Increased Insulin Resistance in Obese Children Who Have Both 972 IRS-1 and 1057 IRS-2 Polymorphisms

The obese patients of cohort 1 came from a previously described cohort (16) originating from Mediterranean and Central European countries. The geographic origin of the patients was assessed through family history and analysis of patronymic names and grandparents’ birthplaces. From this cohort, we selected 174 Caucasian children aged 6–18 years (75 boys, 99 girls) whose onset of obesity had occurred 4–8 years before and who had an oral glucose tolerance test (OGTT) at the time of first evaluation. We selected this duration of evolution because we have previously shown that it corresponds with the establishment of generalized insulin resistance in young obese children (15). We recruited a replication age-matched cohort of 165 obese children (78 boys, 87 girls) from a comparable Caucasian population (cohort 2). A group of 39 age-matched children of normal weight was used for comparison.

The obese and lean children were genotyped at position 972 and 1057 of the IRS-1 and IRS-2 loci, as reported (16). SNP genotypes were determined by the analysis of PCR products. In a control group of lean and healthy Caucasians, the allelic frequency of these variants was 7 and 30% respectively, as reported (4,13).

To minimize experimental variability in measuring insulin levels, we used strictly codified study conditions: after 12 h of overnight fasting following 3 days of standardized diet (caloric and carbohydrate content) in the hospital, circulating insulin was measured in unstressed conditions (an intravenous microcatheter was placed in a peripheral vein 24 h before sampling) (16). The OGTT consisted of the ingestion of 1.75 g/kg² (75 g maximum) in 200 ml lemon-flavored water at 10°C and venous blood sampling at 0, 30, 60, 90, and 120 min after the end of the ingestion. Plasma insulin was measured in each sample in duplicate with standard radioimmunoassay (16). As a reproducibility test, duplicate OGTT measurements were performed on the following days in 20 children. The intra-individual coefficient of variation of insulin values at the various time points was between 13 and 18%. We checked that all children had been gaining weight the month preceding the study to ensure that sampled insulin values truly reflected the natural history of subjects’ β-cell function.

We used insulin (μU/ml) and glucose (mg/dl) values during the OGTT to calculate the composite whole-body insulin sensitivity index (ISI_comp) described by Matsuda (17), according to the formula ISI_comp = 10,000/square root of [insulin 0 × glucose 0 × mean OGTT insulin × mean OGTT glucose].

Comparisons between subgroups used the unpaired Student’s t test, except for the small subgroup of children carrying both IRS-1 and IRS-2 variants, in whom we used Wilcoxon’s sign-ranked test for comparison with other groups. Results are expressed as means ± SE.

The genotypes at position 972 of IRS-1 and 1057 of IRS-2 in the two cohorts are shown in Table 1. The combination of di-allelic polymorphisms at each of these loci leads in theory to nine genotypic categories. In fact, because of the relative rarity of Arg in position 972 of IRS-1, we observed only seven of these categories (Table 1). Neither 972 ArgArg IRS-1/1057 GlyGly IRS-2 nor 972 ArgArg IRS-1/1057 AspAsp IRS-2 individuals were found in our cohorts. Only one obese child was homozygous for 972 Arg IRS-1 and heterozygous for 1057 Asp IRS-2.

The proportions of alleles and genotypes (Table 1) followed the Hardy-Weinberg equilibrium. The genotype distribution was comparable with that reported in Caucasians (4,13) and in a control population of lean Caucasians studied in our laboratory (data not shown). This result suggests that IRS-1 and IRS-2 do not play major roles in the predisposition to juvenile obesity in people of European or Mediterranean descent.

