Gene therapy for haemophilia: a long and winding road


  • K. A. HIGH

    1. Howard Hughes Medical Institute, University of Pennsylvania School of Medicine, The Children’s Hospital of Philadelphia, Philadelphia, PA, USA
    Search for more papers by this author

Katherine A. High, Howard Hughes Medical Institute, University of Pennsylvania School of Medicine, The Children’s Hospital of Philadelphia, 3501 Civic Center Boulevard, 5060 CTRB, Philadelphia, PA 19104, USA.
Tel.: +215 590 4521; fax: +215 590 3660.


Summary. Background: Cure, or improvement of disease phenotype, has been a long-term goal in the treatment of haemophilia. An obvious strategy for achieving this goal is the use of gene therapy. Objectives: This paper summarises prior and current clinical trials of gene therapy for haemophilia A and B, and briefly describes additional strategies in pre-clinical development. Results and Conclusions: Approximately 50 human subjects with severe haemophilia A or B have been enrolled in seven different trials of gene therapy. These have used plasmids, retroviral, adenoviral, and AAV vectors, directed to autologous fibroblasts, skeletal muscle, liver, and other target cell types accessed by intravenous injection of vector. Four separate trials have used AAV vectors, three of these targeting liver. Data from animal models suggest that several different gene replacement strategies may eventually yield long-term expression of factor at therapeutic levels, and that in situ correction of gene defects in hepatocytes may eventually be a therapeutic option.

Clinical gene therapy commenced approximately 20 years ago when a group of investigators at the NIH carried out a trial in which T cells from children with severe combined immunodeficiency due to adenosine deaminase (SCID-ADA) deficiency were genetically modified with a retroviral vector expressing ADA and re-infused into the patients [1]. Although years elapsed before SCID-ADA was cured by a gene therapy approach [2], the work had begun safely, and the focus of the field began to move from tissue culture and animal studies to clinical investigation. The purpose of this review is to summarise progress in gene therapy for genetic disease, particularly as it relates to gene therapy for haemophilia. The focus is on clinical studies, but promising pre-clinical data, including studies that are close to clinical translation, will also be discussed.

Gene therapy for genetic disease: 20 years of clinical investigation begins to bear fruit

Genetic disease represents a substantial burden for humanity and for the health care system, yet for many of the most common inherited diseases, including thalassemia, cystic fibrosis, and sickle cell disease, therapeutic options remain limited. An unstated but hoped for goal of the Human Genome Project was that the identification of disease-causing genes would spur the development of therapies for people with genetic disease. The most obvious method to reach that goal was through gene therapy, the transfer of a normal copy of a gene to an individual with a mutation in that gene. Though the concept is simple, the practice is complex, and gene therapy products are arguably some of the most complex yet developed, with both a protein and a nucleic acid component. However, applications for licensed gene therapy products have now been filed [3], and additional products are in late-phase testing, so true gene therapeutics appear to be arriving. It is worth noting that this timeline, 20 years of clinical investigation preceding the first licensed products, is similar to that observed with other new classes of therapeutics, for example monoclonal antibodies.

