A multigenerational study on phenotypic consequences of the most common causal variant of HNF1A-MODY

Aims/hypothesis Systematic studies on the phenotypic consequences of variants causal of HNF1A-MODY are rare. Our aim was to assess the phenotype of carriers of a single HNF1A variant and genetic and clinical factors affecting the clinical spectrum. Methods We conducted a family-based multigenerational study by comparing heterozygous carriers of the HNF1A p.(Gly292fs) variant with the non-carrier relatives irrespective of diabetes status. During more than two decades, 145 carriers and 131 non-carriers from 12 families participated in the study, and 208 underwent an OGTT at least once. We assessed the polygenic risk score for type 2 diabetes, age at onset of diabetes and measures of body composition, as well as plasma glucose, serum insulin, proinsulin, C-peptide, glucagon and NEFA response during the OGTT. Results Half of the carriers remained free of diabetes at 23 years, one-third at 33 years and 13% even at 50 years. The median age at diagnosis was 21 years (IQR 17–35). We could not identify clinical factors affecting the age at conversion; sex, BMI, insulin sensitivity or parental carrier status had no significant effect. However, for 1 SD unit increase of a polygenic risk score for type 2 diabetes, the predicted age at diagnosis decreased by 3.2 years. During the OGTT, the carriers had higher levels of plasma glucose and lower levels of serum insulin and C-peptide than the non-carriers. The carriers were also leaner than the non-carriers (by 5.0 kg, p=0.012, and by 2.1 kg/m2 units of BMI, p=2.2 × 10−4, using the first adult measurements) and, possibly as a result of insulin deficiency, demonstrated higher lipolytic activity (with medians of NEFA at fasting 621 vs 441 μmol/l, p=0.0039; at 120 min during an OGTT 117 vs 64 μmol/l, p=3.1 × 10−5). Conclusions/interpretation The most common causal variant of HNF1A-MODY, p.(Gly292fs), presents not only with hyperglycaemia and insulin deficiency, but also with increased lipolysis and markedly lower adult BMI. Serum insulin was more discriminative than C-peptide between carriers and non-carriers. A considerable proportion of carriers develop diabetes after young adulthood. Even among individuals with a monogenic form of diabetes, polygenic risk of diabetes modifies the age at onset of diabetes. Graphical abstract Supplementary Information The online version contains peer-reviewed but unedited supplementary material available at 10.1007/s00125-021-05631-z.


Research Design and Methods
The onset of diabetes was defined as the earliest occurrence of i) a diabetic value at a study visit (fasting plasma glucose, FPG ≥ 7.0 mmol/l or 120minute glucose during an oral glucose tolerance test (OGTT) > 11 mmol/l or HbA1c ≥ 48 mmol/mol (6.5%); ii) a self-reported year of the diagnosis, or iii) an ICD code for diabetes in the national registries (the Care Registers for Health Care, HILMO, Finnish institute for health and welfare) from birth until five years after the last study visit (ICD9: 250*; ICD10: E10.*-E14.*, * representing any number, or H28.0, H36.0, N08.3).

Statistical analysis
We controlled the statistically significant results by Welch's t-test and sex-and-age-adjusted linear models (commented only in case the significance changed) using both first and mean adult values (after log-transformation, if necessary). Confirmatory t-testing for corrected insulin response (CIR), occasionally with negative values, was performed without transformation. All sex-specific analyses on the statistically significant observations were at least in the same direction in both sexes (expect for fasting proinsulin, which showed a negative and significant male estimate, whereas the female estimate was marginally positive but insignificant). Proinsulin values below the lowest standard of 1.775 pmol/l were adjusted at 1.775 2 pmol/l. For composite (Matsuda) insulin sensitivity index (ISI), we substituted a 90-min measurement (if unavailable) with a mean of 60 and 120 minutes.The computer model-derived homeostatic model assessment (HOMA) indices (see methods) for insulin sensitivity (HOMA2 IS, an inverse number of HOMA2-IR) and insulin secretion (HOMA2β) applied a conversion factor of 6.945 for insulin (from mU/l to pmol/l) provided by the manufacturer, although similar results were obtained using an alternative conversion factor of 6.00.
The main and clearly significant results were significant also in the three large families alone. Some more subtle results required a larger sample size to reach the statistical significance level. On the other hand, in smaller families alone, the results resembled those seen in the large families (B-D), but possibly because of the low sample size, the analyses did not as often reach the significance levels. No significant differences were seen even if we replaced the measurements at the time of the diagnosis (or as close as possible) with the first or last measured value during the follow-up, with or without limiting to the values before the diagnosis of diabetes. Also, when further performing mixed effects Cox models, none of the models yielded an Akaike information criterion to support that BMI or the indices above would affect the age at onset.

