HNF1A altered serum CRP level in population cohorts led to the hypothesis that inactivating mutations in HNF1A (causing MODY) would be associated with larger differences in CRP, measurable in blood, that could be used as a diagnostic biomarker. Oxford have now shown conclusively that highly-sensitive CRP (hsCRP) levels are significantly lower in HNF1A-MODY than in other forms of diabetes, providing good discrimination of HNF1A-MODY from T2D. These findings were initially demonstrated in UK samples (Owen et al Diabetes Care, 2010), then replicated in large European sample sets (Exeter, Oxford, Lund and Krakow) showing ROC Curve C-statistics ranging from 0.79 - 0.97 depending on the centre and sensitivity/specificity approximately 80 % (Thanabalasingham et al Diabetologia 2011, McDonald et al Diabetes Care 2011). It can also discriminate HNF1A and HNF4A-MODY, but is not a good discriminator from T1D. See more detail in validation section. The CRP result was a very important part of the CEED3 work that has been confirmed and extended in very large and comprehensive collaborative studies with the CEED3 partners. As CRP is widely available and can be measured cheaply in routine labs, this offered the chance of early use in clinical studies and its use in clinical practice has already been described (Besser et al BMJ Case Reports 2012).
(e) Glycan profile: HNF1A is a regulator of plasma protein fucosylation, with common variation associated with alteration of the plasma glycosylation profile. Oxford found the DG9-glycan index performs well in discriminating HNF1A-MODY from both T1D and T2D with a sensitivity and specificity 80-90% and ROC curve C-statistic > 0.9 (Thanabalasingham, Diabetes 2012). On paper therefore it seems to be a better biomarker, but is not yet available outside a research setting so would need more development before wide-scale translation is possible.
(f) 1,5-anhydroglucitol - 1,5-Ag was the first biomarker examined by Krakow in their HNF1A MODY patients. The initial finding by Krakow (Skupien et al., Diabetes Care, 2008), was subsequently replicated by Oxford as a biomarker for T2D. If you corrected for HbA1c it gave excellent discrimination from MODY (Pal et al. 2010). This established 1,5-AG a promising biomarkers for diabetes differential diagnosis. In 2012, Krakow started using an ELISA based kit for 1,5-AG and defined the detailed protocol for this method. This will be useful for future research that was held back by the difficulty in accessing the assay.
(g) Cystatin C: Cystatin C was a candidate biomarker for HNF1A MODY because some pathophysiological premises suggested that its level might be altered in this monogenic diabetes. The initial analysis done by Krakow showed lower cystatin C level in HNF1A MODY than in T1DM, T2DM, and in controls. This was not replicated by Exeter and Oxford. However, the consistent finding was that in HNF1A-MODY, cystatin C-based GFR estimate was higher than the creatinine-based one.
(h) Apo-M - early studies by Surich identified apoM as a candidate marker for HNF1A MODY due to reduced expression in Hnf1a-/- mice. However, subsequent human studies have yielded conflicting results. Therefore Oxford, Krakow and Surich aimed to use a highly specific and sensitive ELISA to evaluate apoM as a biomarker for HNF1A-MODY. This confirmed that ApoM is low in HNF1A-MODY, but is also low in T2D, so therefore offers poor discrimination. However ApoM does discriminate HNF1A-MODY from T1D where levels are normal (Mughal SA et al Diabetic Medicine 2013.). Like glycan profile, though, the ready availability of the assay is an issue for routine clinical practice.
1.2.4 Combining discriminatory clinical and laboratory markers into a diagnostic algorithm to estimate the likelihood of a diagnosis of specific types of monogenic diabetes
Exeter, including samples and data from Oxford and Krakow, has recently developed a model combining clinical features with biomarker results. The combined models show considerable improvement over clinical features or biomarkers alone (e.g. the model for early insulin treated patients discriminating T1D from MODY that includes clinical features, C-peptide and islet autoantibodies increases to an ROC AUC of 0.977 from 0.914 for clinical features alone, with a mean improvement in probabilities of 22 %). This has been presented at Diabetes UK and EASD conferences and written up for publication. We now intend to validate the model prospectively on all referrals arriving in the diagnostic laboratory.
1.2.5 Creation of a clinical application based on the diagnostic algorithm - accessible via the internet
The clinical prediction model for MODY based on clinical features developed by Exeter (Shields et al Diabetologia 2011) is now available as an online calculator at http://www.diabetesgenes.org/content/mody-probability-calculator with a link from the CEED3 website. A beta-version of the new model that incorporates both clinical features and biomarkers (Exeter, Oxford and Krakow) has been developed and is currently undergoing testing. This will be available on the website by March 2013.
1.3 Validation of diagnostic approaches in monogenic diabetes
1.3.1 Testing of the diagnostic algorithm in specific clinical settings to provide validation though sequencing for common monogenic genes
The clinical features algorithm was validated on an external test dataset of 350 patients not included in the original model by Exeter and showed similar high ROC AUCs > 0.94. The MODY calculator is currently being used in the Brussels' Diabase, a database of unselected diabetic patients, to assess if this helps detect novel cases of monogenic diabetes.
Oxford is continuing to evaluate biomarkers / combination of biomarkers for identifying cases of MODY in their patient resources with young-adult onset diabetes. This includes comparison of hsCRP and glycan profile for detecting HNF1A-MODY in young adults with T1/T2D and examining whether hsCRP can identify HNF1A-MODY cases in S Asian subjects with young-onset diabetes.
200 subjects with apparent T2D diagnosed < 30 years have been sequenced for common MODY genes. 7 HNF1A mutations were found. A comparison of the performance of hsCRP and DG9-glycan index in identifying the HNF1A-MODY cases: in this dataset both markers were highly specific (80%), but hsCRP had low sensitivity (60 vs 85 %).
S Asian subjects with diabetes diagnosed <50 years were selected for HNF1A sequencing on the basis of low hsCRP (< 1mg / l). The prevalence of HNF1A mutations was 4 % in those sequenced - a quarter of the number seen in a comparable N European group. Thus screening by hsCRP alone had a low yield of MODY cases and a combination of clinical features and biomarkers are most likely to be successful.
2. Development and validation of potential biomarkers for beta-cell dysfunction
2.1 Beta-cell dysfunction: basic science
2.1.1 Identification of DNA-based variants impacting on beta-cell dysfunction
The aim has been to use re-sequencing approaches to identify low frequency variants influencing risk of T2D. The rationale here is that the GWAS approaches are designed to capture common variant signals, but have relatively limited power to detect signals driven by variants of low-frequency (that is minor allele frequency < 5 %) or those that are rare (MAF < 0.5 %). The expectation has been that some of these variants will have a relatively large effect (by comparison with the modest effect common variants identified by GWAS) and that, if such variants can be identified, they might contribute to improved diagnostic precision.
Oxford has been approaching this in a variety of ways. Some of the interesting variants identified in HNF1A and other genes of specific interest which were identified in some of the early experiments (see previous reports), have been followed up but none of them has credible evidence of an independent contribution to deterioration of beta-cell dysfunction resulting in T2D. This has demonstrated that the medium-scale studies (sample size of approximately 1 000) that we had been planning are likely underpowered, and has shifted our efforts towards larger scale studies.
Through funding from CEED3 and other sources, Oxford has recently completed an appraisal and early implementation of targeted sequencing methods (that is, methods which aim to allow the power of next-generation sequencing to be efficiently deployed to examine regions of limited genomic extent, such as the coding regions of a small set of genes of particular research or diagnostic interest). From a technical perspective they have demonstrated that it is possible to recover robust inventories of sequence variation from pooled samples, and that optimal performance was achieved with Illumina's TSCA and Agilent's Haloplex methods. These efforts, which focus predominantly on the issue of population-scale re-sequencing (getting accurate and cost-effective sequence data from tens of thousands of individuals), are complementary to those of Exeter, which focus on molecular diagnostics (getting highly accurate data from individual patients).
Large-scale targeted sequencing studies have been postponed, pending the results from two large exome sequencing studies that are currently underway, in which Exeter, Oxford and Lund are directly involved. Through NIH and Wellcome Trust funding, the GoT2D and T2D-GENES consortia have completed exome sequencing of approximately 13 600 individuals (half T2D cases, half controls). Final analyses on the full data set will be run in early 2013, and will be used to plan targeted sequencing experiments (following up the strongest low frequency and rare variant hits) to be performed in up to 50 000 samples. Interim analyses of these data (for example in the approximately 3000 exomes from the GoT2D study) have shown evidence of enrichment of rare variant alleles across a set of genes known to be causal for monogenic and syndromic forms of diabetes, in T2D cases compared to controls, but with no single gene or variant showing compelling evidence of causal association at this stage.
Brussels have carried out Illumina sequencing of 5 human islet preparations, leading to the identification of 15 000 genes expressed in islets. Of these, 3 000 genes are modified by cytokine exposure, 1 300 by palmitate. Compared islet RNA expression to other tissues shows a very high correlation within islets, and low correlation with other tissues. > 60 % of T1D candidate genes are expressed in human islets and > 80 % in of T2D candidate genes. Exeter, Oxford and Lund have participated in a new meta-analysis applying the MetaboChip for the identification of novel genetic variants increasing risk of T2D (Morris A et al Nat Genetic 2012) as well as the first exome sequencing study to identify rare variants increasing risk of T2D (Albrechtsen A et al. Diabetologia 2012).
Lund applied a systems genetics approach on gene expression in human pancreatic islets and were able to identify a set of 20 genes, which explained 24 % of the variance in HbA1c (Taneera J et al. Cell Metab 2012). A first GWAS for glucose-stimulated insulin secretion during an OGTT has been performed, identifying variants in the GRB10 gene are not only associated with decreased insulin secretion but also show a clear parent-of-origin transmission. Surprisingly, this was associated with lower rather than increased glucose concentrations. Knock-down of the GRB10 gene using shRNA in human islets shows a strong reduction in glucagon, which most likely explains the paradoxical effect on glucose concentrations.
Though the CEED3 funding period is over, these ongoing studies will continue to inform our efforts to refine diagnostic pathways. The fact that we have needed to extend these discovery efforts to include in excess of 60 000 samples is evidence that there are relatively few, if any, low frequency variants of large impact influencing T2D risk, and that the studies originally planned for CEED3 were underpowered.
