Recognition of maturity-onset diabetes of the young in China

Participants

A total of 76 unrelated families who participated in the “National Registration Project for Monogenetic Diabetes” initiated by the Chinese Diabetes Society from 2013 to 2016 were recruited from 35 hospitals (Table S1). All the families met the generally accepted clinical diagnostic criteria for MODY as follows: a positive familial history of diabetes, with at least two successive generations affected; early-onset hyperglycemia (at least one family member diagnosed aged <25 years); and non-insulin dependence (lack of a need for insulin treatment or a serum C-peptide level of >0.60 ng/mL, even after 3 years of insulin treatment)⁹. The mean age at diagnosis of the probands was 22.30 ± 9.33 years, and the mean body mass index (BMI) was 22.31 ± 5.64 kg/m². A total of 74 pedigrees were of Chinese Han ethnicity, and two were of Tibetan and Manchu ethnicity. The clinical data of all probands are shown in Table S2.

The study conforms to the provisions of the Declaration of Helsinki, and was approved by the Clinical Medical Research Ethics Committee of the Third Affiliated Hospital of Sun Yat-sen University, Approval No. [2013]2-106. All informed consent was obtained from all the participants(s) and/or guardian(s).

Target gene and whole-exome capture sequencing

Genomic deoxyribonucleic acid (DNA) was extracted from peripheral blood leukocytes using the QIAamp DNA Blood Mini Kit （Qiagen GmbH, Hilden, Germany）according to the manufacturer’s manual. The DNA concentration and purity were measured using a NanoDrop 2000 (Thermo Fisher Scientific, Wilmington, DE, USA). A 260 : 280 ratio of ~1.8 is accepted to indicate purity of DNA.

The library was prepared, and the genes of interest or exome were captured using the BGI MGISEQ-2000 （BGI, Shenzhen, Guangdong, China）platform. Briefly, genomic deoxyribonucleic acid was broken into 100–500 bp fragments by an enzyme kit (Segmentase; BGI), and the “A” base was then added at 3′ overhangs of 280–320 bp fragments, which were collected by magnetic beads. After adaptor ligation, the library was amplified by ligation-mediated polymerase chain reaction. Capture enrichment was carried out by array hybridization of the library with specific capture probes designed to target genes of interest or exon regions (Roche NimbleGen, Madison, WI, USA), followed by elution and post-capture amplification. Agilent 2100 bioanalyzer and BMG were used to determine the concentration of the products. Qualified single-chain library products were circularized into DNA nanoballs and finally sequenced on the MGISEQ-2000 platform of BGI.

Mutation screening and bioinformatics analysis

The flow chart of mutation screening is shown in Figure S1. Briefly, 76 probands from unrelated families with MODY were screened with the target gene panel of monogenic diabetes-related disorders (DX0270; BGI-Shenzhen Huada Gene Technology Co Ltd., Shenzhen, China), covering 13 genes previously associated with MODY (HNF4A, GCK, HNF1A, PDX1, HNF1B, NEUROD1, KLF11, CEL, INS, BLK, ABCC8, KCNJ11). All variants with a minor allele frequency <0.01 in the Genome Aggregation Database (GnomAD) v2.1.1 or ExAC (v1.0) database found by panel detection were further validated using Sanger sequencing for both probands and family members. For families without identifying genetic causes in the MODY-target gene panel sequencing, whole-exome sequencing was carried out to broaden the scope of genetic screening.

Functional prediction of new missense variants was carried out by SIFT, PolyPhen-2, Protein Variation Effect Analyzer (PROVEAN) or combined annotation-dependent depletion (CADD). A variant was predicted to be deleterious, with SIFT score <0.05, PolyPhen-2 HumVar score >0.95 and PROVEAN score <−2.5 or CADD Phred score >20.

Missense variants with a positive phyloP score and/or with a PhastCons value closer to 1 were predicted to be conserved¹⁰.

For pathogenic variants, the criteria for pathogenic or likely pathogenic in the American College of Medical Genetics/Association for Molecular Pathology Variant Interpretation Standards and Guidelines¹¹ were applied.

Clinical assessment and laboratory measurement

The following clinical data were obtained from patients and their family members: age, sex, nationality, native place, diagnostic approach, age at diagnosis, BMI, glycated hemoglobin (HbA1c), serum glucose, insulin and C-peptide (CP), lipid profile (total cholesterol, low-density lipoprotein, high-density lipoprotein and triglycerides), past medical history (e.g., hypertension, diabetic microvascular and macrovascular complications), and treatment.

