Since the initial publication of a cognitive GWAS demonstrating an association between KIBRA
SNP rs17070145 and episodic memory(Papassotiropoulos, et al., 2006
), there has been interest in replicating and better understanding the nature of the relationship of KIBRA
with cognition and AD risk. Early follow-up studies on the cognitive associations were promising(Almeida, et al., 2008
,Preuschhof, et al., 2010
,Schaper, et al., 2008
,Vassos, et al., 2010
), but a number of subsequent reports have either failed to demonstrate the association(Need, et al., 2008
), or showed association in the opposite direction(Nacmias, et al., 2008
). Two studies have evaluated the association between KIBRA
rs17070145 T allele with AD, with one study finding a protective effect(Corneveaux, et al., 2008
) and the other no association in their complete series and a risky effect in their oldest age group(Rodriguez-Rodriguez, et al., 2007
). In light of these discrepancies, we sought to investigate the association between episodic memory and LOAD risk with rs17070145 and 14 additional SNPs within and flanking KIBRA
in our series of 2403 AD cases and 2511 controls.
To our knowledge, our study represents the largest LOAD case-control study to date, and about 1.4 times greater in size than all published series combined. It is also the first study to evaluate KIBRA
for cognition and AD risk in a different ethnic group, African-Americans. Despite a slightly protective risk estimate (OR = 0.97), our combined series did not show significant LOAD risk association with KIBRA
rs17070145 T allele. When each case-control series was individually assessed, older African-Americans had a significant, protective association with the rs17070145 T allele, consistent with the Corneveaux et al. study(Corneveaux, et al., 2010
). Two of our older Caucasian series (JS_OLD and RS_OLD) also had protective trends for this allele, as did the combined older Caucasian series, though they did not achieve nominal significance. This marginally protective effect for the T allele was further demonstrated in a meta-analysis of all of our series together with the published data on rs17070145. Finally, KIBRA
message was found to be overexpressed (p=0.06) in the temporal cortex, but not in the pathologically-spared cerebellum (p=0.91), of AD subjects compared with controls, consistent in trend with a previous report(Corneveaux, et al., 2010
). Our expression assay is designed to measure all three known isoforms of KIBRA
. Therefore if there are isoform-specific differences in KIBRA
expression in the brains of AD vs. non-AD subjects, this expression assay will not be able to detect these. Thus, it remains possible that, although we found only a trend for KIBRA
overexpression in the temporal cortex of AD subjects, bigger differences could potentially be identified by utilizing isoform-specific assays. Although, we did not identify any associations between the 15 SNPs and KIBRA
expression levels, there may be other KIBRA
variants that influence its brain expression via influencing differential splicing or transcription factor binding.
These AD association findings for KIBRA
rs17070145 SNP raise several intriguing observations. First, an age effect is evident in both our African-American and Caucasian series. Older subjects at the time of AD diagnosis/evaluation show protective trends for the rs17070145 T allele, whereas our younger subjects show no evidence of this association (except for a trend toward the opposite risky effect in our young autopsy series). A prior study also found a significant association between rs17070145 and LOAD among older adults, although the effect was in the opposite risk direction(Rodriguez-Rodriguez, et al., 2007
). Although the older and younger series were not statistically different based on their overlapping 95% CIs and an insignificant Breslow-Day p value, it would be interesting to re-evaluate all existing published series by their ages of onset to determine the influence of age AD risk association in the largest possible dataset.
Second, we show a nominally significant association between the KIBRA
rs17070145 T allele and LOAD risk among African American adults. This effect is in the same, protective direction as that shown in the largest published Caucasian AD association study(Corneveaux, et al., 2010
), as well as what may be expected by KIBRA
's memory-enhancing effect in cognitive studies(Papassotiropoulos, et al., 2006
,Preuschhof, et al., 2010
,Schaper, et al., 2008
,Vassos, et al., 2010
). While preliminary, this finding suggests that the effects of this gene on AD risk may generalize to different ethnic/racial groups. Given the relatively small sample size of this group this association in African-Americans needs replication by others.