2.1.2. Identification of potential beta-cell enriched secreted proteins using in silico and in vitro approaches in isolated islets for prioritisation of candidate markers
Candidate markers: Barcelona and Brussels developed a bioinformatic pipeline that used a host of islet-cell and pancreatic exocrine and alpha cell datasets and prediction algorithms for secreted proteins which identified several beta-cell enriched candidates, including WNT4, SCG5, 7B2, PCSK1N, FAM3C, FAM3A, ELAVL4, PTPRN, DLK1, and Osteopontin: These candidates that were potential candidates for measuring beta-cell mass, beta-cell function and/or beta-cell was done using a pipeline consisting of:
(1) INS-1E cell Affymetrix microarray data sets, selecting genes with an expression = GCK;
(2) ingenuity pathway analysis to select secreted, cytoplasmic or unknown proteins;
(3) present in islets in BioGPS human tissue array;
(4) Well expressed in beta cells vs alpha, ductal or exocrine T1Dbase;
(5) well-expressed in massively parallel signature sequencing of human islets;
(6) protein expressed in human islets (human protein atlas).
The beta-cell enriched candidates, including WNT4, SCG5, 7B2, PCSK1N, FAM3C, FAM3A, ELAVL4, PTPRN, DLK1, and Osteopontin i and culture medium from human islet preparations was provided to Proteomika for setting up assays.
2.1.3 Identification of potential markers for beta-cell function using T2D mouse models and beta-cell lines
SFRP4: Using a co-expression Lund has identified SFRP-4 as potential marker for inflammation in islets of patients with T2D. In mice, SFRP4 results in impaired insulin secretion. It is a secreted protein and its serum concentration strongly predicted future T2D (Mahdi T et al. Cell Metab 2012).
BACE 1 and BACE2 and the global secretome and sheddome of MIN6 cells: The analysis of proteins released in the media of cell lines or primary cultures is a powerful tool to identify biologically active peptides and facilitate their evaluation as biomarker for cancer, neurodegenerative diseases and diabetes. Surich has employed the technique to explore BACE2 and BACE1 (secreted on beta-cell surface, responsible for Tmen27 shedding in beta-cells), but also to explore the global secretome and sheddome of MIN6 cells and murine isolated pancreatic islets. Together 594 proteins were identified in the conditioned medium of MIN6 cells, 240 of which were specifically detected by enrichment of N-linked glycopeptides. All major islet hormones (insulin, glucagon, somatostatin and pancreatic polypeptide) were detected among other known secreted proteins such as granins (chromogranins and secretogranins), transthyretin (TTR), islet amyloid polypeptide (iAPP) and neuropeptide Y (NPY). In addition several secreted factors were identified of which the roles have yet to be studied in the pancreatic beta-cell, including vitamin D-binding protein (DBP), a multifunctional plasma protein of which altered circulating levels are associated with T1D. The islet sheddome identified here consists predominantly of type I membrane proteins, some of which were previously found to be cleaved in different tissues or cell types. Furthermore, several glycophosphatidylinositol (GPI) - or lipid anchored proteins were detected in addition to other membrane-associated proteins, which are putative targets of ectodomain sheddases. Several secreted and shed proteins were found to be more abundant compared to the putative 'cytosolic contaminants'. Thus, insulin-2 (INS2) and insulin-1 (INS1) as well as protein of the chromogranin (CHG) and secretogranin (SCG) protein family were amongst the most abundant hits.
The functional consequences of ectodomain shedding and protein secretion can be diverse and depend on the function(s) of the particular protein or of the protein cleavage products. We therefore grouped the islet / beta-cell secretome and sheddome into functional categories using the PANTHER protein classification system. Nine major categories were identified including receptor proteins (12 %), hydrolases (12 %), signaling molecules (10 %), extracellular matrix proteins (8 %), cell adhesion molecules (8 %), ensyme modulators (8 %), defense/immunity proteins (7 %), proteases (6 %), and transporters (5 %). This heterogeneity of putative shed substrate proteins suggests that diverse and evolutionarily conserved biological processes are regulated by the activity of beta-cell surface proteases.
2.1.4 Identification of potential markers for beta-cell dysfunction through in vitro studies in human islets
Oxford and Lund have shown that the KCNQ1 locus exhibits a complex imprinting pattern with mono-allelic expression in fetal and bi-allelic expression in adults human islets. Lund has also explored the functional consequences of beta-cell specific genetic variants like TCF7L2 (Shou Y et al. Hum Mol Genetic 2011), GIPR (Lyssenko V et al Diabetes 2011.), PAX6 (Ahlqvist E et al. Diabetologia 2011), PCSK2 (Jonsson A et al. Diabetologia 2012) To discover beta-cell-specific gene products that could be employed as biomarkers, Barcelona and Brussels have carried out high-throughput sequence-based epigenetic and transcriptional maps of human islets, FACS purified beta-cells and pancreatic exocrine cells. This technology superseded our early plans to use exon arrays, which have a much more limited performance for this purpose. By subtracting maps from non-islet tissues, this has led to a catalogue of human islet-selective genes.
Brussels in collaboration with Oxford used modelling of lipofuscin accumulation in human beta-cells to show that the beta-cell population is established by age 20, i.e. human beta-cells are very long-lived and few new cells are being formed (Cnop et al. Diabetologia 2010). Brussels has performed the first genome-wide analysis of DNA methylation changes in human islets from T2D donors. 254 genes show differential DNA methylation in islets in T2D; this is a characteristic of the diseased tissue (islets) as there is no differential methylation in blood The RNA sequencing analysis of human islets from T2D donors is ongoing, and will lead to the identification of (i) beta-cell enriched or specific transcripts; (ii) potentially secreted proteins; (iii) alternative splice variants.
2.1.5 Assay development for measurement of candidate plasma beta-cell proteins
The list of prioritised candidate markers (WNT4, SCG5, PCSK1N, FAM3C, FAM3A, REG1A, REG3G, ELAVL4, PTPRN, DLK1, KIAA1324, Osteopontin) and culture medium from human islet preparations was provided to Proteomika for setting up assays. After completing a development phase, Proteomika successfully produced sandwich-type ELISA immunoassays for 4 markers: PTPRN, SCG5, FAM3A and FAM3C. All these assays were specific and functional. No assay could be produced for WNT4.
2.2 Beta-cell dysfunction: Assessment of diagnostic approach
2.2.1 Initial verification of biomarker assays in serum from subjects selected for extremes of beta-cell mass and function
The assays developed for 4 potential biomarkers (FAM3A, FAM3C, SCG5 and PTPRN) were expected to be secreted by human islet cells. Human islets were isolated and cultured by Brussels under several conditions to promote secretion of the markers. Supernatants of human islets were then collected and sent to Proteomika for marker detection. Determinations performed in such samples would serve mainly as proof of assay functionality before performing larger validation studies. Surprisingly, negative results were obtained for all samples and for all markers despite previously all immunoassays bring successfully validated against marker-overexpressing cell culture lysates. The high protein content on the cell culture media (BSA or FBS) could presumably prevent detection of markers. Since the presence of these proteins (BSA or FBS) was required for cell survival they could not be removed from the culture media.
A second marker validation was performed on serum samples from patients with T1D and with normal beta cell function were collected by Exeter and sent to Proteomika. Results for markers SCG5 and PTPRN did not discriminate the 2 patient groups. In contrast quantifications of FAM3A and FAM3C proteins did show reasonable but not total discrimination of T1D from normals. ROC curves were generated and AUC values of 0.89 for FAM3A and 0.835 for FAM3C were calculated. Combination of both markers in a multivariate model did not significantly improve the outcome (AUC 0.9). These promising values will be taken to a further validation study in a wider (more patient groups) and larger (more patients) sample population to evaluate their diagnostic power.
BACE2 and SES6L2 - loss-and gain-of-function studies of in vitro and in vivo models revealed non-redundant roles of BACE1/2 in ectodomain shedding with BACE1 regulating a broader and BACE2 a more distinct set of beta-cell-enriched substrates whose ectodomains can be measured in mouse and human plasma and may serve as biomarkers for beta-cell loss and dysfunction. From the set of validated putative BACE2 targets, two single-pass type I transmembrane proteins of the seisure 6 protein family were selected for further characterisation because of their BACE2 substrate specificity and enrichment in pancreatic islets. Expression analysis by qPCR in multiple mouse tissues revealed a marked enrichment of Ses6l and Ses6l2 in pancreatic islets. In contrast, Ses6, the third member of the seisure 6 gene family, showed low expression levels in islets and predominant abundance in the central nervous system. Together, this demonstrates that SES6L and SES6L2 are colocalised with BACE2 in pancreatic islets.
MicroRNAs as biomarkers: Barcelona has examined high throughput datasets of ncRNAs, including long non-coding RNAs (ncRNAs) (Moran et al, Cell Metabolism 2012), and small RNAs (Correa et al, unpublished; van de Bunt et al, PloS One, 2013, in collaboration with Exeter and Oxford). These studies have provided a collection of novel cell type-specific biomarkers. In particular, one microRNAs (miRNA) has been shown to be elevated during acute beta cell destruction in mice and humans undergoing human islet transplantation. This marker has been tested in a CEED3 cohort of patients with T1D and patients with neonatal forms of monogenic diabetes, showing that it correlates with beta cell damage. Future studies will test the utility of this marker for detecting ongoing beta cell damage in diverse clinically relevant settings.
Transcription factor 1 (Tcf1; hepatocyte nuclear factor 1a [Hnf1a]) is critical for hepatocyte development and function. We investigated if Tcf1 also regulates hepatic miRNAs. Tcf1-dependent miRNA expression in adult mice in which this transcription factor had been genetically deleted (Tcf1(-/-) ) was analysed using miRNA microarray analysis. The miR-192/-194 cluster was markedly down-regulated in liver of Tcf1(-/-) mice. MiR-192/-194 levels were also decreased in two other tissues that express Tcf1, kidney and small intestine, although to a lesser extent than in liver. In order to identify targets of miR-192/-194 in vivo we combined Affymetrix gene analysis of liver in which miR-192/-194 had been silenced or overexpressed, respectively, and tested regulated messenger RNAs (mRNAs) with multiple binding sites for these miRNAs. This approach revealed frissled-6 (Fsd6) as a robust endogenous target of miR-194. MiR-194 also targets human FSD6 and expression of miR-194 and Fsd6 are inversely correlated in a mouse model of hepatocellular carcinoma (Dgcr8(flox/flox) p53(flox/flox) × Alb-Cre). The results from this study support a role of miR-194 in Tcf1 dependent liver tumorigenesis through its endogenous target Fsd6. These results may have important implications for Tcf1-mediated liver proliferation in patients as HNF1A mutations can result in hepatic adenomas.
2.2.2 Testing the utility of genetic and non-genetic biomarkers for association with measures of beta-cell function. Genetic studies (LUND) will be done whilst awaiting validation of plasma assays
Lund has tested the role of novel biomarkers and / or genetic markers for prediction of T2D and deterioration of beta-cell function in man (Lyssenko V et al. Diab Vas Diab Res 2012; Ferrannnini E et al Diabetes 2012). As part of another European Union (EU) project (ENGAGE) we have identified a set of protein biomarkers with the ability to predict future T2D. This project is now in its replication phase.
