Functional Characterization of LIPA Variants associated with Coronary Artery Disease

Abstract

OBJECTIVE:

Lysosomal Acid Lipase (LIPA) mediates cholesteryl ester hydrolysis, and patients with rare loss-of-function mutations develop hypercholesterolemia and severe disease. Genome-wide association studies of coronary artery disease (CAD) have identified several tightly linked, common intronic risk variants in LIPA which unexpectedly associate with increased mRNA expression. However, an exonic variant (rs1051338 resulting in T16P) in linkage with intronic variants lies in the signal peptide region and putatively disrupts trafficking. We sought to functionally investigate the net impact of this locus on LIPA and whether rs1051338 could disrupt LIPA processing and function to explain CAD risk.

APPROACH AND RESULTS:

In monocytes isolated from a large cohort of healthy individuals, we demonstrate both exonic and intronic risk variants are associated with increased LIPA enzyme activity coincident with the increased transcript levels. To functionally isolate the impact of rs1051338, we studied several in vitro overexpression systems and consistently observed no differences in LIPA expression, processing, activity, or secretion. Further, we characterized a second common exonic coding variant (rs1051339) which is predicted to alter LIPA signal peptide cleavage similarly to rs1051338, yet is not linked to intronic variants. rs1051339 also does not impact LIPA function in vitro and confers no CAD risk.

CONCLUSIONS:

Our findings show common LIPA exonic variants in the signal peptide are of minimal functional significance and suggest CAD risk is instead associated with increased LIPA function linked to intronic variants. Understanding the mechanisms and cell-specific contexts of LIPA function in the plaque is necessary to understand its association with cardiovascular risk.

Graphical Abstract

INTRODUCTION

Atherosclerotic plaque progression is the underlying cause of the majority of cardiovascular diseases including myocardial infarction and stroke, leading to tremendous morbidity and mortality worldwide. Understanding the pathophysiology of plaque formation and progression remains an important area of investigation both scientifically and clinically, serving as the basis for future therapeutics. One avenue of insight is mechanistic characterization of genetic risk loci identified in genome wide association studies of atherosclerosis. Amongst identified risk loci, LIPA, encoding for the cholesterol-ester and triglyceride hydrolase Lysosomal Acid Lipase, stands out as an intriguing lead [1–3]. In the atherosclerotic macrophage, LIPA contributes substantially to lipid metabolism to modulate fundamental pathogenic hallmarks of atherosclerosis including foam cell formation, inflammatory signaling, and overall plaque progression [4,5]. Indeed, individuals with complete functional loss of LIPA develop Wolman disease, characterized by failure to thrive, hepatosplenomegaly, hyperlipidemia, and typically infantile death[6]. Cholesterol ester storage disease is a related disorder with a milder presentation and is caused by mutations in LIPA leading to subtotal loss of function[6,7]. These findings, along with similar observations in animal models with LIPA-deficiency[8], lead to the conclusion that loss of LIPA causes hypercholesterolemia, plaque macrophage dysfunction, and atherosclerosis.

Therefore, the identification of two tightly linked common variants in LIPA (rs1412444 and rs2246833, ~30% allele frequency) strongly associated with an increased risk for CAD was initially not surprising. However, these two lead variants are intronic and not associated with lipid levels in the population, suggesting a mechanism independent of hypercholesterolemia [1,2]. Additionally, these variants are associated with increased LIPA mRNA levels in circulating monocytes which seemed contrary to the expectation that loss of function would be atherogenic. Interestingly, in the same Linkage Disequilibrium (LD) block, several groups have noted the existence of a coding polymorphism, rs1051338, resulting in T16P with similar prevalence and increased risk for CAD[9–11]. rs1051338 results a shift from a polar (Threonine) to nonpolar (Proline) amino acid in the signal peptide sequence, a region that is characteristically nonpolar and crucial for appropriate LIPA maturation and trafficking[12,13]. Therefore, the possibility that this substitution could adversely impact enzyme trafficking, secretion, and overall expression presented an attractive explanation for the association of these LIPA variants with CAD. A recent study by Morris et al. proposed T16P was the causative variant amounting to a reduction in LIPA expression and trafficking to lysosomes[9].

