MODY probability calculator for GCK and HNF1A screening in a multiethnic background population

ABSTRACT Objective We aimed to identify the frequency of monogenic diabetes, which is poorly studied in multiethnic populations, due to GCK or HNF1A mutations in patients with suggestive clinical characteristics from the Brazilian population, as well as investigate if the MODY probability calculator (MPC) could help patients with their selection. Subjects and methods Inclusion criteria were patients with DM diagnosed before 35 years; body mass index < 30 kg/m2; negative autoantibodies; and family history of DM in two or more generations. We sequenced HNF1A in 27 patients and GCK in seven subjects with asymptomatic mild fasting hyperglycemia. In addition, we calculated MODY probability with MPC. Results We identified 11 mutations in 34 patients (32.3%). We found three novel mutations. In the GCK group, six cases had mutations (85.7%), and their MODY probability on MPC was higher than 50%. In the HNF1A group, five of 27 individuals had mutations (18.5%). The MPC was higher than 75% in 11 subjects (including all five cases with HNF1A mutations). Conclusion Approximately one third of the studied patients have GCK or HNF1A mutations. Inclusion criteria included efficiency in detecting patients with GCK mutations but not for HNF1A mutations (< 20%). MPC was helpful in narrowing the number of candidates for HNF1A screening.


INTRODUCTION
T he frequency of monogenic diabetes mellitus (DM) has been underestimated in various populations (1). Most cases are caused by mutations in GCK [glucokinase gene] (GCK MODY [maturity onset diabetes of the young]) or HNF1A [hepatocyte nuclear factor 1-alpha gene] (HNF1A MODY) (2). Their molecular diagnosis is expensive but promotes the improvement of genetic counseling and treatment (3). Strategies to select the ideal subjects to screen for monogenic DM (MDM) in different populations are necessary to establish cost-effective diagnostic algorithms.
Different authors have developed clinical criteria for MODY screening based on age, family history and clinical characteristics (4,5). Although their use appears to be cost-effective (6), there is a concern that the screening based on clinical criteria would either miss part of the affected patients or still result in an excessive number of genetic tests (7). Therefore, authors have pursued optimal strategies for selecting patients. Shields and cols. developed a clinical prediction model that generates a probability of MODY (8) and shows good discrimination between MDM and type 1 (T1DM) or type 2 DM (T2DM) in European patients diagnosed under 35 years. The performance of this calculator in non-Caucasians is unknown. The Brazilian population is very diverse and comprises individuals from multiple ethnic backgrounds, especially Caucasoid and Afro-descendants. There are scarce data about the prevalence of MDM and its optimal screening strategy in this setting. Our aim in this study was to estimate the frequency of MDM due to GCK or HNF1A mutations in patients with suggestive clinical characteristics and to investigate if the MODY probability calculator (MPC) could improve patient detection in this population (8,9).

