Introduction

Homozygous familial hypercholesterolemia (HoFH) is a life-threatening autosomal co-dominant disorder with an estimated prevalence of up to 1 in 160,000–300,000 [1, 2]. It is estimated that more than 90 % of HoFH patients harbor loss-of-function mutations in both alleles of the low-density lipoprotein receptor gene (LDLR). These mutations substantially or fully impair LDLR function, resulting in low-density lipoprotein cholesterol (LDL-C) levels frequently in excess of 500 mg/dL and development of cardiovascular disease (CVD), principally coronary artery and aortic valve disease, during adolescence and death in the 20s, if left untreated [2]. Interestingly, significant phenotypic variability has recently been reported in a cohort of molecularly defined HoFH subjects, with LDL-C levels that extensively overlap with levels clinically associated with heterozygous status [1]. A major determinant of the severity of the HoFH phenotype is the residual LDLR activity associated with a given mutation. Based on in vitro assays in cultured fibroblasts, HoFH subjects can be classified as receptor-negative (<2 % residual LDLR activity) or receptor-defective (2–25 % residual LDLR activity) [3]. Receptor-defective FH subjects have significantly lower mean LDL-C levels and develop symptomatic coronary artery disease almost a decade later than their receptor-negative counterparts [3, 4]. Furthermore, they respond better to conventional and novel therapies. In the TESLA B study with the proprotein convertase subtilisin/kexin type 9 (PCSK9) monoclonal antibody evolocumab, receptor-defective HoFH subjects experienced a ∼25 % LDL-C reduction, in contrast with the lack of effect seen in receptor-negative subjects [5]. Taken together, these findings suggest that even a small amount of LDLR activity is beneficial compared to no receptor activity at all.

The treatment of HoFH has historically been very challenging due to the extremely elevated baseline LDL-C levels and poor responsiveness to standard lipid-lowering therapies (LLTs) [2]. Statin treatment is associated with a 10–25 % reduction in LDL-C in this population [6], and the addition of ezetimibe results in a further reduction of 10–15 % [7]. Despite a 60 % reduction in mortality due to these treatments [8], mean life expectancy in HoFH remains low at 33 years [8]. Substantial reduction in LDL-C is obtained with LDL apheresis. However, its effect is transitory and availability is limited [9]. The recently approved lipid-lowering agents mipomersen [10], lomitapide [11], and evolocumab [5] produce additional LDL-C reduction on top of standard therapy in HoFH patients; however, issues of tolerability or lack of efficacy in receptor-negative patients will limit the potential of these agents to fully address the serious unmet medical need in this population. Thus, despite significant therapeutic progress over the last few years, HoFH remains a challenging condition to treat for many of these patients.

HoFH is an ideal candidate for liver-directed gene therapy. Approximately 75 % of the body’s LDL receptors are expressed within the liver, making it the most important organ in LDL metabolism [12]. This role is underscored by the dramatic correction and even resolution of the HoFH phenotype with orthotopic liver transplantation [12–14]. The disadvantages and risks of transplantation and long-term immunosuppression limit the viability of this approach in this disease. However, the delivery of a functional LDLR transgene to the liver has been of great interest. In the early 90s, Grossman et al. conducted the first ever gene therapy clinical trial in HoFH subjects, using an ex vivo approach to transfer replication-deficient human LDLR-expressing retroviruses into the livers of five HoFH patients [15]. To achieve this, the patients underwent resection of the hepatic left-lateral segment, hepatocytes were harvested, transduced ex vivo using the retrovirus vector, and autologous transduced cells were then infused back into their donors via the portal vein. In contrast with its effects in a natural rabbit model of HoFH [16], disappointing results were observed in human subjects with few transduced hepatocytes and variable (6–25 % LDL-C reduction in three subjects) but mostly clinically unremarkable transient metabolic effects. These findings precluded further clinical development of the ex vivo approach but confirmed the feasibility and safety of LDLR gene therapy in humans. A number of different vectors and transgene constructs have since been tested in the pre-clinical setting; however, none have achieved stable gene expression with durable metabolic effects [17, 18]. In the last few years, however, several novel approaches have yielded promising results, bringing us to an exciting time in this field. In this review, we discuss the most advanced of these developments, a recombinant adeno-associated virus (AAV) vector which has recently entered clinical testing. We also review approaches in pre-clinical development including gain-of-function LDLR transgene variants, RNA silencing of the HMG-CoA reductase gene (HMGCR), and induced pluripotent stem cell transplant therapy.