Lund and Dresden have used extremes (highest and lowest decentile) of beta-cell function in the population to identify individuals (with serum samples available) for biomarker discovery in Zurich's laboratory. In addition to this, Lund are performing exome sequencing of families with extreme means (lowest and highest 10% of all families within the Botnia study). Seisure6l proteins are specific substrates of Bace2 in pancreatic islets. Ses6l and Ses6l2 levels are depleted in knock-out mice and found to be enriched in islets and brain. These are islet specific and beta cell specific, so were of potential interested as a biomarker for beta-cell mass. Surich used the stable isotope standards and capture by antipeptide antibodies (SISCAPA) method to develop immuno-MRM assays for the proteins of interest. Two of the tested anti-peptide antibodies, one against a human peptide of SES6L and one against a human / mouse conserved region of SES6L2, were found to be efficient in immuno-MRM assays. We demonstrated that the BACE2 substrates SES6L and SES6L2 circulate in plasma. Furthermore, detection of human SES6L in plasma of a humanised mouse model suggests that these proteins, although shed by a miniorgan, can be measured in the blood.
3. Development and validation of potential biomarkers for diabetic complications
Given the major EU funding of an Innovative Medicine Initiative (SUMMIT) with the aim to identify genetic and non-genetic markers for diabetic complications with over 8 times the total funding of CEED3, identification of genetic and non-genetic markers for diabetic complications has been mostly covered by SUMMIT rather than CEED3. Lund is the managing unit for the project which includes a number of other CEED3 partners (Exeter, Oxford, Helsinki).
3.1 Complications: Basic science
3.1.1 Finding genetic variants associated with complications in diabetes
Results from SUMMIT include a first GWAS for diabetic nephropathy (DN) in T1D has been published (Sandholm N, Plos Genet 2012) and several GWAS for diabetic nephropathy and retinopathy as well as cardiovascular disease in T2D have been completed recently.
In CEED3 we have focused on two projects, 1) testing whether known T2D-associated SNPs are associated with increased mortality and 2) a GWAS for free fatty acid (FFA) concentrations, especially during a meal.
Work with Lund and Exeter has shown a variant in the KCJN11 gene was weakly associated with increased mortality, especially mortality from CVD in patients with T2D. Also we have, together with other partners, shown genetic variants associated with T2D increase risk of Diabetic Nephropathy (Fagerholm E et al. Diabetologia 2012; Alkayaali S et al. Diabetologia 2012).
Postprandial FFA concentrations are strong predictors of the metabolic syndrome and CVD complications in T2D. They also show the highest heritability of any measured metabolic trait (Almgren P. Diabetologia 2011). Of the best loci in DGI GWAS of 1100 non-diabetic individuals FFA we replicated seven in Botnia Prevention, Prediction and Prevalence study (Botnia PPP, n=4900) but of these only rs2658404 on chr1 near ROR1 (receptor tyrosine kinase orphan receptor 1) and PGM1 (phosphoglucomutase) genes replicated in the second replication study (Metsim n=7000).
3.1.2 Identification of non-genetic biomarkers involvement in the development of nephropathy from ongoing studies of T1D
Helsinki has been applying NMR metabolomics to investigate the risk factors for T1D complications in 4 studies.
In the first study, we measured the serum metabolic profiles for 326 patients with T1D with normal AER, microalbuminuria or overt kidney disease. The NMR platforms provides information on 14 lipoprotein subclasses, over 20 low-molecular-weight molecules, and additional information on the circulating fatty acid species and their level of saturation. We found that sphingomyelin (odds ratio 2.53 P < 0.001) large VLDL cholesterol (odds ratio 2.36 P < 0.001) total triglycerides (odds ratio 1.88 P < 0.001) omega-9 and saturated fatty acids (odds ratio 1.82 P < 0.001) glucose disposal rate (odds ratio 0.44 P < 0.001) large HDL cholesterol (odds ratio 0.39 P < 0.001) and glomerular filtration rate (odds ratio 0.19 P < 0.001) were associated with kidney disease. Sphingomyelin was a significant regressor of urinary albumin (P < 0.001) in multivariate analysis with kidney function, glycemic control, body mass, blood pressure, triglycerides and HDL cholesterol (Mäkinen VP et al. Metabolomics, 2012).
In the second study, we compared the baseline data from the previously mentioned set of 326 patients with clinical follow-up of 8.2 years (Mäkinen VP et al. J Proteome Res, 2012). We also used the self-organising map in the data analysis, which is able to provide a comprehensive phenotypic profiling of metabolomics data. As expected, we saw the same connection between serum sphingomyelin and advanced kidney disease as before, but we also observed that patients with an unfavourable metabolic profile (high saturated fatty acids, high lipoprotein triglycerides and low HDL cholesterol) progressed to more severe kidney disease on average 10 years before than the patients with a favourable metabolic profile. In conclusion, we were able to show the strong connection between features that are the hallmarks of the metabolic syndrome in the general population and the progression of microvascular complications in T1D. Furthermore, the sphingolipid pathway may be an important player in the later stages when kidney function deteriorates (Makinen et al. J Proteome Res 2012).
The third study focused on the lipoprotein subclasses that were measured for the majority of the patients in the FinnDiane cohort (Mäkinen VP et al. J Intern Med, in press), We analysed the baseline biochemical data for 3 544 patients with T1D, and constructed a self-organising model to see which lipoprotein phenotypes were associated with diabetic kidney disease. We then compared these phenotypes to prospective data on kidney disease progression and mortality. We found a consistent association between elevated triglyceride content in all subclasses and the severity of kidney disease. A similar pattern was also observed for longitudinal changes in kidney status. Furthermore, we replicated the observation from the smaller studies that the metabolic syndrome characteristics were significant predictors of kidney disease progression and mortality. Interestingly, our findings also suggested that poor glycemic control in this cohort was not associated with lack of insulin, and we hypothesise that the same mechanisms and life style factors that contribute to obesity and T2D in the general population are the likely contributors to the progression of microvascular injuries and premature death in T1D.
The previous work was based on the population-based metabolic profiles in a fasting-like state. We also wanted to see if there were any differences with respect to the dynamic metabolic responses between patients with T1D and healthy controls. We measured the NMR metabolomics profiles from 40 cases and 40 controls during a standardised 24-hour meal experiment. Each participant ate an identical breakfast, lunch and dinner, and blood was drawn every two hours during the day. After controlling for the baseline concentrations of metabolites, we did not detect substantial differences between the diabetic and non-diabetic subjects. In particular, we observed substantial meal effects for triglycerides and branched-chain amino acids, but these responses were equally pronounced in both study groups.
3.2. Complications: Assessment in monogenic diabetes
3.2.1 Assessment of long-term complications in patients with different subgroups of monogenic diabetes
As part of the investigation of cardiovascular complication development in monogenic diabetes, Oxford has investigated clot structure and function, which determine predisposition to atherothrombotic disease, in HNF1A-MODY. This study compared clot structure / fibrinolysis and inflammatory markers in subjects with HNF1A-MODY compared to matched T2D and healthy controls and found that HNF1A-MODY subjects had a less dense clot structure than T2D, but greater than healthy controls. Lysis time of the clot was shorter in HNF1A-MODY than T2D. Fibrinogen and PAI-1 were similar in MODY and T2D, but C3 levels were lower in MODY. This indicates HNF1A-MODY subjects have a decreased thrombotic tendency than T2D and thus suggests an intermediate cardiovascular risk between T2DM and healthy controls.
Detailed measurements of complications and risk factors for complications have been assessed in approximately 100 GCK MODY patients, and a set of similar aged controls and young-onset T2D patients by Exeter. GCK patients have a similar prevalence of both microvascular and macrovascular of complications to controls, and a significantly lower prevalence of complications compared with T2D, despite an average of 50 years of raised glucose. Only retinopathy was slightly higher than in controls, and only consisted of mild non-sight-threatening background retinopathy with the majority of cases having < 5 micoraneurysms only. These results have implications for the management of GCK patients, where regular screening for complications is not necessary. In addition, these results also can be extended to show that the current target levels for glycaemia (which GCK MODY patients mimic), are appropriate for limiting complications.
Potential impact:
This work has had, and will continue to have a major impact for patients with monogenic diabetes, particularly the familial form presenting in early adulthood known as MODY. These patients are usually not recognised as having a different subtype of diabetes from the majority with Type 1 or T2D. This is of great importance as it is well established that these patients have a different response to treatment, specifically there is a 'pharmacogenetic' effect of their treatment response. This means that any method that helps doctors and patients to recognise that they do not have type 1 or type 2, but rather have an inherited form of diabetes will result in a marked improvement in their clinical care. As the majority of these cases run in families it will have an impact not only for them, but also for other members of their family. A major benefit for the patients is that many are unsuccessfully managed with insulin and other drugs and find that once a correct diagnosis is made they then can be appropriately treated, in many cases with sulphonylurea tablets or diet alone. Appropriate treatment reduces the amount of medication that the patient is taking which reduces side effects and costs both for the patient and also for the national government.
The impact will therefore be seen in terms of improved health of the patient, as their blood glucose is better controlled and consequently their risk of complications is reduced, and also in financial terms because of the money saved on treatment. For patients with GCK MODY, the treatment is stopped and the blood glucose control remains unaffected. Patients with HNF1A and HNF4A MODY respond particularly well to low dose sulphonylureas which are considerably cheaper than the more modern drugs used in T2D which are not effective in these subgroups. The improvements in diagnosing these patients represent a significant medical advance with major implications for individuals, as well as a socio-economic impact. The advances have involved the integration of clinical characteristics into an online probability calculator for MODY enabling any patient or doctor to go on a public website and enter their details to assess how likely they are to have a monogenic form of diabetes. This calculator has been widely validated and already used by over 5 000 doctors and patients.
The making of a diagnosis of MODY is improved by the establishment of a new biomarker, CRP, that discriminates these patients from T1D and T2D. Other biomarkers used for discriminating MODY from T1D that have been previously described have come from the collaboration of this large European consortium who have been able to be work out how they are to be used and to prove their effectiveness. All of these factors mean that for the clinician, by using just clinical interpretation, routine clinical data and simple investigations available at hospitals and clinics throughout Europe, they now have the tools to more systematically identify these patients. Further work has been carried out to assess the longer term impact of finding these specific types. Important work has been done on the complications of patients with GCK MODY. There was a major question about whether these patients were at increased risk, given that their blood glucoses were slightly raised throughout their life. Definitive studies have now been performed as a part of CEED3.
