An intronic PICALM polymorphism, rs588076, is associated with allelic expression of a PICALMisoform

Phosphatidylinositol binding clathrin assembly protein (PICALM) facilitates clathrin-mediated endocytosis. PICALM binds phosphatidylinositol 4,5- bisphosphate (PIP₂), adaptor protein 2 (AP₂) and clathrin to mediate endocytic clathrin coated vesicle formation at the plasma membrane. Although PICALM is ubiquitously expressed, PICALM expression is more pronounced in microvessels [1, 2]. Previous studies have shown PICALM co-localizes with APP and modulates amyloid beta (Aß) generation [3‐5]. Accumulation of Aß deposits is a hallmark of Alzheimer’s Disease (AD) pathology.

Genome wide association studies in multiple cohorts have identified single nucleotide polymorphisms (SNP)s near the PICALM gene as significantly associated with AD risk [6‐10]. Studies were first conducted with Caucasian populations and then independently verified in several although not all Asian populations [11‐15]. These studies report that the rs3851179 A allele reduces AD risk with an odds ratio of 0.88 [6]. This SNP is located approximately 80 kb 5′ of PICALM.

Understanding how rs3851179 alters PICALM to impact AD risk may lead to novel insights into AD mechanisms and potential treatments. Since rs3851179 is not in linkage disequilibrium (LD) with a SNP that alters a PICALM amino acid (r² < 0.1), we hypothesize that rs3851179 is associated with changes in mRNA transcription or processing. Allelic expression imbalance (AEI), which is an expression difference between allelic transcripts within an individual, has been used to detect cis-regulatory effects [16‐20]. Here, we performed an AEI analysis by comparing allelic expression through the use of two exonic SNPs, rs76719109 and rs592297, in AD and non-AD brain samples. These studies included 35 samples that were heterozygous for rs76719109 and 19 samples that were heterozygous for rs592297. While PICALM expression did not show AEI overall, one individual showed robust PICALM AEI, with an allelic ratio of 0.76. Additionally, significant AEI was detected for the PICALM isoform lacking exons 18 and 19 (D18-19 PICALM). Sequencing and additional genotyping established that rs588076 was robustly associated with this AEI pattern. Interestingly, rs588076 has been associated with blood pressure response to ^Ca++ channel blocking agents [21]. We discuss these overall results in the context that genetic regulation of PICALM isoforms relative to AD risk is highly complex with further work necessary to elucidate the mechanisms modulating genetic risk.

To detect the presence of regulatory cis-acting SNPs in human brain samples, we measured allelic ratios in cDNA from reverse transcribed mRNA. Heterozygosity for an exonic "reporter" SNP provides the means to compare the expression of one allele with another allele within an individual. Our criteria for reporter SNPs for AEI analysis is that the SNPs are present in exons and have a minor allele frequency (MAF) greater than 15%, which allows for sufficient sample numbers for analysis. Only two PICALM SNPs satisfied these criteria, rs76719109 and rs592297 (Figure 1). Rs76719109 has a MAF of 0.44 and resides within exon 17; PCR amplification from exon 17–20 allowed us to measure AEI for total PICALM as well as PICALM splice variants lacking exon 18 or exons 18–19 (Figure 1a). Rs592297 has a MAF of 0.20 and resides in exon 5. PCR amplification from exon 5–6 produced a single PCR product for cDNA (Figure 1b). The AEI assay was validated in two ways. First, we tested the linearity of the assay by generating a cDNA standard curve consisting of five different rs76719109 T:G ratios (Figure 2). Our input T:G ratios ranged from 1:4 to 4:1. We found a robust linear relationship between input and observed T:G ratios. Second, we applied our experimental approach to genomic DNA (gDNA), which represented a positive control with an expected "allelic" ratio of 1.0. Rs76719109 and rs592297 showed gDNA ratios of 1.01 ± 0.03 (mean ± SD, n = 35) and 0.96 ± 0.05 (mean ± SD, n = 19), respectively (Figure 3). Hence, this AEI assay appears robust for detecting and quantifying variations in allelic expression.