Approaches in Clinical Development: Recombinant AAV Vectors

Development of AAV Vectors for Gene Therapy

AAV is a replication-defective, non-pathogenic parvovirus with a linear 4.7-kB single-stranded DNA [19]. In its latent phase, AAV DNA is retained in the host cell nucleus in a circular episomal form with negligible rates of genomic integration allowing for stable gene expression in post-mitotic tissues [20]. The first AAV vector to be developed, AAV2, achieved efficient gene transfer and stable hepatocyte transduction with low immunogenicity in pre-clinical models [20, 21]; in human subjects, however, AAV2 had poor liver tropism and was associated with a limiting adaptive immune response [20, 22]. A number of different AAV vectors have since been developed from other naturally occurring AAV serotypes. These differ from AAV2 in tropism and immunogenicity allowing for broader use of the AAV vector class as gene therapy platforms [23–25]. In addition, bioengineering approaches have generated novel variants that, in pre-clinical testing, evaded cellular and humoral immune responses more effectively and transduced a wide range of cell types more efficiently than their naturally occurring parent serotypes [26•]. In 2006, the first bioengineered AAV vector, AAV2.5 (a chimeric serotype derived from AAV1 and AAV2), entered clinical development in a Duchenne’s muscular dystrophy cohort with mixed results [27].

AAV vectors are now widely used in gene therapy agents and have performed well clinically [25, 26•]. Most notable among these successes is that of the lipoprotein lipase (LPL) expressing AAV1 vector, AAV1-LPL S447X, which became the first gene therapy agent to gain regulatory approval in the Western world in 2012 [28]. In LPL-deficient subjects, AAV1-LPLS447X improved post-prandial triglyceride metabolism with clinically significant reductions in the incidence and severity of pancreatic episodes for up to 2 years after administration of the vector with follow-up ongoing [28, 29]. The factor IX (FIX) expressing liver-targeted AAV8 vector, scAAV2/8-LP1-hFIXco, also performed well in a phase 1/2 hemophilia B trial [30], achieving serum FIX levels of 1–7 % of normal from baseline levels of <1 % with clinically significant phenotypic improvement. This has persisted for up to 4 years in all of the subjects treated [31••]. The results from this trial hold particular significance for the prospects of a liver-targeted recombinant AAV8 agent for HoFH.

Efficacy of Recombinant AAV8 in Mouse Models of HoFH

AAV8 was identified for liver-directed gene therapy in HoFH due to its strong liver tropism and relatively low (38 %) seroprevalence in Western populations [24, 32]. The first AAV8-based HoFH gene therapy study compared the efficacy of a human LDLR-expressing recombinant AAV8 vector to AAV2 and AAV7 in Ldr −/− mice on a high-fat diet. Compared to AAV2, AAV8 was associated with more efficient gene transfer (2.1 vs. 52 gc/cell), hepatocyte transduction (4.2 vs. 81.2 % of all hepatocytes), and an improved lipid profile (total cholesterol 227 vs.1032 mg/dL, p<0.001, at day 56 post-vector administration) [33]. AAV8 was also superior to AAV7 in these endpoints. Moreover, crucially, in AAV8-treated mice, metabolic effects were maintained for up to 20 weeks and the development of atherosclerosis was almost completely attenuated.

The differences in the pre-clinical and clinical results observed with the HoFH ex vivo gene therapy approach highlight important limitations with the use of animal models. This has led to substantial efforts in developing an animal model for HoFH that more accurately reflects the human pathophysiology. By deleting the apolipoprotein B messenger RNA (mRNA) editing catalytic polypeptide-1 (Apobec1) and LDLR genes (Ldlr −/−/Apobec1 −/−), Kassim et al. designed a murine model of HoFH that, on a low-fat chow diet, develops a lipid and atherosclerotic phenotype comparable to that seen in human HoFH subjects [34, 35]. In Ldlr −/−/Apobec1 −/− mice maintained on a chow diet, increasing doses of a recombinant AAV8 vector containing a mouse LDLR transgene driven from a liver-specific thyroxine-binding globulin (TBG) promoter (AAV8. TBG.mLDLR) produced a significant dose-dependent reduction in total cholesterol and non-HDL-C within 7 days of treatment [35]. Interestingly, the lowest dose that produced a significant lipid improvement (total cholesterol −141 mg/dL and non-HDL −130 mg/dL at day 35) was associated with transduction of only 5–10 % of hepatocytes. Complete metabolic correction was achieved with doses at least an order of magnitude higher than this dose and maintained for up to 6 months. Gene transfer and hepatocyte transduction were also dose-dependent and, at the highest dose tested, achieved 60–70 % hepatocyte transduction at 35 days post-vector treatment.

