p-values greater than 2.0 are plotted against genomic order in . At a genome-wide significance level of approximately p<10−8
(based on the number of informative markers in the genome) [15
] there was a single significant region on chromosome 10. Rs477692 was associated with differential response (p<10−8.15
). By defining a region of association with p-values less than 10−5.5
, there were a total of 20 SNPs in a region of high linkage disequilibrium (LD) (r2
>0.64 for all pair-wise contrasts, average r2
>0.90). These SNPs were contained within the O-6-methylguanine-DNA methyltransferase (MGMT
) gene. Annotation of the MGMT gene along with the LD structure in our dataset is shown in , using LocusZoom [16
]. A Q-Q plot of the overall results is shown in .
(A) Whole genome Manhattan plot of temozolomide for all SNPs with NLPVs above 2.0. Negative log10 p-values greater than 5.5 are indicated with red vertical lines. Locus View LD structure for SNPs nearby rs531572 are shown in (B).
Quantile-quantile plot for a random sample of over 20k empirical p-values from association testing. Any loci near MGMT were removed. Panel (A) is of untransformed p-values and (B) is of negative log10 p-values.
To determine if these SNPs are potentially functionally relevant, we next sought to test for association with MGMT transcript levels and examined how these genotypes are related to differential dose response. Within this region, only one SNP was specifically genotyped, rs531572, while the others were imputed. Thus, given the high degree of LD within this block, we focused our functional assessment on this one SNP.
Rs531572 was highly significantly associated with both IC50 values (p<10−6,4, ) and MGMT transcript levels (p<10−25, ) with the A allele associated with that higher IC50 values and greater MGMT transcript levels. There was a direct correlation between MGMT transcript levels and temozolomide response with greater endogenous MGMT transcript levels associated with IC50 (p<0.006, ). However, inter-individual variation in MGMT transcript levels only explained 1.8% of the variation in IC50. These results are illustrated further in a , with a scatter plot between IC50 and adjusted expression values.
Figure 3 (A) Box plots for the estimated IC50 for temozolomide by genotype for rs531572. (B) Boxplots of MGMT transcript levels differ by rs531572 genotype. For each boxplot, outliers were noted as those values that are above or below 1.5 times the interquartile (more ...)
Scatter plot of estimated IC50 values and adjusted expression levels. Genotypes are color coded, where black, red and green are for GG, GA and AA, respectively.
Additionally, we tested this SNP for association with differential methylation in the Cancer Genome Atlas Data. The Illumina Infinium Human DNA Methylation27 platform contains 20 different probes that are associated with MGMT. The first 4 methylation probes associated with MGMT (situated within the promoter region of the gene and most likely to influence expression levels) have median beta values < 2.0, indicating little methylation, while the probes situated on the latter part of the gene have median beta values > 0.8, indicating they are highly methylation. Association of the rs531572 SNP with the beta values over the first 4 methylation probes (A-D) indicate that this SNP is not associated with alteration of methylation in the promoter region of MGMT (p>0.05). The methylation data is shown in .
Boxplots of MGMT methylation beta values for the 4 probes in the promoter region. Probe A is located at Chr10: 131155063, Probe B at Chr10: 131155199, Probe C at Chr10: 131155565 and Probe D at Chr10: 131155686