Strong evidence of linkage exists on chromosome 19 in the AAA1 locus [15, 16], but no specific risk loci have yet been discovered. We identified nine functional positional candidate genes within this interval for further study by genetic association. Two of these, PEPD and CD22, showed evidence of nominal association in multiple SNPs, although only one SNP remained significant after correction for multiple testing. Additionally, significant five- and two-SNP haplotypes were identified for PEPD and CD22, respectively.
Evidence of association in PEPD and CD22, was of considerable interest. Based on gene function, gene expression, and protein expression, these genes are plausible candidates for contributing to the risk of developing an AAA. The role of PEPD in collagen metabolism makes it a strong functional candidate for AAA. Although it is unlikely that complete loss of PEPD function alters AAA risk, given that it results in the severe phenotype of prolidase deficiency, polymorphisms that alter PEPD activity could contribute to AAA formation, particularly if they resulted in increased enzymatic activity . Increased PEPD activity has been associated with increased collagen production, both in in vitro experiments [21, 47] and in keloid scars . Excess collagen production is observed in AAA tissue , and has been suggested to contribute to weakening of the aneurysm wall . PEPD is also regulated by several mechanisms relevant to AAA pathogenesis. Nitric oxide, which may contribute to AAA formation , upregulates PEPD activity . Similar to other metalloproteases, PEPD activity is sensitive to doxycycline , a tetracycline antibiotic with anti-inflammatory properties that have been studied for their therapeutic potential in AAA [53–56]. Estrogen can regulate PEPD in vitro [57, 58], an interesting observation given that AAA is six times more common in men than women . Finally, increased serum activity of PEPD has been associated with diseases such as asthma , coronary artery disease , and fatty liver disease .
CD22 is a sialic-acid binding protein found on cells of B-cell lineage. It can inhibit signaling through the B-cell receptor , suggesting that impairment of gene function might lead to dysregulation of the immune response and either autoimmune disease or increased inflammation. Supporting this hypothesis is evidence of a role for CD22 in systemic sclerosis; anti-CD22 autoantibodies contribute to systemic sclerosis pathogenesis  and there is evidence of association between a SNP in CD22 and limited cutaneous systemic sclerosis . Systemic sclerosis is characterized by immune infiltration and excessive fibrosis, both of which are observed in AAA subtype known as inflammatory AAA . Given that patients with inflammatory AAA are more likely to have a positive family history , CD22 polymorphisms could contribute to AAA risk. On the other hand, Cd22 knockout mice display a minimal phenotype with no evidence of increased autoimmune disease . It is plausible, however, that polymorphisms in CD22 could modulate the immune response in an inflammatory disease such as AAA through changes in the amount of CD22 protein produced in a cell or an alteration of the gene's function in B-cell signalling.
Exon sequencing of PEPD and CD22 showed that both have polymorphic coding regions. Polymorphisms that did not result in amino acid substitution or changes to the 3'-UTR were not considered further, since their functional effects are difficult to predict. After investigation for predicted functional consequences, all of the sequence changes identified in the coding regions or 3'-UTR of CD22 and PEPD appeared to be tolerated.
One limitation of this study was that the sample size used for genetic association provided only modest power. Although we did observe nominal association in several genes, only one remained significant after correction for multiple testing. Our results should, however, be considered in light of the prior evidence of linkage on chromosome 19q13. While this evidence does indicate the presence of variants contributing to AAA risk at this locus, the differences between linkage and association do not make identification of associated variants a given. Linkage is tolerant of allelic heterogeneity, whereas a variant must be sufficiently common to be detected by association. Therefore, a biologically relevant variant in a linked region may exhibit either weak association, as we observed, or no evidence of association (extensive allelic heterogeneity). Furthermore, if there is also locus heterogeneity in genetic risk factors between populations, increasing sample size using subjects from a different population than that originally studied may actually decrease the evidence of association observed. In light of this and our observation of association at multiple SNPs for each gene, we consider our findings for PEPD and CD22 worth additional studies.
It is possible that the associations seen here resulted from spurious associations or from population stratification. A recent association study examining SNPs from the chromosome 19 linkage interval in the Dutch population found several nominally associated SNPs but none remained significant for multiple testing . Interestingly, the SNP showing the strongest evidence of association resided in CEPBG, which is in agreement with our observation of association in the region of CEBPG and PEPD (Figure 1).
Finally, by focusing on functional candidate genes rather than genotyping the entire linkage region, it is possible that linked and associated genes were overlooked. The strategy presented here was based on reducing the number of tests due to the limited power of our sample and the increased ability to interpret associations in genes of known function. Recent technological improvements in high-throughput sequencing provide a potentially more efficient approach to follow-up of linkage results [66, 67]. Deep sequencing at high coverage of linkage intervals in members of linked families has identified novel loci. For late age-at-onset diseases the lack of complete nuclear pedigrees and inability to identify definitively unaffected individuals, will reduce the approach to affected relative pair sequencing.