Rare Variants in PLD3 Increase Risk for Alzheimer’s Disease in Han Chinese

INTRODUCTION

Recent advances in next-generation sequencing technology have made it possible to identify novel functional variants with large effect sizes associated with Alzheimer’s disease (AD) risk. A rare coding variant p.V232M (rs145999145) in phospholipase D3 (PLD3) gene was found to segregate with the disease in two unrelated families with late-onset AD (LOAD) and a two-fold increased AD risk in the cohorts of European descent [1]. This association of p.V232M with AD risk has been successfully replicated in a Dutch population [2]. Another independent resequencing study in a German population validated the association of PLD3 with AD at the gene level, but not the association of p.V232M variant with AD [3]. Other replication studies on familial or sporadic AD failed to find the association of p.V232M variant with AD risk [4–7]. Meanwhile, no p.V232M variant was identified in two previous studies (360 LOAD and 400 controls included in study of Jiao et al., and 18 AD probands with familial disease history in study of Zhang et al.) in Han Chinese population before [8, 9]. In addition to p.V232M, most of other coding variants were located in exon11 of PLD3, including the synonymous variant p.A442A (rs4819), which affected splicing and was reported to be nominally associated with LOAD risk in the original publication [1]. Here, to test whether coding region variants of PLD3 were associated with AD in a large Han Chinese cohort, we performed sequencing to analyze all 13 exons of PLD3 in 960 LOAD patients and 1,330 healthy controls.

METHODS

The present study comprised 2,290 unrelated subjects (960 patients with LOAD and 1330 healthy control subjects matched for gender and age), who were northern Han Chinese in origin. The patients were collected from the Department of Neurology at Qingdao Municipal Hospital and several other hospitals in Shandong province. Clinical diagnosis of probable AD was established following the NINCDS-ADRDA criteria [10] independently by at least two senior neurologists. All patients were defined as sporadic since none of their first-degree relatives had a family history of dementia. The age- and gender-matched healthy controls were recruited from the Health Examination Centers of each collaborating hospital and were confirmed healthy through medical history, general examinations, laboratory examinations, and Mini-Mental State Examination (MMSE) score by physicians. Informed consent was obtained from subjects directly or from their caregivers, and the protocol for this study was approved by the Institutional Ethics Committees of Qingdao Municipal Hospital.

High-throughput sequencing was carried out by Shanghai Genesky BioTech Company as described [11]. Briefly, genomic DNA was extracted from venous blood using the Wizard genomic DNA purification kit (Promega, Madison, WI, USA). The sequences of PLD3 were obtained from the UCSC Genome Browser Database and referred to Human Genome Resources. Specific primers for exons of PLD3 were designed using the primer 3 software. PCR amplification reactions were then carried out in a GeneAmp PCR system (Applied Biosystems, Foster City, CA, USA). PCR products were sequenced on an Illumina MiSeq high-throughput sequencing platform (Illumina, San Diego, CA, USA). The details of protocol are available upon request.

The potential effects of rare coding variants on protein function were predicted using PolyPhen-2 (http://genetics.bwh.harvard.edu/pph2/) [12] and SIFT (http://sift.jcvi.org/) [13]. Bioinformatic prediction analysis of splicing variants was performed using Human Splicing Finder 3.0 software (http://www.umd.be/HSF3/HSF.shtml). Differences in age and MMSE scores between AD patients and controls were detected using Student’s t test. Differences in gender and APOE ɛ4 status between these two groups were assessed using χ² test. Differences in allele distributions between patients and controls were analyzed by Fisher’s exact test, and the p value, odd ratio (OR), and 95% confidence intervals (CIs) were then calculated using SPSS version 13.0 statistical software. Bonferroni correction was used for the adjustment of multiple comparisons. p < 0.05 was considered statistically significant.

RESULTS

The demographic characteristics of patients and controls were summarized in Table 1. We first sequenced exon11 of PLD3 for rare variants that were related to LOAD in our cohort. Two rare variants (c.1067G>A: p.R356H and c.1020-8G>A, present in the heterozygous state) were identified, both of which were only present in different LOAD patients. Of note, a variant c.1020-8G>A close to splice site was significantly associated with LOAD risk (Pcorrected = 0.015, OR = 3.59, 95% CI: 1.29–10.04, Table 2). However, a nonsynonymous mutation p.R356H was not significantly associated with LOAD risk after Bonferroni correction (Pcorrected = 0.465, Table 2). Besides, the previously reported rare variant p.A442A in exon11 in European populations was not detected in our cohorts.