The key goals of gene therapy for genetic disease are: (i) long-term expression, and (ii) expression at levels high enough to ameliorate or cure the clinical phenotype of the disease. There are essentially two strategies for achieving long-term expression. The first is to use an integrating vector to transduce a stem cell; all the daughter cells then carry the donated gene, to give long-lasting expression. The second employs a strategy of transducing long-lived post-mitotic cells, such as cells in the central nervous system, skeletal muscle, cardiac muscle, or even (less long-lived) hepatocytes. In this situation, as long as the donated DNA can be stabilised, expression will last for the lifetime of the cell. Integration into host cell DNA is not required if the transferred DNA can be stabilised episomally. There are now examples of clear-cut clinical successes using both of these strategies. For example, both ADA-SCID [2], and X-linked adrenoleukodystrophy, a rare neurodegenerative disorder presenting in childhood [4] have been treated using retroviral or lentiviral transduction of haematopoietic stem cells (HSCs) in affected children. In the first example, the immune system was reconstituted, and in the second the neurodegenerative process was arrested. For the strategy of transducing long-lived post-mitotic cells, the most convincing successes have been in the setting of Leber’s congenital amaurosis, a retinal degenerative disorder that results in blindness at an early age. Subretinal injection of an adeno-associated viral (AAV) vector expressing the mutant protein (RPE65) in affected adults and children has resulted in improvement in measures of retinal and visual function in > 25 affected individuals [5–7], with trials continuing. Currently, gene therapy for genetic diseases is coalescing around these two approaches, with use of lentiviral vectors to transduce HSCs and other stem cells, and use of AAV to transduce long-lived post-mitotic cell types for which it has tropism, including cells in the CNS, skeletal muscle, cardiac muscle and liver. The former strategy is nearly always ex vivo, with harvesting of bone marrow, isolation of HSCs, transduction in the laboratory, and re-infusion of the genetically modified autologous cells. The second strategy is nearly always in vivo, since the target cells of interest do not lend themselves well to removal and reimplantation. Not surprisingly, these two different approaches have differing profiles of toxicities and safety concerns: integrating vectors carry the potential for insertional mutagenesis [8], and AAV vectors, since they are administered in vivo, must confront the human immune response at close quarters, and thus carry greater risk of triggering responses that may impede long-lasting expression [9,10]. Other immune responses can occur related to the transgene product itself [11], and there can be risks or complications related to the route of delivery, but at a first pass level, the two approaches sort based on the types of complications that investigators have needed to address to make clinical treatment a reality (Table 1).

Table 1.   Potential complications of gene therapy
Immune response and associated toxicity: AAV vectors, adenoviral vectors
Genotoxicity and insertional mutagenesis: integrating vectors
Phenotoxicity: overexpression or ectopic expression of the trangene
Gene silencing
Vertical transmission
Horizontal transmission

Gene therapy for haemophilia: the first round of trials paves the way for the second

From today’s vantage point of 20 years’ experience in clinical gene therapy, some of the early attempts at gene therapy for haemophilia seem somewhat naïve, but nearly all of them have given rise to more sophisticated versions under investigation today (Table 2). Gene therapy as an approach has clear advantages for the treatment of haemophilia, since the natural history of mild and moderately severe haemophilia teaches that continuous expression of clotting factor, even at modest levels, is likely to prevent all life-threatening bleeds, and most other bleeds as well [12]. Haemophilia also has clear advantages as a model system for solving problems related to the application of gene therapy to genetic disease: the endpoints are clear and easy to measure, the levels of the transgene product do not need to be tightly regulated, since any level from 2% of normal to 100% is likely to improve symptoms, there is latitude in choice of target tissue as long as the transgene product can be secreted into the circulation, and there are small [13–16] and large [17–19] animal models of the disease that facilitate screening of strategies for efficacy before they move to clinical trials. These advantages as a model system, and the large size of the haemophilia market, motivated early interest from investigators and biotechnology companies in gene therapy approaches. The first round of clinical trials of gene therapy for haemophilia (Table 2) were all first-in-class, that is the first example of the delivery of a specific vector to a particular target tissue, and were generally safe, that is there were no serious adverse events associated with these trials, but none of these resulted in long-term expression of clotting factor at therapeutic levels [20,21]. Certain shortcomings of these trials are clearer now than they were a decade ago when they were initiated, but results from the early gene therapy clinical studies suggested questions that were subsequently addressed in laboratory experiments that clarified aspects of the work. For example, it was clear that retroviral vectors require a dividing target cell, since the vector cannot reach the nucleus until the nuclear membrane disassembles during mitosis. Thus intravenous infusion of a retroviral vector would be expected to result in transduction of rapidly dividing cells such as haematopoietic cells and cells lining the gastrointestinal tract, if indeed the vector survived intact in the circulation. The clinical trial of a retroviral vector for haemophilia A failed to show sustained expression, although vector DNA was found in WBCs at late time points after transduction [22]. Subsequent work by VandenDriessche et al. [23] and by Xu L et al. [24] showed that intravenous infusion of retroviral vectors into neonatal animals could result in transduction of the liver, since the hepatocytes in a growing animal are rapidly dividing.

Table 2.   First and second generation gene therapy strategies
 First generation trialsSecond generation trials or related pre-clinical strategies
VectorTarget tissueReferenceVectorTarget TissueReference
  1. Gray shaded: clinical trial; blue shade: pre-clinical strategies.