At the same time we have been able to establish that other proposed biomarkers, specifically the genetic biomarkers, even when combined in a test, were not sufficiently powerful to be able to be used at an individual level, although they did show significant differences at the higher level. This is important because it means that the emphasis should be put on the measures of non-genetic biomarkers rather than genetic biomarkers and shows that clinical features clearly outperform. This has shown that good science can provide important information regarding mechanism, but not necessarily inform patient care. Further impact of our work, not directly relating to patient care, is for industry where work by this group has started to identify specific markers of beta-cell presence i.e. beta-cell mass and also beta-cell destruction. These have proved very difficult to assay and hence the data are at very early stages, but could represent the first step towards being able to measure in man, whilst living, how many beta-cells they are producing. This work requires further development, but if such assays were available, it would mean that we would have an immediate way of assessing potential drugs that may be able to maintain beta-cell mass against novel trend of beta-cell loss as diabetes progresses. Also the work has enabled us to show when the cells are likely to be destroyed by therapy using microRNA. These findings, although at a preliminary level, have the potential to help the pharmaceutical industry in their search for assessment of novel drugs and for understanding mechanisms whereby new compounds work.
The other key impact of the CEED3 work has been in the development of the education of physicians throughout Europe and indeed the world. This has been achieved by many media. Clearly the conventional route of producing publications has been extremely successful with 107 produced as a result of work from the CEED3 partners during the period of the grant. These have been published in the very highest journals, particularly the top diabetes journals Diabetes Care and Diabetologia. Over 90 national and international talks have been given by members of the department talking about the identification of patients with MODY and why it is crucial to find these patients. Particularly important has been outreach programs aimed at reaching countries where genetic testing is not available and providing the potential for this and also disseminating information about genetic subgroups. Even when resources are limited, finding these patients can have a massive impact because it rapidly becomes cost effective due to the changes and improvement in therapy. The other key approach has been to provide this information to patients. This has been done by putting our talks on YouTube to enable patients to see things directly and by also having a public website and links to other national websites.
Examples of videos from the project shown on YouTube are shown here: http://www.youtube.com/user/ceed3eu
The exploitation of results is primarily within the healthcare system although further development of the findings from beta-cells will allow exploitation with pharmaceutical companies looking at changes in beta-cell mass or beta-cell dysfunction. This will be dependent on assays produce by our biotech companies. The biotech company Proteomika is working on the assays with a view to commercially developing these in conjunction with the academic partners.
List of websites: English, English / Polish
The 'Collaborative European effort to develop diabetes diagnostics' (CEED3) project brought together European leaders in beta-cell biology, genetics, and clinical diabetes. The main aim of CEED3 was to develop, test and validate genetic and non-genetic biomarkers as diagnostic tests in diabetes.
1. Diabetes diagnostics in monogenic diabetes
The major work was in identifying genetic subgroups particularly maturity-onset diabetes of the young who are frequently misdiagnosed as type 1 or 2 diabetes (T1D or T2D) and need completely different treatment once they are recognised and a genetic diagnosis made. Using shared resources from large patient groups, both novel and conventional biomarkers for the diagnosis of maturity onset diabetes of the young (MODY) were assessed. The most important novel biomarker was C-reactive protein (CRP) that identified HNF1A-MODY from both type 1 and T2D. This was tested in thousands of samples in 3 co-authored publications enabling the precision, sensitivity and specificity of the assay to be established. Proteoglycans was another novel marker that was developed on the basis of genomewide analysis results, this also showed excellent specificity and sensitivity for diagnosing HNF1A MODY but the difficulty in performing the assay limits its utility at present. CEED3 also established that conventional biomarkers, beta-cell antibodies and C-peptide measurement, were excellent at discriminating MODY from T1D.
An important innovation with wide clinical application was the development of an online MODY probability calculator. This allowed physicians or patients to enter basic clinical information and get a validated estimate of the likelihood that a patient has MODY. This allows appropriate patients to be chosen for genetic testing. A more refined model that integrates biomarker information improved the accuracy of the predictions particularly for insulin treated patients.
The CEED3 grouping have been select to write European guidelines on the diagnosis and management of monogenic diabetes enabling this important science to have a direct influence on guidelines.
2. Developing biomarkers to detect beta-cell dysfunction and Beta-cell loss
A variety of sophisticated approaches using human islets, animal models and cell lines have identified important lead compounds secreted by beta-cells either in health or after beta-cell destruction / dysfunction. The development of assays for these compounds was very challenging but initial studies have confirmed that some biomarkers measured in blood are detecting either beta-cell presence or beta-cell destruction. These will go forward for further validation studies. The testing of beta-cell markers will be assisted by the establishment of CEED3 shared resources from patients with type 1, T2D, different subtypes of MODY and at the extremes of distribution of beta-cell function. Genetic biomarkers utilising polymorphisms can predict deterioration of beta-cell function to diabetes but the effect sise is markedly less than the use of age and BMI and is not clinically useful.
3, Biomarkers to detect assess complications
Genetic and non-genetic biomarkers assessed for complications added little to the conventional markers cholesterol, glycaemia and blood pressure for vascular disease and micro-albuminuria measurement for nephropathy. Work in this area is continuing in the EU IMI SUMMIT project which includes many members of CEED3.
4. Education and dissemination
A major role of CEED3 was to disseminate to clinicians the novel and established work on identifying and treating monogenic diabetes. This was done by over 90 talks and lectures, the CEED3 clinician's course and seminar days, two educational forums, publishing over 100 papers, producing online videos and the CEED3 and http://diabetesgenes.org websites.
Project context and objectives:
CEED3 Project context
CEED3 responded to an important clinical need in diabetes care, in aiming to develop diagnostic tools to differentiate specific subgroups of diabetic patients. Diabetes is a major health problem that generates a significant burden, not only as the most common cause of renal replacement and blindness in young people but as a major risk for cardiovascular disease including myocardial infarction, stroke and amputation.
Effective clinical care for diabetic patients is often hindered by delayed or inadequate diagnosis and subsequent inappropriate treatment. Utilising the knowledge from leading European groups in the fields of beta-cell biology, biomarker discovery, diabetes genetics and clinical medicine the consortium aimed to address these challenges in diagnosis. Through the integration of the consortium's combined expertise the CEED3's core objective was the improvement of diagnostic tests able to differentiate patients into subsets or categories. Improved diagnosis would alter the treatment and management of diabetes, guiding the individualisation of therapy and having a direct benefit on patient care.
The timing of the project was highly relevant given the advances in basic and clinical science in the years before the project's inception. The human genetics of diabetes had progressed considerably allowing the ability to define both the rare monogenic causes of diabetes and the predisposing alleles for the common forms of diabetes and complications. The technical advances in polygenic analysis of common disease had also facilitated the use of cheap and effective large scale genome-wide analysis performed in large numbers of well characterised subjects.
CEED3 key objectives
In achieving its aims of individualising treatment across subgroups of diabetic patients, the project identified specific aims and objectives of its intended diagnostic tools:
(a) allow recognition of those who should be considered for molecular genetic testing, and identify patients within specific subgroups of monogenic diabetes;
(b) detect subgroups at high risk of beta cell dysfunction to help tailor treatment, follow-up or even identification of a novel treatment approach;
(c) improve identification of those patients at greatest risk of vascular complications. In all of these clinical areas the ability to differentiate subgroups of patients will alter and ultimately improve the treatment and management of diabetes.
The CEED3 project aimed to ensure progress went rapidly from scientific discovery to clinical advances in the form of validated diagnostic tools by bringing together experts in the laboratory with geneticists and clinical scientists.
The CEED3 objectives were achieved through a clear process of discovery of novel genetic and non-genetic biomarkers, validation within test and cohort data samples, development of clinical application and dissemination of this application.
In order to develop a clinically implemented diagnostic test, the core CEED3 pipeline consisted of:
- Discovery: Given the critical requirement of genetic and non-genetic biomarkers in order to establish effective diagnostic tests, the first steps for CEED3 were studies identifying new potential biomarkers as well as the integration of results from other existing scientific studies to identify all potential biomarkers.
- Validation: Potential markers were evaluated through validation in human samples (predominantly deoxyribonucleic acid (DNA) and plasma) from patients with known diagnostic status.
- Clinical application: Validated markers were then considered for clinical implementation and tested for diagnostic discrimination compared to existing clinical testing. A key consideration at this point was whether such tests were sufficiently robust and user friendly for widespread clinical use.
- Dissemination: Clinical tests that were considered clinically applicable and suitable for widespread use were then disseminated to clinicians internationally through the wide network generated by the leading experts brought together by CEED3.
These objectives provided the framework for CEED3 work packages, focusing the expertise of the consortium within the identified fields but feeding back to the broader aims of diagnostics development.
From this European-wide collaboration, CEED3 has successfully addressed its core objectives including:
(1) Development of an enhanced super national framework for research related to the development of diabetes diagnostic tools that assembles the best researchers, the best samples and data sets for primary discovery and subsequent validation, and the best development platforms for clinical tools to ensure widespread dissemination.
(2) The identification and validation of potential genetic and non-genetic biomarkers that can be used to differentiate between the discrete subtypes of diabetes using information from studies of humans and model systems.
(3) Development of a clinical tool to aid the diagnosis of monogenic diabetes integrating validated biomarkers with clinical phenotype assessment.
(4) The identification and validation of genetic and non-genetic biomarkers that can be used to identify individuals with early beta-cell dysfunction initial development of diagnostic tools based on these for potential use in the clinic.
(5) The identification of genetic and non-genetic biomarkers that can be used in early identification of patients with type 2 and monogenic diabetes at increased risk of developing chronic diabetic vascular complications.
(6) The dissemination of knowledge and provision of training on diabetes differential diagnosis and CEED3 results to all potentially interested patient, professional and societal groups.
Project results:
The main aim of CEED3 was to use a European collaboration to develop diabetes diagnostics. There have been three main areas where we have aimed to develop these diabetes diagnostics:
(a) differentiating subtypes of diabetes, particularly monogenic diabetes from the more common forms of diabetes;
(b) identifying markers of beta-cell dysfunction that would be useful in early diagnosis of diabetes and in monitoring the decline of diabetes; and
(c) markers for the development of complication.
The work carried out and main findings are described in detail below:
1. Differentiating subtypes of diabetes including monogenic diabetes
1.1 Differential diagnosis of diabetes: Basic science
1.1.1 Using genetic analysis of T1D and T2D
A massive international collaborative effort including Exeter, Oxford and Lund has enabled genetic data from large international cohorts, including those from the CEED3 partners to be integrated and analysed together. These have resulted in the identification of over 50 susceptibility variants for T2D. The search for genetic markers for T1D has resulted in over 70 variants being identified.
1.1.2 Using clinical resources and conventional and next generation sequencing technology to define prevalence of MODY and refine testing and develop potential diagnostic tests to T2D
Exeter has worked to identify how common MODY is in children with diabetes diagnosed before 18 years. They examined 391 patients from the 3 955 Swedish children from better diagnosis in diabetes study who did not have any pancreatic autoantibodies at diagnosis. Using Sanger sequencing they tested the common MODY genes and found 58 children with MODY (15 %). This is consistent with at least 1.5 % having MODY. This diagnosis was not known in over 60 % of cases and 45 % were on the wrong treatment with half of these being on unnecessary insulin.
