Introduction

There is a high prevalence of young onset diabetes in India [1]. The prevalence of type 2 diabetes in youth in India varies from 25 to 40% in different studies [2, 3]. Maturity-onset diabetes of the young (MODY) is an important subset of monogenic diabetes with an estimated global prevalence of 1–4% in subjects with age of diabetes onset below 30 years [4]. 3.1% of subjects are clinically classified as MODY in the registry for youth onset diabetes in India (YDR) [3]. Apart from the obvious opportunity for genetic counselling, the diagnosis of MODY has therapeutic implications which include choosing not to treat mild hyperglycaemia (MODY 2) and appropriate selection of therapeutic agent (sulfonylurea in MODY 1 and MODY 3 and insulin in MODY 5) [5,6,7]. Normal BMI, strong family history of young onset diabetes, absent islet cell autoantibodies, absence of ketosis and responsiveness to oral hypoglycaemic agents are clinical pointers toward MODY. The original clinical criterion for a diagnosis of MODY is now known to have lower sensitivity than previously assumed [8, 9]. Proper categorisation of young onset diabetes is challenging in Indian subjects due to lower antibody positivity in type 1 diabetes, higher prevalence of fibro-calculous pancreatic diabetes and a lower BMI at diagnosis in patients with young onset type 2 diabetes [2, 10, 11]. MODY calculators and the biomarker approach with its limitations help in screening patients, but have not been well-validated in the Indian population [12, 13].

Genetic testing by Sanger sequencing alone is being increasingly replaced by multigene NGS panels [14, 15]. The last 10 years has seen significant advances in this direction [4]. Currently, at least 14 genetic loci have been associated with autosomal dominant MODY and the list is growing [4, 16, 17]. Interpretation of genetic variants detected by the aforementioned methods is sometimes challenging. American College of Medical Genetics guidelines 2015 and, ACGS best practice guidelines 2019, and the UK framework are used to arrive at the correct conclusions [18, 19]. Studies from different countries suggest that 10–33% of patients who are clinically categorised as MODY harbour disease-causing variants at the genetic loci studied [20,21,22,23,24,25, 29]. However, many of these variants which were earlier thought to be pathogenic are now classified as benign or variants of unknown significance (VUS). This has been made possible by large population-based genetic data sets published in the last 5 years [26, 27]. The current study describes our attempt at deciphering MODY genetics in the state of Kerala, South India.

Materials and methods

60 patients who had onset of diabetes between 3 and 30 years of age and a positive family history of DM (at least one parent/sibling with diabetes with an age of onset ≤ 35 years of age) were included in the study (Fig. 1). 17 patients were aged below 18 years. Patients with islet cell autoimmunity (positive auto antibodies to antigens-GAD 65, IA-2), clinical or radiological evidence of pancreatitis (absence of abdominal pain, steatorrhea, pancreatic calcification in ultrasound and X-rays), history of diabetic keto-acidosis (DKA) and identifiable forms of syndromic diabetes (lipodystrophy, Klinefelter syndrome, H syndrome, Wolfram syndrome (DIDMOAD), Thiamine-responsive megaloblastic anemia (TRMA), obesity syndromes) were excluded. Insulin use at enrolment was not considered as an exclusion criteria. C-peptide measurement was performed when random blood glucose was above 8 mmol/l. Oral hypoglycaemic agents were discontinued for 24 h prior to C-peptide testing. The family was informed of the need for genetic analysis of the parents or the sibling in the event of a positive result. Informed written consent was obtained from all participants. This study was approved by the Institutional Ethical committee (IRB No: IEC-AIMS-2017-ENDO-426 dated 23.11.2017). A detailed pedigree chart (Fig. 2) and the clinical data were recorded. MODY genetic testing was done using Targeted Next-Generation Sequencing at Christian medical college, Vellore for a comprehensive panel of 13 MODY genes (HNF1A, HNF4A, GCK, PDX1, HNF1B, NEUROD1, KLF11, CEL, PAX4, INS, BLK, ABCC8, KCNJ11) as per a previously published protocol [28]. In short, the targeted MODY genes were amplified using multiplex PCR followed by a library preparation which involved fragmentation of long PCR amplicons, barcoded adaptor ligation and size selection. Equimolar libraries were then further utilised for template preparation using Ion one touch OT2 emulsion PCR and enrichment using Ion ES. Sequencing was done on the Ion Torrent PGM using the Ion PGM 200 Sequencing Kit (Ion Torrent, Life Technologies), 316 chips (multiplex 8–10 samples).

