Glutathione S-Transferase (GST) family of genes have been implicated in multiple neuropsychiatric 1 2 3 4 and neurodegenerative diseases 5 6 7 8 9 10 11 ; where altered levels or function of these enzymes is thought to impact levels of oxidative stress and/or inflammation in a way that contributes to disease susceptibility. A linkage locus on chromosome 10q that has been implicated in both Alzheimer's (AD) 11 12 13 and Parkinson's disease (PD) 13 harbors two GST genes of the omega class: GSTO1 and GSTO2, which are approximately 75 kb apart.

GSTOs have enzymatic activities as thioltransferases and dehydroascorbate reductases that promote antioxidant activity and can also function in metabolism of drugs and toxins 14 . Additionally, GSTO1 was shown to promote activation of the pro-inflammatory cytokine, interleukin-1β (IL-1β) by post-translational processing 15 . Given their location and function, they have been studied as candidate genes in AD and PD 5 6 9 11 14 16 17 18 . Li et al. compared hippocampal gene expression levels in 6 AD vs. 2 control brains and identified significantly lower GSTO1 levels in the AD hippocampi 5 . This group studied AD and PD families that showed linkage to chromosome 10, using the age-at-onset phenotype 13 and identified association of multiple SNPs at the GSTO locus with delayed age-at-onset of both diseases 5 , with the strongest effects observed for GSTO1 rs4925 and GSTO2 rs2297235 SNPs that are in tight linkage disequilibrium (LD). No significant influence was detected for either AD or PD risk in this study.

Since this initial report, several follow-up studies have been published with mixed approaches and results. Kölsch et al. reported association of rs4925 with earlier age-at-onset of AD, thus in opposite direction to the original report 6 , and no effect on AD risk. Lee et al. found modest association of rs4925 with AD risk in Carribean-Hispanic families that show linkage to chromosome 10q 11 as did Capurso et al. in an Italian case-control series 9 , though neither study detected an age-at-onset effect. A case-control study by Wahner et al. was the only report for an effect of GSTO locus on PD risk, with both rs4925 and rs2297235 conferring protection, especially in those with smoking history 16 . Additionally, several studies reported lack of association with age-at-onset or risk of AD 17 18 or PD 14 .

Additional investigation of the GSTO locus is needed to further elucidate the role of these genetic variants in AD and PD, especially given the potential to establish the glutathione metabolism as a molecular pathway that is common to multiple, chronic neurologic diseases. An important shortcoming of most prior reports on the GSTO locus is the modest sample sizes, which could underlie the inconsistent results likely due to lack of power, sample or locus heterogeneity or a combination of these factors. Both AD 19 20 and PD 21 are complex diseases with substantial genetic component. Some of the genetic risk for these diseases has been identified via linkage and association studies and shown to influence age-at-onset 19 20 21 22 23 . More recently, genome-wide association studies (GWAS) of AD 24 25 26 27 28 and PD 29 30 , with sample sizes exceeding 10,000 subjects provide considerably greater power for detection of susceptibility loci. Despite their advantages, GWAS do not explain all of the underlying genetic component of these and other complex diseases, thus necessitating alternative approaches 31 , including analysis of quantitative phenotypes.

In this study, we assessed the GSTO locus for its role in AD and PD, using an in-depth approach aimed at surmounting these challenges. Given the original report of association with delayed age-at-onset of AD and PD 5 7 , and with risk of AD in some follow-up studies 9 11 , we postulate that GSTO locus variants confer risk of LOAD in older age. We have a collection of > 8,000 late-onset AD (LOAD) case-controls, which includes a large series of older subjects ≥ 80 years of age-at-diagnosis/death (clinical/autopsied LOADs) or evaluation (controls). We analyzed two previously reported, coding SNPs in GSTO1 and GSTO2 for association both with disease risk and age-at-diagnosis in the LOAD series, as well as a large PD series. Reduced expression levels of GSTO1 5 and other glutathione metabolism genes 10 32 have been reported in AD. We therefore analyzed the GSTO SNPs for association with brain GSTO1 and GSTO2 levels in > 750 brain samples from autopsied subjects with AD and other brain pathologies to determine whether they influenced disease risk by affecting brain gene expression. In an expression GWAS (eGWAS) testing association of 24,526 transcript levels measured in these brain samples with 213,528 cisSNPs within ± 100 kb of the tested transcript, we identified 686 genes that have significant cisSNP/transcript associations (in-press, PLoS Genetics). We analyzed these genes to discover molecular pathways that are enriched for genes with significant brain cisSNPs, and identified glutathione metabolism as one of the top pathways. Our results suggest that GSTO locus variants influence brain GSTO2 levels and confer AD risk at older age. These findings have mechanistic implications for the GSTO locus and glutathione metabolism genes, which should be explored further in AD and other chronic, neurologic diseases to identify functional variants that influence disease risk by altering brain gene expression levels.

