Prdx6 is a member of the thiol-specific antioxidant protein family and in overexpressing cell and mouse models has been shown to be protective against oxidant stress which null models show sensitivity to oxidants [7, 9, 27]. Thus, PRDX6 is a suitable candidate gene for ALI risk. The extent of genetic variation within PRDX6 remains largely unknown, therefore we performed direct sequencing of the PRDX6 gene, and identified novel variants for future study. We also tested the newly discovered SNPs and tagging SNPs for association with ALI using our trauma cohort, and did not demonstrate an association with trauma-related ALI.
We identified 43 novel variants among African American and European American subjects with either ALI or control status. None of the 43 SNPs identified were in coding regions which may indicate that the Prdx6 protein is highly conserved across phyla. Approximately 19 kb on chromosome 1 was sequenced in order to achieve adequate coverage of the PRDX6 gene and flanking 5' and 3' UTRs. Special attention was given to the GRE2 and ARE1 regions -749 to -737 and -357 to -349, respectively. The ARE1 within the PRDX6 promoter was shown to play a role in regulation of transcription and to be inducible under conditions of oxidative stress  and the GRE2 may be capable of binding transcription factors under oxidative stress conditions . Due to the GC rich content of the region surrounding the ARE1, we were unable to optimize PCR reaction conditions in a way to prime through the secondary structure. The GRE2 region was sequenced, but no variation was noted. The GC rich region within the PRDX6 promoter might warrant further investigation since methylation of DNA cytosine residues are often found in the sequence context CpG. Several new sequencing approaches are emerging that target methylation sites using restriction enzyme treatment followed by sequence by synthesis .
In addition to comparing our results with NCBI's dbSNP, we compared our novel and known SNPs with the resequencing data registered in 1000 Genomes. The 1000 Genomes project aims to find most genetic variants with frequencies of at least 1%. Thus far three sequencing projects contribute to the database, low coverage sequencing of 179 individuals from 4 populations, high coverage sequencing of 2 mother-father-child trios, and exon targeting sequencing of 697 individuals from 7 populations . Although 1000 Genomes aims to identify over 95% of variation in any individual, 27 of our novel SNPs and 1 previously recorded SNP are not present in the database, signifying a need for resequencing of extreme phenotypes, such as ALI cases.
Novel and previously recorded SNPs in the 5' UTR and first intron of PRDX6 were submitted to TESS to determine their likelihood of being in transcription factor binding sites. We found 19 motifs in the reference sequences that are capable of binding known transcription factors and 21 in the alternative sequence. A comparison between the results of the reference sequence search and the alternative revealed that in most cases, the SNP of interest changes the motif enough to cause a different transcription factor to bind that site or can cause a binding site to disappear and vice versa. After comparison with the ENCODE data, we found that our sequences have not yet been shown to bind the three overlapping transcription factors tested in ENCODE experimentally.
Known SNPs validated in the sequencing effort were compared using a Patrocles search query for miRNA target sites within PRDX6 to determine if any of our SNPs were in putative target sites for miRNAs. Three of the eight SNPs returned from the search corresponded with our known SNPs. Only one of the three SNPs was found to have a corresponding known miRNA (miR-942). Some miRNAs are known to control the expression of genes at the posttranscriptional level . However, very limited data are available on miR-942.
We performed an association study for ALI using newly uncovered SNPs and SNPs selected from Hapmap and NCBI's dbSNP and observed no significant association between any of the SNPs in this study and ALI. This lack of association may be due to several causes. First, the detectable effect size is modest because of sample size limitations. We genotyped 513 subjects to test for an association between our selected SNPs and ALI, but this sample size was inadequate to detect relative risks below 1.93 and 1.69 for alleles with MAFs of 0.05 and 0.10, respectively. Second, our analyses were limited to patients with severe trauma. Thus, our study did not evaluate a possible association with other causes of ALI such as sepsis. Finally, it is possible that PRDX6 genetic variation may not modify the risk of ALI.
The genotype data were used to construct haplotype blocks to better assess the PRDX6 gene structure. Haplotype analysis plays an important role in association studies between genotype and phenotype, since SNPs found to be in strong LD can capture most of the genetic variation across fairly large regions . The haplotype blocks constructed from our genotype data did not show strong linkage disequilibrium using confidence intervals, therefore tagging SNP strategies in future studies should be approached with caution.
Our resequencing data did not show any variation in the coding region of PRDX6. Had nonsynonymous SNPs been discovered, it would have prompted us to investigate whether any of these SNPs had any effect on protein structure, which could cause a loss of function in Prdx6. Since we cannot make a connection between coding region SNPs and conformational changes in the protein, we examined regulatory effects. We found several promoter SNPs that change the sequence of potential TFBSs based on conservation data. We were unable to confirm that these sequences were in fact TFBSs due to the lack of available data. However if any of our promoter SNPs showed a significant association with ALI or another phenotype perhaps using a larger sample size, future studies using promoter constructs could offer more information on upregulation of PRDX6. We also found several SNPs in the 3'UTR. It is possible that one or more of these SNPs is responsible for changing an miRNA binding site, thus repressing protein translation.
Our study has several limitations. One potential limitation of this study is the number of genotype call failures. Ten and nine markers for African Americans and European Americans respectively were eliminated from our analysis since they were under the 95% completion rate cut-off. This high rate of genotype failure was due to difficulties with consistent assay performance rather than DNA quality. If these genotypes had been obtained, it is a possible that an association may have been observed. Also, we did not adjust our results for ancestry informative markers (AIMs). Instead our population was stratified based on skin color, which may not be an adequate proxy for population admixture effects. Another possible limitation is a candidate gene approach that focused on a single gene: PRDX6. ALI risk may be considered a complex phenotype, and thus likely is not fully explained by a variation in a single gene . Finally, we only tested for association in patients with ALI from severe trauma. Thus, it is possible that PRDX6 may play a role in the initiation or severity of ALI after other insults, including sepsis, or in determining recovery from ALI.
PRDX6 has been shown to play a role not only in ALI, but other diseases as well. A recent studied demonstrated that PRDX6 promotes lung cancer metastasis and invasion via phospholipase A2 activity in mice . Another publication reported that PRDX6 transfected breast cancer cells metastasized more readily to the lungs when compared with control cells . It is possible that our novel SNPs may function in lung cancer as well as ALI. The interaction between GSTpi and PRDX6 is another interesting subject for future studies. GSTpi expression is elevated in tumors from a variety of cancers, including lung cancer, compared to normal tissue . Testing gene-gene interactions between PRDX6 and GSTpi would be an interesting future direction both in ALI and other diseases such as cancer.