Our multi-step approach identifies the HLA-DOA gene, and the B-lymphocyte in which it is exclusively expressed, as plausible candidates contributing to pediatric LTx rejection. Adding early functional validation to family-based association in the same dataset is especially useful, because it obviates the need for immediate replication in an independent cohort, whose accrual may take several years in the case of rare disease traits. This novel approach was motivated by failure when we first performed genome-wide association with the unmodified TDT applied to all 550,000 SNPs (on-going study and data not shown). All SNPs showing significant p-values failed to remain significant after Bonferroni correction for multiple hypothesis testing.
Because SNP reduction with PAF comparisons and the GC test are independent, the combined p-value for r9296068 is a multiple of values from both procedures (0.041*0.0183=0.00075). However, this would not be significant after multiple-testing correction for 1,774 SNPs. Therefore, transmission characteristics of rs9296068 were confirmed by showing that allele frequencies in Rejectors and Non-Rejectors were respectively, significantly greater (51.4 vs 36.8%, p=0.015), and less (26.8 vs 36.8%, p=0.074) than those in the normal control population of 400 Caucasian children (). Finally, in direct comparisons of allele frequencies between Rejectors and Non-Rejectors, the 5′ flanking UTR of HLA-DOA was represented by five SNPs among 14 top-ranked SNPs (p≤0.01) (), of which rs9296068 achieved the highest p-value (0.002), and another, rs9276994, also showed evidence of over-transmission in Rejectors (p=0.074), and under-transmission in Non-Rejectors (p=0.082), even though parental allele frequencies were similar. Significantly, all five highly-ranked SNPs, beginning with rs6457699 and ending with rs9277015, were present upstream, in an ≈11.6 kb segment, at a distance of ≈4-15 kb from the first HLA-DOA exon (Figures and ). Promoter function has been predicted for a highly conserved ≈10kb region immediately upstream of the first HLA-DOA exon, with Caucasians demonstrating a linkage disequilibrium (LD) block encompassing the rs9296068 locus () (27
). Because 3 of 5 discriminatory SNPs localize to this putative promoter, altered transcription factor binding, and differential regulation of HLA-DOA gene expression can be postulated. Significant repression of the first HLA DOA exon among Rejectors in our study supports this view. Decreased HLA-DOA gene expression is also seen with increasing numbers of the risk allele of rs9296068 for both Caucasian and Yoruba populations in public databases (28
Figure 4 The HLA-DOA gene is transcribed from the minus strand. The upstream 5′ flanking UTR region defined by five discriminatory SNPs in simple association testing of Rejectors vs Non-Rejectors lies between rs9296068 to the right and rs6457699 to the (more ...)
We interpret our observed associations as suggestive of a causal role for the HLA-DOA gene, and for the B-lymphocyte, rather than a causal role for the SNP itself; however causality remains unproven until definitive studies identify a true causal variant, and the mechanism by which it might alter HLA-DOA gene expression and B-cell function. For the association itself, all statistical tests used sequentially in the current study are necessary because neither simple association testing nor family-based association testing generates, by itself, a list of SNPs larger than the expected false positive prediction (p=0.05 times 1,774 SNPs=89 false-positives). The HLA-DOA gene, and its adjacent region, is of interest for several reasons. Compared with one of its 3′ neighbors, HLA-DPB1, which is highly polymorphic, and can influence immunological outcomes in bone-marrow, corneal and renal transplantation, the HLA-DOA gene is relatively non-polymorphic, and inhibits class II antigen presentation in mature B-cells; but its role in organ transplantation is unknown (29
). Our findings lead us to speculate that a missing exon transcript may produce a dysfunctional HLA-DOA gene product, which facilitates rejection by failing to inhibit antigen presentation by B-lymphocytes. The resulting increased B-lymphocyte participation in the rejection process is seen as nearly three-fold higher intragraft B-lymphocyte content in rejecting liver grafts in our study, when the minor allele G of rs9296068 was present, compared with rejecting allografts from children without this allele. Among pediatric LTx, prior supportive evidence also includes greater resistance of activated B-lymphocytes to immunosuppression, as well as a relative excess of B-lymphocytes during the risk period for early rejection (18
). B-lymphocyte-rich infiltrates, as well as B-lymphocyte-specific gene expression products have already been demonstrated in renal allografts with steroid-resistant acute cellular rejection (33
). While our observations are preliminary, they suggest that the HLA-DOA gene and its vicinity be mapped further to identify novel causal variants.
We acknowledge several limitations. The heterogeneous diagnoses precipitating LTx are potential sources of stratification, although no single disease dominated either outcome group. For example, all 5 children with PSC carried the risk allele, with four experiencing rejection, despite statistically similar proportions of PSC in either group (4/37 Rejectors vs 1/43 Non-Rejectors, p=NS). Second, our SNP reduction step which relies on PAF comparisons, is biased by inclusion of eight Non-Caucasians trios, seven in the Rejector group. In accepting rs9296068 as a candidate, an illustrative subanalysis shows that our conclusions would be unchanged if an adequately powered Caucasian sample was available (Appendix 5
). For these reasons, we have also relied on functional studies, coupled with allele-specific decrease in HLA-DOA expression in public databases for both CEU and YRI, to suggest biological relevance for the risk allele. We hope to relate differences in HLA-DOA exon repression to allelic variations at rs9296068 in a larger future sample. The small numbers of Rejectors (n=11) in whom HLA-DOA exon repression was shown, could not be divided further for adequately powered correlations with allelic variants in this study.
We conclude that our combination of methods can identify biologically relevant associations in small populations. We propose to validate our conclusions and address the primary power limitations of this pilot study by extending it to multicenter LTx populations.