Gene Panel Sequencing of Patients With Monogenic Diabetes Brings to Light Genes Typically Associated With Syndromic Presentations

Monogenic diabetes (MgD), mainly represented by maturity-onset diabetes of the young (MODY), is a genetically diverse group of diseases accounting for 2–3% of all diabetes. Identifying the gene has practical consequences for patients in terms of prognostics and therapeutic decisions, management of associated features, and genetic counseling (1). In practice, routine genetic testing of MgD consists of two distinct diagnostic strategies depending on clinical presentation. The first is based on gene panel sequencing (which we will refer to as NGS) for patients presenting a personal and familial history of MODY phenotype (2,3). The second diagnostic strategy is a targeted genetic analysis in case of diabetes with specific extrapancreatic features, e.g., HNF1B screening for patients with a combination of both diabetes and renal morphological abnormalities (4) or m.3243A>G mutation detection for patients with a clinical history highly suggestive of MIDD (maternally inherited diabetes and deafness) (5). However, studies of large cohorts have shown the clinical variability of both diabetes and associated features in these genetic subtypes (6–8) and have demonstrated the obvious bias in their clinical presentation according to the initial selection criteria of patients, in particular for HNF1B-related disease (6,7).

NGS offers the possibility of including rare forms of genetic diabetes or syndromic entities including diabetes. This is now performed in our routine genetic testing of MgD without any phenotypic a priori hypothesis. We report here a large cohort of patients with suspected MgD in whom the unanticipated involvement of genes associated with syndromic presentations shows the importance of testing large panels in MgD.

This study included a total of 1,676 probands who were referred for MODY genetic testing between 2017 and 2020. The study was performed in accordance with the Declaration of Helsinki. All patients provided written informed consent for genetic analyses. Clinical characteristics of the whole cohort can be found in Supplementary Table 1. None of these requests for genetic testing were already targeted toward a known syndrome; genetic analysis was performed solely based on MODY-leaning clinical and biological data (3). Once a molecular diagnosis of unsuspected etiology was established, clinical information was requested from the clinician.

This study cohort was compared with control cohorts of HNF1B (n = 145, previously reported [4]) or m.3243A>G (n = 74) patients identified through targeted analysis between 2014 and 2021.

A panel of 18 MgD genes including the m.3243A>G mutation (Supplementary Table 2) was analyzed from patient sample extracted from blood through the SureSelectQXT (Agilent Technologies) technology (according to manufacturer’s instructions) and run on a MiSeq instrument (Illumina). All regions of interest had 100% coverage with a minimal threshold of 30 reads at each nucleotide position. The targeted regions were screened for single nucleotide variants, indels, and copy number variations thanks to the SEQNEXT software, version v5.1.0 (JSI Medical Systems). Targeted analysis of m.3243A>G was performed as previously described (8). We classified variants following the Monogenic Diabetes Expert Panel (ClinVar) rules, adapted from the American College of Medical Genetics and Genomics guidelines (9).

GraphPad Prism 5.04 software was used for all the statistical analyses. Differences between groups of individuals were analyzed with Fisher exact test if the event was discrete and unpaired t test for quantitative variables.

All mutations identified in the study can be found in Supplementary Table 3. The clinical data generated or analyzed as part of this study are not publicly available due to patient confidentiality but are available from the corresponding authors on reasonable request. Clinical information on individual patients with specific MgD mutations is available on request to assist other laboratories and clinicians with variant interpretation and genetic diagnosis of patients.

Variants were identified in 307 (18.3%) patients (Fig. 1A and Supplementary Table 3). As expected, 72% of the diagnoses were made with one of the three most frequent causes: GCK, HNF1A, and HNF4A (Fig. 1B). The other patients (10.1%) were distributed among least common genes. Surprisingly, 17.9% carried a variant in a gene responsible for a syndromic disease. Given the very specific phenotype of GCK patients (Supplementary Table 4), we chose to compare the phenotypes of the patients bearing variants in a gene typically associated with syndromic manifestations (“syndromic patients”) with those with variants in traditional MgD genes, excluding GCK (“nonsyndromic patients” [Table 1]). Syndromic patients met traditional MgD inclusion criteria in terms of age at diagnosis and absence of obesity. In contrast, they had significantly less family history over three generations than nonsyndromic patients (29% vs. 60%, respectively). Our analysis showed that insulin secretion defect of syndromic patients was more severe, particularly at diagnosis. At that time, their HbA_1c was significantly higher (median 9.85% vs. 7.8%; 84 vs. 62 mmol/mol). This was in agreement with the circumstances of discovery of their diabetes, which was more often symptomatic, with 5 of them (out of 33 with symptoms) having diabetic ketoacidosis at diabetes onset. Likewise, syndromic patients were more often treated with insulin, both at diagnosis (56% vs. 32%) and at referral after a median evolution of 2 years (66% vs. 44%).