To assess the prevalence in an older group and assess the role of novel sequencing methods. Lund and Paris have carried out a key project to sequence (Roche 454) 4 known MODY loci (HNF1A, HNF4A, HNF1B and glycokinase (GCK) in 541 GAD antibody-negative patients with age at onset of diabetes < 35 years from the 'All new diabetics in Scania' (ANDIS) project. This will allow detection of both high penetrance MODY causing mutations and also rare moderate penetrance variants predisposing to T2D. 32 variants are now being taken forward for validation using Sanger sequencing. This work (combined with the BDD) results for this region will allow the regional prevalence of MODY to be assessed in those diagnosed under 35 years
In addition, Lund and Paris have performed exome sequencing of 12 individuals with MODY-like diabetes (autosomal dominant diabetes with onset < 25 yrs of age). After filtering, a number of novel variants/mutations were found. We are now: (i) validating using Sanger sequencing;
(ii) sequencing in additional family members; and
(iii) exploring the functional consequences of the variants on the clinical phenotypes.
1.1.3 Generation of a collaborative cohort of DNA and serum / plasma samples from specific patient subgroups for early discovery and validation
Exeter, Oxford, Lund and Krakow have collaborated to produce large sample collections including samples from patients with monogenic diabetes, as well as Type 1 and T2D. These have been used extensively in biomarker testing and validation (Thanabalasingham et al Diabetologia 2011, McDonald et al Diabetes Care 2011, McDonald et al Diabetic Medicine 2011). Further details are in section '1.2 Assessment of diagnostic approaches in monogenic diabetes '.
1.1.4 Use of available samples for further rounds of proteomic and metabolomic biomarker discovery
Lund has shown, using state of the art metabolomics, that MODY2, due to mutations in the GCK gene, is metabolically a normal condition in contrast to all other types of hyperglycaemia (Spegel P et al. Diabetes 2012).
Helsinki pursued distinct T1D biomarkers by nuclear magnetic resonance (NMR) metabolomics. They initially observed dramatic differences between 800 T1D patients with no nephropathy form the FinnDiane Study and age and sex matched population controls from the Cardiovascular Risk in Young Finns Study. Unfortunately, these differences were later revealed to be strongly correlated with the difference in sample collection protocol between the two cohorts, and so were unable to replicate the biomarkers in a subset with identical sample procedures in cases and controls. Further work on the metabolic differences between T1D, T1D with nephropathy, and healthy controls are ongoing.
Oxford has also performed urine metabolic profiling in HNF1A-MODY, GCK-MODY and T2D, using both NMR and ultra-performance liquid chromatography - mass spectroscopy (UPLC-MS) (Gloyn et al, PLoS One 2012). This failed to detect any robust urinary markers for the MODY subtypes, but it was confirmed that those with HNF1A-MODY have raised urine glucose and relative aminoaciduria (which was no longer observed when correcting for glucose). Raised urine betaine was also observed in the HNF1A-MODY subjects, but a thorough examination of urinary 1-carbon metabolites did not replicate this finding.
1.1.5 Identification of candidate secreted proteins arising from mutations in HNF1A and HNF4A using genetic mouse models
Barcelona has profiled ribonucleic acid (RNA) from Hnf4a and Hnf1a deficient islets and liver from knockout mice to generate over 10 candidate secreted or shed biomarkers markers for testing in diagnostic assays and disease progression studies in HNF4A and HNF1A diabetes (Boj et al, Diabetes 2009, Servitja et al Mol Cell Biol 2009, Boj et al, Plos Genetics, 2010).
1.2 Assessment of diagnostic approaches in monogenic diabetes
1.2.1 Definition of key clinical data to discriminate between subtypes of monogenic diabetes from each other and T1D and T2D
Oxford has performed an evaluation of a set of approximately 250 clinical T1D and approximately 300 clinical early-onset T2D patients to estimate the extent of misdiagnosis of MODY, and to determine whether existing clinical pathways for differential diagnosis should be revised (Thanabalasingham et al, Diabetes Care 2012) Clinical features were defined that we hypothesised would be useful in differentiating HNF1A/4A- and GCK-MODY from T1D and T2D (as listed below). These features were then used as a pre-selection for genetic testing in our collections of young-onset T2D (n=300) and T1D (n=250).
Differentiating HNF1A-MODY from T1D: Subjects with significant C-peptide secretion post-honeymoon period were selected (C-peptide > 0.02 nmol/l or stimulated C-peptide increment > 0.02 nmol/l) and sequenced for HNF1A/4A and 3 HNF1A mutations, 8.5 % (n=21) of the apparent T1D fitted these criteria. Mutations were found in 10 % of those sequenced (1 % of the T1DM cohort).
Differentiating HNF1A-MODY from T2D: Subjects with either (a) age of onset up to 30 or (b) age of onset <45 yrs + absence of metabolic syndrome (IDF criteria) were selected. 28 % (n=82) of the apparent T2D fitted at least one of these criteria and were sequenced for HNF1A/4A. 10 HNF1A and 2 HNF4A mutations were found which was 4% of the overall T2D group (15% of those sequenced).
Differentiating GCK-MODY from T1D or T2D: Subjects with characteristics consistent with the specific features of GCK-MODY (FPG 5.5 - 8.5 mmol / l, HbA1c = 8.5 %, OGTT 2 hour increment < 4.5 mmol / l) and C-peptide positive were selected. 7.6 % (n=42) from the combined T1D and T2D groups were sequenced for GCK-MODY. 1 mutation was found in the T2D group (< 0.5 % of overall T2D, 2 % of those sequenced). Using the extended criteria more than doubled the cases of HNF1A/4A-MODY that would have been identified based on traditional guidelines (Diagnosis age<25, with parental history and evidence of endogenous insulin). Our strategy for finding GCK-MODY was less successful and this may have reflected that the HbA1c of 8.5% was too high studies from Exeter suggest 7.5 not 8.5 %.
In addition, Exeter investigated clinical characteristics (age at diagnosis, parent affected with diabetes, HbA1c, age at referral/recruitment, treatment, time to insulin (if applicable), gender and BMI) to discriminate MODY from type 1 and T2D and developed a clinical prediction model. This model utilises a weighted combination of clinical features to provide a probability on a continuous scale of whether a particular patient has MODY. The model performs well, with an ROC AUC of 0.95 for discriminating MODY from T1D and an AUC=0.98 for discriminating MODY from T2D, with similar AUCs when tested on an external dataset. This algorithm is now available as an online calculator at http://www.diabetesgenes.org/content/mody-probability-calculator with a link from the CEED3 website and has been published (Shields et al. 2012, Diabetologia).
Exeter has also produced a model based on the same clinical features to help identify HNF1A/4A MODY from GCK MODY (abstract accepted for Diabetes UK annual professional conference 2013). This model shows good discrimination (ROC AUC=0.88) and may help determine the order of testing until next generation sequencing is established.
1.2.2 Integration of data from ongoing efforts to identify phenotypic biomarkers with the potential to discriminate between diabetes subtypes
Exeter, Oxford, Lund and Krakow have collaborated to share data and samples from in excess of 3n 000 patients including 801 with MODY (400 HNF1A, 325 GCK, 55 HNF4A, 21 HNF1B), 730 with type 1 and approximately 1 521 with T2D which has allowed the assessment of the diagnostic potential of a number of biomarkers.
1.2.3 Testing of the ability of known and potential novel genetic and non-genetic biomarkers to differentiate subtypes of monogenic diabetes from each other and from T1D and T2D
Known biomarkers
(a) Urinary C-peptide/Creatinine Ratio (UCPCR): Exeter has shown a UCPCR > 0.2 nmol / mmol is an extremely sensitive (97%) and specific (96%) marker for differentiating MODY from T1D in adults outside the honeymoon period (Besser et al., Diabetes Care, 2011). Work in pediatric cases has shown UCPCR>=0.7 nmol / mmol is sensitive (97%) and specific in discriminating MODY and T2D from T1D >2y post-diagnosis (Besser et al., Pediatric Diabetes, 2013)
(b) Islet autoantibodies: Islet autoantibodies (glutamate decarboxylase (GAD) and IA2) were found in 82 % of type 1 patients but in < 1 % of MODY patients (Exeter), suggesting a positive result is very good at ruling out MODY in young-onset patients (McDonald et al., 2011 Diabetic Medicine).
(c) Lipoproteins: Samples from Exeter, Oxford and Krakow have been used to investigate lipoproteins in discriminating MODY from T2D. The most promising finding was HDL>1.12 mmol / L which had 75 % sensitive and 64 % specific for discriminating MODY from T2D (McDonald et al., 2012, Clin Chim Acta).
Novel biomarkers
(d) Highly sensitive CRP: Findings from genome-wide association study (GWAS) that common variation near HNF1A altered serum CRP level in population cohorts led to the hypothesis that inactivating mutations in HNF1A (causing MODY) would be associated with larger differences in CRP, measurable in blood, that could be used as a diagnostic biomarker. Oxford have now shown conclusively that highly-sensitive CRP (hsCRP) levels are significantly lower in HNF1A-MODY than in other forms of diabetes, providing good discrimination of HNF1A-MODY from T2D. These findings were initially demonstrated in UK samples (Owen et al Diabetes Care, 2010), then replicated in large European sample sets (Exeter, Oxford, Lund and Krakow) showing ROC Curve C-statistics ranging from 0.79 - 0.97 depending on the centre and sensitivity/specificity approximately 80 % (Thanabalasingham et al Diabetologia 2011, McDonald et al Diabetes Care 2011). It can also discriminate HNF1A and HNF4A-MODY, but is not a good discriminator from T1D. See more detail in validation section. The CRP result was a very important part of the CEED3 work that has been confirmed and extended in very large and comprehensive collaborative studies with the CEED3 partners. As CRP is widely available and can be measured cheaply in routine labs, this offered the chance of early use in clinical studies and its use in clinical practice has already been described (Besser et al BMJ Case Reports 2012).
(e) Glycan profile: HNF1A is a regulator of plasma protein fucosylation, with common variation associated with alteration of the plasma glycosylation profile. Oxford found the DG9-glycan index performs well in discriminating HNF1A-MODY from both T1D and T2D with a sensitivity and specificity 80-90% and ROC curve C-statistic > 0.9 (Thanabalasingham, Diabetes 2012). On paper therefore it seems to be a better biomarker, but is not yet available outside a research setting so would need more development before wide-scale translation is possible.
(f) 1,5-anhydroglucitol - 1,5-Ag was the first biomarker examined by Krakow in their HNF1A MODY patients. The initial finding by Krakow (Skupien et al., Diabetes Care, 2008), was subsequently replicated by Oxford as a biomarker for T2D. If you corrected for HbA1c it gave excellent discrimination from MODY (Pal et al. 2010). This established 1,5-AG a promising biomarkers for diabetes differential diagnosis. In 2012, Krakow started using an ELISA based kit for 1,5-AG and defined the detailed protocol for this method. This will be useful for future research that was held back by the difficulty in accessing the assay.