Fig. 1
figure 1

Study flow

Full size image
Fig. 2
figure 2

Pedigree chart of variant positive patients

Full size image

Bioinformatic analysis

The generated sequencing data were mapped to the human genome reference hg19. The Torrent suit software with v5 (Life Technologies) was used for all analysis. The coverage analysis was calculated using Torrent Coverage Analysis, and potential pathogenic variants were identified using the Torrent Variant Caller and DNA STAR software (DNASTAR, Madison, WI, USA). The Human Gene Mutation Database (HGMD®Professional 2019.4), was utilised to classify the identified variants as reported or novel. Furthermore, the population sequencing database GnomAD was explored to validate the novel variants identified adhering to the latest guidelines by the American College of Medical Genetics. All novel variants were evaluated for sequence conservation and the likelihood of pathogenicity evaluated using Mutation taster, Mutation Assessor, PROVEAN, FATH MM, FATHMM-MKL, META SVM, METAR, Sorting Intolerant From Tolerant (SIFT) and LRT. Final variant classification was based on ACMG guidelines 2015 [18]. Variants with pathogenic/likely pathogenic variant categorisation were taken as clinically significant and positive. Sanger sequencing was performed to confirm all identified mutations and rare variants were recorded.

Biochemical analysis

This was performed including plasma glucose (fasting/postprandial), HbA1C, stimulated C-peptide, alanine transaminase and fasting lipids (cholesterol, triglyceride, LDL and HDL cholesterol). Biochemical tests were done using a Cobas C 8000 auto analyser from Roche diagnostics (Germany). Glycosylated haemoglobin was measured by the ion exchange high-performance liquid chromatography (HPLC) method (Bio-Rad 2 Variant II turbo glycated haemoglobin (HbA1c) analyser; CV 0.69%). GAD65 antibodies were quantitatively measured using ELISA (Euroimmune kit CV 4.7%).

Statistical tools

Statistical analysis was done using IBM SPSS software (Version 21.0, Chicago, IL, USA). For descriptive statistics, categorical variables were expressed as numbers and percentages.

Continuous variables were expressed using mean, median and standard deviation (SD).

Results

The baseline characteristics of the study population are summarised in Table 1. In total, NGS analysis came out with 29 variants in 27 individuals. Variant segregation data were recorded when family genetic data were available. The mean read depth of these samples was > 300X with > 99% with 20× coverage. More importantly, the variants were confirmed by Sanger sequencing. The assessment of pathogenicity was made by application of ACMG 2015 guidelines and confirmed with the Clingen pathogenicity calculator and Varsome. Genome aggregation database (GnomAD)-based population allele count and frequency were recorded for each variant.

Table 1 Baseline characteristics of the study cohort
Full size table

Pathogenic or likely pathogenic variants

Four out of sixty (6.6%) were positive for variants classifiable as positive (P/LP): one patient with HNF4A c.691C > T, (p.Arg231Trp), two with HNF 1Ac.746C > A, (p.Ser249Ter) and c.1340C > T, (p.Pro447Leu), and one with ABCC8 c.4544C > T, p.Thr1515Met). All the three variants in MODY 1 and MODY 3 genes were documented in paediatric age group (< 18 years). Three out of 17 paediatric patients (17.6%) had P/LP variants. Genetic characteristics of these patients are depicted in Tables 2 and 3.