Haemophilia APlasmidAutologous fibroblastsRoth et al. [25]LentiviralEx vivo transduction of haematopoietic stem cellsYarovoi et al. [75]
Shi et al. [76]
RetroviralIntravenousPowell et al. [22]
Haemophilia AHelper-dependent adenoviral vector Helper-dependent adenoviral vectorLiverBrunetti-Pierri et al. [91]
Haemophilia BAAV-2MuscleKay et al. [42]AAVSkeletal muscle by intravascular approachArruda et al. [40]
Haemophilia BAAV-2Liver via hepatic artery infusionManno et al. [9]AAV2Liver via hepatic artery infusion
AAV8Liver via IV infusion

In another trial, sponsored by Transkaryotic Therapies, autologous fibroblasts were transduced with a plasmid expressing B domain-deleted (BDD) F.VIII, and after selection, a single clone expressing at high levels was identified, and cells were expanded and transplanted onto the omentum [25]. At the higher cell doses, F.VIII expression was detected at least transiently in some subjects, but the cumbersome nature of this ex vivo administration route diminished enthusiasm for moving it forward. (It should be noted though that selection of clones for those least likely to undergo transformation in a battery of in vitro assays appears prescient in retrospect). Finally, a single subject enrolled on a trial of a fully-deleted adenoviral vector, sponsored by Genstar, experienced self-limited thrombocytopenia and transaminase elevation. The brisk immune responses triggered by adenoviral vectors have reduced enthusiasm for using these in the setting of genetic disease where long-term expression is critical. Each of these strategies has a descendant in on-going work; for example, in terms of integrating vectors, lentiviral vectors (which do NOT require a dividing target cell) are now being used to transduce haematopoietic stem cells, to generate platelets that express F.VIII [26–28]. This type of ex vivo vector administration can also be considered an heir to the trial that used genetically modified autologous fibroblasts to express F.VIII. Advances in vector manufacturing and in routes of delivery to liver have allowed Brunetti-Pierri et al. [29,30] to achieve longer-term expression of F.VIII in large animal models using fully deleted adenoviral vectors.

A distinctive feature of the initial gene therapy trials for haemophilia B was that, in contrast to the studies for haemophilia A, the former were based on extensive pre-clinical data in the only large animal model of haemophilia, the dog. Many of the pre-clinical studies in haemophilia A were undertaken before the canine F8 cDNA sequence was available [31], and there was reluctance on the part of the small biotechnology companies sponsoring the studies to go back and rework the pre-clinical studies.

The initial AAV-F.IX studies are of interest since they are the only ones of the original group that have moved forward in the current period with only modest changes. The first study was based on administration of an AAV2 vector to skeletal muscle. Long-term expression of a transgene from an AAV vector in an immunocompetent animal had first been described following administration of vector to skeletal muscle [32,33], forming a logical starting point for these studies. These initial descriptions were an exciting breakthrough at the time, since, in contrast to retroviral vectors, AAV vectors did not require a dividing target cell, and could effect long-term expression of the donated gene without requiring integration into the host cell genome. Using an AAV vector expressing human F.IX in an immunodeficient mouse (to prevent immune responses against the human protein), Herzog et al. [34] showed long-term expression of F.IX at levels of 5%–7% of normal; this was followed by similar studies using an AAV vector expressing canine F.IX in the haemophilia B dog model [35] that showed sustained expression of canine F.IX at levels of approximately 1%–2% of normal. This seemed a straightforward procedure to move to the clinic, and in fact was the first clinical trial of AAV vectors administered to skeletal muscle. Findings that were emerging from simultaneous on-going large animal studies however, shaped the clinical trial; for example, IM administration of the vector to a haemophilic dog model with a small deletion resulting in an early stop codon [19] resulted in inhibitor formation [36] so the trial had to be limited to individuals with disease due to missense mutations. A second study in haemophilic dogs showed that even in dogs with disease due to a missense mutation, escalation of the vector dose, particularly the dose administered to a single site, increased the risk of inhibitor formation [37]. This finding drove a requirement that the dose/site be limited to 1.5 × 1012 vg, which meant that dose escalation required the use of ever-increasing numbers of injections. Thus even at a dose that was 5-fold lower than that predicted to be therapeutic based on dog studies, approximately 100 injections were required for an average adult male; escalation to the predicted therapeutic dose did not seem practical given these limitations. A ‘quick fix’ through the use of a serotype with greater tropism for skeletal muscle [38] consistently resulted in inhibitor formation even in missense mutation dogs, so this did not solve the problem. Subsequent work has focused on intravascular delivery techniques that allow delivery of vector to much larger areas of muscle, with resulting higher circulating F.IX levels [39,40]. However this approach also required short-term immunosuppression to block inhibitor formation; this technique for vector delivery has been explored most extensively in animal models of muscular dystrophy, but has not yet been applied clinically [41].