To understand the effects of a LIPA variant on enzyme function, it is crucial to appropriately understand and evaluate LIPA trafficking, which is typical of enzymes of the lysosomal lumen. Following signal peptide cleavage in the ER, nascent LIPA in the ER/golgi exists as a 56 kDa propeptide which can be secreted or routed to the lysosome via the mannose-6-phosphate receptor trafficking system[12]. In the lysosome, acidity and/or proteases result in propeptide cleavage to yield a ~41 kDa mature, active peptide[12–15]. Therefore, the relative expression of these forms carries a great deal of information regarding subcellular distribution and processing, and both forms of LIPA can be observed in human monocyte derived macrophages[16] (in supplement). Morris et al. had evaluated the impact of rs1051338 on LIPA function using a COS7 transient overexpression model and monocytes/macrophages from human subjects[9], concluding that LIPA T16P interferes with lysosomal trafficking to lead to a net loss of function for the locus. However, analyses in the Morris study were restricted to a single LIPA band of unspecified molecular mass making interpretation of many experiments problematic regardless of which band was analyzed. Further, monocyte/macrophage analyses therein were restricted to n=3–4 individuals with the non-risk versus risk exonic genotypes, which is highly underpowered to assess typically smaller effect sizes associated with single nucleotide polymorphisms (SNPs), especially given the already known effect of this locus on increasing LIPA mRNA [1,2,17].

Therefore, in the present study we sought to evaluate the functional impact of LIPA CAD risk variants on LIPA function in monocytes isolated from a sufficiently large patient cohort (n=114). Further, to isolate the functional impacts of the exonic variant without the confounding influence of intronic variants, we studied its impact in vitro on expression, maturation, and secretion taking into consideration the discussed relevance of LIPA forms and subcellular distribution in multiple systems. Last, we characterize a similar exonic variant (rs1051339) previously speculated to associate with Wolman’s disease[18] that results in the amino acid change G23R with relevance to which SNP’s may be causative in the LIPA CAD risk LD block.

MATERIALS AND METHODS

Data Availability

The authors declare that all supporting data are available within the article and its online supplementary files.

Subject Recruitment and Human Monocyte Isolation

All protocols were approved by the Washington University in St. Louis Institutional Review Board. Healthy, young to middle aged subjects (n=32 male and 82 female) with no known history of CAD/MI were recruited from the St. Louis Area. Whole blood was collected from 32 male and 82 female subjects in tubes containing EDTA (BD Vacutainer). Peripheral Blood Mononuclear Cells (PBMC’s) were isolated via centrifugation with Ficoll-Paque PLUS (GE Healthcare). A portion of PBMC’s was saved for DNA Extraction and SNP genotyping. CD14+ monocytes were isolated using the EasySep Human Monocyte Enrichment Kit (19059, StemCell Technologies).

DNA Extraction and SNP Genotyping

DNA was extracted from PBMC samples using the DNeasy Tissue and Blood DNA Extraction kit (Qiagen #69506). PCR-based SNP genotyping was performed using Taqman Human SNP genotyping assays (ThermoFisher #4351379) for rs1051338 (Assay ID: C___8870360_20), rs1412444 (Assay ID: C___8870364_10), and rs2246833 (Assay ID: C___2734272_10). For analyses of human monocyte mRNA and activity segregated by exonic genotype (Figure 1 C–D), individuals with dissociated exonic/intronic genotypes were analyzed separately (Supp. Fig IA–B) because we observed an independent effect of the intronic variant which obscures the relationship between the exonic variant and mRNA/activity in these individuals.

Figure 1. LIPA SNP variants associate with higher Human Monocyte mRNA and Activity.

A. Intron-Exon Map of LIPA highlighting an exonic variant (rs1051338) and the lead intronic variant (rs1412444) associated with CAD. B. SNP alleles, minor allele frequencies, and associations with CAD in the CardioGRAM study. C. LIPA mRNA expression in human monocytes in individuals with given genotype (n = 38, 37, 8) D. LIPA enzymatic activity in human monocytes in individuals with given genotype (n=44, 46, 9). E. Western Blot of LIPA expression in individuals with given genotypes (n=10,8). F. Quantification of LIPA expression normalized to Ponceau loading control. Data are mean +/− SEM. * Indicates statistical significance using linear regression (C,D) or students unpaired t-test (F).

RT-qPCR and mRNA analyses

Total RNA was extracted using the Ambion Purelink RNA Kit (Invitrogen) and reverse transcribed using the Superscript Vilo cDNA synthesis kit (Invitrogen). qPCR was performed using a ViiA-7 RT-PCR system and SYBR-SelectMaster Mix (Applied Biosystems). Pre-validated primers were used for amplification (Supplementary Table I). Assays were performed in duplicate and normalized to ribosomal protein 36B4 mRNA levels (in vitro studies) or β-actin (human monocyte studies). LIPA mRNA (Transcripts per Kilobase Million) in human monocyte-derived macrophages was analyzed using the publicly available RNA-seq Immunpop Dataset [19] available at http://www.immunpop.com

Lysosomal Acid Lipase Activity Assay

LIPA activity was determined using the fluorogenic substrate 4-methylumbelliferone oleate (4-MUO) as previously described[20]. Briefly, 10 μL protein lysates (~1 ug/uL) or conditioned media were added to 150 uL 200 μM Sodium Acetate Buffer, pH 5.5, and preincubated for 10 minutes in the presence or absence (control vehicle DMSO) of Lalistat-2 (final concentration 10 μM) a kind gift from Dr. Paul Helquist, University of Notre Dame). 4-MUO was reconstituted in DMSO (100 mg/ml) and diluted 1:100 in 4% Triton X-100, and 50 uL added to the assay. Samples were incubated 30 minutes at 37C, reaction halted with 1M Tris pH 8 (100 uL) and fluorescence intensity was measured using a fluorometer (excitation 360 nm/ emission 460 nm). Values were calculated as the difference between each sample incubated with and without Lalistat-2. Activity values were normalized to exact lysate protein concentration.

