Exome Sequencing in Children with Clinically Suspected Maturity-Onset Diabetes of the Young (MODY)

Abstract

OBJECTIVE:

Commercial gene panels identify pathogenic variants in as low as 27% of patients suspected to have MODY, suggesting the role of yet unidentified pathogenic variants. We sought to identify novel gene variants associated with MODY.

RESEARCH METHODS AND DESIGN:

We recruited 10 children with a clinical suspicion of MODY but non-diagnostic commercial MODY gene panels. We performed exome sequencing (ES) in them and their parents.

RESULTS:

Mean age at diabetes diagnosis was 10 (± 3.8) years. Six were females; 4 were non-Hispanic white, 5 Hispanic, and 1 Asian. Our variant prioritization analysis identified a pathogenic, de novo variant in INS (c.94G>A, p.Gly32Ser), confirmed by Sanger sequencing, in a proband who was previously diagnosed with “autoantibody-negative type 1 diabetes (T1D)” at 3 y/o. This rare variant, absent in the general population (gnomAD database), has been reported previously in neonatal diabetes. We also identified a frameshift deletion (c.2650delC, p.Gln884AsnfsTer57) in RFX6 in a child with a previous diagnosis of “autoantibody-negative T1D” at 12 y/o. The variant was inherited from the mother, who was diagnosed with “thin type 2 diabetes” at 25 y/o. Heterozygous protein-truncating variants in RFX6 gene have been recently reported in individuals with MODY.

CONCLUSION:

We diagnosed two patients with MODY using ES in children initially classified as “T1D”. One has a likely pathogenic novel gene variant not previously associated with MODY. We demonstrate the clinical utility of ES in patients with clinical suspicion of MODY.

Introduction:

Maturity-onset diabetes of the young (MODY), the most prevalent type of monogenic diabetes, is an autosomal dominant form of diabetes characterized by impaired insulin secretion, usually diagnosed before the age of 25 years (1). MODY is comprised of a heterogeneous group of disorders with variable genetic etiologies and clinical characteristics. Since its first formal description in 1975 (2), at least 14 genes (Table 1) have been reported to be associated with the MODY phenotypes to date (3, 4), with MODY2 and MODY3 being the most prevalent forms (5). Together with MODY1 and MODY5, these account for more than 90% of genetically confirmed cases (6).

Table 1:

The list of genes that have been reported to be associated with MODY.

Type	Gene	Gene name	Type	Gene	Gene name
MODY1	HNF4A	Hepatocyte nuclear factor-4-alpha	MODY8	CEL	Carboxyl-ester lipase
MODY2	GCK	Glucokinase	MODY9	PAX4	Paired box
MODY3	HNF1A	Hepatocyte nuclear factor-1-alpha	MODY10	INS	Insulin
MODY4	PDX-1	Pancreas/duodenum homeobox protein-1	MODY11	BLK	B-lymphocyte specific tyrosine kinase
MODY5	HNF1B	Hepatocyte nuclear factors-1-beta	MODY12	ABCC8	ATP-binding cassette, subfamily C, 8
MODY6	NEUROD1	Neuronal differentiation-1	MODY13	KCNJ11	Inward rectifier potassium channel 11
MODY7	KLF11	Kruppel-like factor 11	MODY14	APPL1	Adaptor protein, phosphotyrosine interacting with PH domain and leucine zipper 1

Misdiagnosis of individuals with MODY leads to suboptimal treatment regimens and surveillance strategies as well as delays in diagnosis of other family members (7–9). For instance, certain subgroups (i.e., MODY 1 and MODY 3) can be treated with sulphonylureas, (9) which target the specific molecular defect, resulting in improved glycemic control, lower risk of hypoglycemia and positive impact on the patient’s quality of life. In another subgroup (i.e., MODY2), correct diagnosis prevents unnecessary treatments and routine screening tests because of the low risk of diabetes-associated complications (9). As opposed to type 1 diabetes (T1D), there is no need for surveillance of autoimmune thyroid or celiac diseases in individuals with MODY. Additionally, a MODY diagnosis relieves the anxiety related to diabetes complications and other associated diseases. Thus, a definitive diagnosis of MODY significantly informs patient care, and would shed light on the natural history of MODY. Despite these remarkable benefits, studies have shown that only 5–7% of confirmed MODY cases were initially correctly classified as MODY (5, 6). For example, out of 47 individuals whose MODY diagnoses were genetically confirmed (MODY1, MODY2 or MODY3) in the SEARCH for Diabetes in Youth Study, only 6% of the individuals had been previously diagnosed correctly with clinical MODY (5). The majority of the misdiagnosed individuals are treated with insulin inappropriately.