GSTO1 and GSTO2, which are evolutionarily conserved genes 14 , previously implicated in AD 5 6 9 11 and PD 5 16 , have diverse attributed functions including antioxidant activity via generation of ascorbate (Vitamin C) 14 34 35 ; biotransformation of inorganic arsenic 14 34 ; modulation of ryanodine receptors and thus calcium release and apoptosis 36 ; and post-translational processing of the pro-inflammatory cytokine, IL-1β 15 . Given their functions which are relevant for the pathophysiology of neurodegenerative diseases and their location in linkage regions for AD 11 12 13 and PD 13 , GSTO locus variants have previously been studied for their association with risk and age-at-onset of AD and PD with mixed results 5 6 9 11 14 16 17 18 .

In this study, we assessed two coding polymorphisms, rs4925 (Ala140Asp) in GSTO1 and rs156697 (Asn142Asp) in GSTO2 in a large LOAD series of > 8,000 subjects, ~3,000 of whom were older (> 80 years) and in a PD series of > 1,300 subjects including both familial and sporadic cases. We found significant LOAD risk association for the minor allele of rs156697 in older subjects and a suggestive trend for delayed age-at-diagnosis. These results are consistent with the original 5 and some of the follow-up reports on this locus 9 11 , and suggest that the reported delay in age-at-onset is likely to be due to an increased risk conferred in older subjects. Given the age-dependent decline in key glutathione metabolism components and their role in mitigating oxidative stress 32 , the postulate that risky GSTO variants lead to increased risk in older LOADs due to accumulation of oxidative damage with increasing age, is a plausible scenario. It should be emphasized that our study utilized age-at-diagnosis as a surrogate for age-at-onset and unrelated case-controls, rather than family-based series. These differences could underlie the marginal age association in our study, in comparison to the original study 5 .

Given the tight LD (r2 = 0.73, D' = 0.94 in HapMap3) 37 between the two coding SNPs tested for AD and PD risk association in this study, we did not correct for multiple testing. If corrected, the AD association in the older ADs would no longer be significant (p = 0.076). Furthermore, GSTO locus variants were not reported to have significant or suggestive association with AD risk in the recent, large GWAS 26 27 28 . Although, these findings could collectively suggest that the AD risk association in our study is a false positive, there are alternative explanations: First, the effect conferred by GSTO2 rs156697 is age-specific based on our results, and others 5 . Additionally, unlike the older series in our study, the younger LOAD series had significantly heterogeneous results for the rs156697 SNP. Thus, the large LOAD GWAS need to be re-analyzed focusing on the different age groups and also for age-at-onset or diagnosis association. Second, the effect of the GSTO2 variant is likely modest for LOAD risk, despite strong effects on brain gene expression. Third, although GSTO2 rs156697 has the strongest effect on brain expression of this gene in that locus, it may still not be the functional variant, thus leading to weak or heterogeneous effects on LOAD risk. Our results in LOAD risk and brain gene expression provide support for functional variant discovery efforts in the GSTO2 region and screening of such variants for their effects in transcriptional assays.