Regulatory Potential Analyses of LIPA SNP’s

For regulatory potential analyses of intronic LIPA SNP’s, candidate associated SNPs were included according to those in LD (R²>0.8) with lead SNP rs1412444 in both European and South Asian populations (separately) according to data from phase 3 (version 5) of the 1000 Genomes project using the National Cancer Institute LDlink tool available at https://ldlink.nci.nih.gov/. Candidate variants were further parsed by exclusion of variants with low predicted regulatory potential (RegulomeDb score <4). A description of the meanings of scores (scaled 1–7 where 1a indicates highest potential for regulatory influence) can be found at http://www.regulomedb.org/help#score.

SignalP Analyses

All analyses were performed using the SignalP v4.1 tool available at http://www.cbs.dtu.dk/services/SignalP/. Signal peptide “S-Score” values indicate which regions are likely to be included as part of the signal peptide, and Y-score values indicating the most probable cleavage sites based on amino acid properties were obtained through using the long output format with default D cutoffs and the no transmembrane region settings.

Transient Transfections and Lentiviral Transductions

Transient transfection vectors containing LIPA variants with a C-terminus FLAG tag were transfected into HEK293T cells with Lipofectamine 3000 (ThermoFisher L3000008). After 12h, media was changed to normal media and analyses performed 48h post transfection.

Lentiviral vectors expressing LIPA or empty vector control were packaged in HEK-293T cells and target 293T, 3T3, and J774 cells were transduced. Cells were selected in puromycin (4 μg/mL) until all control nontransduced cells were dead and cells were grown in normal media for >48h prior to experiments. Two pooled populations for each variant were matched for mRNA expression and equivalent comparison. Data for stable transduction experiments are expressed versus each respective LIPA 16T control. All protein lysates used for western blotting and activity assays, conditioned media, and mRNA was collected from the same wells.

LIPA Stability and Turnover Assays

Cells were transfected with LIPA T16 or P16 variants. 24h post-transfection, untreated cells were collected. For 6h, cells were either left untreated or administered cycloheximide (100 ug/mL), bortezimib (10 nM), or both as indicated.

Cholesterol Efflux and Ac-LDL loading

Cholesterol efflux was assessed using the BODIPY-cholesterol approach as previously described[21–23]. Briefly, J774 cells were loaded with BODIPY-Cholesterol labeled Ac-LDL (final concentration 50 μg/mL, Kalen Biomedical 770201–7) in full media for 12h and equilibrated in DMEM + 0.2% BSA for four hours. Cells were further incubated for 6h in control (0.2% BSA) or efflux (5% FBS + 0.2% BSA) media over 6h. Cells were rinsed in PBS, detached, fixed, and analyzed using flow cytometry (BD LSR-II and FlowJo) measuring BODIPY-Cholesterol Signal (MFI) in at least 10,000 cells.

Free and esterified cholesterol were measured using a commercially available kit (Thermo Fisher A12216) and standard protocol [24] with or without acetylated low-density lipoprotein loading (50 μg/mL, Kalen Biomedical 770201–7) for 12h.

Western Blotting

Cells were lysed in RIPA buffer and protein concentration assessed using BCA Protein Assay Kit (Pierce, Cat# 23225). Equal amounts of protein were separated by SDS-PAGE and transferred onto nitrocellulose membranes. LIPA Protein expression was detected using the appropriate primary antibody: Human LIPA (OriGene, Cat# TA309730, 1:1,000) and corresponding secondary antibody (1:4,000). Signals were visualized by SuperSignal™ West Pico PLUS Chemiluminescent Substrate (Thermo Fisher Scientific, Cat# 34577) or LiCOR IRdye secondary fluorescent antibodies to allow detect of FLAG/LIPA within the same blot, and imaged using either a Biorad Chemidoc MP, LiCor Odyssey Classic system, or via film exposure. Band intensities were quantified relative to Ponceau-S total protein staining used as a loading control.