SUBJECTS AND METHODS
In this cross-sectional observational study, we selected patients clinically defined with monogenic diabetes from two specialized centers in Rio de Janeiro, between March 2012 and June 2015. MODY screening is not part of the routine laboratory panel of either center. We analyzed thirty-four unrelated probands from Brazilian families for mutations in GCK and HNF1A.
The inclusion criteria were age of DM diagnosis ≤ 35 years, body mass index (BMI) ≤ 30 kg/m 2 or 95th percentile at onset, negative anti-glutamic decarboxylase antibody (anti-GAD) and anti-islet antigen 2 antibody (anti-IA2) antibodies and family history of diabetes in at least two generations, excluding the generation of the index case, and/or two or more first-degree relatives at the same side of the family. We excluded patients with T1DM; past diabetic ketoacidosis; clinical signs of insulin resistance (acanthosis nigricans, increased abdominal circumference and obesity); and secondary causes of diabetes.
The Ethics and Research Committee of the Clementino Fraga Filho University Hospital and State institute of Diabetes and Endocrinology of Rio de Janeiro approved this study protocol. We informed all participants about the aim of this study and provided verbal and written consent.
We calculated the positive predictive value (PPV) for MODY based on the MPC for each patient (8) and divided the patients into two groups. The GCK group included patients with fasting hyperglycemia (100-154 mg/dL); increased glycaemia after 75 g anhydrous dextrose <54 mg/dL and HbA1c < 7.5% (58 mmol/ mol); and evolutionarily stable disease (even without antidiabetic drugs), most often asymptomatic and with hyperglycemia since birth (10). The HNF1A group included all other cases that met the inclusion criteria and that did not have the profile for the GCK group.
We isolated genomic DNA from peripheral blood leukocytes using QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany). Also, we purified PCR products using Clean Sweep PCR Purification Reagent (Applied Biosystems, Vilnius, Lithuania). Then, we performed screening of the entire coding sequence of GCK and HNF1A genes through bidirectional Sanger sequencing using the Big Dye Terminator Kit v3.1 (Applied Biosystems, Austin, TX, USA), conducted on an ABI 3130 Automatic Genetic Analyzer (Applied Biosystems). Primers sequences are available upon request. We confirmed all mutations by bidirectional sequencing of a second PCR reaction. Then, we estimated the serum levels of anti-GAD and anti-IA2 antibodies by means of an enzyme-linked immunosorbent assay method (ELISA) using a EUROIMMUN kit.
We evaluated the differences between patients with mutations and others with Student's t-test and chi-square tests. In addition, we performed statistical analysis using SPSS software (version 22.0).
None of the patients in the GCK group used insulin, and one used OAD. Most patients (70.3%) in How to identify multiethnic MODY mutations?

Genetic tests
We found MODY mutations (GCK or HNF1A) in 11 patients (32.3%). Fifty per cent of patients with nonclassified DM had mutations in GCK or HNF1A. We fond mutations in 16.6% of those previously classified as T1DM and in none classified as T2DM. The only patient classified as GDM had a GCK mutation.
Among the 11 mutations, eight had already been described, and three were novel mutations. The mutation p.Tyr61Asp (c.181T>G), found in exon 2 of the GCK gene of one patient, is a missense mutation classified as pathogenic. The other two novel mutations occurred in the HNF1A gene (exons 2 and 6). We considered the missense mutation p.Val133Glu (c.398T>A) and the frameshift insertion p.Thr433Hisfs*116 (c.1296_1297insC) pathogenic because they alter the codon reading frame due to the insertion of a nucleotide ( Table 2).  We recruited the family of all three probands with novel mutations, as presented in the pedigrees (Figure 1). We recruited four family members of the patient with the p.Tyr61Asp mutation in GCK. Then, we tested all four individuals with diabetes. The brother and sister had the same p.Tyr61Asp mutation, and we observed that it was inherited from the mother with diabetes. The father with recent onset DM (after 50 years of age) did not have the mutation. We recruited the mother with DM of the patient with the mutation p.Thr433Hisfs*116. She also had the same mutation. Four family members of the patient (three sisters and the mother) with the mutation p.Val133Glu in HNF1A were recruited. The only sister with diabetes also had the same mutation of the patient, and they inherited the mutation from their healthy mother. We believe it may be a case of incomplete penetrance uncommon with HNF1A mutations but already observed by other authors (14,15). In addition, the mutation p.Val133Glu was absent in the two healthy sisters.

GCK group
Six patients had PPV ≥ 75%, and 1 had PPV between 50 and 75%. We detected GCK mutations in 5 patients with PPV ≥ 75% and in patients with PPV between 50 and 75%, but not in 1 individual with PPV ≥75%.

DISCUSSION
In this study, we identified patients with phenotypes suggestive of MDM and performed mutation screening for GCK and HNF1A genes. We found mutations in 32.3%.

