Permanent Alteration of PCSK9 With In Vivo CRISPR-Cas9 Genome Editing

Abstract

Rationale

Individuals with naturally occurring loss-of-function PCSK9 mutations experience reduced blood low-density lipoprotein cholesterol (LDL-C) levels and protection against cardiovascular disease.

Objective

The goal of this study was to assess whether genome editing using a clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated (Cas) system can efficiently introduce loss-of-function mutations into the endogenous PCSK9 gene in vivo.

Methods and Results

We used adenovirus to express Cas9 and a CRISPR guide RNA targeting Pcsk9 in mouse liver, where the gene is specifically expressed. We found that within three to four days of administration of the virus, the mutagenesis rate of Pcsk9 in the liver was as high as >50%. This resulted in decreased plasma PCSK9 levels, increased hepatic LDL receptor levels, and decreased plasma cholesterol levels (by 35%–40%) in the blood. No off-target mutagenesis was detected in 10 selected sites.

Conclusions

Genome editing with the CRISPR-Cas9 system disrupts the Pcsk9 gene in vivo with high efficiency and reduces blood cholesterol levels in mice. This approach may have therapeutic potential for the prevention of cardiovascular disease in humans.

INTRODUCTION

Among the best-established causal risk factors for cardiovascular disease, the leading cause of death worldwide, is the blood concentration of low-density lipoprotein cholesterol (LDL-C), and pharmacological therapy that reduces LDL-C levels—namely the statin drugs—has proven the most effective means of reducing the risk of coronary heart disease (CHD). Yet even with the use of statin therapy, there remains a large residual risk of CHD, and a substantial proportion of patients are intolerant of statin therapy. Thus, there is a critical need to develop new strategies for the reduction of LDL-C.

Proprotein convertase subtilisin/kexin type 9 (PCSK9) has emerged as a promising therapeutic target for the prevention of CHD. A gene specifically expressed in and secreted from the liver, and believed to function primarily as an antagonist to the LDL receptor (LDLR), PCSK9 was originally identified as the cause of autosomal dominant hypercholesterolemia in some families, with gain-of-function mutations in the gene driving highly elevated LDL-C levels and premature CHD.¹ In subsequent studies, individuals with single loss-of-function mutations in PCSK9 were found to experience a significant reduction of both LDL-C levels (~30%–40%) as well as CHD risk (88%).^2,3 Notably, even individuals with two loss-of-function mutations in PCSK9—resulting in ~80% reduction in LDL-C levels—appear to suffer no adverse clinical consequences.^4,5 This observation suggests that therapies directed against PCSK9 would offer cardiovascular benefit without any accompanying undesirable effects. Just 10 years after the discovery of PCSK9, PCSK9-targeting monoclonal antibodies are being evaluated in clinical trials.⁶ Yet even if these antibody-based drugs prove effective, their effects on LDL-C are short-lived, and patients will have to receive injections of the drugs every few weeks, which will limit their use as preventative therapy.

The ability to permanently alter the human genome has been made possible by the technology now commonly known as “genome editing.” Recently published clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated (Cas) systems use Streptococcus pyogenes Cas9 nuclease that is targeted to a genomic site by complexing with a synthetic guide RNA that hybridizes a 20-nucleotide DNA sequence (protospacer) immediately preceding an NGG motif (PAM, or protospacer-adjacent motif) recognized by Cas9.^7,8 CRISPR-Cas9 generates a double-strand break that is usually repaired by non-homologous end-joining (NHEJ), which is error-prone and conducive to frameshift mutations resulting in knock-out alleles of genes.

In light of the observed high efficiencies of CRISPR-Cas9 in mammalian cells in vitro,^7,8 we assessed whether CRISPR-Cas9 can be used to disrupt the mouse Pcsk9 gene in vivo with high efficiency. A proof of principle that the gene can be targeted in mammalian hepatocytes in vivo would suggest that the approach might be viable in humans.

METHODS

We hypothesized that CRISPR-Cas9 would disrupt the mouse Pcsk9 gene in hepatocytes in vivo to a sufficient degree that plasma PCSK9 levels and cholesterol levels would be reduced. We further hypothesized that these would be specific effects, such that the use of CRISPR-Cas9 would affect neither plasma triglyceride levels nor ALT levels.