A 75-g oral glucose tolerance test (OGTT) was carried out in probands and their family members without a prior history of diabetes to determine glucose tolerance and evaluate islet function. A steamed bread meal test (SBMT) was carried out for those with previously diagnosed diabetes. The steamed bread was made of 100 g flour containing carbohydrates equivalent to approximately 75 g glucose. Patients began fasting after dinner the day before the test. On the test day, after collecting fasting blood samples, patients ate the steamed bread completely within 15 min, and venous blood samples were collected at 30 and 120 min. SBMT- and OGTT- stimulated blood glucose, serum insulin and C-peptide are equivalent¹². In China, SBMT is often used as a simulation to replace OGTT to observe the postprandial changes in blood glucose, and islet function in patients with diabetes for safety and comfort reasons.

Glucose-related indices were calculated, including incremental 2-h glucose (∆2h-G) during the OGTT or SBMT, the area under the curve (AUC) of glucose (AUC_Glu), the AUC of CP (AUC_CP), AUC_CP/AUC_Glu (AUC_CP/G) and CP30’ increment/glucose30’ increment (∆CP30’/∆Glu30’). The AUC was calculated using the trapezoidal principle

Homeostatic model assessment of insulin resistance (HOMA-IR) and homeostatic model assessment of β-cell function (HOMA-B) were calculated using the following formulas: fasting insulin (µIU/mL) × fasting plasma glucose (FPG; mmol/L) / 22.5 and 20 × fasting insulin (µIU/mL) / (fasting plasma glucose; mmol/L − 3.5) (%), respectively.

Statistical analysis

Statistical analysis was carried out using Statistical Package for Social Sciences (SPSS for Windows, version 25.0, SPSS Inc., Chicago, IL, USA). Clinical data and laboratory parameters are expressed as the mean ± standard deviation. Comparisons of numerical variables between or among groups were carried out by Student’s t-test or the Kruskal–Wallis test; the χ²-test was used for categorical data. The P-value was corrected by the Bonferroni correction. All tests were two-sided. In all statistical analyses, P < 0.05 was considered to show statistical significance.

Mutation detection rate

Among 76 pedigrees, 32 mutations in six genes (Table 1) that were either previously described or novel likely pathogenic were identified in 31 pedigrees (Figure S2), accounting for 40.79% of MODY cases. The mutations in each proband, including 15 GCK mutations (MODY2), 12 HNF1A mutations (MODY3), two HNF4A mutations (MODY1), one KLF11 mutation (MODY7), one PAX4 mutation (MODY9) and one NEUROG3 mutation, are listed in Table 1. All mutations were inherited genetic alterations, except for two de novo mutations (HNF4A-MODY1 p.Cys808Ser and HNF1A-MODY3 p.Lys120Glu). Proband 16 carried compound GCK mutations inherited from her father. Thus, the frequency of MODY subtype was 18.42% for MODY2, 15.79% for HNF1A, 2.63% for HNF4A, and 1.32% for PAX4, KLF11 and NEUROG3. In addition, an ABCC8 variant is suspected to be causal of MODY12 (Table 1). All of these mutations are absent from or have a minor allele frequency <0.01 in the GnomAD and ExAC databases, and all missense mutations except HNF 1A p.Ile618Met affect a highly conserved residue with a positive PhyloP score and PhastCons score equal to or close to 1 (Table 1).

MODY2 mutations

A total of 12 different GCK mutations were identified (Table 1), including six previously described missense mutations. Four of the six new GCK mutations found in the present study are inactivating, as they caused truncation of proteins due to the introduction of a premature stop codon (p.Tyr289Metfs*5 and p.Tyr413Ter) or caused nonsense proteins due to the modification of the reading frame (p.Asp341-Ser418del and p.Leu415Alafs*44). Two new missense mutations (p.Asp160Glu and p.Lys169Glu) were predicted to be deleterious by SIFT, PolyPhen-2 and PROVEAN, with a CADD Phred score >20 (Table S3). The Arg191Trp, a previously reported and functionally verified mutation in white and Asian populations, was identified in four families (Table 1), making it the most common GCK mutation in the present study.

MODY3 and MODY1 mutations

Nine different HNF1A mutations were identified, all of which have been reported previously. Three known mutations (p.Arg54Ter, p.Arg271Trp, p.Pro291fs*26) were detected in two probands each. In addition, three mutations (p.Arg54Ter, p.Pro291fs*26, p.Val380Cysfs*39) are truncation mutations that represented 41.67% of the HNF1A mutations identified in our MODY3 population.

Two new HNF4A mutations were identified. p.Ser296Thrfs*31 is a truncation mutation, and p.Cys80Ser is a de novo mutation and suspected to be a harmful variant according to SIFT, PROVEAN and PolyPhen-2, with a CADD Phred score of 26.4 (Table S3).

