In a follow-up to their recent study reporting a locus that influences age-at-onset in Alzheimer’s disease (AD) and Parkinson’s disease (PD) on chromosome 10q, Pericak-Vance and colleagues now report the results of what they refer to as their “genomic convergence” method to identify the gene responsible. This method involves first obtaining data from gene expression studies of hippocampi from several AD patients and controls and then specifically looking for changes in genes that map to the AD/PD age-at-onset linkage peak on chromosome 10. It needs to be pointed out from the outset that the gene on chromosome 10 being pursued in this study is reported to affect only age-at-onset and was found not to be associated with risk for getting AD (affection status). In contrast, in three studies published in 2000 in Science, linkage was detected to genes influencing either one’s “risk” for getting AD (based on linkage to “affection status”) or to plasma levels of Aβ42. The composite data of those three papers indicated the presence of at least one (more likely two) AD genes on chromosome...  Read more

In a follow-up to their recent study reporting a locus that influences age-at-onset in Alzheimer’s disease (AD) and Parkinson’s disease (PD) on chromosome 10q, Pericak-Vance and colleagues now report the results of what they refer to as their “genomic convergence” method to identify the gene responsible. This method involves first obtaining data from gene expression studies of hippocampi from several AD patients and controls and then specifically looking for changes in genes that map to the AD/PD age-at-onset linkage peak on chromosome 10. It needs to be pointed out from the outset that the gene on chromosome 10 being pursued in this study is reported to affect only age-at-onset and was found not to be associated with risk for getting AD (affection status). In contrast, in three studies published in 2000 in Science, linkage was detected to genes influencing either one’s “risk” for getting AD (based on linkage to “affection status”) or to plasma levels of Aβ42. The composite data of those three papers indicated the presence of at least one (more likely two) AD genes on chromosome 10; our study supporting the candidacy of the gene for insulin-degrading enzyme IDE (Bertram et al., 2000) and the other two (Ertekin-Taner et al., 2000 and Myers et al., 2000) detected linkage roughly 20-30 Mb proximal. to IDE, perhaps suggesting an additional AD locus on chromosome 10. It is currently unclear whether the age-at-onset locus reported by Li et al., 2002; Li et al., 2003) will explain any of the chromosome 10 AD gene(s) localized in the three AD linkage papers published earlier in Science. Minimally, testing across samples will be necessary to resolve this issue.

In the initial gene expression studies, Li et al reported four genes that significantly differ in their expression between AD and control hippocampi and map to the chromosome 10 AD/PD age-at-onset linkage region. These genes are stearoyl-CoA desaturase; NADH ubiquinone oxidoreductase 1 beta complex 8; protease serine 11; and glutathione S-transferase, omega-1 or GSTO1. In the expression analyses, Li et al started with 22,000 genes and then went on to select 1,072 genes (demonstrating the strongest signals) for comparison in AD and control brain samples. They then found that 52 out of the 1,072 genes demonstrated "significant differences in gene expression levels between AD and controls." Interestingly, if one uses p = 0.05 as a cutoff for statistical significance, one would expect to randomly find ~54 positive genes from a total of 1,072; here, they report 52. Thus, one could argue that the initial data emerging from the “expression component” of the convergence strategy could have been obtained purely by chance. However, it is certainly interesting that four of these 52 genes mapped near the AD/PD age-at-onset locus on chromosome 10.

These four genes were then tested for genetic association with AD (using 2,814 samples; 1,773 AD cases) and PD (1,362 samples; 635 cases). Allelic association was found for two members of the GST omega family (GSTO1 and GSTO2) and the further distally located PRS11. These findings are particularly intriguing, given the potential roles of the GSTO proteins in the post-translational modification of interleukin-1β and the inflammation process. For GSTO1, the authors report association of age-at-onset for both AD and PD with same allele, Asp-140, of the polymorphism Ala140Asp. In the discussion, the authors make a biological argument that the Asp-140 allele of GSTO1 responds differently from the Ala allele, depending on the substrate (e.g., IL-1β) provided, implying that this SNP may have functional significance in both AD and PD. However, it is puzzling that the significance of the results of the combined analyses of AD and PD are not stronger than those obtained with the individual disease samples, which would be expected if the same allele, Asp-140, were involved in both diseases. Perhaps the authors could comment on this.

In complex genetic studies like this where the significance of association cannot be reliably determined, it would also be helpful if the authors could provide some estimate of the effect size underlying the association, i.e., by how many years the onset of AD and PD was actually delayed, and what were the confidence bounds of these estimates. It would be particularly interesting to compare the degree of onset age delay across the two populations.

And finally, it would be helpful if the authors addressed a pressing question that often arises when progressing from an initial linkage result (Li et al., 2002) to an association result (Li et al., 2003): Does the association of age-at-onset with GSTO1 account for the original linkage peak for age-at-onset on chromosome 10?

In summary, these are very intriguing findings implicating a biologically compelling candidate gene that potentially affects age-at-onset in both AD and PD at the level of regulating inflammatory responses. As with any novel gene-disease association, confirmation by other groups testing independent samples will be essential for ultimate validation. In the spirit of healthy and constructive scientific discourse, we welcome the authors’ replies to this commentary.