Immunofluorescence Analyses

Cells were fixed with 4% paraformaldehyde, blocked and permeabilized (1% BSA, 0.2 % milk powder, 0.3% Triton X-100 in TBS; pH 7.4), and incubated with antibodies sequentially. Specificity of staining was tested in control experiments by omitting primary antibodies. The following primary antibodies were used in 1:250 dilutions: LIPA (Origene Cat# TA309730) and Lamp-1 (Santa Cruz Cat# sc-20011). Species-specific fluorescent secondary antibodies were obtained from Invitrogen/Life Technologies (1:250). A Zeiss LSM-700 confocal microscope was used for image acquisition and images quantified using ZEN microscope software (Carl Zeiss AG). At least 50 cells per group spanning 3 separately stained coverslips were imaged and analyzed. For each analyzed cell, LIPA foci were analyzed by a blinded observer for colocalization to lysosomes (Lamp1+) and scored as the % of Lamp1+ LIPA foci for each cell.

Lysosomal Inhibitor Treatment

Cells were transiently transfected with the LIPA 16T variant as described. 36h post transfection, cells were treated with the lysosomal inhibitors Bafilomycin (50 nM) or Chloroquine (50 uM) for 12h to demonstrate inhibition of LIPA propeptide cleavage in the lysosome.

Statistics

Results are expressed as mean +/− S.E.M. Data were examined for non-normality using the Shapiro-Wilk test. Comparisons between two groups were performed using two-tailed Student’s t test or Wilcoxon’s signed rank test for parametric and non-parametrically distributed data, respectively. Comparisons amongst three genotypes were performed with linear regression. P values < 0.05 were considered to indicate statistically significant differences.

RESULTS

Intronic and Exonic LIPA risk alleles are associated with increased monocyte LIPA mRNA expression and Enzyme Activity

Genome wide association studies for coronary artery disease have reproducibly identified LIPA SNP rs1412444 and rs2246833 as lead SNPs out of many in moderate to tight linkage disequilibrium (LD) at the LIPA risk locus. However, nearly all of these risk variants are intronic and perplexingly associated with higher LIPA mRNA expression, contrary to what would be expected based on the many known pathologies associated with loss of LIPA function [6,8]. Using data from the CardioGRAMplusC4D genome-wide association studies (GWAS) of CAD [11], we and others had identified a sole exonic variant in this LD block, rs1051338, as a candidate for explaining the pathogenicity of this locus. rs1051338 is in fairly strong LD with rs1412444 (R²=0.859 in all individuals), and here we focus on rs1412444 as a representative surrogate for many tightly linked intronic variants including rs2246833. Our analyses reveal rs1051338 is associated with CAD to a degree similar to rs1412444 in pooled data from the CardioGRAM metaanalysis. Genomic locations, risk alleles, minor allele frequencies, and association with CAD presented in Figure 1 A–B.

We sought to study whether rs1051338 could negatively impact LIPA function in humans, potentially explaining CAD risk. To this end, we isolated and studied CD14+ classical monocytes as a highly disease-relevant cell type from a large cohort of healthy young to middle-aged individuals with no known history of cardiovascular disease. First, rs1051338 is dose-dependently associated with higher human monocyte LIPA mRNA expression (Figure 1C), consistent with the known cis-expression quantitative trait loci (eQTL) effect at this locus[1,2,17]. Contrary to any possibility of overt loss of function, we demonstrate for the first time that the risk exonic genotype also dose-dependently associates with a gain of function in monocyte LIPA enzyme activity (Figure 1D). Complementary to this data, individuals with homozygous risk versus nonrisk alleles have significantly increased LIPA expression (Fig 1E., quantified in 1F) in proportion to the increase in activity. This suggests that any subtle effects of LIPA risk SNPs in this locus on overall protein translation or degradation are largely trumped by the eQTL effect, which directly translates to a straightforward increase in enzyme activity.

Dissecting which variants, in isolation or in combination, may be responsible for these factors is a challenge in any functional genetics study. In addition, it has been proposed that the observed mRNA increase could be compensating for a loss of enzyme function caused by rs1051338. However, our data constitute strong evidence that rs1412444 or closely associated intronic variants are instead responsible for the eQTL effect. Amongst all individuals homozygous for the major (nonrisk) allele for rs1051338 (the exonic SNP), those with a dissociated (heterozygous) genotype for rs1412444 (the intronic SNP) displayed significantly higher LIPA mRNA and activity versus the major homozygous individuals for rs1412444 (Supp. Fig I A,B). RegulomeDb, a tool that identifies DNA features and regulatory element alterations by SNPs, suggests that multiple intronic SNPs in high LD with the GWAS lead SNP rs1412444 are likely to be regulatory (Supp. Fig. I C), providing premise for future functional genomic interrogation of the causal SNPs underlying the GWAS and eQTL association[25]. Last, LIPA expression is known to be induced upon monocyte-to-macrophage differentiation[16], potentially lessening the relative impact of this locus on increasing expression in macrophages. However, using publicly available RNA-seq data, we observe the intronic variant rs1412444 is associated with similarly increased LIPA mRNA levels in human monocyte-derived macrophages[19].