As another barrier to genetic confirmation of MODY, the yield of genetic testing utilizing only well-characterized MODY-associated genes has been reported as low as 27% (10–14). This low positivity rate suggests the roles of other genes in MODY pathogenesis that are yet to be identified. Here, we aimed to identify novel genetic variants via exome sequencing (ES) in patients with a clinical suspicion of MODY but for whom a commercial MODY gene panel has been non-diagnostic.

Materials and Methods:

Recruitment and study participants

We identified individuals with clinical suspicion of MODY, and negative genetic test results on a commercial MODY gene panel using Texas Children’s Hospital (Houston, Texas) Electronic Medical Records (EMR) Population Health registry and individual referrals from pediatric endocrinologists at Texas Children’s Hospital. Patients with diabetes at Texas Children’s Hospital routinely undergo genetic testing for MODY based on clinical characteristics suggestive of MODY that include the following criteria: negative islet autoantibodies, family history of diabetes, diagnosis of diabetes at <25 years of age, and lack of typical features of type 1 or type 2 diabetes. Family history of diabetes was considered positive based on the presence of a family member in both first-degree and/or second-degree relatives with diabetes. Typical features of type 1 diabetes include relatively younger age at onset, presence of islet autoantibody(ies), beta-cell failure characterized by low C-peptide levels, diabetic ketoacidosis at diagnosis, lifelong exogenous insulin dependence, and lack of acanthosis nigricans. However, children with type 2 diabetes usually present with overweight/obesity, post-pubertal onset, acanthosis nigricans, presence of a family member with diabetes, lack of islet autoantibodies, high/normal C-peptide levels, and ability to manage their diabetes with lifestyle interventions and an oral medication (i.e., metformin) (15). We extracted clinical data on all patients with diabetes who were seen at Texas Children’s Hospital Pediatric Endocrinology Clinics between April 2016 and March 2019. Together with individual referrals from pediatric endocrinologists at our institution, we found 35 eligible candidates who had a clinical suspicion of MODY but a negative genetic test results on a commercial gene panel. During the study period, the most commonly used genetic panel was the monogenic diabetes 5-gene panel of Athena Diagnostics, that includes HNF4A, GCK, HNF1A, PDX-1, and HNF1B (MODY1 through MODY5, respectively). Less common was genetic testing through GeneDx (5-gene panel including the same 5 genes) and Baylor Genetics (25 genes).

The study was approved by the Institutional Review Board at Baylor College of Medicine and written, informed consent, and assent (when appropriate) was obtained from all participants in the study. The study enrolled 28 participants (10 families): 9 full parent-proband trios, and one singleton (proband only).