Candidate guide RNAs were designed to target exon 1 or exon 2 of the Pcsk9 gene, transfected into 3T3-L1 cells, and assessed for efficacy with Surveyor assays. Adenoviruses either expressing green fluorescent protein (GFP) or coexpressing Cas9 plus a guide RNA targeting Pcsk9 exon 1 (CRISPR-Pcsk9) were generated. In a pilot experiment, a total of four 11-week-old male C57BL/6 mice were used, two each for the GFP and CRISPR-Pcsk9 adenoviruses. In a second, more comprehensive experiment, a total of 15 5-week-old female C57BL/6 mice were used, five each for the GFP and CRISPR-Pcsk9 adenoviruses and five with no virus. Following the virus administration, after three days (in the first experiment) or four days (in the second experiment) the mice were sacrificed after overnight fasting, and the livers were harvested and terminal blood samples collected.

For the second experiment, we tested the null hypotheses that upon CRISPR-Cas9 genome editing, each of four plasma analytes—PCSK9, triglyceride, cholesterol, and ALT—did not differ among the groups of mice. The levels of the analytes were each compared among three groups—mice that received no virus (N = 5), mice that received GFP virus (N = 5), and mice that received CRISPR-Pcsk9 virus (N = 5). Initially, the Kruskal-Wallis test was performed for each of the four analytes, with a statistical significance threshold of P < 0.0125 to account for multiple testing (Bonferroni correction for four tests). Two analytes (PCSK9 and total cholesterol) were found to reach statistical significance. For each of these two analytes, the Mann-Whitney U test was performed for each possible pairwise comparison among the three groups of mice, with a statistical significance threshold of P < 0.00833 to account for multiple testing (Bonferroni correction for two sets of three pairwise comparisons performed). All statistical analyses were performed using GraphPad Prism 6 for Mac OS X.

An expanded Methods section is available in the Online Supplement.

RESULTS

We initially screened candidate CRISPR guide RNAs targeting sequences in exon 1 and exon 2 of the mouse Pcsk9 gene in 3T3-L1 cells. We found that the guide RNA targeting exon 1 displayed ~50% mutagenesis at the on-target site in Pcsk9, as judged by the Surveyor assay (Figure 1A, Online Figure I). We made an adenovirus coexpressing Cas9 and this guide RNA (CRISPR-Pcsk9), using an adenovirus expressing GFP as a control.

Figure 1. On-target and off-target effects in mouse cells and livers receiving CRISPR-Cas9.

A shows Surveyor assays performed with genomic DNA from 3T3-L1 cells transfected with Cas9 and a guide RNA targeting Pcsk9 exon 1 (gRNA-1) or a guide RNA targeting Pcsk9 exon 2 (gRNA-2). B shows Surveyor assays performed with genomic DNA from liver samples taken from mice three days after receiving a control adenovirus expressing GFP (A, B) or an adenovirus expressing Cas9 and gRNA-1 (CRISPR-Pcsk9) (C, D). C shows Surveyor assays performed with genomic DNA from liver samples taken from mice four days after receiving no virus (1–5), the GFP virus (6–10), or the CRISPR-Pcsk9 virus (11–15). D shows Surveyor assays performed with liver genomic DNA from the “A” and “C” mice. The Pcsk9 exon 1 on-target site and ten genomic sites deemed to be the most likely off-target sites for CRISPR-Cas9 activity (OT1–OT10; see Online Supplement for site sequences) were assessed. Arrows show the cleavage products resulting from the Surveyor assays; the intensity of the cleavage product bands relative to the uncleaved product band corresponds to the mutagenesis rate.

In a pilot experiment, the CRISPR-Pcsk9 virus and the GFP virus were administered to two 11-week-old male mice each. After three days, we sacrificed the mice in order to harvest liver tissue. Whereas there was no evidence of mutagenesis in the control mice, the CRISPR-Pcsk9 mice displayed substantial levels of mutagenesis, with one of the mice showing ~50% mutagenesis in the Surveyor assay, consistent with alteration of at least half of the Pcsk9 alleles in the liver (Figure 1B, Online Figure I). Analyzing liver DNA from that mouse, we found that a wide variety of indels were produced in Pcsk9, ranging from 1 bp to 228 bp, with the possibility of larger indels that were not detected by PCR analysis (Online Figure I). The most frequent indels were a 1-bp insertion and a 2-bp deletion. We assessed for off-target mutagenesis at the 10 sites deemed most closely matched to the on-target site and most likely to harbor off-target effects (six sites with three mismatches to the on-target site and the four highest-scoring sites with four mismatches to the on-target site; see Online Supplement for site sequences). We found no evidence of significant off-target mutagenesis, within the limit of detection by the Surveyor assay (Figure 1D).