There was no significant association of GSTO SNPs with disease risk in the combined PD series. This may not be surprising given the difference in sample size and therefore power between the LOAD and PD series. Whereas our older LOAD series (1,338 LOAD vs. 1,604 controls) have ~61% power to detect the effect of the GSTO2 rs156697 SNP (OR = 1.14), the combined PD series (661 PDs vs. 702 controls) tested for this SNP, has ~32% power to detect this effect at α = 0.05. There was, however, association with decreased risk in the familial PD cases for the GSTO1 rs4925 minor allele. Although consistent with one other study in PD 16 , this finding requires further replication. It is intriguing to note that this variant also conferred a protective effect in the LEAPS-PD GWAS, which assessed PD sib-pairs in its first stage 38 39 . The opposite direction of association in the familial PDs (and some of the younger LOAD series) vs. the older LOAD series could have several explanations including the tested SNPs not being functional themselves but marking different functional variants of opposing effects; heterogeneity due to different gene-gene or gene-environment interactions in different groups; and false positivity in some of the tested series.

Although both GSTO SNPs are in coding regions, they do not lead to any change in the enzymatic activities of GSTO1 34 35 or GSTO2 34 . While their effects on LOAD and PD could be due to other, untested alterations in protein function, another potential mechanism of action is influencing levels of gene expression. Indeed, both SNPs had highly significant effects on brain gene expression levels of GSTO2, but not GSTO1. Amongst the 22 cisSNPs tested for association with GSTO levels in our brain eGWAS, rs156697 had the strongest effect, where the risky minor allele was associated with lower brain GSTO2 levels. These results strongly suggest that the risk conferred by the GSTO locus is most likely due to variants which influence GSTO2 levels in the brain. These findings are biologically compatible with the very high antioxidant function of GSTO2, where its dehydroascorbate reductase activity was found to be 70-100% greater than that of GSTO1 34 .

Brain expression levels of other key enzymes of glutathione metabolism are also significantly influenced by genetic variants, as was identified from pathway analysis of our significant brain eGWAS results. Given our findings with GSTO2 and other studies implicating glutathione metabolism genes in neurodegenerative diseases 8 10 32 , it will be important to analyze these additional glutathione metabolism genes with high brain regulation, for variants that influence risk of AD and other neurodegenerative diseases.

In summary, our results support GSTO2 as a risk gene for older LOAD subjects, where risky genotypes reduce brain levels of this gene, which likely leads to accumulation of oxidative damage worsening with increasing age. These findings have implications for disease mechanism, as well as the search for genetic risk variants in AD and other neurodegenerative diseases. First, it will be important to analyze the existing large LOAD and PD risk GWAS by different age-strata and also using age-at-onset as the outcome, where available. Second, GSTO2 should be sequenced for variants that may influence gene expression and thereby disease risk. Third, association with expression levels provides a unique opportunity to identify the actual disease gene at the linkage or association locus. Fourth, individual or combined assessment of glutathione pathway genes that are regulated in the brain, may uncover additional neurodegenerative risk variants. Further establishment of GSTO2 and other glutathione metabolism genes in AD and PD awaits discovery and mechanistic studies of functional genetic variants.

N. Graff-Radford, M.D. has served as a consultant to Codman and received grant support from Elan Pharmaceutical Research, Pfizer Pharmaceuticals, Medivation, and Forrest. R.C. Petersen, M.D., Ph.D. has been a consultant to GE Healthcare and Elan Pharmaceuticals, has served on a data safety monitoring board in a clinical trial sponsored by Elan Pharmaceuticals, and a safety monitoring board for Wyeth Pharmaceuticals.

MA carried out SNP genotyping assays, executed the disease risk and age at onset statistical analysis and helped to draft the manuscript. FZ participated in the organization and design of the gene expression study. HSC, CSY and AAN significantly contributed to the analysis of the gene expression data. RM participated in the collection of the gene expression data and carried out SNP genotyping assays. JEC and VSP significantly contributed to the design of the gene expression study and statistical analysis. MMC led the design, collection and analysis of SNP genotyping data used for the gene expression study. CNR and RP carried out SNP genotyping assays. TN, LM, KGM, GB and AIO provided technical expertise in the preparation, organization and maintenance of the RNA and DNA samples. SM, SM and CG participated in the analysis of the gene expression data. DS and FR participated in the collection of gene expression data. CPK and JJ supervised the collection of gene expression data. SBS, JOA, MB, RJU, ZKW, OAR, RCP, NRG and DWD collected and provided the case-control samples used in these studies. SGY significantly contributed to the conception and organization of the study. NET conceived of the study, and participated in its design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.