Introduction

Atherosclerotic cardiovascular disease (ASCVD) is the leading cause of morbidity and mortality in the developed world [1]. The development of ASCVD is strongly influenced by environmental factors including smoking, sedentary lifestyle, obesity, and alcohol consumption, and also by genetic factors, which contribute to the medical risk factors associated with ASCVD development including hypertension, diabetes, elevated low-density lipoprotein cholesterol (LDL-C) and reduced high-density lipoprotein cholesterol (HDL-C). LDL-C is intimately involved in each step in the cascade to cardiovascular disease development: macrophage engorgement and foam cell formation, arterial wall injury, inflammation and smooth muscle cell proliferation, and reduced LDL-C concentrations are associated with protection from ASCVD [2]. In spite of the aggressive use of LDL-lowering medications such as statins, many patients cannot achieve the therapeutic LDL-C levels recommended by clinical guidelines and are thus at risk of developing ASCVD [3]. There is therefore a need for novel therapies to reduce plasma LDL-C and prevent the development of ASCVD and its sequela of myocardial infarction (MI).

Initially, much of the insights into lipid metabolism came from the study of rare Mendelian disorders of high and low LDL-C (Table 1) [4]; however, much of the variation in plasma lipid levels within a population cannot be explained by the rare mutations associated with Mendelian lipid disorders. Genome-wide association studies (GWAS), which involve genotyping hundreds of thousands of single nucleotide polymorphisms (SNPs) in thousands of individuals phenotyped for lipid traits, have evolved as an unbiased tool to identify common variants that contribute in a causal manner to lipid variations in the population [5, 6, 7•, 8•]. Importantly, GWAS aims to identify many genes and loci of generally small effect size that contribute to population variations in a trait, instead of single gene defects that have dramatic effects on phenotype, like those seen in Mendelian disorders.

Table 1 Summary of Mendelian disorders of elevated and reduced LDL-C
Full size table

GWAS for lipid traits have already identified common variants in genes known to cause Mendelian lipid disorders, including LDLR, PCSK9, APOB, LDLRAP, and IBCG5/8 for LDL-C. GWAS have also identified genes that are drug targets for LDL-lowering therapies, including NPC1L1, the molecular target of ezetimibe, and HMGCR, the molecular target of statin drugs [8•]. GWAS have also identified a number of novel genes and loci that contribute to population variability in LDL-C. The locus in the human genome with the strongest association to LDL-C is a novel 1p13 locus, which harbors the genes SORT1, PSRC1, and CELSR2. Importantly, the 1p13 locus is associated not only with the intermediate phenotype of LDL-C, but also with the hard clinical endpoint of cardiovascular disease. This review focuses on the initial identification of the 1p13 locus as a genome-wide significant and important determinant of LDL-C levels and cardiovascular disease risk in the population, early mechanistic studies to determine how the 1p13 locus affects LDL-C and cardiovascular disease risk, and the current controversy surrounding the disparate mechanistic studies that have emerged.

Initial Identification of the SORT1 Locus

Before its association with LDL-C levels in humans, the 1p13 locus was identified in 2007 in the Welcome Trust Case Consortium (WTCCC) as an important risk factor for coronary artery disease Table 2 [9]. The association between the 1p13 locus and cardiovascular disease was replicated in other larger studies, including the Global Lipids Genetics Consortium, which included more than 100,000 individuals of European decent [8•], and most recently in the CARDIoGRAM study, which included more than 140,000 individuals of European decent [10•] . The association between the 1p13 locus and cardiovascular disease has been replicated in non-European populations, as well [8•, 11, 12]. The 1p13 locus has also been linked to MI, and homozygosity for the minor alleles at 1p13 is associated with a 40% reduction in MI risk [9, 13•].

Shortly after the 1p13 locus was found to be associated with cardiovascular disease, a large-scale GWAS with 9,000 individuals of European decent and a replication study in 18,000 individuals of European decent identified the 1p13 locus as a genome-wide significant determinant of LDL-C levels in humans [5]. Importantly, this study demonstrated that homozygosity for the minor allele is associated with a 16 mg/dL lower systemic LDL-C as compared to homozygotes for the major allele, which is an unusually large effect size for a common variant. These results have been confirmed in other European and non-European populations [14, 15•, 16]. All subfractions of LDL were reduced in individuals of the minor haplotype; however, the most dramatic reductions occurred in the small, dense LDL fractions, which are derived from the lipolysis of triglyceride-rich VLDL1 and thought to be the most atherogenic [17•].