To test the hypotheses that genome editing would result in reduced plasma PCSK9 and cholesterol and no differences in triglycerides and ALT, we next performed a more comprehensive experiment in which the CRISPR-Pcsk9 virus and the GFP virus were administered to five 5-week-old female mice each, with an additional group of five mice receiving no virus. After four days, the CRISPR-Pcsk9 mice all displayed substantial levels of mutagenesis, in some cases exceeding 50%, with no mutagenesis observed in any of the mice in the two control groups (Figure 1C). We compared plasma PCSK9 levels at four days by enzyme-linked immunosorbent assay (ELISA); the CRISPR-Pcsk9 mice displayed substantially lower PCSK9 levels compared to each of the control groups of mice [2597 pg/mL with CRISPR-Pcsk9 virus vs. 26461 pg/mL with GFP virus (P = 0.0079) and vs. 21734 pg/mL with no virus (P = 0.0079); N = 5 per group; Figure 2A].

Figure 2. Effects of CRISPR-Cas9 genome editing on mice.

A shows the results of ELISAs for PCSK9 protein, measurements of triglyceride levels, and measurements of total cholesterol levels in plasma samples from mice four days after receiving no adenovirus, GFP adenovirus, or CRISPR-Pcsk9 adenovirus (N = 5 mice for each group). B shows the full plasma lipoprotein cholesterol profiles of pooled plasma samples from each group of mice. C shows the plasma ALT levels in mice four days after receiving virus (N = 5 mice for each group) and hematoxylin/eosin stains of liver sections from representative mice. D shows the results of Western blot analysis of liver samples taken from mice four days after receiving virus. For A and C, P values were determined by the Kruskal-Wallis test among all three groups (in red); if statistically significant, the Mann-Whitney U test between each pair of groups was performed (P values in black). Error bars show s.e.m.

Whereas there was no significant difference in plasma triglyceride levels among the three groups at four days, CRISPR-Pcsk9 mice had significantly lower levels of total plasma cholesterol, with 35%–40% reduction compared to the control groups [101 mg/dL with CRISPR-Pcsk9 virus vs. 157 mg/dL with GFP virus (P = 0.0079) and vs. 161 mg/dL with no virus (P = 0.0079); N = 5 per group; Figure 2A]. We performed complete lipoprotein profiling of pooled plasma samples from each group, observing reduced HDL and LDL fractions in CRISPR-Pcsk9 mice (Figure 2B), consistent with prior observations in Pcsk9 knockout mice.⁹ (Of note, reduction of PCSK9 in humans is not expected to reduce plasma HDL cholesterol levels, as observed in these experiments in mice, due to differences between human and mouse HDL metabolism.) No significant difference in blood ALT levels at four days among the three groups was observed, and hematoxylin/eosin staining of liver sections from representative mice that received either the GFP virus or the CRISPR-Pcsk9 virus showed no inflammation (Figure 2C). Finally, we assessed LDLR levels in liver by Western blot analysis. PCSK9 functions to downregulate LDLR;⁹ consistent with this relationship, the CRISPR-Pcsk9 mice had higher levels of LDLR protein than the control groups of mice (Figure 2D).

DISCUSSION

In this proof-of-principle study, we found that CRISPR-Cas9 could disrupt the mouse Pcsk9 gene in vivo with high efficiency and result in decreased circulating PCSK9 levels, increased hepatic LDLR levels, and decreased plasma cholesterol levels. The 35%–40% lower cholesterol levels in the CRISPR-Pcsk9 mice compared to control mice is consistent with the 36%–52% lower levels previously observed in Pcsk9 knockout mice compared to wild-type mice.⁹ Thus, this approach may have therapeutic potential for the prevention of cardiovascular disease in humans.