Despite the limitations of the AAV-F.IX muscle trial, it was clear from biopsies of injected subjects that transduction had occurred, since Southern blots on DNA extracted from muscle tissue were positive for the donated DNA, and immunohistochemistry showed F.IX being expressed in skeletal muscle [42,43]. Moreover there had been no safety signals following administration of even fairly large doses of vector (2 × 1012 vg kg−1) to adult men with severe haemophilia B. Pre-clinical studies in the haemophilic dog model suggested at least two important advantages for a liver-directed approach, first, that F.IX was much more efficiently secreted into the circulation from liver than from skeletal muscle, so that administration of similar doses of vector resulted in as much as a 10-fold higher plasma F.IX level [44,45]; and second, that expression of any gene from an AAV vector introduced into hepatocytes tended to promote tolerance to the expressed gene [46]. The AAV2-F.IX liver trial thus commenced in August 2001 with high hopes for success. Again, this was the first trial in which an AAV vector had been introduced into liver, so the first doses allowed by the regulatory agencies were considerably below those that had been predicted to be efficacious based on pre-clinical studies in dogs, but these were safely infused into the hepatic artery, and at the third dose tested, 2 × 1012 vg kg−1, F.IX was detected in the circulation at levels of 10%–12% beginning 2 weeks after vector infusion. In contrast to results from extensive studies in animals however, these levels began to gradually drop starting about 4 weeks after vector infusion; the decline in F.IX levels was accompanied by a rise in serum transaminases, with no change in bilirubin levels. Beginning about 5 weeks after vector infusion, the liver enzymes peaked and then gradually returned to normal without medical intervention; over the same time period, the F.IX levels continued to decline, so that 12 weeks after vector infusion, the liver function tests had returned to normal, and the F.IX level was once again < 1% [9].

The leading hypothesis to explain this disappointing turn of events was that an immune response to the vector capsid had resulted in immune-mediated destruction of the transduced hepatocytes. Since humans are among the few natural hosts for infection by wild-type AAV, which occurs normally only in the context of a helper virus infection such as adenovirus, (since AAV is naturally replication-defective and requires a helper virus in order to replicate), it was theorised that the response depended upon memory T cells that had been primed to AAV capsid proteins years earlier, and that were found in humans but not other species. The clinical protocol was amended to allow collection of peripheral blood mononuclear cells (PBMCs) from subjects infused with vector; when these cells were interrogated by IFN-γ ELISpots in a subsequently infused patient, a response to capsid, but not to F.IX was documented, with an initial CD4+ response followed by a rise in the population of capsid-specific CD8+ T cells. The kinetics of the rise and fall of this population of cells corresponded to the rise and fall in serum transaminase levels [47,48].

On-going trials of gene therapy for haemophilia B: results evolve rapidly

The results in the first AAV liver trial framed ensuing efforts; they demonstrated that AAV vectors could transduce human liver and lead to therapeutic levels of F.IX expression, but that long-term expression was abrogated by an immune response to vector capsid. It should be stressed that this latter statement was a hypothesis developed on evidence from only a few subjects, and that multiple other hypotheses were advanced, including that the response was actually to peptides derived from alternate open reading frames of the wild-type F.IX sequence [49,50]; or that the immune response could be avoided by the use of alternate serotypes of AAV that were postulated to transduce dendritic cells poorly and thus fail to trigger the responses seen with AAV2 [51]. Nonetheless on the assumption that the initial hypothesis was correct, two subsequent trials were proposed; the first used the same AAV2-F.IX vector, but now co-administered with a short course of immunosuppression consisting of mycophenolate mofetil (MMF) and rapamycin [52]. This regimen had been tested in non-human primates in conjunction with an AAV vector and did not result in toxicity or alteration in the characteristics of vector transduction [53]. This stood in contrast to results in another regimen where use of MMF, rapamycin and daclizumab together broke tolerance to human F.IX in the non-human primate model [53]. This regimen has been tested in one human subject, who showed no evidence of transaminase elevation despite a vector dose of 1.2 × 1012 vg kg−1. Subsequent experience with co-administration of an AAV vector (into skeletal muscle) with a short course of immunosuppression (12 weeks of MMF and cyclosporine) in patients with inherited lipoprotein lipase deficiency demonstrated that this manoeuvre was safe in over 15 subjects receiving this regimen [54].