Based on the above positive findings in exon11 of PLD3, we further extended our study by sequencing the other exons of PLD3 in our cohort. As shown in Table 2, a total of 13 rare variants which are all present in the heterozygous state were identified, including 10 nonsynonymous mutations and 3 other variants. All these variants were from separate individuals, and no individual harbored two or more variants. Although many rare variants were more frequent in LOAD patients than in controls, only p.I163M was significantly associated with the risk of LOAD, as its minor (G) allele significantly increased LOAD susceptibility after Bonferroni correction (Pcorrected = 0.03, OR = 12.47, 95% CI: 1.58–98.25, Table 2). Besides, since many variants in PLD3 were infrequent, a pooled chi-squared analysis of rare variants was used to assess the overall burden of variants between LOAD cases and controls. This analysis also revealed statistically significant difference (p = 4.21×10^{- 6}). Notably, the previously reported LOAD-associated p.V232M variant was identified (only one carrier) in our LOAD group of Han Chinese cohort (Table 2).

DISCUSSION

Recent next-generation sequencing studies had identified PLD3 as a new risk gene, with risk variants doubling the risk for developing LOAD in the cohorts of European descent [1]. However, some follow-up replication studies questioned that PLD3 might not be so important in AD [2, 4–7]. Based on a query with “PLD3” at the AlzData.org webserver, which aims to provide an in-depth integrating system to integrate multiple data of AD [14], we found a relatively low convergent functional genomics ranking score, suggesting a potentially weak effect of this gene in AD. In addition to the genetic association evidence, Cruchaga et al. [1] demonstrated that PLD3 is down-regulated in AD brain tissues and functions in AβPP processing. PLD3 was highly expressed in hippocampus and cerebral cortex which are critical regions for AD pathology. And it was shown that senile plaques in the frontal cortex containing dystrophic neurites have the intense accumulation of PLD3 [15]. Preliminary studies of our group have found that PLD3 variants were significantly associated with CSF Aβ_{1 - 42} levels and influenced AD-related neuroimaging, which further provided multiple lines of evidence for the possible roles of PLD3 in AD pathogenesis [16–18].

In this study, using high-throughput sequencing techniques, we showed for the first time that a rare coding variant p.I163M in exon7 and an intronic variant c.1020-8G>A in exon11 were associated with an increased risk of LOAD in Han Chinese population. Because of their rarity (MAFs were 0.22% and 0.17% respectively in our total sample), this finding needs to be further confirmed in a larger and independent Han Chinese cohort. According to Polyphen-2 and SIFT score, p.I163M has possibly damaging impacts on protein function of PLD3, and further studies are required to investigate how it affects PLD3 functions and confers risk for LOAD. Bioinformatic prediction analysis of splicing variants was performed using Human Splicing Finder 3.0, which predicted that the c.1020-8G>A variant has probably no impact on splicing.

Interestingly, the previously reported rare variant p.V232M in European cohorts was not identified in Han Chinese population before [8, 9]. However, we found the occurrence (only one carrier) of p.V232M in our AD group, but no association of p.V232M with LOAD risk in our current study. This might be explained by the genetic background differences or limited sample size of the Chinese cohorts. In addition, we identified a rare mutation p.R356H that was present only in patients with LOAD, but was not associated with LOAD risk after Bonferroni correction. The Bonferroni correction is a method used to adjust the p value to counteract the problem of multiple comparisons and reduce the chance of a type I error. It should be noted that p.R356H variant had been identified in a European cohort previously [3], but was also not significantly related to LOAD susceptibility due to its rarity. Considering its probably damaging impacts on functions of PLD3 based on Polyphen-2 score, further functional study might also be needed.

In summary, our results indicated that rare variants of PLD3 were likely to play an important role in LOAD in northern Han Chinese population. All these results require further confirmation in Han Chinese by future studies with larger samples, and functional studies are needed to investigate the underlying mechanism by which these rare variants influence the susceptibility to LOAD.