Background Microsatellites are not amenable to high-throughput genotyping and have been excluded from GWAS. However, several microsatellites affect expression of nearby genes and those in the asporin and BMP5 genes have been associated with susceptibility to OA. In a recent analysis of functional microsatellites we found association of the -794 CATT microsatellite in the MIF gene with hip OA in 1782 patients compared with 1878 healthy controls of European ancestry. Replication was hampered by the lack of a method to obtain the genotypes without performing the laboratory tests.
Objectives We aimed to develop an imputation methodology for this microsatellite using the genotypes of SNPs in linkage disequilibrium and to apply it for validation.
Methods In our previous study, we had included 1090 samples that were also in the arcOGEN GWAS. Genotypes of the 130 SNPs in the LD region of the microsatellite in these samples were used as source of haplotype information. Imputation was done with Impute2 with modifications allowing estimation of the probabilities for the number of copies of each microsatellite allele and their posterior combination. Performance of this procedure was evaluated in 10 replicates of training, with 90% of the samples, and testing in the remaining 10% of the samples with known microsatellite genotypes. Once concordance of imputed genotypes was established, the approach was applied for imputation of the MIF microsatellite in 5667 population controls from Wellcome Trust Case-Control Consortium (WTCCC) and in 2466 hip OA cases from arcOGEN (all of them of European UK ancestry). In addition, we have validated the functional effect of the MIF microsatellite on the plasma levels of MIF in 361 healthy control samples by ELISA (R&D Systems) from subjects that were either homozygous for the 5 repeat or the 6 repeat alleles.
Results The genotypes of the three alleles with frequency >1.0% were imputed with sufficient accuracy (91.6%) and call rate (98.8) in the reference samples. However, other microsatellites showed that the procedure will require further refinements to attain this performance for microsatellites with more alleles. Application of this procedure to the WTCCC and arcOGEN samples produced genotypes for 99.0% of the samples. Comparison of the allelic frequencies showed significant differences between hip OA and controls in women (figure 1) following the same pattern found in our previous study, with the five repeats allele less frequent in the patients than in the controls (OR =0.88, [95% CI] 0.79 to 0.98, P=0.018).
Analysis of MIF in plasma of healthy controls showed higher levels in the homozygous for the 5 repeat allele (3.6 ng/mL) than for the 6 repeat allele (2.7 ng/mL; P=0.00025) following the direction previously reported by other authors and without differences between women and men.
Conclusions A new method to impute genotypes of microsatellites based on SNP genotypes has been developed. It has shown good call rate and accuracy. This procedure allowed us to validate the association of the MIF microsatellite with hip OA in a large number of cases and controls. Higher MIF levels were found in the subjects with the hip OA protective allele. These results contribute to define the role of MIF in OA.
Acknowledgements Instituto de Salud Carlos III (Spain), grants PI11/01048, PI12/01909 and RD12/0009/0008 that are partially financed by the European Regional Development Fund
Disclosure of Interest None declared