To evaluate whether the AD-associated SNP rs3851179 was associated with unequal allelic PICALM expression, we performed AEI analysis with rs76719109 and rs592297on a total of 54 samples. Twelve of these 54 samples were heterozygous for both SNPs. Hence, we analyzed PICALM for AEI in a total of 42 unique individual samples. This effort analyzed 4.2 million sequences for rs76719109 and 1.4 million sequences for rs592297. If rs3851179 modulated total PICALM expression, we expected to see significant AEI in individuals heterozygous for rs3851179, but not in individuals homozygous for rs3851179. When we analyzed the results for the exon 17 SNP, rs76719109, significant AEI was observed in only a single sample, termed AD40 (see below). To evaluate whether a subtle difference in allelic expression may be present and associated with rs3851179, we compared the mean allelic expression between the rs3851179 homozygous and heterozygous groups. This approach did not discern a significant difference between the two groups, i.e., the allelic ratios for rs3851179 homozygous and heterozygous groups was 0.94 ± 0.08 (n = 17) and 0.97 ± 0.06 (n = 18), respectively (Figure 4a, p = 0.24, t-test). Hence, rs3851179 did not appear associated with PICALM AEI. To confirm this finding and extend the analysis to additional samples, we analyzed allelic expression by using the exon 5 SNP, rs592297. Significant AEI was not observed in any sample, noting that the sample with the significant AEI result from the rs76719109 analysis was homozygous for rs592297 and not suitable for evaluation. When the allelic ratios were analyzed by t-test to evaluate whether rs3851179 was associated with allelic expression, the findings confirmed that rs3851179 was not associated with robust AEI; the allelic ratios were 0.93 ± 0.05 (n = 4) for rs3851179 homozygotes and 1.01 ± 0.07 (n = 15) for heterozygotes (Figure 4b, p = 0.06). Hence, we found that the total PICALM allelic ratio for each person was remarkably consistent with both reporter SNPs; rs3851179 is not associated with overall allelic PICALM expression.

Interestingly, cDNA from the individual termed AD40 showed significant unequal allelic mRNA expression with a T:G ratio of 0.76 (Figure 4a, Additional file 1: Table S1). This ratio is based on a total of four experiments that detected a total of 68141 copies of the T allele and 88400 copies of the G allele. This AEI was not due to genomic normalization as the genomic T:G ratio was 1.01. These data are based upon the rs76719109 SNP because this individual was homozygous for rs592297. A similar result was obtained when each of four separate replicates were analyzed individually, i.e., when each replicate was analyzed individually, the T:G ratio was 0.73 ± 0.04 (mean ± SD). Hence, significant AEI was observed in a single individual among the 42 unique samples.

Having considered overall PICALM allelic expression, we proceeded to apply this AEI analysis to PICALM isoforms. The exon 17 to exon 20 PCR amplicons captured three different PICALM isoforms because exons 18 and 19 are variably spliced. These isoforms were termed full length PICALM (contains exons 18–19), D18-PICALM (lacked exon 18) or D18-19 PICALM (lacked both exons 18 and 19). We analyzed each of these isoforms for the presence of AEI as a function of rs3851179 heterozygosity. For the full length PICALM and D18-PICALM isoforms, significant AEI was observed only in the AD40 sample. To evaluate whether a subtle difference in allelic expression may be present, we compared the allelic expression of the full length PICALM isoform between the rs3851179 homozygous and heterozygous groups by using a t-test. However, no difference was observed as the average T:G ratio in the rs3851179 homozygous and heterozygous groups was 0.93 ± 0.07 (n = 17) and 0.93 ± 0.06 (n = 18), respectively (Figure 5a, p = 0.81). Likewise, for the D18-PICALM isoform, the rs3851179 homozygous and heterozygous groups showed mean T:G ratios of 0.95 ± 0.09 (n = 17) and 1.01 ± 0.12 (n = 18), respectively (Figure 5b, p = value = 0.13). When we evaluated the D18-19 PICALM isoform, significant AEI was detected in multiple samples (Figure 5c). We analyzed these results in two ways. First, we compared the frequency of samples with significant AEI between the rs3851179 homozygous and heterozygous individuals by using a Fisher’s exact test; a significant difference between groups was not detected (Figure 5c, p = 0.44). Second, we compared the mean T:G ratio between rs3851179 homozygous and heterozygous individuals by using a t-test. However, the rs3851179 homozygous and heterozygous individuals showed similar values that did not achieve significance, i.e., 1.19 ± 0.16 and 1.11 ± 0.22, respectively (p = 0.08). Hence, rs3851179 heterozygosity was not associated with AEI for these PICALM isoforms.