Sample collection, exome sequencing, and variant interpretation

Blood samples were collected using PAXgene Blood DNA Tubes following manufacturer guidelines. Genomic DNA was extracted from blood according to standard procedures. Whole exome sequencing was performed at the Human Genome Sequencing Center (HGSC) at Baylor College of Medicine. Exome sequencing was performed according to previously described methods (16). Briefly, dual indexed paired end libraries were pooled and hybridized to the HGSC Vcrome2.1 plus Spike-In design. These capture pools were sequenced in a multiplexed format (70 samples/lane) using the NovaSeq 6000 S4 flow cell reagent kit (300 cycles) and NovaSeq Xp 4-lane (NovaSeq V1.0 chemistry / NCS V1.6 instrument software). ES samples yielded an average of 12.7Gb and achieved 97.8% of the targeted bases covered at a depth of 20X or greater. Sequencing reads were aligned and mapped to the human genome reference sequence (GRCh37) with the in-house bioinformatics Mercury pipeline (17). Single nucleotide variants (SNVs) and short Indels were identified using Platypus (18). Variants were annotated using the Annovar software (19), including variant type (stop gain or loss, amino acid changes, splice site variants), minor allele frequencies in the gnomAD population database (20), and in silico prediction of variant deleteriousness based on CADD (21) and REVEL (22) algorithm scores. Gene-centric annotations included pLI scores from gnomAD (probability of being loss-of-function intolerant), and Domino database AD score (probability of a gene to be associated with autosomal dominant condition) (23). We performed three different variant filtering and prioritization strategies utilizing the parents’ medical history information and available trio data. In tier one, for probands with no family history of MODY or atypical diabetes and with parental ES data available, we prioritized de novo variants and rare (MAF < 1%) homozygous, and rare compound heterozygous potentially pathogenic variants (PPV: loss of function variants (LOF) (stop gain and stop loss, splice site SNVs, frameshift indels), and missense variants with REVEL score greater than 0.7). In tier two, for probands where one of the parents had a diabetes diagnosis with similar sub-phenotype to the proband, we prioritized rare (MAF < 0.1%), potentially pathogenic variants (PPVs) inherited from that parent, in genes with autosomal dominant inheritance as annotated in the OMIM database. In tier three, we prioritized rare (MAF < 0.1%), potentially pathogenic variants (PPVs) in 70 diabetes candidate genes (Table 2) in all 10 probands regardless of the inheritance pattern in the trio or known disease inheritance annotation in the OMIM database. Each candidate variant was visually verified using IGV viewer (24). Pathogenicity of each variant was also evaluated according to the American College of Medical Genetics and Genomics (ACMG) guidelines for variant interpretation (25).

Table 2:

Neonatal diabetes and MODY candidate genes (N=70).

ABCC8	EIF2S3	LMNA	RFX6
AGPAT2	FOXP3	LPL	SIRT1
AIRE	GATA4	LRBA	SLC19A2
AKT2	GATA6	MNX1	SLC29A3
APPL1	GCK	MTTL1	SLC2A2
BLK	GLIS3	NEUROD1	STAT1
BSCL2	GLUD1	NEUROG3	STAT3
CDKN1C	HADH	NKX2	STAT5B
CEL	HNF1A	NKX2–2	TNFAIP3
CISD2	HNF1B	PAX4	TRMT10A
COQ2	HNF4A	PAX6	WFS1
COQ9	IER3IP1	PCBD1	ZBTB20
CP	IL2RA	PDX1	ZFP57
CTLA4	INS	PIK3R1
DCAF17	INSR	PLIN1
DNAJC3	ITCH	POLD1
DUT	JAK1	PPARG
DYRK1B	KCNJ11	PPP1R15B
EIF2AK3	KLF11	PTF1A

Candidate variants validation

Candidate variants and their inheritance were validated by Sanger sequencing. Target amplicons were amplified from genomic DNA using conventional PCR (HotStarTaq DNA polymerase, QIAGEN) and PCR amplification products were analyzed by Sanger sequencing using established methods.

Results:

Study participants

Ten probands who had clinical characteristics suggestive of MODY but negative genetic test results on a commercial gene panel were enrolled in the study. Mean age at diabetes diagnosis was 10 (± 3.8) years. Six were females; 4 were non-Hispanic white, 5 Hispanic, and 1 Asian. Previously assigned diabetes types were T1D (n=7), unknown (n=2) and ketosis-prone diabetes (n=1). All of the participants had negative test results for at least 3 islet autoantibodies including glutamic acid decarboxylase-65 (GAD-65), ICA-512 and insulin autoantibodies. Five of the ten participants were also tested for zinc-transporter-8 (ZnT8) autoantibodies with negative results. All but one had a positive family history of diabetes. Eight children had a gene panel completed through Athena Diagnostic Laboratories. The demographic and clinical characteristics are summarized in Table 3. In two out of 10 probands, our exome variant prioritization strategy identified potentially pathogenic variants causing MODY.

