PLD3 variants in population studies.

Abstract

Cruchaga et al.1 report that rare genetic variants in PLD3(phospholipase D3) are associated with increased Alzheimer’s disease(AD) risk. They showed that PLD3 is involved in amyloid-β precursor protein (APP) processing and overexpressed in brain tissue from AD patients. However, even the key variant PLD3-Val232Met, did not pass genome wide significance. This observation raises the question if the genetic association of PLD3 with AD replicates. We associated PLD3-Val232Met with AD in 3 large population-based studies and 3 case-control studies. In total, we meta-analyzed results from 1914 AD cases and 8021 controls of European descent. Additionally we searched for other coding PLD3-variants in sequence data of 1067 AD cases and 1553 controls. Carrier frequencies of PLD3-Val232Met in controls ranged from 0.34-1.42%, consistent with 0-1.17% reported by Cruchaga et al.1 (Table 1). Likewise, the frequencies of PLD3-Val232Met in cases ranged from 0.66-2.19% compared to 0.7-2.6%.1 We note that the range of carrier frequencies overlaps between cases and controls, such that in some population based cohorts, the carrier frequency in controls (e.g., 1.28% in FHS) is higher than that of cases in other cohorts (e.g., 0.68% in AGES). Within each cohort frequencies of PLD3-Val232Met were higher in cases than controls (Table 1), but in none of the populations the case carrier frequency for PLD3-Val232Met was significantly increased. However, pooled analyses showed a 1.94 fold increased risk of AD of carriers compared to non-carriers (Odds Ratio [OR] 1.94, adjusted for age and sex, 95% confidence interval=1.05 to 3.57) that was marginally significant (p-value = 0.03)(Table 1). Of note, the crude ORs often differed considerably from the age and sex adjusted estimates. With the exception of ADNI, the ORs were higher after adjustment for age and sex, suggesting that many a-symptomatic carriers were relatively young compared to cases and age needs to be controlled for in the analyses as a putative confounder. Table 1 Association of PLD3-Val232Met with Alzheimer’s disease We further associated other coding variants in PLD3 with AD and performed a gene-based test using sequence data from two studies encompassing 1067 AD cases and 1553 controls. We meta-analyzed results of whole genome sequence data of the ADNI study, 499 AD cases and 293 controls, with results of a combined cohort of 568 Dutch AD cases and 1260 Dutch controls. We observed 21 rare polymorphic coding variants and 1 splice site variant. Of the 20 observed PLD3-variants detected by Cruchaga et al.1, we observed 9 (S63G, P76A, V232M, N284S, C300Y, A442A, G452E, D447G and R488C). Five variants showed the same direction of effect as seen by Cruchaga et al.1. PLD3-A442A was one of the variants that showed a same direction of effect(Odds Ratio [OR] 1.24, 95% confidence interval = 0.74 to 2.06, p-value = 0.41). After correcting the p-value for multiple testing, none of the variants observed in our study conferred a significant increase in AD risk. Gene-based analysis also did not show significant association of PLD3-variants with AD risk (SKAT-O p-value = 0.61 and burden test OR 1.27 95% confidence interval [CI] =0.85 to 1.9, p-value = 0.24). In conclusion, the carrier frequencies of PLD3-Val232Met in our data set are consistent with those reported by Cruchaga et al.1 and we showed a nominally significant association of PLD3-Val232Met with AD. This is in contrast to findings presented in companion papers by Heilmann et al., Lambert et al. and Hooli et al. However, in contrast to Cruchaga et al.1 we found no significant association of other PLD3-variants with AD in the single variant or gene based analyses. Therefore, in our analyses PLD3 does not yet meet the criteria proposed by MacArthur et al.2 to be implicated in AD. Hence, our data do not strongly support an important contribution of rare PLD3-variants in the etiology of AD. The most notable findings in our study are the need to control for age as a confounder in rare variant analyses and the high variability of the frequency of PLD3-Val232Met across populations. The latter finding highlights the need for careful matching of cases and controls for ethnic background when investigating rare variants.