We considered the subset of samples that showed significant AEI further. Among these, AD40 showed significant AEI for each of the isoforms, with the D18, D18-19 and full length PICALM isoforms having allelic T:G ratios of 0.77, 0.53 and 0.77, respectively. As noted above, this finding is consistent with the hypothesis that a rare SNP acts to alter global PICALM allelic expression in this individual. Additionally, we noted that the D18-19 PICALM isoform showed significant AEI for multiple individuals (Figure 5c). In these eight individuals, the rs76719109T allele was expressed more than the G allele with the ratio ranging from 1.30 to 1.48. This is in contrast to AD40 where the T allele was expressed less than G allele at a ratio of 0.53 (Table 1).

Since the D18-19 PICALM isoform but not overall PICALM showed AEI in multiple samples, we hypothesized that a local SNP influences this splice pattern. To identify this SNP, we sequenced 6400 bp of genomic DNA between exons 17–20 in C11 and AD33, the two individuals showing the highest AEI ratio (Table 1). We identified several SNPs that were heterozygous in these samples including rs588076, rs645299, and rs618629. The MAF for these SNPs in CEU range from 20-31% (Table 2). Additionally, these SNPs are in strong LD with each other (Table 2). Rs588076 and rs618629 reside in intron 17 while rs645299 is within intron 18. We genotyped all samples for rs588076. The frequency of samples showing significant AEI was significantly associated with rs588076 heterozygosity (Figure 5d, p = 9.54 × 10^-5, Fisher’s exact test). We interpret these results as strongly suggesting that rs588076, or a SNP in strong LD with rs588076, is functional by modulating PICALM exon 18–19 splicing.

Since rs588076 is associated with D18-19 PICALM AEI and rs3851179 has been robustly associated with AD, we evaluated the extent that rs588076 is associated with AD risk. We evaluated 1789 AD and 2529 non-AD individuals from the Mayo Clinic cohort. Rs588076 was significantly associated with AD risk when analyzed in a logistic regression model (odds ratio = 0.8413, 95% confidence intervals (0.7151-0.9898), p = 0.0372). However, when rs3851179 was added to this logistic regression model, only rs3851179 was significantly associated with AD (Table 3). Haplotypic analysis confirmed that the haplotype containing both rs588076 and rs3851179 report the same association with AD as the haplotype containing only rs3851179 (data not shown). Hence, rs588076 is associated with D18-19 PICALM AEI but not AD risk.

The primary findings of this report include (i) overall PICALM expression shows no evidence of global AEI even when parsed by AD-associated SNPs, (ii) robust global AEI was detected in one sample, suggesting the existence of a rare SNP that strongly modulates PICALM expression, and (iii) eight individuals show AEI for the D18-19 PICALM isoform that is associated with rs588076. However, rs588076 was not associated with AD risk when considered in a model that also included rs3851179. In summary, analysis of allelic expression has proved a useful tool for the evaluation of cis-acting regulatory polymorphisms and AD risk.

