Association of hGrb10 Genetic Variations With Type 2 Diabetes in Caucasian Subjects

The genes contributing to type 2 diabetes are mostly unknown (1). Grb10 is an adapter protein that, in target tissues (2,3), interacts with the insulin receptor (3–9), thus affecting downstream signaling (10–12) and insulin action (2,7,10–16). We tested the hypothesis that variants in the Grb10 gene modulate the risk for type 2 diabetes, one of the most frequent outcomes of insulin resistance.

Resequencing of coding and immediately flanking sequences of hGrb10 (17,18) identified six single nucleotide polymorphisms (SNPs), five of which had a minor allele frequency >5%. Based on their physical location and their mutual linkage disequilibrium, these five SNPs (i.e., rs1800504, rs2715128, rs2072235, rs4947710, and rs3807550) could be grouped into two clusters. The rs1800504 and rs4947710 SNPs (each belonging to different clusters) were analyzed for association with type 2 diabetes in 764 diabetic patients and 323 unrelated control subjects from the east coast of central Italy (19,20). No significant association with type 2 diabetes was observed for rs1800504 (data not shown). By contrast, the genotype distributions of rs4947710 (i.e., G/G, G/A, and A/A) were significantly different between case and control subjects (87.2, 12.3, and 0.5% vs. 61.9, 35.8, and 2.3%, respectively, P < 0.0001), with A allele carriers showing a reduced risk of type 2 diabetes (unadjusted odds ratio 0.239 [95% CI 0.17–0.33], P = 0.0001; age-, sex-, and BMI-adjusted odds ratio 0.235 [95% CI 0.15–0.36], P = 0.0001). A potential biological relevance of rs4947710 was suggested by an in silico analysis (ESEfinder; available at http://exon.cshl.edu/ESE/), which indicated that the G-to-A substitution of rs4947710 may cause the disruption of a putative consensus motif for the human Ser/Arg-rich proteins SF2/ASF.

To replicate this association, we studied 731 type 2 diabetic case and 358 nondiabetic control subjects, all being Caucasians from the Boston area (20). In contrast to what was observed in the Italian population, the genotype distributions of rs4947740 were similar in case and control subjects (G/G = 83.9%, G/A = 15.2%, and A/A = 1.0% and G/G = 86.0%, G/A = 14.0%, and A/A = 0.0%, respectively, P = 0.15).

In conclusion, a significant association between the hGrb10 rs4947710 SNP (whose biological function on differential splicing is suggested by in silico analysis) and type 2 diabetes was found in Caucasian subjects from Italy but not in those from the U.S. Lack of replication of genotype-phenotype associations is not an uncommon event in the study of complex disorders (21–23) and can arise from the original result being a false-positive because of bias or chance or from the second result being a false negative because of insufficient power. Such explanations, however, do not seem to account for our conflicting findings. The population of the original study was relatively homogeneous, making the possibility of population stratification remote, and the P value for association with type 2 diabetes was highly significant, making chance an unlikely explanation of the association finding. The replication study had close to 100% power to detect the odds ratio observed in the original study. Thus, lack of replication in our study is likely to result from differences in the genetic and/or environmental background of the populations studied, highlighting the need for large, collaborative studies providing sufficient power to investigate gene-gene and gene-environment interactions and their differences among populations.