Table 2 In silico analysis of pathogenic/likely pathogenic MODY variants
Full size table
Table 3 Variant pathogenicity
Full size table

Patient M1 (ABCC8:c.4544C > T(p.Thr1515Met) is a 35-year-old male with onset of diabetes at 20 years of age and on treatment with insulin. He had a BMI of 21.2 kg/m2 and C-peptide level of 0.65 nmol/l. Non-proliferative diabetic retinopathy had been documented at 31 years of age. His mother and brother both developed diabetes at around 33 years of age. Near-identical gene variants have been described previously in association with both congenital hyper-insulinism and MODY [29, 30]. He was planned for a trial of sulfonylurea; however, modest stimulated C-peptide, long-standing diabetes and poor follow-up dissuaded us.

Patient M2 is a boy with diabetes from 13 years of age carries a nonsense mutation HNF 1Ac.746C > A,

9p.Ser249Ter). He had a good response to sulfonylurea therapy and insulin could be stopped.

Patient M3 is a 12-year-old girl with diabetes from the age of 11 years carried a digenic cis variant HNF1 c.1340C > T (p.Pro447Leu /ABCC8 c.2152G > A (p.Gly718Ser). This MODY 3 variant is reported to be pathogenic [17, 31]. She had been maintaining good glycaemic control with very low dose sulfonylurea. Her mother, who has the same variant, developed diabetes at 15 years of age and has been maintaining adequate glycaemic control with a once daily dose of glimepiride 0.5 mg.

M4 is an 11-year-old boy on insulin since diagnosis one year prior to presentation. He had poor glycaemic control on insulin which improved with sulfonyl therapy. He carried the HNF4c.691C > T p.R231W variant. His mother had onset of diabetes at the age of 21 years and harboured the same variant. This variant has been reported in HGMD [32].

Variants classifiable as VUS and benign

Besides the pathogenic and likely pathogenic variants, other variants classifiable as VUS or benign were identified in classical genes (MODY 1 to MODY 3) and in non-classical genes like PAX4, Neuro D1, BLK1, PDX1, KLF 11 and CEL. Five patients had PDX1 c.670G > A(p.Glu224Lys) and one individual had PDX1 c.97C > A(p.Pro33Thr), which are currently considered as benign or VUS [33, 34]. The details of these variants are depicted in Table 4.

Table 4 Benign variants or variants of unknown significance (VUS)
Full size table

Discussion

The current study looked at the genetic confirmation rate in young subjects with multigenerational diabetes and reported rates which are lower than those reported previously. The genetically proven MODY detection rate based on the four common MODY types (HNF1A, HNF4A, HNF1B and GCK) has been shown to be lower in south Asians (SA) in a large population database from the UK [20]. Only 12.6% were genetically proven compared to 29.1% in the white Caucasian (WC) population. The detection rate in children was, however, similar (26.7% in SA vs 32.6% in WC). In the absence of variant description, it is unclear whether all the variants found in this study would be considered as pathogenic based on ACMG criteria. The genetic confirmation rates in recent studies from France, Ukraine, Greece and Turkey vary from 16 to 33% [22, 24, 25, 29].

Pathogenic/likely pathogenic changes were seen only in 6.6% of patients in the current study, despite using an NGS panel looking at 13 genes. Two out of nine patients (22.2%) aged less than 13 years were genetically proven to have MODY in our study, which is similar to the paediatric detection rate in a UK-based study [20]. All the variants in children were in classical genes (either MODY 1 or MODY 3).

Several Indian studies in the past that looked at MODY genetics were limited in nature, by virtue of the number of MODY genes screened [34,35,36]. There are two Indian studies, both from the neighbouring state of Tamil Nadu, which looked at the MODY genetic profile. Both studies had an inclusion strategy very similar to the current study, except that the study by Chapla et al. recruited subjects with an age of onset of diabetes less than 35 years compared to 30 years in the study by Mohan et al. In the study by Chapla et al., 56 patients clinically diagnosed to have MODY underwent NGS with a detection rate of 19.6% [21]. However, many of the variants considered pathogenic at that time are reclassified as benign/VUS now in light of ACMG2015 guidelines and new population data (PDX1 c.670G > A), HNF1A c1501G > T, NEUROD1 c723C > G) [33]. This study used an NGS panel incorporating 10 genes which did not include ABCC8. A more recent study (2018) from Chennai (Mohan et al.) included 152 subjects with a MODY diagnosis (age less than 30 years and satisfying Fajans’ clinical criteria) with a reported genetic confirmation rate of 15% [37]. 7.2% of these were noted in HNF1A and 3.3% were in ABCC8. There was one patient each with pathogenic variants in GCK and HNF1B genes and the rest were in non-classical genes like BLK, CEL, KLF11, PDX1 and KCNJ11 and newer gene variants (RFX6, WFS1, AKT2, NKX6-1) which are being recognised as contributing to early-onset diabetes in the heterozygous state. It should be stressed that, despite being a larger study both in terms of number of patients involved and the gene variants screened, the pathogenicity detection rate was lower than that reported in the first study. The role of KLF 11, PAX4 and BLK as MODY genes has been either disputed or refuted in recent years [38].