Twenty-five patients were found to carry the mitochondrial m.3243A>G mutation, accounting for 8% of all patients with a molecular etiology. These patients were compared with a cohort of 74 MIDD patients who had been referred for the analysis of the mitochondrial variant (Supplementary Table 5). The clinical and biological characteristics of their diabetes were not statistically different. Family history of the fortuitous patients with the m.3243A>G mutation showed that 56% of them were not evocative because incidence of diabetes was present in both maternal and paternal families, only reported in the father’s family, or not known (Fig. 2A). Deafness had been reported in 5 of 25 cases (Fig. 2B). Follow-up after molecular diagnosis showed that actually 14 of 25 had hearing impairment (56%) and that among the remaining patients, only 1 had a recent hearing test showing no impairment; the rest were not evaluated by hearing test. Retrospective interrogation revealed that six mothers also had hearing impairment (Fig. 2A). Macular dystrophy was not reported in any of these cases, and the sole ophthalmic defect was proliferative retinopathy in two patients. Among the other symptoms associated with MIDD, retrospective clinical evaluation identified cardiac anomalies in five patients (two with hypertrophic cardiomyopathy, two dilated cardiomyopathy, and one repolarization disorder), psychiatric features in three, stroke in two, and muscle weakness in three.

Given the high number of MIDD patients in our cohort, we screened a replication cohort of 536 patients negative for a previously analyzed panel of seven MgD genes, not including MT-TL1 (3). In this analysis we identified 15 additional fortuitous cases. Clinical characteristics replicated the findings from the initial cohort (Supplementary Table 5).

HNF1B was also prominently mutated in this series, representing 6% of patients with MgD. They were compared with a previously reported series (7) of 145 positive patients who had been directly referred for HNF1B analysis (Supplementary Table 6). HNF1B patients who were screened by NGS because their clinical history was not typically suggestive of HNF1B disease were found to have diabetes characteristics similar to those of patients whose analysis was targeted from the start. They had significantly more family history of diabetes, but the reported family history of kidney disease was not significant (Fig. 2C). They also had significantly less extrapancreatic features (kidney, genital tract, and pancreas morphological abnormalities [Fig. 2D and Supplementary Table 6]).

Finally, three additional genes involved in syndromic diabetes were found mutated in this study: WFS1, INSR, and PLIN1. Phenotypic data are given in Table 2. Nine patients with recessive WFS1 variants, either compound heterozygous or homozygous, were identified. As expected for a recessive disease, the homozygous cases were reported in patients with an ethnicity where consanguinity is a known feature (Africa, Middle East, and Ashkenazi Jews). Median age at diagnosis was 12 years (interquartile range 3–20). All WFS1 patients but two were insulin dependent from diagnosis. The only feature that could have suggested a Wolfram syndrome associated with recessive WFS1 mutations was a cataract reported in two of nine patients. Following molecular diagnosis, other extrapancreatic symptoms were described in a total of six of nine WFS1 patients. Lastly, INSR and PLIN1 were found mutated in three patients who all turned out to display insulin resistance features after postdiagnosis clinical evaluation.

This study brought to light the cumulative importance of very uncommon forms of genetic diabetes. Most importantly, it underlined that 17.9% of the patients with an MgD variant carried a defect in a gene associated with a syndromic form. It showed that the mitochondrial m.3243A>G mutation and variants in HNF1B or WFS1 were the most frequent causes of MgD after GCK, HNF1A, and HNF4A. These results underlie the undisputable benefits of NGS strategies.