(g) Cystatin C: Cystatin C was a candidate biomarker for HNF1A MODY because some pathophysiological premises suggested that its level might be altered in this monogenic diabetes. The initial analysis done by Krakow showed lower cystatin C level in HNF1A MODY than in T1DM, T2DM, and in controls. This was not replicated by Exeter and Oxford. However, the consistent finding was that in HNF1A-MODY, cystatin C-based GFR estimate was higher than the creatinine-based one.
(h) Apo-M - early studies by Surich identified apoM as a candidate marker for HNF1A MODY due to reduced expression in Hnf1a-/- mice. However, subsequent human studies have yielded conflicting results. Therefore Oxford, Krakow and Surich aimed to use a highly specific and sensitive ELISA to evaluate apoM as a biomarker for HNF1A-MODY. This confirmed that ApoM is low in HNF1A-MODY, but is also low in T2D, so therefore offers poor discrimination. However ApoM does discriminate HNF1A-MODY from T1D where levels are normal (Mughal SA et al Diabetic Medicine 2013.). Like glycan profile, though, the ready availability of the assay is an issue for routine clinical practice.
1.2.4 Combining discriminatory clinical and laboratory markers into a diagnostic algorithm to estimate the likelihood of a diagnosis of specific types of monogenic diabetes
Exeter, including samples and data from Oxford and Krakow, has recently developed a model combining clinical features with biomarker results. The combined models show considerable improvement over clinical features or biomarkers alone (e.g. the model for early insulin treated patients discriminating T1D from MODY that includes clinical features, C-peptide and islet autoantibodies increases to an ROC AUC of 0.977 from 0.914 for clinical features alone, with a mean improvement in probabilities of 22 %). This has been presented at Diabetes UK and EASD conferences and written up for publication. We now intend to validate the model prospectively on all referrals arriving in the diagnostic laboratory.
1.2.5 Creation of a clinical application based on the diagnostic algorithm - accessible via the internet
The clinical prediction model for MODY based on clinical features developed by Exeter (Shields et al Diabetologia 2011) is now available as an online calculator at http://www.diabetesgenes.org/content/mody-probability-calculator with a link from the CEED3 website. A beta-version of the new model that incorporates both clinical features and biomarkers (Exeter, Oxford and Krakow) has been developed and is currently undergoing testing. This will be available on the website by March 2013.
1.3 Validation of diagnostic approaches in monogenic diabetes
1.3.1 Testing of the diagnostic algorithm in specific clinical settings to provide validation though sequencing for common monogenic genes
The clinical features algorithm was validated on an external test dataset of 350 patients not included in the original model by Exeter and showed similar high ROC AUCs > 0.94. The MODY calculator is currently being used in the Brussels' Diabase, a database of unselected diabetic patients, to assess if this helps detect novel cases of monogenic diabetes.
Oxford is continuing to evaluate biomarkers / combination of biomarkers for identifying cases of MODY in their patient resources with young-adult onset diabetes. This includes comparison of hsCRP and glycan profile for detecting HNF1A-MODY in young adults with T1/T2D and examining whether hsCRP can identify HNF1A-MODY cases in S Asian subjects with young-onset diabetes.
200 subjects with apparent T2D diagnosed < 30 years have been sequenced for common MODY genes. 7 HNF1A mutations were found. A comparison of the performance of hsCRP and DG9-glycan index in identifying the HNF1A-MODY cases: in this dataset both markers were highly specific (80%), but hsCRP had low sensitivity (60 vs 85 %).
S Asian subjects with diabetes diagnosed <50 years were selected for HNF1A sequencing on the basis of low hsCRP (< 1mg / l). The prevalence of HNF1A mutations was 4 % in those sequenced - a quarter of the number seen in a comparable N European group. Thus screening by hsCRP alone had a low yield of MODY cases and a combination of clinical features and biomarkers are most likely to be successful.
2. Development and validation of potential biomarkers for beta-cell dysfunction
2.1 Beta-cell dysfunction: basic science
2.1.1 Identification of DNA-based variants impacting on beta-cell dysfunction
The aim has been to use re-sequencing approaches to identify low frequency variants influencing risk of T2D. The rationale here is that the GWAS approaches are designed to capture common variant signals, but have relatively limited power to detect signals driven by variants of low-frequency (that is minor allele frequency < 5 %) or those that are rare (MAF < 0.5 %). The expectation has been that some of these variants will have a relatively large effect (by comparison with the modest effect common variants identified by GWAS) and that, if such variants can be identified, they might contribute to improved diagnostic precision.
Oxford has been approaching this in a variety of ways. Some of the interesting variants identified in HNF1A and other genes of specific interest which were identified in some of the early experiments (see previous reports), have been followed up but none of them has credible evidence of an independent contribution to deterioration of beta-cell dysfunction resulting in T2D. This has demonstrated that the medium-scale studies (sample size of approximately 1 000) that we had been planning are likely underpowered, and has shifted our efforts towards larger scale studies.
Through funding from CEED3 and other sources, Oxford has recently completed an appraisal and early implementation of targeted sequencing methods (that is, methods which aim to allow the power of next-generation sequencing to be efficiently deployed to examine regions of limited genomic extent, such as the coding regions of a small set of genes of particular research or diagnostic interest). From a technical perspective they have demonstrated that it is possible to recover robust inventories of sequence variation from pooled samples, and that optimal performance was achieved with Illumina's TSCA and Agilent's Haloplex methods. These efforts, which focus predominantly on the issue of population-scale re-sequencing (getting accurate and cost-effective sequence data from tens of thousands of individuals), are complementary to those of Exeter, which focus on molecular diagnostics (getting highly accurate data from individual patients).
Large-scale targeted sequencing studies have been postponed, pending the results from two large exome sequencing studies that are currently underway, in which Exeter, Oxford and Lund are directly involved. Through NIH and Wellcome Trust funding, the GoT2D and T2D-GENES consortia have completed exome sequencing of approximately 13 600 individuals (half T2D cases, half controls). Final analyses on the full data set will be run in early 2013, and will be used to plan targeted sequencing experiments (following up the strongest low frequency and rare variant hits) to be performed in up to 50 000 samples. Interim analyses of these data (for example in the approximately 3000 exomes from the GoT2D study) have shown evidence of enrichment of rare variant alleles across a set of genes known to be causal for monogenic and syndromic forms of diabetes, in T2D cases compared to controls, but with no single gene or variant showing compelling evidence of causal association at this stage.
Brussels have carried out Illumina sequencing of 5 human islet preparations, leading to the identification of 15 000 genes expressed in islets. Of these, 3 000 genes are modified by cytokine exposure, 1 300 by palmitate. Compared islet RNA expression to other tissues shows a very high correlation within islets, and low correlation with other tissues. > 60 % of T1D candidate genes are expressed in human islets and > 80 % in of T2D candidate genes. Exeter, Oxford and Lund have participated in a new meta-analysis applying the MetaboChip for the identification of novel genetic variants increasing risk of T2D (Morris A et al Nat Genetic 2012) as well as the first exome sequencing study to identify rare variants increasing risk of T2D (Albrechtsen A et al. Diabetologia 2012).
Lund applied a systems genetics approach on gene expression in human pancreatic islets and were able to identify a set of 20 genes, which explained 24 % of the variance in HbA1c (Taneera J et al. Cell Metab 2012). A first GWAS for glucose-stimulated insulin secretion during an OGTT has been performed, identifying variants in the GRB10 gene are not only associated with decreased insulin secretion but also show a clear parent-of-origin transmission. Surprisingly, this was associated with lower rather than increased glucose concentrations. Knock-down of the GRB10 gene using shRNA in human islets shows a strong reduction in glucagon, which most likely explains the paradoxical effect on glucose concentrations.
Though the CEED3 funding period is over, these ongoing studies will continue to inform our efforts to refine diagnostic pathways. The fact that we have needed to extend these discovery efforts to include in excess of 60 000 samples is evidence that there are relatively few, if any, low frequency variants of large impact influencing T2D risk, and that the studies originally planned for CEED3 were underpowered.
2.1.2. Identification of potential beta-cell enriched secreted proteins using in silico and in vitro approaches in isolated islets for prioritisation of candidate markers
Candidate markers: Barcelona and Brussels developed a bioinformatic pipeline that used a host of islet-cell and pancreatic exocrine and alpha cell datasets and prediction algorithms for secreted proteins which identified several beta-cell enriched candidates, including WNT4, SCG5, 7B2, PCSK1N, FAM3C, FAM3A, ELAVL4, PTPRN, DLK1, and Osteopontin: These candidates that were potential candidates for measuring beta-cell mass, beta-cell function and/or beta-cell was done using a pipeline consisting of:
(1) INS-1E cell Affymetrix microarray data sets, selecting genes with an expression = GCK;
(2) ingenuity pathway analysis to select secreted, cytoplasmic or unknown proteins;
(3) present in islets in BioGPS human tissue array;
(4) Well expressed in beta cells vs alpha, ductal or exocrine T1Dbase;
(5) well-expressed in massively parallel signature sequencing of human islets;
(6) protein expressed in human islets (human protein atlas).
The beta-cell enriched candidates, including WNT4, SCG5, 7B2, PCSK1N, FAM3C, FAM3A, ELAVL4, PTPRN, DLK1, and Osteopontin i and culture medium from human islet preparations was provided to Proteomika for setting up assays.
2.1.3 Identification of potential markers for beta-cell function using T2D mouse models and beta-cell lines
SFRP4: Using a co-expression Lund has identified SFRP-4 as potential marker for inflammation in islets of patients with T2D. In mice, SFRP4 results in impaired insulin secretion. It is a secreted protein and its serum concentration strongly predicted future T2D (Mahdi T et al. Cell Metab 2012).
BACE 1 and BACE2 and the global secretome and sheddome of MIN6 cells: The analysis of proteins released in the media of cell lines or primary cultures is a powerful tool to identify biologically active peptides and facilitate their evaluation as biomarker for cancer, neurodegenerative diseases and diabetes. Surich has employed the technique to explore BACE2 and BACE1 (secreted on beta-cell surface, responsible for Tmen27 shedding in beta-cells), but also to explore the global secretome and sheddome of MIN6 cells and murine isolated pancreatic islets. Together 594 proteins were identified in the conditioned medium of MIN6 cells, 240 of which were specifically detected by enrichment of N-linked glycopeptides. All major islet hormones (insulin, glucagon, somatostatin and pancreatic polypeptide) were detected among other known secreted proteins such as granins (chromogranins and secretogranins), transthyretin (TTR), islet amyloid polypeptide (iAPP) and neuropeptide Y (NPY). In addition several secreted factors were identified of which the roles have yet to be studied in the pancreatic beta-cell, including vitamin D-binding protein (DBP), a multifunctional plasma protein of which altered circulating levels are associated with T1D. The islet sheddome identified here consists predominantly of type I membrane proteins, some of which were previously found to be cleaved in different tissues or cell types. Furthermore, several glycophosphatidylinositol (GPI) - or lipid anchored proteins were detected in addition to other membrane-associated proteins, which are putative targets of ectodomain sheddases. Several secreted and shed proteins were found to be more abundant compared to the putative 'cytosolic contaminants'. Thus, insulin-2 (INS2) and insulin-1 (INS1) as well as protein of the chromogranin (CHG) and secretogranin (SCG) protein family were amongst the most abundant hits.