The T2D-PRS and the onset of diabetes
To increase the power of the Cox proportional hazards models, we also included the year of birth as a covariate, because the younger participants have undergone more active screening for diabetes than the older ones. Inclusion of the birth year as a covariate in a Cox proportional hazards model increased the statistical power of the analyses: 1. In a model with both carriers and non-carriers, T2D-PRS had β = 0.352 (HR 1.42) with p = 0.00021, HNF1A carrier status β = 3.07 (HR 21.6) with p < 2×10⁻ 16 , and year of birth β = 0.031 (HR 1.03) with p = 9.4×10⁻ 9 2. In a model with only carriers, T2D-PRS had β = 0.318 (HR 1.38) with p = 0.0019, and the year of birth β = 0.028 (HR 1.03) with p = 6.7×10⁻ 7 Similarly, a linear model without any covariate but only to include the carriers born since 1930, the result was of the same magnitude (for 1 SD unit increase of T2D-PRS, predicted age at diagnosis decreased by 3.0 years, p = 0.0254, r 2 = 0.037).
Among the carriers, those diagnosed with diabetes before the age of 20 years had a significantly higher T2D-PRS than those remaining free of diabetes at 20 years (mean 0.54 vs 0.070; MWU estimator -0.50; CI -0.89, -0.13; p = 0.0091). The carriers belonging to the highest tertile of T2R-PRS had an earlier onset of diabetes than the carriers in the lowest tertile (mean 24 yrs vs 31 yrs; MWU estimator -6.1 yrs, CI -13.0, -1.1, p=0.016).

Urine threshold for glucose
HNF1A variant carriers have been reported to have lower renal glucose threshold than related non-carriers [ESM1]. We compared urine glucose excretion (dipstick data detecting U-glucose ≥5.5 mmol/l) during the OGTT in carriers and non-carriers without measurable fasting urine glucose. We found no difference in the relationship between the urine glucose and P-glucose AUC (ESM Fig. 3) or peak P-glucose (data not shown) between groups. In three carriers, P-glucose peaked relatively high (> 8.5 mmol/l) during the OGTT without positive U-glucose after the OGTT. Also, negative fasting U-glucose result was observed despite FPG as high as 13.4 mmol/l in one carrier, and between 7.4 and 9.9 mmol/l in 12 carriers. As the results from the four insulin assays as well as the three C-peptide assays correlated strongly with each other (with r 2 at 96% or more), respectively, the measurements for insulin and C-peptide were transformed to cohere with those obtained from the electrochemiluminometric immunoassay. We excluded data for samples from earlier (1990s) visits analysed by less specific methods without reliable conversion factors. P, plasma; S, serum; U, urine. Mean adult levels of glucose, insulin, C-peptide, proinsulin and lipids at fasting and during an OGTT besides anthropometric measurements, and the mean age at assessment. The fasting levels are shown separately for the whole group and for the subgroup who participated in an OGTT (see methods). In t-testing, both FICR and CIR lost their statistical significance but were not normally distributed. N (% female), total number of individuals (the proportion of women); MWU, Estimator of the Mann-Whitney U test; CI, 95% confidence interval; LM est, linear model estimate; CIR, corrected insulin response; ISI, composite insulin sensitivity index (Matsuda index) a Linear model estimate adjusted for age and sex.

ESM
b Including the 0 min values reported for the OGTT.
c One outlier was excluded for CIR and ten for fasting insulin:C-peptide ratio. d CIR (at 30 min) in 100 pmol/l (mmol/l) -2 , ISI in 10 000 (mg/dl) -1 (mU/l) -1 ESM Estimates and p values for coefficients from linear models using the carrier status, plasma glucose level and their interaction term as predictors.
The models with a significant interaction are in bold.
a Including all fasting samples.
ESM Fig. 3. The area under the curve (AUC) of plasma glucose during OGTT (x-axis) plotted against glucose excreted in urine during the OGTT (yaxis) in those carriers (red square boxes) and non-carriers (blue circles) of the HNF1A p.(Gly292fs) variant, whose urine sample before the OGTT was negative for glucose in a semiquantative dipstick analysis. Using an estimate of glucose concentration by a dipstick analysis after the OGTT, the urine glucose is calculated as ( ) ⋅ ( ).