Table 3:

Demographic and clinical characteristics of 10 probands with suspected MODY diagnosis. Participants in bold were found to have a pathogenic variant in the present study.

Case ID	Age at diagnosis (y)	Sex	Race & Ethnicity	Diabetes type per clinical diagnosis	Islet antibodies	Random C-peptide at diagnosis (ng/mL)	Random C-peptide post-diagnosis (ng/mL)	Previously negative MODY panel	Significant past medical history	Family History of Diabetes
1	13.7	M	Hispanic	T1D	Neg x4	0.39	1.66 (1-y pd) and 0.84 (2-year pd)	Athena 5-gene panel		Mother, maternal aunt, MGM, MGF
2	12.7	M	Hispanic	Atypical KPD	Neg x3	1.73	2.92 (3-wk pd)	BGL 25-gene panel		Mother, sister, MGM, PGM
3	12.1	F	Asian	T1D	Neg x4	1.67		Athena 5-gene panel	Pancreatic pseudocyst, necrotizing pancreatitis, hypertriglyceridemia	Mother, father, PGM, MGF
4	6.6	F	Hispanic	Unknown	Neg x4	1.16	0.49 (3-y pd)	Athena 5-gene panel	Depression/anxiety	Mother, sister, MGM
5	15.6	F	Hispanic	Unknown	Neg x4	1.04	0.19 (9-mo pd)	Athena 5-gene panel	Depression/anxiety	Mother, father, sister, MGM, MGF, PGM
6	7.4	F	NHW	T1D	Neg x3	0.21	1.02 (15-mo pd)	Gene Dx 5-gene panel		Mother, MGGM
7	12.3	F	NHW	T1D	Neg x3	2.21		Athena 5-gene panel		Mother, father, PGF, PGGM
8	3.7	M	Hispanic	T1D	Neg x3	0.39	0.62 (4-y pd)	Athena 5-gene panel		MGM, MGF
9	7.7	F	NHW	T1D	Neg x4	0.4	1.5 (3-mo pd)	Athena 10-gene panel		None
10	8.0	M	NHW	T1D	Neg x3	0.67	0.36 (2-y pd)	Athena 5-gene panel		MGM

Abbreviations: y, years; wk, weeks; mo, months; M, male; F, female; NHW, non-Hispanic white; T1D, type 1 diabetes; KPD, ketosis-prone diabetes; pd, post-diagnosis; Neg x4, negative for 4 islet autoantibodies including GAD-65, ICA-512, insulin and ZnT8 autoantibodies; Neg x3, negative for 3 islet autoantibodies including GAD-65, ICA-512 and insulin autoantibody; pd, post-diagnosis; BCM, Baylor Genetics Laboratory; MGM, maternal grandmother; MGF, maternal grandfather; PGM, paternal grandmother; PGF, paternal grandfather; MGGM, maternal great grandmother; PGGM, paternal great grandmother

Family A

The proband in family A (MODY case 8) was diagnosed with T1D at 3 years of age and enrolled in our study at 11 years of age. At the time of diagnosis, he had tested negative for islet autoantibodies including glutamic acid decarboxylase-65 (GAD-65), islet cell antigen-512 (ICA-512) and insulin autoantibodies. He had lean body habitus (age- and sex-adjusted BMI: 19^th percentile) at diagnosis of diabetes. He presented with marked hyperglycemia (serum glucose of 430 mg/dL; reference range (RR) in fasting: 70–99 mg/dL) but he did not have diabetic ketoacidosis. His random C-peptide at diagnosis was 0.39 ng/mL (RR: 0.81–3.85 ng/mL) and a follow-up random C-peptide 4-years post-diagnosis was 0.62 ng/mL (RR: 0.81–3.85 ng/mL). His hemoglobin A1c (HbA1c) at diagnosis was >14%, ranged 5.5–7.4% on follow-up visits with a recent increase to 9.6% at his last office visit at 13 years of age. His diabetes had been managed with multiple daily insulin injections including both short- and long-acting insulin analogs since diagnosis. He had several screening tests completed for other autoimmune diseases (i.e., hypothyroidism, celiac disease), all with normal results. His family history was significant for type 2 diabetes in both maternal grandmother and maternal grandfather. His maternal grandmother was diagnosed with diabetes at 58 years of age and her diabetes was initially managed with an oral agent followed by transition to exogenous insulin therapy. However, his maternal grandfather was diagnosed with diabetes at 56 years of age and has been managed with an oral agent only.