In a further study, a more advanced humanized model was designed using Ldlr −/−/Apobec1 −/− mice engineered to express the human apoB100 transgene (Ldlr −/−/Apobec1 −/−/human ApoB transgenic) [36••]. In these mice, an AAV8 vector expressing the human LDLR transgene (AAV8.TBG.hLDLR) produced comparable metabolic effects to that seen in the previous study [36••]. Of interest, experiments performed in heterozygous FH (HeFH) mice in this study showed that, although AAV8.TBG.hLDLR also produced a dose-dependent reduction in serum lipid levels in HeFH mice, the minimum effective dose in this cohort was an order of magnitude higher than that recorded in the setting of the LDLR knockout mice.

Atherosclerotic regression is an important therapeutic endpoint for lipid-lowering therapies used in HoFH patients. Atherosclerotic progression was assessed in a cohort of Ldlr −/−/Apobec1 −/− mice maintained on a high fat for 2 months [36••]. At this point, these mice received a tail-vein injection of AAV8.TBG.mLDLR and were followed for an additional 2 months while remaining on the high-fat diet. Within 7 days of receiving the vector, a significant reduction in mean total cholesterol was observed in these mice (1555 to 266 mg/dL) that normalized by the end of the follow-up period (67 mg/dL). This was associated with an 87 % reduction in the aortic plaque burden. Histological analysis of these lesions showed smaller lesions, fewer inflammatory cells, and favorable plaque remodeling [36••].

Taken together, these studies demonstrate that a recombinant LDLR-expressing AAV8 vector produces efficient gene transfer, stable hepatocyte transduction, significant lipid correction, and atherosclerotic regression in humanized mouse models of HoFH. However, it has been shown that AAV vectors transduce human hepatocytes less efficiently than murine hepatocytes [24, 37]. Recombinant FIX-expressing AAV8 produced substantially greater serum FIX elevations in mouse models of hemophilia B [38] than it did in human hemophilia B subjects [30]. Crucially, however, the levels achieved in the human hemophiliacs although objectively low (1-7 % of normal), produced clinically significant effects that have persisted for more than 3 years [31••]. This is of relevance to the use of AAV8 in HoFH because patients carrying LDLR mutations associated with >2 % residual LDLR activity (receptor-defective) have a significantly better prognosis than patients with LDLR mutations associated with <2 % residual LDLR activity (receptor-negative) [4].

Safety Profile of Recombinant AAV8 Vectors

Published studies in mice have not provided evidence of toxicity associated with the LDLR-expressing AAV8 vector [33, 35, 36••]. However, caution must be taken in applying these findings to humans because substantial differences exist between the murine and human immune systems [39, 40]. AAVs are naturally present in the environment [41], so it is expected that a percentage of the population have been exposed to them and will have developed anti-AAV8 neutralizing antibodies (nAbs) [32, 41] and possibly also AAV memory T cells [42]. Furthermore, as demonstrated by previous gene therapy trials in LPL deficient [28] and hemophilia patients [30, 31••, 43], administration of AAV vectors induces both cellular and humoral adaptive immune responses in human subjects.

Humoral immunity poses a challenge to gene therapy with AAV vectors for a number of reasons. First, approximately 30 % of patients from Western populations have pre-existing anti-AAV8 neutralizing antibodies [32] and would be precluded from treatment with these vectors. Pre-existing antibodies against other serotypes such as AAV2 are even more prevalent [32]. In non-Western populations, such as the Chinese where anti-AAV8 sero-prevalence is >80 % [44], many more could be ineligible for AAV treatment [41]. Second, immunity that develops as a result of AAV8 administration precludes re-administration of the transgene with the same vector should transgene expression falls below therapeutic levels.