B.
How to identify multiethnic MODY mutations?
Arch Endocrinol Metab. 2020;64/1 Although a high frequency of mutations in these two genes have been reported in individuals with a clinical diagnosis of MDM from several populations, especially from Northern Europe (16,17), others (including Southern Europeans, Asians and Brazilians) have found a lower frequency of mutations, as true with our findings (18)(19)(20)(21).
In some populations, other types of diabetes might share clinical features with MDM more frequently than others. Alternatively, other genes implicated in the pathogenesis of MODY could be more frequent in these groups, such as HNF4A, insulin promoter factor-1 (IPF-1), HNF1B, NeuroD1 and others. In the other two studies with Brazilian cohorts (20,21), approximately 60% of patients with clinical suspicion of MDM did not have mutations in HNF1A, GCK or HNF4A genes, but we did not test other MODY genes.
In those with mild fasting hyperglycemia, we found GCK mutations in 85.7%, which suggests a high sensitivity for the clinical criteria. The frequency of GCK mutations in this study was slightly higher than previously reported for most populations (42.4 to 61%), including Brazilians (21-23). We found mutations in HNF1A, the most common gene for non-GCK MDM, in 18.5%, which was much like the data obtained by Santana and coworkers in the Brazilian population but lower than reported in other populations (16)(17)(18)(19)24). Therefore, the clinical criteria for the selection of patients for GCK testing seem to be adequate in most populations, including ours. Surprisingly, we found the opposite for non-GCK MODY. This suggests that an improvement in the clinical criteria to adequately select patients for screening is necessary or that, alternatively, other genes should also be investigated.
We have evaluated if MPC (8) could help identify a group of patients that would be more suitable for screening. All patients with HNF1A mutations had PPV > 75%. If we had selected only subjects above this cutoff, 60% of the patients would not have been tested. This strategy would make the screening more cost-effective. However, the frequency of mutations in other genes in patients with a clinical diagnosis of non-GCK MODY is still not known for our population. It is possible that patients without HNF1A mutations could have alterations in other MODY genes, which would be overlooked with this strategy.
In this study, a cutoff for MODY probability of > 75% and >62%, based on MPC, was found in all patients with HNF1A and GCK mutations, respectively. Although in UK the current pick-up rate for MODY testing is PPV > 25% (9), other authors have found a good specificity and negative predictive value in higher cutoff values (> 62.5%) for detecting MODY in non-Caucasian population (25). Our findings suggest that higher cutoff values should be considered for MODY screening in non-Caucasian populations.
This study has some limitations. First, we included a limited number of patients. Secondly, we tested only two MODY genes for financial reasons. Additionally, we used only one methodology (Sanger sequencing) to investigate mutations. This method is unable to detect copy number variations, large deletions and duplications that can represent up to 3% of all genetic alterations of the GCK, HNF1A and HNF4A (26). Another concern is related to the absence of a control group. The selection of the studied population was done in a pragmatic way when the patient did not present clinical criteria of T1 or T2 DM. An additional potential limitation is the lack of C-peptide measurement as a screening tool for MODY. The strengths of our study included finding three novel mutations not previously reported and providing new information about the screening of monogenic diabetes in individuals with multiethnic backgrounds. For future studies, we aim to bypass these limitations and to perform functional genomic studies to confirm novel mutations as pathogenic ones.
In conclusion, we investigated MODY mutations in patients with clinical features suggestive of MDM from a multiethnic background. Approximately one third of patients with clinical features suggestive of MDM from a multiethnic background had GCK or HNF1A mutations. While clinical criteria were efficient for detecting patients with GCK mutations, we found HNF1A mutations in less than 20% of the cases. Although MPC has not been validated for non-Caucasians, its use as a screening tool for selecting patients to test for HNF1A mutations, using a cutoff of 75%, would reduce the number of tests in 60% and increase the percentage of positive cases to 45%. These data suggest that the use of the MPC could be a costeffective strategy for selecting patients to screening for non-GCK MODY mutations, but it is important to consider the possible role of non-HNF1A mutations in non-Caucasian populations such as ours.