The 1p13 locus has been replicated in larger GWAS for lipid traits in European populations, including a study with 40,000 individuals, and the largest GWAS for lipid traits, the Global Lipids Genetics Consortium (GLGC), which included over 100,000 individuals of European descent. The GLGC study named the 1p13 locus as the most strongly associated with LDL-C in European populations, with a P value of 1 × 10−170 (95% CI, 5.24–6.06 per allele copy) [8•]. Importantly, the 1p13 locus has been replicated as an important determinant of LDL-C in non-European populations as well [8•, 16, 18, 19].

Dissecting the Locus and the Association Signal

The challenge of the 1p13 locus is that the SNPs most strongly associated with LDL-C and ASCVD risk lie in a non-coding region between two genes, PSRC1 and CELSR2, and lie extremely close to the SORT1 gene, making it difficult to name the gene responsible for the association. Expression quantitative trait loci (eQTL) data suggest that homozygosity for the minor allele is associated with elevated expression of SORT1, PSRC1, and CELSR2 in human liver, with SORT1 and PSRC1 expression most strongly influenced by genotype with 10- to 12-fold elevations in expression with homozygosity for the minor allele. Interestingly, the expression changes seen with homozygosity for the minor allele were not present in subcutaneous or omental adipose, suggesting a liver-specific effect of the 1p13 locus on LDL-C levels and ASCVD risk [17•].

Further complicating the problem is that none of these genes have an obvious link to LDL metabolism: PSRC1 encodes the protein proline-serine rich coiled-coil domain-1, which is a downstream target of p53 and is believed to be involved in cell cycle progression and formation of the mitotic spindle. CELSR2 encodes the protein cadherin EGF LAG seven-pass G-type receptor 2, which is involved in neuronal cell migration and adhesion. Neither PSRC1 nor CELSR2 are well-characterized proteins. This contrasts with SORT1, which encodes the protein sortilin, which is fairly well characterized as a cell surface receptor as well as an intracellular trafficking receptor.

Sortilin is a multi-ligand sorting receptor originally identified by receptor-associated protein (RAP) affinity chromatography [20]. Sortilin is one of five members of the mammalian VPS10 domain family and harbors an N-terminal propeptide with a furin cleavage site, an extracellular VPS10 domain for ligand binding, a single transmembrane domain, and a C-terminal cytoplasmic tail with strong homology to the cation-independent mannose-6-phosphate receptor (CI-M6PR), bearing two endolysosomal sorting motifs: a dileucine-based sorting motif and a tyrosine-based sorting motif [20].

Sortilin localizes primarily to the Golgi apparatus and has been shown to bind a variety of ligands there and traffic them to the lysosome (Fig. 1). Sortilin also localizes to the plasma membrane where it can act as a signalling receptor for pro-neurotrophins [21], function as an uptake receptor [22•], or be cleaved by the metallopeptidase TNF alpha-converting enzyme (TACE) in a mitogen-activated protein kinase (MAPK)/protein kinase C (PKC)-dependent manner to generate a soluble fragment comprising the sortilin luminal domain [23]. It has recently been reported that sortilin is cleaved intracellularly and at the cell surface by the protease ADAM10 to generate the soluble luminal domain [24]. Sortilin binds to a number of ligands involved in lipid metabolism including lipoprotein lipase [25], apolipoprotein E [26], and apolipoprotein AV [27], which are involved in triglyceride metabolism. It is unclear, however, how these interactions might explain the strong association of the SORT1 locus with LDL-C.