The functional consequences of ectodomain shedding and protein secretion can be diverse and depend on the function(s) of the particular protein or of the protein cleavage products. We therefore grouped the islet / beta-cell secretome and sheddome into functional categories using the PANTHER protein classification system. Nine major categories were identified including receptor proteins (12 %), hydrolases (12 %), signaling molecules (10 %), extracellular matrix proteins (8 %), cell adhesion molecules (8 %), ensyme modulators (8 %), defense/immunity proteins (7 %), proteases (6 %), and transporters (5 %). This heterogeneity of putative shed substrate proteins suggests that diverse and evolutionarily conserved biological processes are regulated by the activity of beta-cell surface proteases.
2.1.4 Identification of potential markers for beta-cell dysfunction through in vitro studies in human islets
Oxford and Lund have shown that the KCNQ1 locus exhibits a complex imprinting pattern with mono-allelic expression in fetal and bi-allelic expression in adults human islets. Lund has also explored the functional consequences of beta-cell specific genetic variants like TCF7L2 (Shou Y et al. Hum Mol Genetic 2011), GIPR (Lyssenko V et al Diabetes 2011.), PAX6 (Ahlqvist E et al. Diabetologia 2011), PCSK2 (Jonsson A et al. Diabetologia 2012) To discover beta-cell-specific gene products that could be employed as biomarkers, Barcelona and Brussels have carried out high-throughput sequence-based epigenetic and transcriptional maps of human islets, FACS purified beta-cells and pancreatic exocrine cells. This technology superseded our early plans to use exon arrays, which have a much more limited performance for this purpose. By subtracting maps from non-islet tissues, this has led to a catalogue of human islet-selective genes.
Brussels in collaboration with Oxford used modelling of lipofuscin accumulation in human beta-cells to show that the beta-cell population is established by age 20, i.e. human beta-cells are very long-lived and few new cells are being formed (Cnop et al. Diabetologia 2010). Brussels has performed the first genome-wide analysis of DNA methylation changes in human islets from T2D donors. 254 genes show differential DNA methylation in islets in T2D; this is a characteristic of the diseased tissue (islets) as there is no differential methylation in blood The RNA sequencing analysis of human islets from T2D donors is ongoing, and will lead to the identification of (i) beta-cell enriched or specific transcripts; (ii) potentially secreted proteins; (iii) alternative splice variants.
2.1.5 Assay development for measurement of candidate plasma beta-cell proteins
The list of prioritised candidate markers (WNT4, SCG5, PCSK1N, FAM3C, FAM3A, REG1A, REG3G, ELAVL4, PTPRN, DLK1, KIAA1324, Osteopontin) and culture medium from human islet preparations was provided to Proteomika for setting up assays. After completing a development phase, Proteomika successfully produced sandwich-type ELISA immunoassays for 4 markers: PTPRN, SCG5, FAM3A and FAM3C. All these assays were specific and functional. No assay could be produced for WNT4.
2.2 Beta-cell dysfunction: Assessment of diagnostic approach
2.2.1 Initial verification of biomarker assays in serum from subjects selected for extremes of beta-cell mass and function
The assays developed for 4 potential biomarkers (FAM3A, FAM3C, SCG5 and PTPRN) were expected to be secreted by human islet cells. Human islets were isolated and cultured by Brussels under several conditions to promote secretion of the markers. Supernatants of human islets were then collected and sent to Proteomika for marker detection. Determinations performed in such samples would serve mainly as proof of assay functionality before performing larger validation studies. Surprisingly, negative results were obtained for all samples and for all markers despite previously all immunoassays bring successfully validated against marker-overexpressing cell culture lysates. The high protein content on the cell culture media (BSA or FBS) could presumably prevent detection of markers. Since the presence of these proteins (BSA or FBS) was required for cell survival they could not be removed from the culture media.
A second marker validation was performed on serum samples from patients with T1D and with normal beta cell function were collected by Exeter and sent to Proteomika. Results for markers SCG5 and PTPRN did not discriminate the 2 patient groups. In contrast quantifications of FAM3A and FAM3C proteins did show reasonable but not total discrimination of T1D from normals. ROC curves were generated and AUC values of 0.89 for FAM3A and 0.835 for FAM3C were calculated. Combination of both markers in a multivariate model did not significantly improve the outcome (AUC 0.9). These promising values will be taken to a further validation study in a wider (more patient groups) and larger (more patients) sample population to evaluate their diagnostic power.
BACE2 and SES6L2 - loss-and gain-of-function studies of in vitro and in vivo models revealed non-redundant roles of BACE1/2 in ectodomain shedding with BACE1 regulating a broader and BACE2 a more distinct set of beta-cell-enriched substrates whose ectodomains can be measured in mouse and human plasma and may serve as biomarkers for beta-cell loss and dysfunction. From the set of validated putative BACE2 targets, two single-pass type I transmembrane proteins of the seisure 6 protein family were selected for further characterisation because of their BACE2 substrate specificity and enrichment in pancreatic islets. Expression analysis by qPCR in multiple mouse tissues revealed a marked enrichment of Ses6l and Ses6l2 in pancreatic islets. In contrast, Ses6, the third member of the seisure 6 gene family, showed low expression levels in islets and predominant abundance in the central nervous system. Together, this demonstrates that SES6L and SES6L2 are colocalised with BACE2 in pancreatic islets.
MicroRNAs as biomarkers: Barcelona has examined high throughput datasets of ncRNAs, including long non-coding RNAs (ncRNAs) (Moran et al, Cell Metabolism 2012), and small RNAs (Correa et al, unpublished; van de Bunt et al, PloS One, 2013, in collaboration with Exeter and Oxford). These studies have provided a collection of novel cell type-specific biomarkers. In particular, one microRNAs (miRNA) has been shown to be elevated during acute beta cell destruction in mice and humans undergoing human islet transplantation. This marker has been tested in a CEED3 cohort of patients with T1D and patients with neonatal forms of monogenic diabetes, showing that it correlates with beta cell damage. Future studies will test the utility of this marker for detecting ongoing beta cell damage in diverse clinically relevant settings.
Transcription factor 1 (Tcf1; hepatocyte nuclear factor 1a [Hnf1a]) is critical for hepatocyte development and function. We investigated if Tcf1 also regulates hepatic miRNAs. Tcf1-dependent miRNA expression in adult mice in which this transcription factor had been genetically deleted (Tcf1(-/-) ) was analysed using miRNA microarray analysis. The miR-192/-194 cluster was markedly down-regulated in liver of Tcf1(-/-) mice. MiR-192/-194 levels were also decreased in two other tissues that express Tcf1, kidney and small intestine, although to a lesser extent than in liver. In order to identify targets of miR-192/-194 in vivo we combined Affymetrix gene analysis of liver in which miR-192/-194 had been silenced or overexpressed, respectively, and tested regulated messenger RNAs (mRNAs) with multiple binding sites for these miRNAs. This approach revealed frissled-6 (Fsd6) as a robust endogenous target of miR-194. MiR-194 also targets human FSD6 and expression of miR-194 and Fsd6 are inversely correlated in a mouse model of hepatocellular carcinoma (Dgcr8(flox/flox) p53(flox/flox) × Alb-Cre). The results from this study support a role of miR-194 in Tcf1 dependent liver tumorigenesis through its endogenous target Fsd6. These results may have important implications for Tcf1-mediated liver proliferation in patients as HNF1A mutations can result in hepatic adenomas.
2.2.2 Testing the utility of genetic and non-genetic biomarkers for association with measures of beta-cell function. Genetic studies (LUND) will be done whilst awaiting validation of plasma assays
Lund has tested the role of novel biomarkers and / or genetic markers for prediction of T2D and deterioration of beta-cell function in man (Lyssenko V et al. Diab Vas Diab Res 2012; Ferrannnini E et al Diabetes 2012). As part of another European Union (EU) project (ENGAGE) we have identified a set of protein biomarkers with the ability to predict future T2D. This project is now in its replication phase.
Lund and Dresden have used extremes (highest and lowest decentile) of beta-cell function in the population to identify individuals (with serum samples available) for biomarker discovery in Zurich's laboratory. In addition to this, Lund are performing exome sequencing of families with extreme means (lowest and highest 10% of all families within the Botnia study). Seisure6l proteins are specific substrates of Bace2 in pancreatic islets. Ses6l and Ses6l2 levels are depleted in knock-out mice and found to be enriched in islets and brain. These are islet specific and beta cell specific, so were of potential interested as a biomarker for beta-cell mass. Surich used the stable isotope standards and capture by antipeptide antibodies (SISCAPA) method to develop immuno-MRM assays for the proteins of interest. Two of the tested anti-peptide antibodies, one against a human peptide of SES6L and one against a human / mouse conserved region of SES6L2, were found to be efficient in immuno-MRM assays. We demonstrated that the BACE2 substrates SES6L and SES6L2 circulate in plasma. Furthermore, detection of human SES6L in plasma of a humanised mouse model suggests that these proteins, although shed by a miniorgan, can be measured in the blood.
3. Development and validation of potential biomarkers for diabetic complications
Given the major EU funding of an Innovative Medicine Initiative (SUMMIT) with the aim to identify genetic and non-genetic markers for diabetic complications with over 8 times the total funding of CEED3, identification of genetic and non-genetic markers for diabetic complications has been mostly covered by SUMMIT rather than CEED3. Lund is the managing unit for the project which includes a number of other CEED3 partners (Exeter, Oxford, Helsinki).
3.1 Complications: Basic science
3.1.1 Finding genetic variants associated with complications in diabetes
Results from SUMMIT include a first GWAS for diabetic nephropathy (DN) in T1D has been published (Sandholm N, Plos Genet 2012) and several GWAS for diabetic nephropathy and retinopathy as well as cardiovascular disease in T2D have been completed recently.
In CEED3 we have focused on two projects, 1) testing whether known T2D-associated SNPs are associated with increased mortality and 2) a GWAS for free fatty acid (FFA) concentrations, especially during a meal.