The other on-going haemophilia B trial postulated that administration of a more efficient vector could result in F.IX expression at doses sufficiently low that the immune response would be avoided altogether. This seemed a reasonable hypothesis since there had been no evidence of immune responses to vector (as evidenced by transaminase elevation) at the lower doses tested in the first AAV2-F.IX trial (8 × 1010 vg kg−1), and 2/3 subjects infused at 4 × 1011 vg kg−1 also showed no detectable clinical signs of a T cell response to capsid. To generate a more efficient AAV vector, Nathwani and colleagues introduced three modifications to the original vector; these included switching to serotype 8, which has a tropism for liver [55]; use of a self-complementary vector design, which obviates the need for second strand synthesis and leads to more efficient expression [56]; and use of a codon-optimised vector to enhance translational efficiency [57]. This trial commenced in London in March 2010 and as presented at the American Society of Hematology meetings in December 2010, has led to long-term expression of F.IX in two of the first four subjects infused [58]. After vector infusion it was subsequently shown that the two subjects who failed to express initially had low titre neutralising antibodies to AAV8 that had not been detected on the initial (less sensitive) screening assay. A more sensitive assay for neutralising antibodies, now in place, should allow investigators to select those subjects who are likely to benefit from the vector infusion, but this does raise the question of how to effect gene transfer in the 20%–30% of individuals [59] with detectable titres of neutralising antibodies to AAV8. Both plasmapheresis and pharmacologic approaches to decrease titres of circulating IgG have been considered.

Analysis of T cell responses to AAV8 capsid after infusion has shown that these are detectable in PBMCs in the second dose cohort, but not in the lower dose cohort, and are measureable in the IFN-γ ELISpot assay at levels at least as high as those seen in the earlier AAV2-F.IX liver trial [9,60]. This suggests that continued dose escalation may result in detectable immune responses as seen in the AAV2-F.IX trial; clearly the most important question to be answered as this trial progresses will be determination of the highest dose that can be safely administered. This question may have greater relevance for F.VIII than for F.IX however, since levels as high as 4% have been achieved in a dose that has been administered without incident [61]. Pre-clinical studies suggest that higher doses may be required to achieve therapeutic levels of F.VIII [62]; however, for both F.VIII and F.IX, the use of higher-specific activity variants [63] or of codon-optimised constructs [64] may allow attainment of therapeutic levels at vector doses that are well-tolerated clinically.

Although these are early days, the initial results from the second AAV-F.IX liver trial suggest that it will be possible to achieve long-term expression of F.IX in subjects with severe haemophilia B, and attention is now turning to issues such as how to achieve efficient transduction in those with pre-existing antibodies to AAV [59], and whether there will be late side effects related to the low but finite level of integration into the genome that occurs with AAV administration [65,66].

Will there be gene therapy approaches for those with severe liver disease?

The prevalence of infection with hepatitis B, hepatitis C, or both, is high in the adult haemophilia population, due to viral contamination of plasma-derived concentrates that were used prior to 1985 [67]. The trials of AAV-F.IX have included subjects with a history of hepatitis, but have required that such subjects undergo liver biopsy prior to vector administration and have excluded subjects with evidence of moderate or marked fibrosis or inflammation (Metavir score of 3 or greater) [9,68]. Given that vector integration into the genome is documented to occur [66,69,70], and that an increase in hepatocellular carcinomas (HCC) has been described in at least one mouse model following AAV vector administration to liver [65,71], it will be important to address any additional risk of HCC from vector in a population already at risk for this complication. Thus several other gene therapy strategies have been explored for the subset of haemophilia patients for whom liver-directed gene therapy is not likely to be an option. There are promising pre-clinical studies that use either skeletal muscle or haematopoietic cells as targets and these will be briefly discussed.