Fig. 1
figure 1

Sortilin, the protein product of the SORT1 gene, is synthesized as a proprotein and is cleaved by furin in the trans-Golgi network to produce the mature sortilin protein. Very low-density lipoprotein (VLDL) synthesis begins in the rough endoplasmic reticulum (RER, pink) with synthesis and lipidation of its primary protein component, apolipoprotein B (apoB, yellow circle). apoB is further lipidated and modified in the Golgi apparatus (blue) to form a mature VLDL particle. VLDL can either undergo pre-secretory degradation or can be secreted. VLDL is lipolyzed in the periphery to generate LDL, which is taken up by liver primarily by the LDL receptor. a Work by Linsel-Nitschke et al. [15•] suggests that Sort1 overexpression reduces LDL-C by enhancing uptake of low-density lipoprotein (LDL) particles in vitro. b Work by Musunuru et al. [17•] suggests that Sort1 expression reduces LDL-C by reducing production/secretion of the LDL precursor VLDL cholesterol. c Work by Kjolby et al. [28•] suggests that Sort1 overexpression increases LDL cholesterol by facilitating secretion of the LDL precursor VLDL

Full size image

Identification of SORT1 as the Causal Gene and Early Mechanistic Studies

One year following identification of the 1p13 locus as a genetic determinant of LDL-C levels and ASCVD risk, the first mechanistic study was published. Linsel-Nitschke et al. [15•] replicated the association of the 1p13 locus with LDL-C and CAD risk and further demonstrated that homozygosity for the minor allele was associated with elevated expression of Sort1 in peripheral blood with no change in PSRC1 or CELSR2 expression. The directionality was consistent with the eQTL data that associated minor allele homozygosity with elevated SORT1 expression in liver and supported a role for SORT1 as the causal gene at the 1p13 locus responsible for the LDL association.

Table 2 Genome-wide association of the 1p13 SORT1 locus with LDL-C and cardiovascular disease
Full size table

To determine whether SORT1 overexpression affects LDL-C levels by enhancing LDL catabolism, Linsel-Nitschke et al. [15•] transfected HEK293 cells with a cDNA encoding human SORT1 and assayed for LDL uptake. They incubated cells with 125I-LDL for 4 h in the presence or absence of competitive inhibitors followed by counting intracellular radioactivity. Consistent with the lower plasma LDL-C seen in humans with elevated Sort1 expression, SORT1 overexpression in vitro caused a small but statistically significant increase in LDL uptake [15•]. This study supported SORT1 as the causal gene at the 1p13 locus and suggested a mechanism by which elevated SORT1 expression in human liver could reduce LDL-C and ASCVD risk, namely by increasing LDL uptake in liver. However, this work fell short of identifying the mechanism by which SORT1 expression could enhance LDL uptake.

Identification of the Causal SNP and the Mechanism of Sort1 Overexpression

Homozygosity for the minor allele at the 1p13 locus is associated with a 10- to 12-fold increase in SORT1 and PSRC1 expression with a significant eQTL in human liver; however, when the GWAS was published, the mechanism for the profound effect on gene expression was unknown. A study published in 2010 by Musunuru et al. [17•] identified the genetic mechanism of the expression changes. Musunuru et al. [17•] looked at expression of all nearby genes in the 1p13 locus (SARS, CELSR2, PSRC1, MYBPHL, SORT1, PSMA5, and SYPL2) in liver, subcutaneous adipose and omental adipose in individuals genotyped for the associated haplotype block at the 1p13 locus and found that homozygosity for the minor allele is associated with a 10- to 12-fold increase in expression of SORT1 and PSRC1 in human liver with heterozygotes for the minor allele having an intermediate phenotype. CELSR2 expression also increased with homozygosity for the minor allele; however, the magnitude of the expression change was far less robust. Importantly, homozygosity for the minor alleles at 1p13 did not lead to an expression change in SORT1, CELSR2, or PSRC1 in subcutaneous or omental fat, suggesting a liver specificity to the expression changes seen with the protective minor alleles [17•].

Through fine mapping studies, Musunuru et al. [17•] identified a 6.1-kB segment of genomic DNA that contained the causal variant that explains the 1p13 LDL-C association. Musunuru et al. [17•] identified bacterial artificial chromosomes (BACs) containing either the major or minor haplotype and cloned them into a firefly luciferase expression vector, transfected the construct into a hepatic cell line, and showed that the BAC harboring the minor allele haplotype was able to drive higher luciferase expression than the major allele haplotype, suggesting that the 6.1-kB region contained the genetic regulatory element responsible for increased expression at the 1p13 locus conferred by the minor allele. Through a series of deletion and mutagenesis constructs, Musunuru et al. [17•] were able to identify the causal SNP that alone conferred the expression changes of the minor allele. Through a series of dominant negative C/EBP transfection studies, studies in cells lacking CCAAT-enhancer-binding protein (C/EBP), electrophoretic mobility shift assay (EMSA) studies, and chromatin immunoprecipitation (ChIP) experiments, Musunuru et al. [17•] were able to demonstrate that the minor allele at rs12740374 creates a C/EBP transcription factor binding site that is not present in the major allele. As C/EBP is a liver-enriched transcription factor, this could explain the specificity of the expression changes to liver.

