Systemic lupus erythematosus (SLE) is a severe multisystem autoimmune disease of unknown etiology. Genetic factors clearly play a role in susceptibility, and a number of genetic loci have been implicated in the disease [1
]. Despite the successes of recent genetic association studies, only a fraction of the genetic liability for SLE has been explained to date. SLE is a heterogeneous disease clinically, and there is strong evidence that the molecular pathogenesis of the condition varies considerably between individuals as well. For example, specific autoantibodies are formed in some patients and not others, and these autoantibody specificities have been associated with clinical features of the disease [2
]. In addition, approximately half of adult patients with SLE demonstrate overactivity of the interferon alpha (IFN-α
) pathway in their peripheral blood [2
]. Interestingly, high IFN-α
and SLE-associated autoantibodies are heritable as traits in SLE families and can be found in family members who are not affected by SLE [5
]. Autoantibodies can be found in sera for many years prior to the clinical diagnosis of SLE [7
], and it is thought that some of the autoantibodies may be themselves directly pathogenic. IFN-α
is a cytokine involved in viral defense, capable of bridging the innate and adaptive immune systems [8
]. Interestingly, when recombinant human IFN-α
has been given as a treatment for chronic viral hepatitis, some treated individuals have developed de novo SLE, which frequently resolves upon discontinuation of the IFN-α
]. These data support the concept that both IFN-α
and SLE-associated autoantibodies represent causal factors in human SLE. Additionally, both IFN-α
and SLE-associated autoantibodies are heritable within SLE families supporting a genetic contribution, and thus the idea that these molecular measurements could be used as a phenotype in genetic studies.
In previous work, we have begun to map genetic variants which are associated with high IFN-α
and with the presence of particular autoantibodies in SLE patients [11
]. Some well-established genetic risk factors for SLE have been associated with one or both of these molecular phenotypes [14
]. In addition, we have performed a genome-wide association study (GWAS) using these two molecular traits as phenotypes to enable discovery of novel genetic variants associated with IFN-α
and SLE-associated autoantibodies [19
]. A number of novel genes have been validated from this screen to date [19
], although much of the variance in both serum IFN-α
and the presence or absence of particular autoantibodies remains to be explained.
In prioritizing genetic variants to be followed up in our GWAS scan, we used gene ontogeny and expert literature search to prioritize variants which were in or near genes related to immune responses. This was based upon the supposition that SLE is an autoimmune disease, and many of the well-validated loci which have emerged from unbiased studies to date encode genes with immune function. This approach has some limitations, as genetic variations which were not near known genes were not prioritized, nor were those which did not have known function within the immune system. It is clear that genetic variants can sometimes impact the expression of a gene which is not nearby, and these genetic variants may be assigned to irrelevant nearby genes in gene ontogeny analysis. Additionally, many genes which could be critical to human disease pathogenesis may still be unstudied and unknown, and thus unlikely to be prioritized in follow up candidate studies.
To address these possibilities in our GWAS validation, we searched our top 200 SNPs in a public database which links genome-wide SNP data from the HapMap project to genome-wide gene expression data from the HapMap lymphoblastoid B-cell lines (SCAN) database, [21
]. Genes which are disease associated are more commonly associated with alternate gene expression than genes which are not disease associated [22
], and thus genes from our top 200 which were strongly associated with differences in gene expression should be more likely to be true associations. In this study, we leverage gene expression data sets to prioritize additional candidates from our trait-stratified GWAS for validation in an independent cohort. We found eleven SNPs which were significantly associated with alternate gene expression of multiple transcripts in public databases, and had not been prioritized for followup in our initial GWAS screen. Four of these eleven SNPs were significantly associated with the important molecular subphenotypes IFN-α
and SLE-associated autoantibodies in our independent validation cohort, validating this method of genetic discovery.