As mentioned earlier, our lab had some preliminary data indicating an association between CASP8 SNPs and CRC risk. In order to further investigate the potential role of inherited CASP8 variants in affecting CRC risk, the following was performed; firstly, rs3834129 (an ins/del variant in the promoter region of CASP8 previously implicated in CRC risk) (Sun et al. 2007) was genotyped in 1193 CRC cases and 1388 controls from Sheffield, Leeds, Dundee and Utah. Secondly, the CASP8 coding region, in addition to intron/exon boundaries, the promoter region and 3‟UTR were
sequenced in 94 CRC cases from the Sheffield cohort.
The SNP rs3834129 was genotyped using TaqMan assays. Several QC assessments were performed to validate the TaqMan assay and the genotyping data. In summary, the QC measures indicated a reasonable quality of the genotyping results. Chi-square association test was then performed and the results have shown that there was no significant difference in genotype frequencies between cases and controls. These results contradict the previous results from the Chinese population (Sun et al. 2007) which showed an association between rs3834129 and multiple cancer risks including CRC. However, the results are in concordance with the published data on the European and multi-ethnic American populations which found no effect of rs3834129 on CRC risk (Haiman et al. 2008, Pittman et al. 2008). Based
on the size of the analysed cohort, we had ~98% power (α=0.05) to detect an effect
of similar size (OR=0.75/ copy of deletion allele) as described by Sun and colleagues. Possible explanations for the discrepancy between our results (and the European and the American studies) in comparison to the original study in the Chinese population could be due to different frequencies and effects of genetic and environmental modifier factors between the different populations, or due to the initial association being a false positive. Despite the lack of association overall, rs3834129 genotyping results are currently being used in a more in depth analysis of CASP8 inherited variants and CRC risk. This further analysis is being performed both to fine map associations and to investigate various subgroups of CRC cases (Curtin et al. manuscript in preparation).
CASP8 sequencing was performed to try and identify rare coding variants that might affect CRC risk. Association studies are usually performed using a “tagging SNP”
approach which depends on common and known SNPs. Therefore, rare or unknown SNPs could have been easily missed. Sequencing 94 cases yields an 85% probability of identifying variants with a minor allele frequency (MAF) of 0.01. Normal DNA extracted from the peripheral blood of 94 CRC cases from the Sheffield population was sequenced and six novel variants were identified. The first variant (- 565 G>A) was within the promoter region of CASP8, however, it did not affect any of the known functional sequences identified in that area. Moreover, in silico analysis did not predict the introduction of any transcription factors binding sites or the presence of any regulatory regions. The genomic location was not conserved across 5 mammalian species. Therefore, it was not predicted to have an effect on CASP8 function. The second variant was identified only in one case and was a 115 bp deletion in exon 3 (c.1-8338Del115). Despite this region having a potential regulatory role, as predicted by the presence of DNase cluster region and histone marks which
following observations led to the prediction that it was unlikely to be functionally important. Firstly, CASP8 has 8 different isoforms and exon 3 is only expressed in the G isoform precursor. The deletion results in an in-frame deletion of the first 11 amino acids of isoform G precursor which is not part of the mature protein. Additionally, the deleted 11 amino acids are not part of the known conserved functional domains of CASP8 protein. Furthermore, in silico analysis of the 11 amino acids sequence using the database of protein domains, families and functional sites Prosite (http://www.expasy.ch/prosite/, accessed July, 2011) did not predict any highly conserved and functional domains. Finally, isoform G is not one of the predominantly expressed CASP8 isoforms and it is not detected at the protein level (Scaffidi et al. 1997). Therefore, this variant did not seem a suitable candidate for further investigation at this stage. The four other variants were intronic, and were not predicted to have an effect on CASP8 mRNA splicing using 5 different online splice sites prediction tools. Apart from the intron 3 variant c.1-7982A>G, none of the variants were predicted to be in a region with a potential regulatory role. However, the variant c.1-7982A>G was in the same region as the exon 3 115 bp deletion and it was not predicted to be functionally important for the same reasons discussed above.
CNVs were recently shown to play an important role in several human inherited diseases and cancer (Beckmann et al. 2007). Moreover, they have been used successfully to predict cancer related genes (Venkatachalam et al. 2010). Although the above results do not suggest a role for CASP8 inherited variants in CRC risk, we decided to investigate the role of CNV23598, reported on the database of genomic variants, because of its physical proximity to the known breast cancer D302H risk variant and due to its unknown frequency.
The genotyping results of 47 selected breast cancer cases and controls suggested a strong association between the insertion allele and the CASP8 risk haplotype.
However, when 288 CRC cases and controls were investigated, the genotype frequencies differed markedly from HWE with excess heterozygotes. Several molecular techniques (Gel-based and fluorescent GAP PCR, in addition to DNA sequencing) were used in an attempt to validate the genotyping results. The combined results showed some differences from the published reference sequence, and suggested the presence of multiple insertion copies of CNV23598. DNA sequencing results and the use of in silico tools indicated that the genotyping assays were specific to the target region. Multi-allelic CNVs are known to occur in the human genome, which could explain the presence of the 3rd copy (McCarroll and Altshuler 2007). Moreover, reported CNVs in the genomic databases usually lack specificity at their breakpoints. Several more CNVs were recently reported on the database of genomic variants which overlap CNV23598, but with variable breakpoints at both ends. This could explain some of the difficulties encountered in the genotyping experiments. Thus, our attempts to accurately genotype CNV23598 using conventional molecular techniques have so far failed. The 47 breast cancer samples described in this project were recently sequenced using a 2nd generation sequencing platform as part of a different project. The analysis of these data may reveal the exact sequence of the CNV flanking regions and resolve some of the inconclusive observations for the data.