Proof of Principle in Mice: Sortilin Overexpression Reduces Plasma Lipids and VLDL Secretion

To take the story further, Musunuru et al. [17•] used adeno-associated virus (AAV) serotype 8 to drive liver-specific expression of a murine Sort1 cDNA in mice and looked at the effects on plasma lipids in a variety of “humanized” mouse models of lipoprotein metabolism, including ApoBEC−/−; APOB transgenic, ApoBEC−/−; LDLR+/−; APOB transgenic, ApoBEC−/−; LDLR−/−; APOB transgenic, and ApoBEC−/−; LDLR−/− mice. These mouse models have lipid profiles more similar to humans, as human liver does not express ApoBEC whereas mouse liver does, and wild-type mice are predominantly HDL animals, whereas humans have much more LDL-C in circulation. In all of these mouse models, sortilin expression consistently reduced plasma cholesterol ranging from 30% to 70%. Musunuru et al. [17•] also used AAV8 to express Psrc1 in mouse liver in LDLR−/− mice, and showed no effect on plasma lipids, again supporting SORT1 as the causal gene at the 1p13 locus.

To address the mechanism, Musunuru et al. [17•] performed in vivo VLDL secretion studies in Sort1-expressing mice and showed a 50% reduction in the VLDL secretion rate with Sort1 expression. They also performed S35-labelling experiments in primary hepatocytes isolated from control and Sort1-expressing mice and showed a 40% reduction in apoB secretion with Sort1 expression, suggesting that Sort1 reduces plasma cholesterol by reducing the VLDL secretion rate [17•].

Proof of Principle: Sortilin Knockdown Increases Plasma Lipids and apoB Secretion

As a complementary approach to their overexpression studies, Musunuru et al. [17•] used short-interfering RNAs (siRNAs) directed against Sort1 prepared as a lipidoid formulation to knock down Sort1 expression specifically in liver in ApoBEC−/−; APOB transgenic, ApoBEC−/−; LDLR+/−; APOB transgenic, and ApoBEC−/−; LDLR−/− mice. These siRNAs were successful in reducing Sort1 expression 70% to 90% in liver with no effect on Sort1 expression in adipose. Consistent with the human data and with their overexpression data, liver-specific Sort1 knockdown consistently increased plasma cholesterol 20% to 125% [17•].

To determine whether Sort1 knockdown leads to increased apoB secretion, primary hepatocytes were isolated from control and siRNA-injected Sort1 knockdown mice and S35-labelling experiments were performed. Consistent with the overexpression studies, Sort1 knockdown in primary hepatocytes led to a 75% increase in apoB secretion [17•].

Disparate Findings in Sort1−/− Mice

Shortly after the Musunuru et al. [17•] study was published, a second study by Kjolby et al. [28•] using a Sort1−/− mouse was published. After 6 weeks of western-type diet feeding, Kjolby et al. [28•] showed that Sort1−/− mice had 20% lower plasma cholesterol compared to wild-type mice, and Sort1−/−; LDLR−/− double knockout mice had 30% lower cholesterol than LDLR−/− mice. Kjolby et al. [28•] went on to show that Sort−/−; LDLR−/− double knockout mice fed a western-type diet for 8 months had a 60% reduction in atherosclerotic lesion formation. Kjolby et al. [28•] also found that overexpression of a human Sort1 cDNA in liver by adenovirus in wild-type and Sort1−/−; LDLR−/− double-knockout mice fed a western-type diet increased plasma cholesterol [28•]. This contrasts with the human GWAS data and the mechanistic studies done by Linsel-Nitschke et al. [15•] and Musunuru et al. [17•], which all predict that increased Sort1 expression should reduce LDL-C levels deficiency should increase plasma cholesterol and the burden of atherosclerotic disease.

