Genotype-phenotype associations between risk alleles and disease subtypes may give insight into disease etiology and mechanisms. Recent results show that rs7574865, a variant allele of STAT4,
confers an increased risk for both SLE and rheumatoid arthritis (RA) 
, suggesting the involvement of common pathways of pathogenesis among these two autoimmune diseases. STAT4
-deficiency is associated with accelerated renal disease and increased mortality 
in a murine lupus model, but with protective effects for arthritis in knockout mice 
. Since SLE is an extremely heterogeneous disease, with multiple correlated subphenotypes, we sought to investigate whether or not STAT4
appears to contribute to this phenotypic heterogeneity in human SLE. We believe that our data provide compelling evidence that STAT4
is associated with more severe SLE manifestations, particularly with nephritis and with the production of autoantibodies to double-stranded DNA. In contrast, other recently-discovered SLE risk polymorphisms do not appear to be strongly associated with severe disease manifestations 
There have been recent successes in the study of genotype-phenotype associations in SLE and other autoimmune diseases. For example, PDCD1
has been shown to be associated with lupus nephritis and anti-phospholipid antibodies in ethnic subgroups 
, and PTPN22
is primarily associated with anti-cyclic citrullinated peptide (anti-CCP) 
and rheumatoid factor (RF) 
autoantibody positive RA. The STAT4
gene has been shown to be associated with both anti-CCP positive and negative RA 
; it has not yet been investigated in the context of SLE subphenotypes.
Replication of genotype-phenotype associations can be challenging 
; a strength of our study is the inclusion of four independent case series. Other strengths include the availability of two overlapping genotype sets in the STAT4
region for most of the subjects, including genome-wide data to facilitate ancestry analysis, and of course the availability of detailed phenotype data on all four of the case series.
A limitation of our study is that the subjects are of self-reported European ancestry and primarily female. It could be insightful to look at these associations in other populations, particularly since SLE has higher prevalence among African-Americans and other non-European populations 
. The STAT4
gene has recently been shown to be associated with RA in a Korean population 
; however significant associations with subphenotypes – namely age at onset, radiographic progression, and serologic status – were not found.
Another limitation is the inherent difficulty in obtaining accurate phenotype data. Differences between our 4 SLE cohorts may be true differences in patient characteristics, perhaps as a result of differences in selection, but could also be influenced by different methods of assessment and accuracy of individual records. However, although some of the phenotypes we examined are related to disease activity, and may fluctuate naturally or as a result of treatment, we classified SLE patients according to a history of these specific phenotypes. We are encouraged by the fact that our results were quite homogeneous across the different cohorts. Also, any misclassification would presumably be non-differential with respect to genotypes, thus diluting our results rather than causing type I error.
Finally, it is important in genetic studies to protect against false associations due to undetected population substructure. Indeed there were some subjects in our cohort with sizeable non-European ancestry, in spite of being self-reported European, and those had a higher minor allele frequency for rs7574865. However, reanalysis of a more homogeneous subset of subjects of primarily northern European ancestry was very consistent with our overall results. There is even stronger evidence in this subset for relationships between the STAT4 rs7574865 SNP and nephritis subphenotypes, and for an inverse relationship with oral ulcers.
Since the subphenotypes having the strongest risk conferred by rs7574865 were highly correlated, we included clinical variables based on principal components (PC) analysis to investigate the possibility of common underlying effects. The first PC, associated with the severe manifestations of anti-dsDNA antibodies, nephritis and immunologic abnormalities, had similar associations as those of its components. Severe nephritis was consistently the most strongly associated subphenotype. The second PC, associated with the milder skin disease manifestations of malar rash, photosensitivity, and discoid rash, was not significantly associated with rs7574865 in any analysis.
In summary, our study has identified multiple correlated subphenotypes that are strongly associated with the STAT4 gene, including nephritis, autoantibodies to double-stranded DNA, and early age at diagnosis. The next challenge is identifying how these correlated features fit into causal pathways, and therefore to help elucidate the complex etiology of SLE.