The cellular immune response poses the challenge of limiting transgene expression through the destruction of transduced cells [42]. It has been shown that a capsid and, less commonly, a transgene-specific T cell response are elicited by recombinant AAV vectors targeted to skeletal muscle [27, 28, 45] and the liver [30, 43]. Focusing on the liver data, in the first liver-directed hemophilia B trial with AAV2, serum FIX levels of 10–12 % normal were achieved at the highest dose tested in one patient; however, within 2–4 weeks of vector administration, levels declined and returned to baseline by 10 weeks [43]. This was accompanied by a capsid-specific T cell response and a rise in transaminases which was asymptomatic and reversible. In this study, the T cell response was attributed to expansion of AAV2-specific memory T cells [42, 43]; however, alternative theories exist [40, 46–48]. The capsid-specific T cell response was subsequently shown to be dose-dependent, associated with other AAVs and amenable to immunosuppression in the AAV8 hemophilia B trial [30, 31••]. In this study, capsid-specific T cells were detected in the intermediate and higher dosing groups and accompanied by a reversible transaminitis in four out of the six subjects in the high-dose cohort. As in the previous study, FIX levels declined in association with the cellular response; however, in the highest dosing group, mean steady state FIX levels of up to 7 % of normal were achieved and maintained [31••]. Importantly, a course of prednisolone was administered to subjects that experienced a rise in transaminases. The decline in serum FIX levels was more pronounced when the initiation of prednisolone was delayed for >2 days [31••]. Of note, a dramatic 50–70 % reduction in serum FIX levels was also observed in association with modest transaminase elevations (peak ALT 36–202 IU/L), highlighting the detrimental effect of the cellular immune response on gene expression, and the importance of early initiation of immunosuppression [31••]. Crucially, the capsid-specific T cell responses in these trials were subclinical and no transgene-specific T cell responses were detected.

Based on this preliminary safety and efficacy data, a first-in-human phase 1/2a trial has been announced with the recombinant AAV8 vector AAV8.TBG.hLDLR (NCT02651675; see Table 1). Its primary outcome measure is the number of participants that experience vector-related adverse effects by 52 weeks following administration. The total follow-up of subjects in this study is up to 5 years.

Table 1 Gene therapy agents for homozygous familial hypercholesterolemia in recent development
Full size table

Recombinant AAV8 in HoFH: Future Directions

The development of strategies to improve transgene expression and minimize the immune response is ongoing. Bioengineering approaches are being used to develop AAV variants with attenuated immunogenicity and improved transgene transfer and hepatocyte transduction efficacies [40, 49]. One AAV2 variant developed with this strategy was almost 100-fold more resistant to neutralization by nAbs specific for its parent AAV2 capsid in vitro with consistent findings in vivo [50]. AAV2.5 is thus far the only engineered AAV2 variant to undergo clinical testing. In the first-in-human study with this agent, six boys with Duchenne’s muscular dystrophy received an intramuscular injection of one of two doses of a recombinant AAV2.5 vector carrying a minidystophin transgene (AAV2.5-minidystrophin) under corticosteroid cover [27]. Vector genome transfer was confirmed in all of the subjects treated; however, gene expression was poor overall. Both humoral and adaptive immune responses to the AAV capsid were detected in this study, although pre-existing seropositivity for AAV2.5 was shown to be ∼25 % lower than for AAV2. These clinical findings support the safety of engineered AAV serotypes in humans; however, the mixed efficacy and immunological results suggest that further work is needed to develop variants with clear superiority to native AAV vectors.

Based on the successes of the LPL [28] and FIX [51, 52] gain-of-function variants, a further approach in development are gain-of-function LDLR transgene variants. Using site-directed mutagenesis of hLDLR cDNA, Somanathan et al. developed three gain-of-function LDLR transgene variants that expressed LDLR proteins resistant to PCSK9 and/or inducible degrader of LDLR (IDOL)-mediated LDLR degradation. These variants (carried by AAV8 vectors) produced significantly greater cholesterol corrections in vivo and in vitro compared to the wild-type LDLR transgene in the setting of PCSK9 or IDOL overexpression (see Table 1) [53•]. In vivo studies with physiological levels of PCSK9 and IDOL are needed to verify these findings. Moreover, it is possible that these variants could reduce responsiveness to the PCSK9 inhibitors so this should also be addressed in future studies.

