Effect of inbreeding on type 2 diabetes-related metabolites in a Dutch genetic isolate

Autozygosity, meaning inheritance of an ancestral allele in the homozygous state is known to lead bi-allelic mutations that manifest their effects through the autosomal recessive inheritance pattern. Autosomal recessive mutations are known to be the underlying cause of several Mendelian metabolic diseases, especially among the offspring of related individuals. In line with this, inbreeding coefficient of an individual as a measure of cryptic autozygosity among the general population is known to lead adverse metabolic outcomes including type 2 diabetes (T2DM), a multifactorial metabolic disease for which the recessive genetic causes remain unknown. In order to unravel such effects for multiple metabolic facades of the disease, we investigated the relationship between the excess of homozygosity and the metabolic signature of T2DM. We included a set of heritable 143 circulating markers associated with fasting glucose in a Dutch genetic isolate Erasmus Rucphen Family (ERF) of up to 2,580 individuals. We calculated individual whole genome-based, exome-based and pedigree-based inbreeding coefficients and tested their influence on the T2DM-related metabolites as well as T2DM risk factors. We also performed model supervised genome-wide association analysis (GWAS) for the metabolites which significantly correlate with inbreeding values. Inbreeding value of the population significantly and positively correlated with associated with risk factors of T2DM: body-mass index (BMI), glucose, insulin resistance, fasting insulin and waist-hip ratio. We found that inbreeding influenced 32.9% of the T2DM-related metabolites, clustering among chemical groups of lipoproteins, amino-acids and phosphatidylcholines, whereas 80 % of these significant associations were independent of the BMI. The most remarkable effect of inbreeding is observed for S-HDL-ApoA1, for which we show evidence of the novel DISP1 genetic region discovered by model supervised GWAS, in the ERF population. In conclusion, we show that inbreeding effects human metabolism and genetic models other than the globally used additive model is worth considering for study of metabolic phenotypes.


29
Autozygosity, meaning inheritance of an ancestral allele in the homozygous state is 30 known to lead bi-allelic mutations that manifest their effects through the autosomal recessive 31 inheritance pattern. Offspring of related individuals are at an increased risk of inheriting two 32 copies of recessive deleterious alleles, which would expose the offspring to the full (normally 33 compensated) deleterious effects of those alleles, hence decreasing the fitness of the 34 offspring. Consanguineous marriages between close relatives as a result of assertive mating 35 is known to cause severe congenital metabolic consequences in the off-spring 1 . In addition to 36 that moderate inbreeding due to isolation in populations has been shown to cause unfavorable 37 outcomes among with cardio-metabolic and neuropsychiatric parameters 2,3 . Inbreeding was 38 shown to associate with an increase in fasting glucose, blood pressure, body mass index 39 (BMI), waist-hip ratio (WHR) and decrease in high-density lipoprotein cholesterol (HDL-C), 40 intelligence quotient (IQ) and height 4,5 . We have previously shown that some metabolites can 41 well be regulated by genetic variants following the recessive genetic model 6   involved in the pathophysiology of disease 7,8 . In line with this, several circulating molecules 50 have been found associated with T2DM: such as phospholipids, branch-chain amino-acids and 51 lipoprotein subclasses [9][10][11] . In order to find target endophenotypes for researching the 52 recessive genetic effects, we studied the influence of inbreeding over a selected list of metabolites that are known to be related to changes in glucose. Figure 1 shows the outline 54 of the step-wise analysis design.

72
The detailed description of the ERF study and related measurements were reported 73 previously 12 . Baseline type 2 diabetes was defined according to the fasting plasma glucose ≥ 74 7.0mmol/L and/or anti-diabetic treatment, yielding 212 cases and 2,564 controls, totaling up 75 to 2,776 individuals. The follow-up data collection of the ERF study took place in May 2016 (9 76 to 14 years after baseline visit). During the follow-up, a total of 1,935 participants' records 77 were scanned for the incidence of type 2 diabetes in general practitioner's databases.

78
Additionally, a questionnaire on type 2 diabetes medication surveyed on 1,232 participants in

153
The outlying metabolite values that were more than four times standard deviation 154 away from the mean were excluded from the analysis. Non-normally distributed 155 measurements were natural logarithm transformed, or rank transformed accordingly. As 156 described previously 23 , in brief, we assessed the pairwise partial correlation between each 157 metabolite and each glycemic trait (i.e., fasting glucose, fasting insulin, homeostatic model 158 assessment for insulin resistance (HOMA-IR), BMI and WHR) in the non-diabetic participants.

159
The association of metabolites and T2DM were assessed by logistic regression with T2DM 160 status as the dependent variable. We included age, sex and lipid-lowering medication as covariates and adjusted the models by familial relatedness by using the polygenic residuals 162 extracted by the "polygenic" function of R package GenABEL. We applied a Bonferroni

174
We estimated the heritability (H 2 ) for the 151 unique metabolites as well as fasting 175 glucose as a reference in ERF pedigree using SOLAR software 25 with age and sex adjusted.  The inbreeding coefficients for ERF population and their comparisons to the ones 214 calculated for the (outbred) Rotterdam Study population are given in Figure 3A. The family-based ERF study has much higher skewness value than the population-based Rotterdam study.

216
The separation between the distributions is the most remarkable for the FEXOME values with 217 skewness = 0.85 in ERF study and skewness = 0.16 in the Rotterdam Study. Figure 3B shows

232
Among those 12 metabolites were common (P-value < 0.05) in both whole genome-

248
Eight metabolites remained statistically significant after correcting for multiple testing (P-

340
One limitation of our study is that S-HDL-ApoA1 is not measured in any other cohort to our

Figure 4
The effect of inbreeding values on the metabolite profiles and relation to T2DM risk factors. Figure 4A shows the correlation between the inbreeding coefficients and the metabolites and T2DM-related risk factors. Figure 4B correlation between the inbreeding coefficients and the metabolites after adjusting for BMI. Metabolites with at least one suggestive P-value of association (<0.05) are shown with ".".
Significance after correction for multiple testing (P-value < 2.8 × 10 -3 ) is marked with a "* " . Purple: positive association. Green: negative association. The depth of purple and green presents the value of correlation coefficient. F-GW: whole genome array based inbreeding coefficient. F-EXOME: the combination of exome chip and exome sequence based inbreeding coefficient. F-PED: Pedigree based inbreeding coefficient.