AAV vectors readily transduce skeletal muscle, and long-term expression, for periods of up to 10 years, has been documented in muscle biopsies from individuals who have received intramuscular injections of AAV vectors [72,73]. In the initial clinical trials of AAV-F.IX injected into skeletal muscle, circulating levels of F.IX were generally subtherapeutic, that is < 1% after administration of doses up to 2 × 1012 vg kg−1 [42,43], and switching to serotypes with higher levels of expression in muscle resulted in inhibitor formation in haemophilic dog models [38], consistent with the notion that high local levels of expression predisposed to inhibitor formation [36]. To achieve distribution of vector over a wider area of muscle, we have used an intravascular delivery technique that was initially developed as a route of administration for use in muscular dystrophy [39,40], where it is critical to transduce extensive areas of muscle. The technique uses intravascular hydrostatic pressure to extravasate vector across the endothelium and into the interstitium; briefly, a tourniquet is placed at the groin, and vector is infused rapidly under increased hydrostatic pressure (approximately 300 mm Hg, achieved using a sphygnomanometer placed around the bag containing vector solution) via the saphenous vein in a large volume of saline (20 mL kg−1). Doses of AAV2-F.IX of 3 × 1012 vg kg−1 resulted in sustained circulating levels of 1.5%–5% in haemophilia B dogs, over periods of observation of up to 5 years. Among eight dogs treated by this method, the number of bleeds requiring treatment over a cumulative 392 months of observation was reduced from an expected number of 180 (the dogs bleed on average 5.5 times per year [74]) to an observed number of 7 [40], suggesting an excellent phenotypic improvement following administration via this route. It should be noted that this route of administration did require short-term administration of immunosuppression to prevent inhibitor formation. This route of administration has been used successfully for vector administration in large animals, but has not yet been used in humans. Although it has not been formally tested for F.VIII, it seems unlikely that a molecule as large as F.VIII could be efficiently secreted from a site of synthesis in skeletal muscle.

Another potential alternative target cell for vector administration in those with liver disease is the haematopoietic stem cell. Yarovoi et al. [75] first proposed introduction of an integrating vector expressing F.VIII under the control of a platelet-specific promoter. The hypothesis was that F.VIII would be packaged and stored in the platelet a-granules, and released at the site of a bleed when platelets were activated. It was quickly appreciated that this approach might also effect haemostasis in animals with inhibitors, since the F.VIII would be protected from antibodies as it travelled through the circulation within the platelet. The efficacy of this approach has since been substantiated by two independent groups using both transgenic and lentiviral approaches in mice [28,76,77], and has also been shown to improve haemostasis in haemophilic mice with inhibitors [26,78]. Recent data suggest that the clots formed from platelet-released F.VIII may be more likely to embolise [79]. Approaches that rely on lentiviral vectors carry the risk of insertional mutagenesis, a safety concern that would need to be addressed if this strategy moves forward clinically. Sadelain and colleagues have evaluated a similar approach in which F.IX is expressed in RBCs [80]. Experiments in mice demonstrate that this approach can mediate circulating F.IX levels in the range of 5%–7% of normal, and that it seems to promote tolerance to the F.IX transgene product. Their data demonstrate that F.IX expressed in erythrocyte precursors is biologically active. Of course, utilisation of either of these approaches will require preparative regimens to ensure engraftment of the transduced autologous cells. The degree of myeloablation required for engraftment of genetically modified autologous cells is still under investigation and will likely vary according to the disease setting.

Can gene therapy be used to treat patients with inhibitors?