A consistent pattern of AEI in overall PICALM expression was not detected. This was unexpected since Xu et al. reported consistent and robust PICALM AEI [22]. The reason for different results in these two studies is unclear. The studies are similar in that both used rs76719109 as a reporter SNP and similar although not identical PCR primers. The studies differ in that Xu et al. used an Asian population while this report studied Caucasians. One explanation that would account for the difference in the studies was the presence of a confounding SNP in the Asian population in the genomic primer sequence because much of the AEI in Xu et al. was due to correction for imbalance in gDNA [22], although such a SNP has not yet been reported. We previously reported that the AD-associated SNP rs3851179 was associated with a modest difference in PICALM expression when analyzed relative to cell-type specific mRNAs; the minor rs3851179A allele appeared to be expressed modestly higher than the G allele [2]. A similar difference was not observed here. One possible interpretation of these findings is that rs3851179 or its proxy AD SNP acts in a cell-type specific fashion that was discernible in our analysis that included cell-type specific markers. The current AEI study had smaller sample size because only the samples that were heterozygous for rs76719019 or rs592297 were suitable for analysis. However, this would not be expected to affect the AEI results because they rely upon an intra-individual analysis. We interpret these results overall as suggesting that the AD-associated SNP, or its functional proxy, acts in a cell-type specific fashion to modulate PICALM expression. This cell-type specific action was not detectable in this AEI study of mRNA derived from multiple cell types.

The second major finding was that robust AEI was detected for all PICALM isoforms in one individual, arguing for the existence of a rare functional SNP that strongly modulates total PICALM expression. For this individual, the rs76719109G allele was consistently more abundant than the T allele for each PICALM isoform. We hypothesize that AD40 is unique among the 42 samples in showing AEI because this sample is heterozygous for a causal SNP. If this causal allele is present in the heterozygous state in 1 of 42 people, this SNP has a minor allele frequency of ~1.2%. Although current sequencing studies of the PICALM promoter region have not yet identified candidate functional SNPs for AEI in this sample, these studies are on-going and a SNP that strongly modulates PICALM expression would be expected to be a robust AD risk factor.

The third major finding was that the D18-19 PICALM isoform showed robust AEI. There was a strong skew towards increased expression of the rs76719109T allele. Sequencing identified several candidate SNPs including rs588076, which is 509 bp downstream of exon 17. This SNP was found to be robustly associated with AEI for D18-19 PICALM. There are three possible ways rs588076 could influence D18-19 PICALM splicing efficiency: (i) rs588076 is in high LD with a functional SNP that modulates splicing, (ii) rs588076 and other SNPs influence D18-19 PICALM splicing in a cooperative manner, and/or (iii) rs588076 directly influences splicing. Further studies are necessary to discern among these possibilities.

The biological significance of the rs588076 association with D18-19 PICALM is complex. D18-19 PICALM transcripts account for 1-2% of total PICALM expression [2]. Thus rs588076 is significantly associated with AEI for a PICALM isoform that is relatively rare in brain. Exons 18 and 19 encode a total of 27 amino acids that are part of the carboxyl terminal region required for clathrin binding and endocytosis [25]. Hence, the protein encoded by D18-19 PICALM is likely to have reduced function [25]. However, rs588076 was not associated with AD risk and did not enhance the logistic regression model for the rs3851179 association with AD. This leads us to conclude that the rs588076 and D18-19 PICALM isoform may be too rare in the brain to influence AD pathogenesis. Interestingly, rs588076 was recently associated with the blood pressure response to ^Ca++ channel blocking agents [21]. Since rs588076 is associated only with D18-19 PICALM, we speculate that this isoform may be more abundant in other tissues and rs588076 actions upon D18-19 PICALM mediate this systolic blood pressure phenotype.

In summary, analysis of allelic expression has shown that compelling PICALM AEI was not observed in most brain RNA samples. Strong global AEI was documented in one sample, suggesting the existence of a rare PICALM regulatory SNP. A pattern of AEI was clearly discerned for the D18-19 PICALM isoform and rs588076 was significantly associated with this pattern. Rs588076 was not associated with AD risk although this SNP has been associated with a blood pressure-related phenotype. Allele-dependent expression studies may provide further insights into additional AD-associated polymorphisms.