Exonic LIPA risk SNP rs1051338 does not alter LIPA enzyme trafficking or function in isolation.

Out of many SNP’s in the LIPA risk locus for CAD, rs1051338 stood out as a strong candidate for explaining pathogenicity. rs1051338 causes a coding amino acid change from threonine to proline at amino-acid 16 “T16P”. Several aspects of this change pointed to a possibility for disruption of enzyme function or trafficking which may not be reflected in our whole monocyte enzyme activity analyses. First, T16P lies in the signal peptide domain of LIPA. This region is crucial for proper sorting and trafficking to LIPA’s main subcellular destinations, the secretory pathway and the lysosome. Second, T16P results in a shift from a polar amino acid (Thr) to a nonpolar one (Pro) in a region which is characteristically nonpolar.

To experimentally assess the isolated impact of rs1051338/T16P on LIPA maturation and trafficking, we studied a transfection model with overexpression of either LIPA variant with a C-Terminus FLAG tag. We observed no absolute or relative differences in proLIPA (ER/Golgi, ~56 kDa), mature LIPA (Lysosomal, ~41 kDA), or FLAG expression (Fig 2A, quantified in Fig. 2B), in cells well matched for degree of mRNA expression (Supp Fig II-A). Whole cell LIPA activity (Fig 2C) and LIPA secretion as assessed by LIPA activity in supernatant media (Fig 2D) were also unaffected. We further confirmed the dependence of LIPA propeptide cleavage on lysosomal hydrolytic activity by treating cells with the lysosomal inhibitors bafilomycin and chloroquine, demonstrating impaired LIPA propeptide cleavage with either treatment (Supp Fig II-B).

Figure 2. rs1051338 does not impact LIPA in transfected cells.

A. Western Blot of HEK-293T cells transduced with either FLAG-tagged LIPA variant. Each lane represents a transfection replicate with a single representative experiment B. Quantified Densitometry of LIPA expression normalized to loading control C. LIPA activity in whole-cell lysates D. LIPA secretion measured in conditioned media. E. Pulse-chase analyses of LIPA stability in exonic variants. All parameters (expression, activity, secretion, mRNA) are measured within the same well, and are representative of 3 or more independent experiments performed on separate occasions. Data are mean +/− SEM.

If the risk variant were to have reduced stability or mistrafficking, pulse-chase assays could reveal such a defect. We assessed the stability, turnover, and degradation of LIPA variants using a cycloheximide pulse-chase assay (Fig 2E). We also applied a proteasome inhibitor, bortezimib, to assess whether any mistrafficked LIPA would be subject to proteasomal degradation. At 6 hours, cycloheximide treatment nearly completely abolished propeptide expression similarly in either variant as expected given this is the nascent form that is rapidly trafficked to the lysosome. Mature LIPA expression was modestly reduced with cycloheximide but clearly had a longer half-life than the propeptide and/or was preserved due to existing propeptide maturation that balanced turnover. Overall, T16P did not result in lower LIPA stability. Additionally, the addition of bortezimib did not preserve LIPA stability in the absence or presence of cycloheximide in either variant, indicating LIPA is not subject to cytosolic mistrafficking or proteasomal degradation (Figure 2E).

A potential limitation of transfection models is the high degree of overexpression which could saturate pathways involved in LIPA trafficking and mask modest impairments caused by variants like T16P. Typical transfection experiments showed a somewhat favored relative accumulation of pro-LIPA versus empty vector transfected control (Supp Fig II-C). Despite obvious increases in both pro-LIPA and LIPA in overexpression models, the C terminus FLAG tag only marked pro-LIPA (Supp Fig I-A, II-C), likely due to instability of FLAG tags in the lysosome.

To study T16P in a system using a more modest degree of overexpression and confirm findings suggesting minimal function impacts, we generated HEK293T and NIH-3T3 cell lines stably overexpressing either variant in two cell types. In this system, fractionally more overexpressed LIPA was trafficked to the lysosome, but still no differences with T16P were observed in either pro-LIPA or LIPA expression in either cell type (Fig. 3A, quantified in Fig 3B). No difference in cellular LIPA activity (Fig . 3C) or secretion (Fig. 3D was observed, and cell lines were well matched for degree of overexpression (mRNA, Supp Fig III-A). Last, immunofluorescent colocalization analyses revealed the vast majority of LIPA foci to be colocalized to lysosomes (Lamp1+) with either variant (Supp Fig III-B,C)

Figure 3. rs1051338 does not impact LIPA in stable overexpression systems.

