• Editorial
• Open Access

• Published: 9 July 2009

Suarez-Gestal and colleagues [1] selected single variants from either systemic lupus erythematosus (SLE) genome-wide association studies (GWAS) or large candidate-gene-based association studies. Three of the markers tested were identified in all three European-American GWAS. Four more were discovered in a single GWAS; either the International Consortium for Systemic Lupus Erythematosus Genetics (SLEGEN) studies [2–4] or a Swedish study [5]. The remaining three variants were found in a smaller candidate-gene or regional association study. This study by Suarez-Gestal and colleagues represents an important step forward in SLE genetics because it uses a Spanish population to confirm most of the predominantly European-American associations. It appears that the SLE genetics community have learned valuable lessons in GWAS design from the experiences of other complex trait geneticists where replication studies were less consistent and yielded a multiplicity of false positives and negatives [6]. The lack of replication in other complex diseases may reflect a number of flaws in their GWAS design, including inadequate sample sizes, population stratification, poor quality control of genotyping and substandard matching of cases and controls [7]. Independent replication is an important step in endorsing novel susceptibility genes because it both confirms the validity of the GWAS design as well as verifies the contribution of a given locus to disease pathogenesis.

Suarez-Gestal and colleagues confirmed nine of the ten previous associations. This paper represents a major step forward for SLE because each of the loci containing these single nucleotide polymorphisms (SNPs) can now be examined in more detail to determine biological mechanisms and to gain a greater understanding of their roles in the pathogenic process of SLE. The remaining variant, in the coding region of LY9, a gene that forms part of the signalling-lymphocyte-activation molecule (SLAM) locus [8], did not replicate. Possible reasons for this are population-specific differences in allele frequency or haplotype structure between the northern and southern European populations [9] or, alternatively, too little power to detect it. The advent of HAPMAP phase 3, with increased numbers for the original four populations, increases the reliability for the tag-SNP selection as well as adds an additional seven population groups for better correlation between the population used for tag-SNP selection and a wider number of study cohorts. This would minimise the effect of population-specific differences in haplotype structure by allowing for the presence of multiple risk haplotypes carrying a particular risk allele(s) while minimising the type I or II errors due to population-specific differences in allele frequency for single variants.

The ideal replication study design would be to perform a meta-analysis of the existing SLE GWAS data and then select the variants showing the strongest association. Suarez-Gestal and colleagues chose four variants, each identified by a single GWAS, since there is no publically available meta-analysis for all the current GWAS. The lack of reproducibility between GWAS may reflect an overall lack of power for each individual study, since each published GWAS was of intermediate size (1,100 cases or less) and there was some sample duplication between them. The other factor that may contribute to the diversity of strongly associated SNPs between different studies is the heterogeneity of the disease phenotype, which may reflect the underlying genetic heterogeneity. Consequently, there is clear need for a much larger additional GWAS, in both Europeans and non-Europeans, with clearly defined clinical criteria and at a much greater density of markers. This scale of experiment, which is proving successful for other complex diseases, would offset the loss of power associated with the heterogeneity of lupus.

In complex autoimmune diseases such as lupus, a large number of variants are expected to make a small contribution to the overall genetic risk. Furthermore, there may be multiple susceptibility alleles within a given gene, so that the global risk for a given locus will be a combination of the individual risks for each susceptibility allele in that gene. Further complexity arises because not only do particular individuals carry different combinations of risk alleles, but the genotyping chips do not carry every susceptibility allele for a given locus. Interpretation of whether a particular locus is associated with disease may, therefore, depend on the linkage disequilibrium between the genotyped SNP and the functional allele(s). Patterns of linkage disequilibrium between associated SNPs in different studies will have to be taken into consideration before either claiming replication or the lack of it.

The replication of the variants in the paper by Suarez-Gestal and colleagues therefore represents the tip of the iceberg because it has confirmed the association of a number of common risk alleles of moderate disease risk identified from intermediate-sized GWAS. To test rarer mutations (minor allele frequency <0.1%), such as those that will be generated by the 1000 Genomes Project, it may be necessary to both design custom chips, since even the latest generation of GWAS chips [10] will not carry rarer mutations, and also genotype them in very large populations. However, we anticipate that there will be a larger number of genes with a smaller effect size (odds ratio 1.1 to 1.2), so that we will need an increased number of samples in the study cohort to gain sufficient power to find a significant association. Hence, GWAS funded by the Wellcome Trust Case Control Consortium include studies in the range of 5,000 to 10,000 samples; a study of this size is needed in SLE. These larger population sizes will give sufficient power for the analysis of subphenotypes. Nevertheless, the data presented by Suarez-Gestal and colleagues provide a core series of independently validated loci that, together with additional targets generated by larger GWAS, can be used to piece together the key pathways involved in lupus pathogenesis, with each pathway constructed of a number of interacting proteins making an individual contribution to disease susceptibility.