Kjolby et al. [28•] also looked at apoB and triglyceride secretion in primary hepatocytes and live animals and found that Sort1 deficiency was associated with reduced apoB secretion in primary hepatocytes and reduced apoB and triglyceride secretion in mice. Kjolby et al. [28•] also demonstrated that Sort1-deficient mice had increased intracellular apoB content, and this was reduced with adenovirus-mediated Sort1 overexpression. Kjolby et al. [28•] went on to use immunofluorescence, immunoprecipitation, and surface plasmon resonance to show that sortilin binds apoB with extremely high affinity, with a Kd of 1–2 nM and does not bind to apoB48. They suggested that Sort1 serves as a facilitator of hepatic apoB100 export, possibly by direct binding to apoB100.

Reconciling the Differences

The disparate findings of Musunuru et al. [17•] and Kjolby et al. [28•] bring into question the true biology of Sort1 as well as the therapeutic strategy to harness the LDL-modulating power of the 1p13 GWAS locus. The human GWAS data are in agreement with the studies published by Musunuru et al. [17•] and Linsel-Nitschke et al. [15•] and suggest that therapies that raise hepatic Sort1 expression will reduce plasma cholesterol and protect against ASCVD either by reducing VLDL secretion, facilitating LDL uptake, or a combination of both. Additionally, work by Ai et al. [29•] has shown that pharmacologic treatments that reduce Sort1 expression and mouse models that have reductions in Sort1 expression in liver have increased apoB/VLDL secretion, whereas restoration of Sort1 expression in mouse liver using AAVs and mouse models with increased Sort1 expression have reduced VLDL/apoB secretion, consistent with the studies published by Musunuru et al. [17•]. But, is there important biology being shown to us through the findings in the knockout mice?

The human GWAS data suggest that liver-specific increases in Sort1 expression are protective. The studies by Musunuru et al. [17•] used liver-specific manipulations to determine the effect of Sort1 overexpression and knockdown on plasma lipids and VLDL secretion, whereas the studies by Kjolby et al. [28•] used a total body Sort1−/− mouse. Sort1 is expressed both in adipose and small intestine, which are critical organs in lipid metabolism, and Sort1 deficiency in these tissues may be contributing to the lipid and secretory phenotypes observed in the study by Kjolby et al. [28•].

There are also differences in timing, mechanisms of Sort1 manipulations, diet, and mouse models that could explain the disparate phenotypes observed by the two groups (Table 3). Musunuru et al. [17•] performed acute somatic manipulations of Sort1 expression using siRNA and AAV, whereas in the Kjolby mouse model Sort1 has been deficient since conception. Compensatory mechanisms could be in place, which could explain the disparate phenotypes. Furthermore, whereas Kjolby et al. [28•] used a model completely lacking Sort1, the siRNAs used by Musunuru et al. [17•] reduced hepatic Sort1 expression 70% to 90%, meaning there was still residual Sort1 expression. There may be a parabolic relationship between Sort1 concentration and VLDL secretion, explaining the disparate findings in the knockout versus the knockdown models.

Table 3 Summary of the in vivo data suggesting that Sort1 is an important regulator of low-density lipoprotein metabolism
Full size table

Also of note is a recent paper using the Sort1−/− mouse used by Kjolby that suggests that the mouse is not a complete knockout mouse, but a deletion mutant [22•]. The Kjolby mouse has a disruption at exon 14 [30], which leads to a deletion in the VPS10 domain, but still allows a protein to be made. This protein is misfolded and does not traffic to the cell surface; however, this aberrant protein could potentially explain the disparate results.

Conclusions

The 1p13 locus and the causal gene SORT1 represent a promising target for LDL lowering and ASCVD risk reduction. Human GWAS data and the functional studies of alleles identified through GWAS performed by Linsel-Nitschke et al. [15•] and Musunuru et al. [17•] suggest that the therapeutic goal should be to raise Sort1 expression in liver. However, Kjolby et al.’s [28•] findings in the Sort1−/− mouse call into question the therapeutic strategy for Sort1 targeting, and suggest that it is Sort1 inhibition that will reduce plasma cholesterol and ASCVD risk. All research published to date support an unequivocal role for Sort1 in lipid metabolism and ASCVD and suggest that the power of the 1p13 locus can be harnessed to dramatically affect plasma cholesterol and ASCVD risk in humans. Future studies should aim to clarify the directionality of sortilin’s effect on plasma lipids by standardizing the knockout, knockdown, and overexpression studies in terms of diet and mouse model to see if the disparate findings can be reconciled.