A. Western Blot analyses of NIH 3T3 and HEK-239T cells transduced with either LIPA variant. “-” indicates empty vector transduction control and T₁ P₁ T₂ P₂ represent paired sets of cell lines transduced with either variant and matched for mRNA expression. All data is expressed relative to respective T16 variant control. B. Quantified Densitometry of LIPA expression C. LIPA activity in transduced cell lines D. LIPA secretion measured in conditioned media from transduced cell lines. All parameters (expression, activity, secretion, mRNA) within each replicate are measured within the same well and were assessed on two separate occasions. Data are mean +/− SEM.

Exonic SNP rs1051339 disrupts the LIPA signal peptide but is not functionally significant

A key challenge in identifying causative variants within genomic risk loci is the frequently strong linkage disequilibrium amongst variants. As an independent test of whether a signal peptide variant like T16P could disrupt LIPA function or associate with CAD independent of linkage with intronic variants, we identified the nearby SNP rs1051339 as a functionally similar variant meeting these criteria. This SNP is exonic and also causes a functional shift from glycine (nonpolar) to arginine (charged, polar) at amino acid 23 (G23R), and several aspects of this SNP are informative in relation to the LIPA risk locus for CAD and evaluating causality of rs1051338/T16P. Further, rs1051339 has been found in compound homozygosity with other detrimental mutations in Wolman’s patients (annotated therein as “G5R”) and it has been suggested that G23R results in major defects in LIPA secretion[18].

We first used SignalP, an in silico computational tool to contrast the functional impacts of T16P and G23R. The shift in S-score, an index of likelihood of an amino acid residue’s inclusion in the signal peptide, was nearly identical between these two variants (Fig 4A). Similarly, the Y-score, an index taking into account local amino acid properties to predict cleavage site, suggested T16P and G23R both shift the cleavage site identically (Fig 4A). We also sought to test the functional impacts of G23R experimentally using a similar LIPA-FLAG overexpression system. Like T16P, G23R had no impact on LIPA expression or trafficking as assessed by western blot expression of proLIPA (ER/golgi localized), FLAG, and mature (lysosomal) LIPA (Fig 4B, quantified in 4C). G23R also did not impact enzyme activity (Fig 4D.) or secretion (Fig 4E), in cells that were well matched for degree of expression (Supp Fig IV-A). We also analyzed the CardioGRAM GWAS study for CAD risk and found that rs1051339 is common (maf=0.116), yet not in linkage disequilibrium with rs1051338 or other intronic SNP’s (rs1412444) at the LIPA risk locus (LD R² <0.09). Most importantly, despite these similarities with T16P, G23R carries no association with CAD (Fig 4F.), strongly implying the mild signal peptide disruption associated with either variant is of minimal functional significance, and that CAD risk associated with T16P is merely as a result of its association with the intronic LIPA variants.

Figure 4. LIPA SNP rs1051339 does not affect expression, trafficking, activity, or secretion, and is not associated with CAD.

A. SignalP v. 4.1 in silico analyses of LIPA signal peptide property alterations with G23R and T16P. S (Signal Peptide) Score indicates the probability of a region’s inclusion in the signal peptide. Y score takes into account local amino acid properties to predict the signal peptide cleavage site (C.S.) and the predicted C.S. shift with G23R and T16P is shown. Note that the 23R/16P haplotype does not occur in humans. B. Western Blot of HEK-293T cells transfected with either FLAG-tagged LIPA variant. Each lane represents an independent transfection replicate C. Quantified Densitometry of LIPA expression D. LIPA activity in whole-cell lysates E. LIPA secretion measured in conditioned media. All parameters (expression, activity, secretion, mRNA) are measured within the same well and represent at least 2 independent experiments. Data are mean +/− SEM. F. rs1051338 and rs1051339 coding amino acid changes, predicted signal peptide cleavage sites, minor allele frequencies, and association with intronic SNP’s and CAD.

With the conclusion that the LIPA CAD risk locus is associated with increased expression, we sought to conduct a preliminary evaluation of whether an increase in LIPA expression could modulate cholesterol metabolism. An important downstream phenotype of LIPA loss of function is reduced cholesterol efflux which has been demonstrated in both fibroblasts from cholesterol ester storage disease fibroblasts, mouse peritoneal macrophages, and human induced pluripotent stem-cell derived macrophages [16,26,27]. One way through which this occurs is reduced expression of the cholesterol efflux transporter ABCA1, which could be rescued to baseline levels with recombinant enzyme addition. No differences were observed in monocyte ABCA1 mRNA expression across individuals segregated by LIPA rs1412444 genotype (Supp Fig V-A). To evaluate whether modest increases in LIPA expression (as is seen in the CAD risk haplotype) could modulate cholesterol metabolism, we overexpressed LIPA in J774 cells (Supp Fig V-B). Small increases in LIPA expression did not increase cholesterol efflux (Supp Fig V-C), or affect levels of free or esterified cholesterol with or without acetylated-LDL loading state (Supp Fig V-D).