View all comments by Lars Bertram
View all comments by Rudy Tanzi

We appreciate the opportunity to expand on our recent paper identifying GSTO1 as a modifier of age at onset (AAO) in neurological disease. We agree with Bertram and Tanzi that one or more risk genes for AD exist on chromosome 10q. In our previous age-at-onset genomic screen paper (Li et al., 2002), we addressed the linkage findings of the chromosome 10q region and the differences between the risk and age-at-onset genes in this region for AD. Like other reports for risk genes, we have consistently found evidence for a risk gene(s) on 10q in our data and reported this in several publications (Pericak-Vance et al., 1998, Pericak-Vance et al., 2000, Haines and Pericak-Vance, 2001). The 10q linkage region for risk genes is about 47 cM proximal to our linkage region for age at onset. The only overlapping region between the risk and age-at-onset effects was found in one (the NIMH)...  Read more

We appreciate the opportunity to expand on our recent paper identifying GSTO1 as a modifier of age at onset (AAO) in neurological disease. We agree with Bertram and Tanzi that one or more risk genes for AD exist on chromosome 10q. In our previous age-at-onset genomic screen paper (Li et al., 2002), we addressed the linkage findings of the chromosome 10q region and the differences between the risk and age-at-onset genes in this region for AD. Like other reports for risk genes, we have consistently found evidence for a risk gene(s) on 10q in our data and reported this in several publications (Pericak-Vance et al., 1998, Pericak-Vance et al., 2000, Haines and Pericak-Vance, 2001). The 10q linkage region for risk genes is about 47 cM proximal to our linkage region for age at onset. The only overlapping region between the risk and age-at-onset effects was found in one (the NIMH) dataset (Li et al., 2002). Our recent report of GSTO1 gene (Li et al., 2003) showed that GSTO1 modifies age at onset in AD and Parkinson's disease (PD), but we did not see any evidence of risk effect. Clearly, the risk genes on chromosome 10q remain to be found. A number of candidate risk genes in this region have been examined by others, and we are actively investigating these and other candidate genes in our data, as well.

We fear that Bertram and Tanzi may have overinterpreted the differences in p-values for the GSTO1 SNP7 association in the separate and combined datasets. The p-values found in the combined dataset for both SNP7 and SNP9 were <0.008 for the orthogonal model (OM) and <0.02 for the Monks-Kaplan (MK) method, while the p-values were 0.007 (SNP7) and 0.005 (SNP9) from the OM method, and 0.023 (SNP7) and 0.024 (SNP9) from the MK method for AD alone. Clearly, equivalent p-values were obtained from AD and the combined datasets. The inclusion of PD data did not diminish the evidence of GSTO1 association. Although it is true that a larger sample size may enhance the power of a test, several other factors, including the mean and variance of AAO also affect the p-value. Any reported p-value is simply a point estimate of the true significance, and there is expected variation around that estimate. Thus, unless the differences in p-values are (significantly) large, no additional interpretation should be assigned to their differences. The important message from our analyses is that the evidence of GSTO1 association trends in the same direction in the AD, PD, and the combined AD and PD datasets. Therefore, we believe that GSTO1 plays a common role in both AD and PD. We understand the interest in knowing the delay in age at onset of AD and PD related to GSTO1. However, the within-family nature of these tests makes it unclear what relevant measurement should be used to quantify this difference. We are actively investigating methods to quantify the age differences in families.

In our report, SNP7 in GSTO1 shows significant association with AAO. To answer the question of whether SNP7 contributes to the original linkage signal of chromosome 10q, we performed the same linkage analysis (Li et al., 2002) with the addition of SNP7 under the same model. The LOD score remains similar. However, when SNP7 was treated as a covariant for a linkage analysis, we observed a substantial (50 percent) reduction in LOD score to <1.0, strongly suggesting that SNP7 contributes significantly to our observed linkage effect on 10q that we identified in our previous genomic screen (Li et al., 2002).—Yi-Ju Li, Jonathan Haines, and Margaret Pericak-Vance.

References:
Haines JL, Pericak-Vance MA (2001) A Genomic Search for Alzheimer's Disease Genes. In Iqbal K, Sisodia SS, Winblad B (eds) Alzheimer's Disease: Advances in Etiology, Pathogenesis and Therapeutics. London: John Wiley & Sons.

Li YJ, Oliveira SA, Xu P, Martin ER, Stenger JE, Scott WK, et al. Glutathione S-transferase modifies age-at-onset of Alzheimer disease. Human Molecular Genetics 2003;(12)24:1-9.

Li YJ, Scott WK, Hedges DJ, Zhang F, Gaskell PC, Nance MA, Watts RL, Hubble JP, Koller WC, Pahwa R, Stern MB, Hiner BC, Jankovic J, Allen FA Jr, Goetz CG, Mastaglia F, Stajich JM, Gibson RA, Middleton LT, Saunders AM, Scott BL, Small GW, Nicodemus KK, Reed AD, Schmechel DE, Welsh-Bohmer KA, Conneally PM, Roses AD, Gilbert JR, Vance JM, Haines JL, Pericak-Vance MA. Age at onset in two common neurodegenerative diseases is genetically controlled. Am J Hum Genet. 2002 Apr;70(4):985-93. Epub 2002 Mar 01. Abstract

Pericak-Vance MA, Bass ML, Yamaoka LH, Gaskell PC, Scott WK, Terwedow HA, Menold MM, Conneally PM, Small GW, Saunders AM, Roses AD, Haines JL. Complete genomic screen in late-onset familial Alzheimer's disease. Neurobiol Aging. 1998 Jan-Feb;19(1 Suppl):S39-42. Abstract

Pericak-Vance MA, Grubber J, Bailey LR, Hedges D, West S, Santoro L, Kemmerer B, Hall JL, Saunders AM, Roses AD, Small GW, Scott WK, Conneally PM, Vance JM, Haines JL. Identification of novel genes in late-onset Alzheimer's disease. Exp Gerontol. 2000 Dec;35(9-10):1343-52. Abstract

View all comments by Yi-Ju Li