Although the use of adenovirus allows for efficient delivery to the liver and sustained expression of the CRISPR-Cas9 system, it is not the optimal therapeutic vehicle due to the immune response to the virus. Indeed, inflammation and acute phase responses could potentially have affected the plasma cholesterol levels in the mice receiving adenovirus. However, we included a control group that did not receive any adenovirus, and the plasma PCSK9, triglyceride, cholesterol, and ALT levels were very similar to the levels observed in the control group that received the GFP virus. There was no apparent inflammation in the liver within the time frame of the experiments. Thus, we did not see any evidence of confounding due to the use of adenovirus.

While the use of adeno-associated virus would be preferable, the gene encoding Streptococcus pyogenes Cas9 (~4.2 kb) in combination with a CRISPR guide RNA-expressing cassette (~500 bp) is too large to fit into standard liver-targeting adeno-associated virus vectors (e.g., AAV2/8). Furthermore, the rapidity with which robust alteration of the Pcsk9 gene occurred in our experiments—up to >50% mutagenesis in just three to four days—suggests that a single brief pulse of CRISPR-Cas9 expression would be sufficient to achieve a therapeutic effect. Thus, a virus-free delivery method that transiently expresses CRISPR-Cas9 (e.g., RNAs in lipid nanoparticles) might be optimal. A recent study showed that hydrodynamic tail vein injection of DNA vectors encoding CRISPR-Cas9 successfully targeted a liver gene (Fah) with no apparent long-term adverse effects.¹⁰

A possible barrier to therapeutic CRISPR-Cas9 applications is the issue of off-target mutagenesis. In our study, we did not observe significant off-target mutagenesis at a number of potential off-target sites, but we cannot rule out low-frequency events in vivo. Strategies to greatly reduce off-target mutagenesis without impairing on-target mutagenesis are being developed and can be adapted for use in therapeutic applications.

With cardiovascular disease being the number one killer worldwide, a safe and effective PCSK9-genome-editing therapy could have a significant impact on human health. A single administration could confer the benefits of naturally occurring PCSK9 loss-of-function mutations—a permanent reduction in LDL-C levels and CHD risk, equivalent to taking statins every day for the rest of one’s life but without the need for chronic therapy. It could represent a paradigm shift in thinking about cardiovascular therapeutics: a one-shot, long-term solution—not unlike a vaccination—rather than a pill to be taken every day or an injection to be received every few weeks. It could also open the door to a whole new class of therapies, where one might be able to target not just PCSK9 but a number of other potential therapeutic genes; indeed, given the multiplexing capacity of CRISPR-Cas9,^7,8 it might be feasible to efficiently target multiple genes simultaneously with a single therapy.

Novelty and Significance.

What Is Known?

• Clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated (Cas) systems allow for high-efficiency genome editing, particularly gene knock-out, in vitro.

• PCSK9 loss-of-function mutations confer reduced blood cholesterol levels and protection against coronary heart disease (CHD) in humans.

What New Information Does This Article Contribute?

• When delivered by adenovirus, CRISPR-Cas9 produces efficient knock out Pcsk9 in mouse hepatocytes in vivo.

• Knock out Pcsk9 by CRISPR-Cas9 in mouse liver results in reduced plasma PCSK9 protein and cholesterol levels.

Individuals with naturally occurring loss-of-function PCSK9 mutations experience reduced blood cholesterol levels and protection against CHD without any known adverse consequences. Genome editing allows for permanent alteration of genes in mammalian cells. We report that a single administration of the CRISPR-Cas9 genome-editing system to mice produced a high proportion of loss-of-function alleles of the Pcsk9 gene in the liver in vivo, resulting in substantially reduced cholesterol levels (35%–40%). This approach may be therapeutically useful in humans, potentially taking the form of a one-time vaccination that would permanently reduce cholesterol levels and protect against CHD.

Nonstandard Abbreviations and Acronyms

LDL

low-density lipoprotein cholesterol

CHD

coronary heart disease

PCSK9

proprotein convertase subtilisin/kexin type 9

LDLR

low-density lipoprotein receptor

CRISPR

clustered regularly interspaced short palindromic repeats

Cas

CRISPR-associated

PAM

protospacer-adjacent motif

NHEJ

non-homologous end joining

GFP

green fluorescent protein

ELISA

enzyme-linked immunosorbent assay