DISCUSSION

The finding that LIPA was a risk locus for Coronary Artery Disease highlighted the need to better understand the functional impacts of risk variants on LIPA expression and function, and in turn, better understand the function of LIPA in CAD. In the present study, we demonstrate the LIPA risk haplotype associates with both increased LIPA mRNA and enzyme activity in primary human CD14+ monocytes. We found rs1051338, an exonic variant in tight linkage disequilibrium with lead intronic SNP’s, to be of minimal functional significance. Similarly, rs1051339, a nearby exonic SNP also affecting the signal peptide region, did not affect LIPA function. Overall, our results provide strong support for a model wherein the LIPA risk haplotype is associated with a gain of function in LIPA likely driven by one or more intronic SNP’s.

In our large human cohort, we replicated key findings demonstrating the LIPA risk haplotype associated with increased mRNA, and crucially extend these findings to demonstrate for the first time that this translates to increased LIPA enzyme activity. Such functional cis-eQTL effects are common in many studies associating GWAS loci and gene expression with disease. However, we discovered that an exonic variant, rs1051338, was in tight linkage disequilibrium with the intronic variants and provided a plausible candidate mechanism for a loss of function that may not be reflected in whole cell enzyme activity.

rs1051338 causes the amino acid substitution T16P in the signal peptide region of the LIPA protein. T16P is predicted to be disruptive due to the polar to nonpolar amino acid shift in a region that is characteristically nonpolar. This could affect several aspects of signal peptide function such as the cleavage site with implications for downstream trafficking to LIPA’s two main subcellular destinations: the secretory pathway and lysosome. Following signal peptide cleavage, proLIPA is trafficked to the lysosome, where the propeptide is cleaved to produce a mature, active enzyme[12,28]. This type of trafficking is nearly universal to enzymes of the lysosomal lumen[29], and specific evidence for these forms of LIPA includes 35S-Met pulse chase labeling and propeptide cleavage site mutagenesis studies[12,14,15]. Similarly, we provide support for this concept by demonstrating inhibition of LIPA propeptide cleavage with treatment of the lysosomal acidity inhibitors bafilomycin and chloroquine. We also noted that C-terminal FLAG tags selectively marked pro-LIPA, likely due to cleavage of these exposed residues on the mature peptide in the acidic environment of the lysosome.

To evaluate the functional impacts of rs1051338/T16P on LIPA enzyme trafficking in the absence of potentially confounding intronic variants, we studied allele-specific overexpression in vitro and examined the overall impacts on LIPA activity and trafficking and found that T16P had no detrimental impact on overall LIPA expression or activity. As evidenced by the identical expression of the LIPA 41 kDa band in isolation or considered relative to the proprotein, T16P does not adversely impact enzyme maturation or trafficking to the lysosome, nor did it impact secretion. While it is difficult to entirely exclude a subtle impact of T16P, any effect clearly has minimal relevance in relation to the overall increased LIPA activity we observed in humans with the risk haplotype.

As an independent line of evidence for whether T16P could be the causative SNP in the LIPA risk locus, we characterized the functional impacts of rs1051339, a common SNP also lying in the signal peptide that results in the amino acid substitution G23R. This SNP is predicted to disrupt signal peptide function similarly to T16P, yet is not in linkage disequilibrium with intronic variants, allowing evaluation of whether this type of disruption can impact enzyme function or cause CAD in isolation. Additionally, a previous study of Wolman’s patients (LIPA loss of function) observed G23R (annotated therein as “G5R”) in compound homozygosity with another mutation and demonstrated that G23R alone profoundly disrupted LIPA secretion when expressed in insect cells[18]. Overall, we observed no differences in LIPA expression, activity, trafficking, or secretion with G23R studied in vitro, suggesting this SNP is also of minimal functional relevance. This discrepancy from findings in insect cells may be due to inappropriate glycosylation and oligosaccharide processing pathways in insect cells[30] and emphasizes the importance of studying human, or at least mammalian cells in this regard. The other identified homozygous Wolman’s variant (G87V, annotated therein as G60V ) in the Zschenker study in compound homozygosity in conjunction with G23R has been observed to be sufficient to cause Wolman’s elsewhere[31]. Most importantly, rs1051339 carries no association with CAD, implying in turn that rs1051338 association with CAD is merely a result of linkage with intronic variants.