Third, the marginal significance (p=0.07) for the rs17070145 T allele obtained in our meta-analysis of 4436 AD cases and 4334 controls likely reflects the lack of power to detect the relatively small effect conferred by this variant. Indeed, even with the current aggregate sample of nearly 9000 subjects in our meta-analysis, there is ~50% power to detect an association at α=0.05, given the minor allele frequencies observed in the cases and controls. Furthermore, if rs17070145 is not the functional KIBRA polymorphism but a tagging SNP for the functional variant, then different underlying LD patterns in the different series could lead to additional heterogeneity and lack of a discernable association signal in some series, whereas others may show an association signal. Different gene-gene, gene-environment influences in different series, and presence of multiple underlying functional variants within the same gene are additional sources for potential false-negative associations.
We did not observe a significant association for the KIBRA
rs17070145 SNP with episodic memory scores, although several other SNPs showed Bonferroni-corrected association at p<0.05 in one of our series (JS_OLD) .The lack of a clear association between rs17070145 and our two episodic memory phenotypes stands in contrast to the original findings from Papassotiropoulos and colleagues(Papassotiropoulos, et al., 2006
). It should be noted that although there are reports of both AD risk and cognitive associations with the KIBRA
rs17070145 SNP, to our knowledge positive associations for both phenotypes in the same series has not been reported. There could be several explanations for the lack of episodic memory associations with rs17070145 in our series: 1) KIBRA
rs17070145 SNP may influence episodic memory and AD risk differentially depending on age (i.e., a stronger effect on memory early on, but on AD risk later in life). 2) The effect of this SNP on episodic memory may be influenced by factors that have not been taken into account in this study. For example, in a recent study that also employed both the AVLT and Logical Memory tests, the authors failed to find a direct association with delayed recall scores, although they were able to demonstrate an association between AVLT delayed recall and rs17070145 after controlling for initial learning in multiple regression analyses(Bates, et al., 2009
). 3) Different KIBRA
variants may influence memory and AD risk. 4) Our memory association results for rs17070145 may represent false negatives due to the modest effect conferred by KIBRA
and lack of sufficient statistical power. 5) It remains possible that some of the positive associations reported by other investigators may represent false positives. Investigating both AD risk and cognition concurrently in additional series will be necessary to explore these possibilities.
There are a number of significant strengths to our investigation. First, our case-control series represents the largest KIBRA
association study reported thus far. Second, we present data on a cohort of African American adults, and analyze their association to LOAD and episodic memory uniquely with appropriate adjustments (for instance, taking into account their younger age when compared to Caucasian-Americans). Third, we did not limit our study to a single SNP (rs17070145) but considered additional SNPs that may bear on the relationship between KIBRA
and LOAD or episodic memory. Fourth, we pursued meta-analysis of the rs17070145 T allelic association using our entire series and the available published data to maximize the statistical power to find effects on this SNP. Fifth, we investigated brain expression levels of KIBRA
similar to one other study(Corneveaux, et al., 2010
In summary, genetic data from a subset of our series provides suggestive evidence for a role of KIBRA SNPs in AD risk and memory, although significant questions remain. KIBRA appears to confer at most a very modest effect in AD risk and memory, and even sample sizes in excess of 9000 subjects may be underpowered to detect this effect. Thus, clearly larger studies in both Caucasians and other ethnic groups are needed for stronger evidence in support of KIBRA's role in LOAD and memory. Moreover, targeted re-sequencing efforts to identify functional variants that can subsequently be tested in both functional paradigms and association studies will provide definitive answers to some of the questions raised by this and other studies. Use of multiple phenotypes, including broad assessment across cognitive domains including and beyond episodic memory, AD risk, and gene expression levels concurrently in the same series may bring additional understanding to the genetic complexity underlying KIBRA.