Our exome variant prioritization strategy identified a heterozygous de novo, missense variant c.94G>A, p.Gly32Ser (NM_001042376.3) in the Insulin (INS) gene. The variant was confirmed in the proband and was not detected in the parents by Sanger sequencing. This variant has been reported as pathogenic in ClinVar (VCV000021122.3), with strong pathogenicity prediction by multiple algorithms (REVEL score = 0.959, CADD score = 33). In addition, the variant has not been observed in the general population (gnomAD database).

Family B

The proband in family B (MODY case 7) was diagnosed with T1D at 12 years of age and enrolled in our study at 17 years of age. Her BMI at diagnosis was in the 43^rd percentile for age. She did not have acanthosis nigricans on physical examination. She was tested negative for four islet autoantibodies including GAD-65, ICA-512, insulin and ZnT8 autoantibodies. She presented with marked hyperglycemia (serum glucose 578 mg/dL, RR in fasting: 70–99 mg/dL) but she did not have diabetic ketoacidosis. Her random C-peptide at diagnosis was 2.21 ng/mL (RR: 0.81–3.85 ng/mL). Her HbA1c at diagnosis was 13.7%, ranged 7.7–13.1% on follow-up visits with HbA1c of 9.3% at her last office visit at 18 years of age. Her diabetes had been managed with multiple daily injections including both short- and long-acting insulin analogs since diagnosis. She had several screening tests completed for other autoimmune diseases (i.e., hypothyroidism, celiac disease), which all resulted normal except mildly elevated anti-thyroglobulin levels (2–7 IU/mL, RR: ≤ 1 IU/mL). The proband’s mother was diagnosed as a lean adult at 25 years of age with “type 2 diabetes”, which was initially managed with an oral agent followed by exogenous insulin treatment. The father was diagnosed at age 42 years and has been managed with an oral agent only since diagnosis, and paternal grandmother had a history of type 2 diabetes.

Our exome sequencing variant analysis identified a heterozygous frameshift deletion c.2650delC, p.Gln884AsnfsTer57 (NM_173560.4) in exon 19/19 in Regulatory Factor X, 6 (RFX6) gene. This variant was inherited from the mother, and confirmed in the proband and mother by Sanger sequencing. This variant has not been observed in the general population (gnomAD database). Although this variant would currently be classified as a variant of unknown significance based on strict ACMG variant interpretation criteria (VUS: PVS1_moderate, PM2, PP4), heterozygous protein truncating variants in RFX6 gene have been recently reported in patients with MODY (26, 27).

In the remaining eight probands, exome sequencing analysis did not identify any variants that could explain the proband’s diabetes phenotype. All variants identified in the three tiers of the exome data analysis are reported in Table 4.

Table 4:

All variants identified in 3 tiers of exome data analysis in 10 probands.