Our conclusions contrast with another recent report evaluating the functional impacts of rs1051338 on LIPA function which suggested that T16P is a significant loss of function mutation and acts by mistrafficking LIPA away from the lysosome[9]. In the present study, we highlight the need to consider expression of pro-LIPA and mature (lysosomal) LIPA as crucial indices of LIPA subcellular distribution and trafficking. The identity of the single western blot bands analyzed throughout the study of Morris et al is unspecified and problematic for their interpretations drawn. If mature/lysosomal 41 kDa was the focus of analysis in their study, any differences with T16P observed in their enriched lysosomal fraction should be of similar magnitude in whole cell lysates. However, this was not apparent and may therefore be an artifact associated with their crude lysosomal enrichment protocol. Alternatively, if the 56 kDa proprotein band was analyzed, (likely the dominant form in a transfection-overexpression system) the conclusions regarding lysosomal trafficking are hard to justify. Here, we demonstrate C-terminal FLAG-tagged LIPA nearly exclusively marks pro-LIPA in overexpression systems. As LIPA and other enzymes of the lysosomal lumen are exquisitely folded and glycosylated to withstand lysosomal acidity and proteases, we interpreted this as meaning that FLAG tags were unstable in the lysosome. Therefore, the many FLAG blots used in the Morris study are also likely to be inappropriate for drawing conclusions regarding LIPA lysosomal trafficking. Finally, the analyses of human monocytes/macrophages (n=3–4) in the Morris study were underpowered to observe the significant increase in LIPA mRNA and activity we observed in a much larger cohort.

A key conclusion of our study is the exclusion of the possibility that rs1051338 causes a loss of function relevant to overall LIPA expression or activity, especially in relation to the obvious increase in mRNA and activity driven by tightly associated intronic variants. One hypothesis regarding increased mRNA expression at this locus was that loss of function caused by rs1051338 caused compensatory LIPA mRNA upregulation. Our data in individuals with dissociated intronic and exonic genotypes suggests this is not the case and rs1412444 or other tightly associated intronic variants are sufficient to explain increase mRNA and activity. The association of intronic variants with eQTL effects amounting to disease risk is not surprising and is quite common across many discovered loci for complex diseases. Finally, evidence regarding the relationship between loss of LIPA function and CAD risk has been gleaned from studies of the Wolman’s disease LIPA Exon 8 Splice Junction Mutation (E8SJM). Whereas homozygosity leads to a near total loss of function and overt pathology, heterozygosity (resulting in ~50% loss of function [32–34]) carries no association with serum lipids, CAD, or myocardial infarction risk in large GWAS cohorts[35]. This suggests that a small magnitude loss of function in LIPA hypothetically caused by T16P, which we did not observe regardless, would still be very unlikely to confer disease risk.

Attention should now be focused on evaluating the other possibilities through which the LIPA locus could be causative in mediating CAD risk, including increased expression. While LIPA deficiency markedly disrupts cholesterol ester hydrolysis and cholesterol efflux[16,27], we observed no clear changes with mild overexpression in J774 cells. Thorough investigation using in vivo models and taking into account the unique context of the atherosclerotic plaque will be required to fully evaluate whether increased LIPA expression observed in the CAD risk haplotype could be causal. One possibility is that secreted LIPA (along with other lysosomal enzymes) can induce LDL modifications such as aggregation and oxidation [5]. Increases in extracellular LIPA are observed in atherosclerotic lesions, and some have proposed that a very significant portion of aggregated LDL hydrolysis occurs extracellularly in atherosclerotic macrophages[36,37]. A second possibility is that increased LIPA function generates free cholesterol exceeding the capacity of systems for lysosomal efflux and reverse cholesterol transport. Excess free cholesterol in the lysosome is known to increase lysosomal pH and lead to a more general lysosomal dysfunction[5,20] amongst many other cytotoxic effects[38]. Overall, further work is required to characterize whether the increased expression or other yet to be appreciated possibilities could causally explain the CAD risk associated with this locus.

HIGHLIGHTS.

• LIPA risk SNP’s previously identified as conferring risk for Coronary Artery Disease are associated with increased LIPA enzyme expression and activity

• The sole exonic SNP in this linkage disequilibrum block (rs1051338 resulting in T16P) does not impair LIPA expression, activity, lysosomal trafficking, or secretion.

• A functionally similar exonic SNP (rs1051339 resulting in G23R) also does not impact LIPA function, is not in linkage with risk variants, and does not associate with Coronary Artery Disease

• Intronic variants likely drive increased LIPA expression and highlight the need to explore how increased LIPA may be pathogenic in this context.

Abbreviations and Acronyms

CAD

Coronary Artery Disease

LIPA

Lysosomal Acid Lipase

GWAS

Genome-Wide Association Studies

SNP

Single Nucleotide Polymorphism

eQTL

Expression Quantitative Trait Loci

PBMC

Peripheral Blood Mononuclear Cells