Work with Lund and Exeter has shown a variant in the KCJN11 gene was weakly associated with increased mortality, especially mortality from CVD in patients with T2D. Also we have, together with other partners, shown genetic variants associated with T2D increase risk of Diabetic Nephropathy (Fagerholm E et al. Diabetologia 2012; Alkayaali S et al. Diabetologia 2012).
Postprandial FFA concentrations are strong predictors of the metabolic syndrome and CVD complications in T2D. They also show the highest heritability of any measured metabolic trait (Almgren P. Diabetologia 2011). Of the best loci in DGI GWAS of 1100 non-diabetic individuals FFA we replicated seven in Botnia Prevention, Prediction and Prevalence study (Botnia PPP, n=4900) but of these only rs2658404 on chr1 near ROR1 (receptor tyrosine kinase orphan receptor 1) and PGM1 (phosphoglucomutase) genes replicated in the second replication study (Metsim n=7000).
3.1.2 Identification of non-genetic biomarkers involvement in the development of nephropathy from ongoing studies of T1D
Helsinki has been applying NMR metabolomics to investigate the risk factors for T1D complications in 4 studies.
In the first study, we measured the serum metabolic profiles for 326 patients with T1D with normal AER, microalbuminuria or overt kidney disease. The NMR platforms provides information on 14 lipoprotein subclasses, over 20 low-molecular-weight molecules, and additional information on the circulating fatty acid species and their level of saturation. We found that sphingomyelin (odds ratio 2.53 P < 0.001) large VLDL cholesterol (odds ratio 2.36 P < 0.001) total triglycerides (odds ratio 1.88 P < 0.001) omega-9 and saturated fatty acids (odds ratio 1.82 P < 0.001) glucose disposal rate (odds ratio 0.44 P < 0.001) large HDL cholesterol (odds ratio 0.39 P < 0.001) and glomerular filtration rate (odds ratio 0.19 P < 0.001) were associated with kidney disease. Sphingomyelin was a significant regressor of urinary albumin (P < 0.001) in multivariate analysis with kidney function, glycemic control, body mass, blood pressure, triglycerides and HDL cholesterol (Mäkinen VP et al. Metabolomics, 2012).
In the second study, we compared the baseline data from the previously mentioned set of 326 patients with clinical follow-up of 8.2 years (Mäkinen VP et al. J Proteome Res, 2012). We also used the self-organising map in the data analysis, which is able to provide a comprehensive phenotypic profiling of metabolomics data. As expected, we saw the same connection between serum sphingomyelin and advanced kidney disease as before, but we also observed that patients with an unfavourable metabolic profile (high saturated fatty acids, high lipoprotein triglycerides and low HDL cholesterol) progressed to more severe kidney disease on average 10 years before than the patients with a favourable metabolic profile. In conclusion, we were able to show the strong connection between features that are the hallmarks of the metabolic syndrome in the general population and the progression of microvascular complications in T1D. Furthermore, the sphingolipid pathway may be an important player in the later stages when kidney function deteriorates (Makinen et al. J Proteome Res 2012).
The third study focused on the lipoprotein subclasses that were measured for the majority of the patients in the FinnDiane cohort (Mäkinen VP et al. J Intern Med, in press), We analysed the baseline biochemical data for 3 544 patients with T1D, and constructed a self-organising model to see which lipoprotein phenotypes were associated with diabetic kidney disease. We then compared these phenotypes to prospective data on kidney disease progression and mortality. We found a consistent association between elevated triglyceride content in all subclasses and the severity of kidney disease. A similar pattern was also observed for longitudinal changes in kidney status. Furthermore, we replicated the observation from the smaller studies that the metabolic syndrome characteristics were significant predictors of kidney disease progression and mortality. Interestingly, our findings also suggested that poor glycemic control in this cohort was not associated with lack of insulin, and we hypothesise that the same mechanisms and life style factors that contribute to obesity and T2D in the general population are the likely contributors to the progression of microvascular injuries and premature death in T1D.
The previous work was based on the population-based metabolic profiles in a fasting-like state. We also wanted to see if there were any differences with respect to the dynamic metabolic responses between patients with T1D and healthy controls. We measured the NMR metabolomics profiles from 40 cases and 40 controls during a standardised 24-hour meal experiment. Each participant ate an identical breakfast, lunch and dinner, and blood was drawn every two hours during the day. After controlling for the baseline concentrations of metabolites, we did not detect substantial differences between the diabetic and non-diabetic subjects. In particular, we observed substantial meal effects for triglycerides and branched-chain amino acids, but these responses were equally pronounced in both study groups.
3.2. Complications: Assessment in monogenic diabetes
3.2.1 Assessment of long-term complications in patients with different subgroups of monogenic diabetes
As part of the investigation of cardiovascular complication development in monogenic diabetes, Oxford has investigated clot structure and function, which determine predisposition to atherothrombotic disease, in HNF1A-MODY. This study compared clot structure / fibrinolysis and inflammatory markers in subjects with HNF1A-MODY compared to matched T2D and healthy controls and found that HNF1A-MODY subjects had a less dense clot structure than T2D, but greater than healthy controls. Lysis time of the clot was shorter in HNF1A-MODY than T2D. Fibrinogen and PAI-1 were similar in MODY and T2D, but C3 levels were lower in MODY. This indicates HNF1A-MODY subjects have a decreased thrombotic tendency than T2D and thus suggests an intermediate cardiovascular risk between T2DM and healthy controls.
Detailed measurements of complications and risk factors for complications have been assessed in approximately 100 GCK MODY patients, and a set of similar aged controls and young-onset T2D patients by Exeter. GCK patients have a similar prevalence of both microvascular and macrovascular of complications to controls, and a significantly lower prevalence of complications compared with T2D, despite an average of 50 years of raised glucose. Only retinopathy was slightly higher than in controls, and only consisted of mild non-sight-threatening background retinopathy with the majority of cases having < 5 micoraneurysms only. These results have implications for the management of GCK patients, where regular screening for complications is not necessary. In addition, these results also can be extended to show that the current target levels for glycaemia (which GCK MODY patients mimic), are appropriate for limiting complications.
Potential impact:
This work has had, and will continue to have a major impact for patients with monogenic diabetes, particularly the familial form presenting in early adulthood known as MODY. These patients are usually not recognised as having a different subtype of diabetes from the majority with Type 1 or T2D. This is of great importance as it is well established that these patients have a different response to treatment, specifically there is a 'pharmacogenetic' effect of their treatment response. This means that any method that helps doctors and patients to recognise that they do not have type 1 or type 2, but rather have an inherited form of diabetes will result in a marked improvement in their clinical care. As the majority of these cases run in families it will have an impact not only for them, but also for other members of their family. A major benefit for the patients is that many are unsuccessfully managed with insulin and other drugs and find that once a correct diagnosis is made they then can be appropriately treated, in many cases with sulphonylurea tablets or diet alone. Appropriate treatment reduces the amount of medication that the patient is taking which reduces side effects and costs both for the patient and also for the national government.
The impact will therefore be seen in terms of improved health of the patient, as their blood glucose is better controlled and consequently their risk of complications is reduced, and also in financial terms because of the money saved on treatment. For patients with GCK MODY, the treatment is stopped and the blood glucose control remains unaffected. Patients with HNF1A and HNF4A MODY respond particularly well to low dose sulphonylureas which are considerably cheaper than the more modern drugs used in T2D which are not effective in these subgroups. The improvements in diagnosing these patients represent a significant medical advance with major implications for individuals, as well as a socio-economic impact. The advances have involved the integration of clinical characteristics into an online probability calculator for MODY enabling any patient or doctor to go on a public website and enter their details to assess how likely they are to have a monogenic form of diabetes. This calculator has been widely validated and already used by over 5 000 doctors and patients.
The making of a diagnosis of MODY is improved by the establishment of a new biomarker, CRP, that discriminates these patients from T1D and T2D. Other biomarkers used for discriminating MODY from T1D that have been previously described have come from the collaboration of this large European consortium who have been able to be work out how they are to be used and to prove their effectiveness. All of these factors mean that for the clinician, by using just clinical interpretation, routine clinical data and simple investigations available at hospitals and clinics throughout Europe, they now have the tools to more systematically identify these patients. Further work has been carried out to assess the longer term impact of finding these specific types. Important work has been done on the complications of patients with GCK MODY. There was a major question about whether these patients were at increased risk, given that their blood glucoses were slightly raised throughout their life. Definitive studies have now been performed as a part of CEED3.
At the same time we have been able to establish that other proposed biomarkers, specifically the genetic biomarkers, even when combined in a test, were not sufficiently powerful to be able to be used at an individual level, although they did show significant differences at the higher level. This is important because it means that the emphasis should be put on the measures of non-genetic biomarkers rather than genetic biomarkers and shows that clinical features clearly outperform. This has shown that good science can provide important information regarding mechanism, but not necessarily inform patient care. Further impact of our work, not directly relating to patient care, is for industry where work by this group has started to identify specific markers of beta-cell presence i.e. beta-cell mass and also beta-cell destruction. These have proved very difficult to assay and hence the data are at very early stages, but could represent the first step towards being able to measure in man, whilst living, how many beta-cells they are producing. This work requires further development, but if such assays were available, it would mean that we would have an immediate way of assessing potential drugs that may be able to maintain beta-cell mass against novel trend of beta-cell loss as diabetes progresses. Also the work has enabled us to show when the cells are likely to be destroyed by therapy using microRNA. These findings, although at a preliminary level, have the potential to help the pharmaceutical industry in their search for assessment of novel drugs and for understanding mechanisms whereby new compounds work.
The other key impact of the CEED3 work has been in the development of the education of physicians throughout Europe and indeed the world. This has been achieved by many media. Clearly the conventional route of producing publications has been extremely successful with 107 produced as a result of work from the CEED3 partners during the period of the grant. These have been published in the very highest journals, particularly the top diabetes journals Diabetes Care and Diabetologia. Over 90 national and international talks have been given by members of the department talking about the identification of patients with MODY and why it is crucial to find these patients. Particularly important has been outreach programs aimed at reaching countries where genetic testing is not available and providing the potential for this and also disseminating information about genetic subgroups. Even when resources are limited, finding these patients can have a massive impact because it rapidly becomes cost effective due to the changes and improvement in therapy. The other key approach has been to provide this information to patients. This has been done by putting our talks on YouTube to enable patients to see things directly and by also having a public website and links to other national websites.
Examples of videos from the project shown on YouTube are shown here: http://www.youtube.com/user/ceed3eu
The exploitation of results is primarily within the healthcare system although further development of the findings from beta-cells will allow exploitation with pharmaceutical companies looking at changes in beta-cell mass or beta-cell dysfunction. This will be dependent on assays produce by our biotech companies. The biotech company Proteomika is working on the assays with a view to commercially developing these in conjunction with the academic partners.
List of websites: English, English / Polish