Approaches in Pre-Clinical Development

RNA Silencing of 3-Hydroxy-3-Methyl-Glutaryl-CoA Reductase

In a 2010 paper, Hibbitt et al. showed that combining statin therapy with a LDLR transgene construct containing LDLR genomic regulatory elements increased transgene expression up to fivefold [54]. Building on this work, they piloted an approach that combined RNA silencing of 3-hydroxy-3-methyl-glutaryl-CoA reductase (HMGCR), which encodes the enzyme target of statins, with a plasmid that contained the LDLR transgene driven by LDLR genomic regulatory elements (pLDLR-LDLR) [55•]. Using a luciferase reporter gene driven by LDLR regulatory elements, they showed that small interfering RNA oligonucleotides and microRNA (miRNA) specific for murine HMGCR transcripts increased LDLR promoter activity by 300-fold and 12-fold, respectively [55•]. Furthermore, when they co-transfected four HMGCR-specific miRNA plasmids with pLDLR-LDLR into Ldr −/− mice, they observed significantly lower LDL-C levels (−∼1.5 mmol/L, p = 0.0036) in these mice as compared with controls that only received the pLDLR-LDLR plasmid. Thus, RNA-mediated knockdown of HMGCR provides an effective complementary strategy for enhancing LDLR transgene expression. How this approach would compare to combining LDLR gene transfer with regular statin therapy, which is the standard of care in HoFH, is an important question to address in the development of this approach.

Induced Pluripotent Stem Cell-Based Cell Transplant Therapy

Induced pluripotent stem cell-based cells (iPSCs) are somatic cells generated by overexpression of specific transcription factors. As orthotopic liver transplantation is curative in HoFH [2], autologous transplantation of genetically corrected cells derived from iPSCs has been of interest as a way of harnessing the benefits of transplantation while avoiding its undesirable effects. It has been shown that iPSCs derived from patients with familial hypercholesterolemia (FH) can be differentiated into hepatocyte-like cells that closely recapitulate the FH phenotype in vitro [56, 57]. Building on this, Fattahi et al. showed that transfecting iPSCs derived from a HoFH patient with a plasmid containing the LDLR transgene generated hepatocyte-like cells that expressed LDLR mRNA and were able to internalize LDL under feedback regulation from extracellular LDL-C [58]. In a subsequent study, Ramakrishnan et al. showed that LDL internalization into such cells was also responsive to extracellular lovastatin and sterol, consistent with the restoration of physiological LDL metabolism [59•].

In vivo data in HoFH pre-clinical models are not yet available; however, corrected iPSCs that were transplanted into the liver of fumarylacetoacetate hydrolase (FAH)-deficient mice, another inherited metabolic disease of the liver, reversed the disease phenotype [60]. Of note, however, genetically corrected hepatocytes in FAH mice possess a unique survival advantage to native hepatocytes and expand robustly to repopulate the liver. Conversely, engraftment and repopulation of the liver are known to be less effective with iPSCs than with mature human hepatocytes [61]. Translation of iPSC technology to the clinic will also require significant refinement of the iPSC technology itself to ensure the genetic stability of derived cells, production of high-quality iPSCs, and efficient and reproducible differentiation to fully functional hepatocyte-like cells [61]. In addition, the risk of tumorigenesis associated with retained stem cells and inadvertently activated oncogenes during reprogramming must be mitigated in cells that are to be used therapeutically [62]. Thus, although appealing, the iPSC technology remains very much experimental and is unlikely to be a viable therapeutic for HoFH in the near future.

Conclusion

After more than two decades, gene therapy for HoFH has returned to clinical testing. The AAV vector class which has been widely successful in clinical trials in other Mendelian diseases is the backbone on which this agent has been developed. The recently announced first-in-human clinical trial is expected to recruit a dozen HoFH subjects, who will receive a recombinant LDLR-expressing AAV8 vector with follow-up of up to 5 years. It is worth noting that none of the AAV gene therapy agents that have been clinically successful in the past decade have produced complete reversal of the disease phenotype. As such, the recombinant LDLR-expressing AAV in testing is unlikely to prove curative but might produce clinically significant LDL-C reduction and improved response to lipid-lowering agents such as PCSK9 inhibitors. Nevertheless, a “cure” remains the ultimate goal of any gene therapy approach. Achieving this will depend on achieving sufficient and sustained levels of stable LDLR expression. Bioengineering strategies have been employed to develop gain-of-function LDLR transgenes and AAV vectors adapted to enhance transgene transfer, transduction efficacy, and minimize immunogenicity compared to their natural occurring AAV counterparts. Combining LDLR gene therapy with RNA silencing of HMGCR is a further strategy intended to enhance LDLR gene expression. These approaches are still largely in pre-clinical development; however, with promising early results, it is hoped that they will advance us further toward a genetic cure for HoFH in the future.