This study represents the first comprehensive evaluation of all of the top PD-GWAS implicated SNPs on genome-wide transcription. In our analysis of the effects of SNPs on the expression of nearby probes (within 250 kb of the either end of the full region of the risk locus), we identified genes at both the MAPT and HLA loci whose expression is significantly associated with the PD GWAS SNPs. These findings provide novel insight into the genes that are potentially responsible for influencing PD risk at these loci.
It is particularly difficult to interpret the PD GWAS findings for the MAPT
locus, which is the second most strongly implicated region in the US Parkinson Disease GWAS Consortium 
. The region is complicated by the large and well characterized inversion which distinguishes the common “H1” haplotype associated with increased PD risk from the less common and PD-protective “H2” haplotype. The inverted region spans approximately 1.5 Mb 
and contains more than twenty transcribed genes. Finally, the inversion creates an extraordinarily long stretch of high LD, which impedes the identification of the specific gene(s) responsible for PD risk.
Our studies reveal several important findings for the MAPT
region. Of particular interest is the association between the PD SNPs and two probes in a duplicated region distal to MAPT
. Little is known about the more strongly associated probe, targeting the hypothetical protein gene LOC644246
(located within an intron of KIAA1267
). However, previous studies 
examining association between PD associated SNPs and expression of nearby genes observed a strong association between the SNP rs393152 (in strong LD with rs439945, R2
0.87) and LRRC37A
in 133 neurologically normal frontal cortex samples. In addition, a genomewide study of expression and methylation QTLs in 143 normal cortical samples showed significant association between rs439945 and LRRC37A
. We saw no significant association between the SNPs in the region and the expression of MAPT
itself. While it is notable that other genes showed a stronger effect than MAPT
, limitations in power due to the size of the study prevent us from actually ruling out potential associations to any genes in the region.
Finally, a probe targeting the gene DCAKD (43,100,706–43,112,509) located 800 kb proximal to MAPT (43,971,750–44,105,697) showed an interesting pattern of association with the PD risk SNPs. In this case, PD samples with the risk allele (the major allele in this case) showed increased expression of this gene, while controls with the PD risk allele showed decreased expression. This finding was unique among the cis-effects in the observation of opposite effects of the SNP on probe expression, and highlights the need for studies of affected samples to identify important disease specific associations, which would go unobserved in healthy samples. Although little is known about the DCAKD gene, its kinase function and the disease-specific nature of the association may indicate the greatest potential for translational impact among the MAPT region findings.
In the HLA region, while the associated SNPs (rs2395163 at 32,387,809 showed the strongest association to gene expression and had the strongest association to PD in the region in the original meta-analysis) are located near the HLA-DRA
gene, altered expression is seen for the genes HLA-DQA1
(at 32,605,183–32,611,428) and HLA-DQA2
(at 32,709,163–32,715,219), located 217 to 321 kb away. While it is not unusual to see regulatory effects which are fairly distant from the gene [See 17 for review]
, these findings emphasize the difficulty inherent in determining which gene is associated with disease risk in gene dense GWAS implicated regions. Although this eQTL relationship has not been previously reported in any studies of cortical samples, examination of the University of Chicago's expression quantitative trait loci data repository (http://eqtl.uchicago.edu/cgi-bin/gbrowse/eqtl/
) indicated that association between rs2395163 and HLA-DQA1
has also been observed in LCLs 
and monocytes 
In addition to our evaluation of cis
-acting SNP effects, we also examined trans
-acting SNP effects across the entire transcriptome. We observed significant associations between PD risk SNPs in the SNCA
region and several genes across the genome. In each case, the significant association was driven by a strong increase in gene expression in the PD cases, with little to no effect observed in the controls. These results suggest that the genes implicated by this eQTL study may be involved downstream in a disease pathway initiated by SNCA
. Interestingly, of the 18 probes showing significant modification by PD risk SNPs, the probe targeting LY6K
(associated with the SNCA
SNP rs2583975) was the only probe to also show significant (after Bonferroni correction) differential expression between cases and controls (p
0.001, Table S1
In contrast to the pattern of increased expression for probes associated with SNCA SNPs, the trans effects observed for the SNP in the RIT2 region showed decreased expression in the control samples and little to no effect in the PD cases. While the immediate significance of this pattern is not clear, it may indicate a mechanism in which the regulation of these genes is disrupted in PD.
One limitation to using microarray data for this study is the potential for SNPs within the probe sequence to lead to false positive results in cis
analyses. We examined the 5 probes showing significant cis
-associations and identified two, A_24_P852756 (HLA-DQA2
) and A_24_P58331 (DCAKD
), with known SNPs located in the targeted region. A_24_P852756 had two known SNPs within the microarray probe, one of which was rare, but the other (rs9276439) had a MAF of 11.6% and therefore may have been present in our sample. Unfortunately, we do not have genotyping of this SNP (rs9276439) available in the microarray samples, but were able to determine that there is no LD (r2
0.012) between this SNP and the identified eSNP rs2395163 in HapMap samples, suggesting that the SNP in the probe is not driving the eSNP association.
In the A_24_ P58331 probe (DCAKD), there were 4 known rare SNPs with estimated minor allele frequencies ranging from 0.02% to 1.4%. In this case we were able to directly assess the sequence of the probe in 39 of the 50 microarray samples (17 cases/22 controls) and, as expected based on the allele frequencies, none of the SNPs (nor any novel variants) was observed, suggesting that within-probe SNPs are unlikely to be driving the association. However, we can certainly not rule out the possibility that additional known or unknown variants in the other associated probes may have either contributed to or masked SNP associations. This study is also limited to only those probes that were included on the used microarray and is, therefore, unable to evaluate all genes and differentiate all paralogs of transcripts of potential interest.
It should also be noted that this study is not intended to indicate an absence of either cis- or trans-acting effects of any of these SNPs on specific probes, a task which is more suited to larger control samples. Our goal is instead to identify the associations between SNPs and gene expression, which may be most relevant to PD by focusing on disease specific tissue (for which no larger arrayed PD sample exists) and expression relationships.
These studies suggest that, at the MAPT locus, the genes LOC644246, LRRC37A and its duplicate LRRC37A2, and the DCAKD gene may influence PD risk. It cannot be ruled out that more than one of these genes at the MAPT locus influence PD risk. In addition, these studies suggest that either HLA-DQA1 or HLA-DQA2, or both, at the HLA locus may influence PD risk. Either sequencing studies or functional knock-down studies of these genes may lend further insights into whether mutations in them or variants that influence their expression may influence PD risk and whether they may be therapeutic targets for PD treatments.