MODY Case #	Genomic Change (GRCh37)	HGVS variant annotation	Gene	Variant type	gnomad MAF (%)	REVEL	Inheritance
Tier One
8	1:152130386 C>G	NM_001122965.1:c.-20–1G>C	RPTN	splicing	0		de novo
8	2:189917641 G>T	NM_000393.5:c.2657C>A, p.Pro886His	COL5A2	missense	0	0.731	de novo
8	11:2182108 C>T	NM_001042376.3:c.94G>A, p.Gly32Ser	INS	missense	0	0.959	de novo
10	11:103306670 G>C	NM_001080463.2:c.12388–1G>C	DYNC2H1	splicing	0		de novo
Tier Two
1	5:1201932 G>A	NM_001003841.3:c.167G>A, p.Trp56Ter	SLC6A19	stopgain	0		mother
1	8:20074794 G>A	NM_001693.4:c.1225G>A, p.Gly409Arg	ATP6V1B2	missense	0	0.76	mother
3	1:35227042 G>A	NM_153212.3:c.187G>A, p.Val63Ile	GJB4	missense	0.02	0.728	mother
3	19:35530036 C>A	NM_001037.5:c.464C>A, p.Ala155Glu	SCN1B	missense	0	0.752	mother
4	15:48786451 T>C	NM_000138.5:c.2678A>G, p.Asp893Gly	FBN1	missense	0.01	0.862	mother
5	1:24646237 GGTGA>G	NM_198174.3:c.17+183_17+186delAGTG	GRHL3	splicing	0		mother
5	2:121736086 A>C	NM_005270.5:c.1445A>C, p.Lys482Thr	GLI2	missense	0.03	0.893	mother
6	5:110462529 T>C	NM_139281.3:c.2636T>C, p.Leu879Pro	WDR36	missense	0	0.79	mother
6	10:102748553 C>T	NM_021830.5:c.586C>T, p.Arg196Ter	TWNK	stopgain	0		mother
6	18:19378050 AGG>A	NM_020774.4:c.1099_1100delGG, p.Gly367Ter	MIB1	frameshift deletion	0		mother
Tier Three
7	6:117252529 TC>T	NM_173560.4:c.2650delC, p.Gln884AsnfsTer57	RFX6	frameshift deletion	0		mother

Discussion:

In a cohort of 10 probands who were suspected to have MODY but had negative test results on a commercial gene panel, we found two with potential molecular diagnoses of MODY using ES. We identified a de novo, missense variant in the INS gene (c.94G>A, p.Gly32Ser) and a frameshift deletion (c.2650delC, p.Gln884AsnfsTer57) in the RFX6 gene in two children previously diagnosed with T1D.

Variants in INS have emerged as a relatively common cause of neonatal diabetes (28, 29). Stoy et al reported 10 heterozygous INS variants associated with this phenotype, including five individuals with the same c.94G>A (p.Gly32Ser) variant who were all diagnosed in infancy (median age at diagnosis: 24 weeks) (28). Edghill et al also reported individuals from two families who were diagnosed in the first year of their life with the same INS variant (29). To the best of our knowledge, our case is the first report of c.94G>A variant in INS in association with monogenic diabetes beyond infancy. Delayed diagnosis does not explain the later age of detection in our case as he did not demonstrate the early-onset marked hyperglycemia or diabetic ketoacidosis seen with the INS c.94G>A (p.Gly32Ser) variant (28, 29). Molven et al also described two families with other INS variants (c.137G>A in a proband, his father and paternal aunt, and c.163C>T in a proband and her mother) diagnosed with diabetes in relatively older ages (i.e., 10–20 years of age) (30). The older presentation in our case and the cases in Molven et al’s report suggests the effect of yet to-be-determined genetic and/or environmental protective factors delaying the presentation of clinical diabetes. Understanding these protective factors may have implications in clinical care beyond individuals with variants in INS. It is also an example of phenotype variability and emphasizes the need to have a broad differential diagnoses.

In regard to the pathogenesis, as demonstrated in the mouse model, most of the INS variants are predicted to affect disulfide bonds and cause accumulation of misfolded proinsulin in the endoplasmic reticulum (ER), resulting in ER stress and beta-cell dysfunction (28, 31). Although the amino acid change in our case, p.Gly32Ser, does not appear to involve disulfide bridges directly, it blocks the formation of a nearby disulfide bond and thus impairs proinsulin folding, and induces ER stress (28, 32).