Given the very specific phenotypes traditionally associated with these etiologies, special attention was given to certain features to determine whether these patients actually constituted new clinical entities. Deafness was thus found to be prominent at follow-up (56% of patients) among the m.3243A>G patients of this series. Lack of deafness signaling at referral is multifactorial. This feature may be subclinical, with hearing loss in high frequencies that do not impede the patient’s functioning. Patients might simply have neglected mentioning this information, supposing it unrelated to the diabetes they are consulting for. For some patients, deafness was actually indicated at time of referral, but, either because it was associated with an alternate cause or because other elements of the patient’s history did not match a MIDD diagnosis, this diagnosis was ignored. Indeed, family history of diabetes or deafness was uninformative in most cases. These findings emphasize the clinical variability associated with the m.3243A>G mitochondrial disease (10) and the fact that patients are certainly underdiagnosed considering the frequency of m.3243A>G mutation in the general population (11). The extent of these diagnoses cannot be attributed to a better sensitivity of the NGS technology, detecting cases with very low heteroplasmy. Indeed, the median heteroplasmy in blood for these patients was 27% (interquartile range 19.5–34), above the detection limit of our traditional technique (8). The previous underestimation of m.3243A>G cases, rather, lies in a total absence of this target from the sequencing panels of most of the reported MgD cohorts (3,12,13), except for one Korean study on 109 patients revealing 5 cases with the mitochondrial variant (14).

Comparison of HNF1B patients with those identified via HNF1B-targeted analysis shows that they escaped the specific HNF1B pipeline because they did not have as many associated features. Conversely, their family history was significantly enhanced (93.3% vs. 52.1%), since this inclusion criterion is requested for an MgD panel analysis, while it was not for HNF1B-targeted analysis, given the high proportion of de novo variants in this gene. The important share of HNF1B patients in our cohort cannot be explained by an underestimation of this etiology because of a moderate renal dysfunction that would be related to the gene deletion. Indeed, the proportion of patients with a whole gene deletion was not statistically different from that of the HNF1B-targeted cohort (7).

Syndromic forms of MgD previously assumed to be rare, particularly those related to WFS1, actually account for 4% of MgD diagnoses. They remain to be clinically characterized, as in differential diagnosis type 1 diabetes had been considered for some of these patients and type 2 diabetes for others. It should be noted that the p.Arg558Cys WFS1 variant identified in homozygosity in one patient was already reported as associated with mild forms of Wolfram syndrome and, in heterozygosity, with a higher risk of type 2 diabetes (15,16). Interestingly, biallelic WFS1 variants were recently reported in Chinese patients with nonautoimmune type 1 diabetes (17). Further studies on large cohorts should be undertaken to confirm whether restricted phenotypes or incomplete penetrance of associated manifestations are related to these syndromic genetic etiologies (18,19).

Restricting the analysis of syndromic genes to patients with the respective specific phenotypes would miss those with partial presentations. While it is undoubtedly important to identify those cases for patient care and family counseling, they highlight new challenges for the molecular biologist and for the clinician. Indeed, the reporting of these cases has potential major consequences echoing the incidental findings issue. Among the syndromic cases, those with WFS1 mutations are definitely the most delicate situations. Referral clinicians have expressed reluctance as to informing patients who had only sought medical opinion about their diabetes. When consent is given for molecular analyses, if large panels are analyzed, it should be made clear to the patient that the result might imply that diabetes is only one feature of a broader syndrome. Altogether this also highlights the importance of detailed clinical phenotyping, including extrapancreatic features. It helps the interpretation in genes still scarcely known. The knowledge of an uncommon feature might strengthen the plausibility of the pathogenicity of an unknown variant and is therefore crucial to the establishment of the molecular diagnosis.

Acknowledgments. The authors thank all the patients who participated in the study; the members of the Monogenic Diabetes Study Group of the Société Francophone du Diabète (Supplementary Material) for providing clinical data; F. Bellanger, S. Clauin, G. Leroy, C. Lemaitre, and P. Pellet (Department of Genetics, Pitié-Salpêtrière Hospital, Assistance Publique–Hôpitaux de Paris) for molecular genetic analyses; and Dr. Johanne Le Bihan (Department of Biochemistry, Pitié-Salpêtrière Hospital, Assistance Publique–Hôpitaux de Paris) for assistance with Prism software.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. C.S.-M. performed variant interpretation, collated and interpreted data, performed statistical analyses, and wrote the manuscript. D.B. performed variant interpretation, collated clinical data, and reviewed the manuscript. M.B. performed genotyping. C.B.-C. designed the study and reviewed and edited the manuscript. The Monogenic Diabetes Study Group of Société Francophone du Diabète referred patients for molecular analysis and provided clinical information. C.B.-C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts the m.3243A>G data were given as an oral presentation at the 7th Meeting of the European Association for the Study of Diabetes Study Group on the Genetics of Diabetes (EASD SGGD), Prague, Czech Republic, 16 –18 May 2019. These results were also introduced as a poster presentation at the Precision Diabetes Medicine 2020 virtual meeting, 8–10 April 2021.