Other MODY mutations

Other MODY mutations (MODY-others) included those in KLF11 (MODY7), PAX4 (MODY9) and a novel mutation in NEUROG3 (Table 1). None of these mutations have been described previously. KLF11 p.Lys453del causes a single amino acid deletion, which might have a deleterious effect on the function of the protein, and has damaging predictions by PROVEAN (Table S3). PAX4 p.Pro111Leu is predicted to be harmful by SIFT, PolyPhen-2 and PROVEAN, with a high CADD (Phred score = 28.7; Table S3). NEUROG3 p.Arg55Glufs*23 changes the reading frame, causes premature termination of translation and is considered a disease-causing mutation.

Suspected causal variant

ABCC8 p.Thr1115Met was found in a pedigree. This variant is absent from the GnomAD and ExAC databases of East Asia, is evolutionally conserved (Table 1), and is predicted to be damaging by SIFT, PolyPhen-2 and PROVEAN, and has a CADD phred score >30 (Table S3). Therefore, although the co-segregation between mutation and phenotypes could not be clearly defined (Figure S2), we treat it as a suspected causal variant of MODY in our study.

Clinical characteristics of MODY subtypes and MODYX

The participants with MODY subtypes included probands and their mutation-positive relatives, whereas participants with MODYX included only probands. As shown in Table 2, the carriers of MODY mutations had a mean age of 34.46 ± 17.91 years, a mean age at diagnosis of 28.79 ± 16.21 years, a mean BMI of 20.7 ± 3.58 kg/m² and a mean HbA1c of 6.65 ± 1.20%.

MODY2 showed mild fasting hyperglycemia (6.76 ± 0.90 mmol/L) and markedly elevated Glu120’ levels in OGTT or SBMT (11.76 ± 2.99 mmol/L). The mean 2-h glucose increment of OGTT or SBMT (∆2h-G) for MODY2 was 5.05 ± 2.99 mmol/L. A small increase in mean HbA1c was found in MODY2 (6.35 ± 0.44%). After dividing MODY2 participants into two groups on the basis of median ∆2h-G (5.61 mmol/L), those with ∆2h-G <5.61 mmol/L were significantly younger, younger at diagnosis, and had a shorter diabetes duration, lower BMI and higher ∆CP30’/∆G30’ than those with ∆2h-G ≥5.61 mmol/L (Table S4). Among MODY2 participants, 62.07% were treated with diet alone, 37.93% with oral hypoglycemic agents (OHA) and 10.34% with insulin (Table 2). The mean duration of MODY2 was 3.52 ± 3.90 years, and no diabetic vascular complications were reported (Table 2). Further analysis showed that MODY2 patients treated with OHA were significantly older, older at diagnosis and had a longer duration with a relatively lower HOMA-B value than patients treated with diet alone (Table S5).

MODY 3/1 participants had the longest duration (9.52 ± 8.25 years), the highest AUC_Glu and the lowest ∆CP30’/∆G30’ after OGTT or SBMT among MODY subtypes. Microvascular and macrovascular complications of diabetes, including diabetic retinopathy, nephropathy, neuropathy and stroke, were only reported in the MODY3/1 group. A total of 25% of MODY3/1 patients were treated with sulfonylureas, and 50% were treated with insulin.

MODY-others had the highest BMI, HOMA-IR and HOMA-B values among the MODY subtypes, and the frequency of overweight in MODY-others tended to be higher than MODY2 and MODY3/1 (Table 2). MODY7 and MODY9 showed a high proportion of overweight or obesity with a mean HOMA-IR of 2.86 ± 0.90 and a mean HOMA-B of 74.24 ± 41.33 (Table S6). The clinical phenotype of NEUROG3 p.Arg55Glufs*23 is characterized by hyperglycemia with mild intermittent abdominal pain (the proband complained of abdominal pain, and diabetes was detected during her visit; her sister also showed intermittent mild abdominal pain and hyperglycemia). There was a lack of abdominal pressure pain, lack of rebound tenderness and no diarrhea in the NEUROG3 mutation carriers, and no abnormality was found on laboratory examination of the abdominal gastrointestinal tract. A total of 55.56% of patients with MODY-others were treated with diet alone, and those treated with OHA and insulin accounted for 22.22% each.

MODYX patients had a significantly younger age and age at diagnosis; higher BMI, HbA1c and triglycerides; and lower high-density lipoprotein than MODY patients, and also showed significantly higher AUC_Glu, AUC_CP, ∆CP30’/∆Glu30’ and HOMA-B (Table 2). A significantly higher percentage of overweight was found in MODYX than in MODY patients. The proportions of MODY subtypes and MODYX patients treated with insulin were 28.79 and 55.81%, respectively (P = 0.005). A total of 16.28% of MODYX patients received initial intensive insulin treatment, whereas just 3.03% of MODY patients did. No statistically significant difference in other clinical features was found between MODY and MODYX patients.