The variant in subject M1 (ABCC8NM_000352.5:c.4544C > T(p.Thr1515Met) was interesting (https://www.ncbi.nlm.nih.gov/snp/?term=rs769989185 has details of alternate transcripts). The likely pathogenic variant of ABCC8 p.Met1514Thr was reported in a recent study from Greece [29]. This is the same variant found in our study with an alternate transcript. Heterozygous variants (ABCC8c.4543A > G p.T1515A variant and ABCC8 c.4547C > T p.Thr1516Met) were previously reported as pathogenic in babies with congenital hyper-insulinism [30]. No known history of hypoglycaemia prior to the onset of diabetes was noted in this patient. However it has been reported that the parents of patients with congenital hyper-insulinism, who have a genetic variant in the heterozygous state, develop early-onset diabetes without a history of preceding manifestations of congenital hyper-insulinism [39]. Congenital hyper-insulinism is also known to evolve into a state of early-onset diabetes, both in the heterozygous and homozygous state [39, 40]. 8% of patients with a clinical diagnosis of MODY, but negative for MODY 1, MODY 2 and MODY 3 gene variants, were found to have pathogenic variants in the ABCC8 gene in a study from the UK [41]. A study from Singapore has also documented pathogenic variants in the ABCC8 gene, suggesting that among the non-classical genes; this is perhaps the most frequent one that contributes to the MODY spectrum. MODY due to ABCC8 variants is known to respond to sulfonylurea therapy, thereby underscoring the therapeutic implications of a precise genetic diagnosis [42]. Non-classical genes, especially ABCC8, have been reported as possible contributors to MODY pathogenesis in studies from Italy and Brazil [42, 43].

Assigning pathogenicity to gene variants is challenging. Many of the gene variants reported as pathogenic in the past are now known to be tolerated variants without phenotypic manifestations [26]. Patient M5 exemplifies this dilemma. This individual had a novel variant inABCC8 (p.V1165M) which had a likely pathogenic variant output in the variant classifier Varsome. However, the allele count of 18 (allele frequency 0.000588) for the South Asian population in the GnomAD database makes this conclusion questionable. Five patients had PDX c.670G > A(p.Glu224Lys), which was earlier thought to be pathogenic but is currently considered to be benign/VUS in view of its high prevalence in controls [33].

This study has several limitations. The limited number of analysed samples makes it impossible to derive prevalence of pathogenic or likely pathogenic positivity rates in subjects clinically diagnosed as MODY. The NGS methodology used in our study could have missed copy number variations [44], intronic and promoter variants. A backup MLPA strategy could have offset these short coming, at least partially [45]. The main strength of the study is the strict adherence to current guidelines for variant classification.

The question which naturally arises is the genetic make-up of the patients without pathogenic variants. Several explanations can be postulated, including an early-onset type 2 diabetes, intronic variants or copy number variations [44] (large deletions and duplications) which may be missed with the currently used sequencing technology or presence of as yet undiscovered genes (MODY X).

Conclusion

The detection rate of MODY-related pathogenic or likely pathogenic gene variants, by careful application of ACMG 2015 guidelines, was lower in this cohort of patients with a clinical diagnosis of MODY compared to the western literature. The contribution from some of the non-classical MODY genes, especially ABCC8, was evident in this study.