Compared with children who have normal BMIs, the studied obese children showed elevated insulin concentrations and comparable glucose values during the OGTT (Fig. 1) and therefore had a decreased sensitivity index (Table 2). Plasma insulin concentrations during the OGTT and insulin sensitivity index were comparable in the children who had either the Arg972 variant of IRS-1 (column 4 in Table 1) or one or two copies of the Asp1057 variant of IRS-2 (columns 2 and 3 in Table 1) and in those with the nonvariant Gly972 and Gly1057 genotypes (column 1). Even children homozygous for the Asp1057 IRS-2 variant showed normal insulin values if they had a nonvariant Gly972 IRS-1 gene (column 3). Clear-cut differences appeared, however, when the studied variants were present simultaneously in the IRS-1 and IRS-2 molecules, resulting in higher insulin and glucose values during the OGTT (Fig. 1) and in a 25–35% lower insulin sensitivity index (columns 5–7 in Table 1).

To summarize our results, we compared insulin sensitivity in the following categories (Table 2): patients having none of the variants in IRS-1 or IRS-2, one variant in IRS-1 or IRS-2, and variants in both IRS-1 and two proteins. Children with one variant in either IRS-1 or IRS-2 showed a slight but significant decrease of sensitivity. Children with variants in both IRS-1 and IRS-2 genes were markedly resistant to insulin.

The only child homozygous for the IRS-1 variant and heterozygous for the IRS-2 variant (column 7 in Table 1) was younger than the average age in the cohort and subgroups. When compared with age-matched obese children having one or two nonvariant forms of the IRS genes, selected from the two study cohorts, this child showed decreased insulin sensitivity (Table 3).

Because IRS-1 and IRS-2 molecules also participate to the regulation of insulin secretion at the β-cell level (14), we questioned whether the studied variants could affect this process. Although the study was not designed specifically for this purpose, we could plot fasting insulin levels versus insulin sensitivity index values (Fig. 2). By comparing the regression equations in group A (no variant) and group C (variant in both IRS-1 and IRS-2 genes), we observed little difference in plasma insulin, except possibly when insulin sensitivity was extremely low. In these conditions, group C patients may have lower insulin secretion, with the caveat of a limited number of observations.

Juvenile obesity can be considered a pre-type 2 diabetic state because ∼25–30% of children with BMI around 170% of normal, such as those studied here, are expected to develop diabetes in adulthood (15,18). Juvenile obesity is an interesting model to search for genetic factors of insulin resistance because this phenotypic trait is recently established (19) and not yet modified by disease evolution and medical interventions. We used the OGTT to evaluate insulin sensitivity because this test shows early alteration in the natural history of obesity and allows calculation of the “composite” index described by Matsuda and DeFronzo (17; see also 20), a reliable estimation of insulin sensitivity.

Insulin sensitivity was decreased in children who had a variant in both IRS-1 and IRS-2 molecules, whereas each of the variants alone had a much smaller effect. This is possibly the consequence of the two IRS molecules playing complementary roles in insulin signaling (1). Even if the variants do not abolish the function of the IRS molecule bearing them, our results are reminiscent of those obtained in knockout mice for IRS-1 and IRS-2 genes (2). Animals lacking IRS-1 (IRS-1^−/−) or IRS-2 (IRS-2^−/−) had surprisingly mild phenotypes, and neither heterozygous IRS-1^+/− nor IRS-2^+/− mice had abnormalities of insulin physiology. In contrast, combined heterozygous IRS-1^+/− IRS-2^+/− mice showed marked insulin resistance, indicating the importance of combined gene dosage effects of the IRS proteins. Based on our findings, the combination of two mild genetic defects in insulin signaling can favor insulin resistance in humans. As observed for the IRS-1 variant in adults (11), the phenotypic expression of these defects is expected to be particularly visible in obese individuals because they are prone to insulin resistance. The physiological relevance of combined IRS-1 and IRS-2 variants in humans of normal weight remains to be studied in cohorts of appropriate size. Given the complementary function of IRS molecules, it is not surprising that the separate analysis of either the 972 IRS-1 or 1057 IRS-2 variant showed no direct association with type 2 diabetes (7,12). Our results suggest that the implication of candidate gene polymorphisms in the control of a quantitative trait, such as insulin sensitivity, is more likely to be demonstrated at the level of the trait itself than the association of these polymorphisms with a complex and genetically heterogeneous disease such as type 2 diabetes (21,22).