Regulatory Factor X6 (RFX6) is a transcription factor expressed in pancreatic islet cells and was previously reported to play a role in beta cell development (33). Homozygous and compound heterozygous variants in RFX6 were shown to cause Mitchel-Riley Syndrome, characterized by neonatal diabetes, hypoplastic pancreas, intestinal atresia and gallbladder hypoplasia (34, 35). However, the effect of heterozygous variants in RFX6 was unknown until a report showing the association of heterozygous, protein truncating variants of RFX6 with MODY in two unrelated families (26). In this cohort, haploinsufficiency of RFX6 was associated with lower gastric inhibitory peptide and beta cell dysfunction. Also, individuals with RFX6 heterozygosity showed variable onset of diabetes (27% at 25 years of age and 78% at 51 years of age) and were diagnosed with diabetes in adulthood (median age at diagnosis: 32 years, the youngest age: 24 years). Recently, Akiba et al reported a family consisting of a child, mother and maternal grandmother with diabetes who shared a heterozygous, protein truncating variant in RFX6 (27). The proband was diagnosed with MODY at 10 years of age while the mother was diagnosed with diabetes at 26 years of age during her pregnancy and the maternal grandmother was ascertained at 50 years of age. Our case expands the limited knowledge about the role of heterozygous, protein truncating variants of RFX6 in the clinical presentation and pathogenesis of MODY. The good therapeutic response to dipeptidyl peptidase-4 (DPT-4) inhibitors in three patients with RFX6 variants warrants further investigation and is an example of personalized medicine in diabetes management (36).

Studies conducted with families fitting MODY criteria but negative for mutations in known MODY genes provide further support for the role of other genes. In a study with 67 French MODY families, 11 families did not have mutations in the known MODY genes (37). Genome wide scans showed putative evidence for linkage to chromosomes 3, 5, 6 and 10 in 23 families from the UK fitting MODY criteria without a pathogenic variant in known MODY genes (38). In a different study examining 21 extended US families with MODY, not caused by known MODY genes, Kim SH et al. identified a novel genetic locus for MODY on chromosome 8p23 and a potential locus on 2q37 (39). Additionally, ES has been shown as a useful tool in detecting mutations in MODY in our study and others (40). Nine probands with MODY recruited from the Norwegian MODY Registry underwent ES following negative result on conventional diagnostic Sanger sequencing of the candidate genes (41). In these patients, ES led to a genetic diagnosis of MODY in at least 3 patients, which demonstrated the improved molecular diagnosis of MODY using this technique. In our cohort of 10 probands with negative results from commonly used commercial MODY panels, we had two (20%) with likely molecular confirmation resulting from the exome analysis. As more comprehensive genetic testing methods become widely available, our study adds to the growing body of evidence for the usefulness of ES in molecular diagnosis of MODY.

The strength of our study is the use of an advanced molecular diagnostic technique (i.e., ES) instead of gene panel approach, and sophisticated data interpretation tools in a racially/ethnically diverse children with clinically distinct diabetes phenotypes. ES and whole genome sequencing (WGS) offer many advantages over gene panels. Gene panels focus on a limited number of genes while ES and WGS assay all protein coding regions and entire genome, respectively. ES can identify ~85% of disease-related genetic variants despite covering only approximately 1% of the entire genome (42). WGS provides additional information due to its ability to detect variants in both coding and non-coding regions (43) and more comprehensive identification of structural variants (44). With increasing availability and reduced cost, both ES and WGS are emerging as clinically useful diagnostic techniques in many different diseases (43, 45, 46). The limitations include a small cohort size and lack of functional studies for confirmation. However, the diabetes phenotypes in both of our cases can be explained with confidence based on the evidence gathered from the literature as outlined above. Exome sequencing analysis will identify only coding SNVs and short indels, while many structural variants, and variants affecting regulatory regions would usually not be detected. Other technologies such a whole genome sequencing would enable the identification of a larger spectrum of sizes and types of SVs (44), as well as identification of noncoding variants and might lead to a diagnosis in the other 8 probands.

In conclusion, ES identified a candidate pathogenic variant in 20% of probands with atypical diabetes suspected of MODY but with negative genetic test results on a commercial MODY gene panel. Our case with c.94G>A (p.Gly32Ser) variant in INS gene is the oldest presentation of diabetes in an individual carrying this variant. Last but not least, our report expands the limited knowledge on heterozygous, protein truncating variants in RFX6 in clinical presentation and pathogenesis of MODY.