For protein-based therapeutics, the two cornerstones of inhibitor treatment are inhibitor eradication by immune tolerance induction (ITI) with F.VIII or F.IX, and treatment of bleeding episodes through the use of bypassing agents. With respect to the latter, we hypothesised based on the success in animal models of AAV-mediated gene transfer to liver for F.IX and F.VIII that expression of F.VIIa at therapeutic levels could also be achieved. The key advance was to generate an engineered F.VIIa molecule that could be secreted as the activated form; we tested several novel variants and were able to show that a variant in which a PACE-furin cleavage site (RKRRKR) had been installed between Arg152 and Ile153 resulted in highly efficient intracellular processing with nearly complete conversion to the two-chain, activated form [81] in a variety of cell lines and in mouse liver. Long-term expression of this engineered F.VIIa resulted in correction of the bleeding diathesis in mice with haemophilia A or B [81]; however, subsequent studies demonstrated that expression of F.VIIa at high levels (average 3150 ng mL−1) resulted in premature mortality in mice, due primarily to thrombi formation in lungs and heart. Such findings were not observed in mice expressing F.VIIa at lower levels (average 1150 ng mL−1); lifespan in these mice was normal with no evidence of histopathologic abnormalities even at 18 months of age [82]. In subsequent studies, we extended the same approach to dogs with haemophilia A or B, and showed complete prevention of spontaneous bleeds in four treated dogs, over a cumulative 79 months of observation [83]. Based on historical data, the expected number of bleeding episodes requiring treatment in these animals over this time period was 36 [74]. The prevention of spontaneous bleeds was perhaps predictable in three of these animals, where circulating levels of F.VIIa ranged from 1.3 to 2.6 mg mL−1, following infusion into the portal vein of doses of AAV8-F.VIIa ranging from 6.25 to 12.5 × 1013 vg kg−1. Perhaps the most surprising finding in the study was that a haemophilic dog dosed at a lower level (2.0 × 1013 vg kg−1), that did not achieve detectable circulating levels of F.VIIa, nonetheless also showed complete prevention of spontaneous bleeding episodes over 34 months of observation, suggesting that even very low levels of continuously expressed F.VIIa may result in amelioration of bleeding episodes in the large animal model of haemophilia. It is unclear why continuously expressed levels of F.VIIa may be more effective than the peaks and troughs that characterise recombinant protein infusion therapy. It will be important to determine whether this is a reproducible finding in haemophilic dogs; if so it may be worth investigating in a clinical study in patients with inhibitors that cannot be eradicated. To this end, it may be critical to use high-specific activity variants of F.VIIa, in order to reduce the dose of vector required [84]. It will also be important to demonstrate that the approach is efficacious in haemophilic dogs with inhibitors, as a precursor to studies in humans.

Experience with the use of gene therapy to eradicate inhibitors is more limited. Certainly, a considerable body of evidence suggests that expression of any transgene from an AAV vector in hepatocytes tends to promote tolerance to the transgene product [46,85]. However, in most of these studies, vector was administered prior to the development of an immune response to the transgene product, whereas the opposite is true in individuals already diagnosed with inhibitors. Earlier studies with F.VIII-expressing retroviral vectors introduced into haematopoietic stem cells had shown that haemophilia A mice that had received bone marrow transplants with cells transduced with a vector expressing F.VIII were less likely to form inhibitors after intravenous exposure to F.VIII than animals that had received non-transduced cells, but again, this has limited clinical relevance [86]. In a recent study, Finn and colleagues showed that inhibitors in dogs with severe haemophilia A could be eradicated using an AAV-mediated gene therapy approach that essentially carries out immune tolerance induction with a single vector injection [87]. In this study, haemophilia A dogs with inhibitors (peak titres ranging from 3.6 to 12 BU) received high doses (2 × 1013 vg kg−1) of each of two vectors, one expressing the F.VIII light chain and the other the F.VIII heavy chain. For three of the dogs, where onset of inhibitor was relatively recent (≤ 8 months), inhibitor eradication occurred within 4–5 weeks, and was followed by detectable expression of canine F.VIII (cFVIII) activity in the range of 1.5%–8%. In the fourth animal, which was from an inhibitor-prone strain of haemophilia A dogs at the colony at Queen’s University in Kingston, Ontario, the duration of the inhibitor before the gene therapy attempt was much longer, 24 months. Infusion of vector resulted initially in a dramatic rise in the inhibitor titre to 216 BU (from a baseline of 3.5 BU), followed by a gradual fall to approximately 0.8–1 BU 80 weeks after vector administration. At that point, canine F.VIII light chain antigen could be detected in the circulation at a level of 12 ng mL−1, but there was still no cF.VIII activity. These initial studies would suggest that, as is the case for protein-based immune tolerance induction, the earlier the treatment is initiated after inhibitor development, the more likely it is to result in eradication of the inhibitor. This approach warrants further investigation in light of the limited options for individuals with inhibitors who fail ITI or who cannot tolerate a prolonged course of IV infusion therapy.

Genome editing: the final frontier? In vivo gene correction as an approach to treating genetic disease

The approaches discussed thus far in this review are all based on transfer of a normal copy of the F8 or F9 gene into an individual bearing a mutation in one of these genes. A longstanding goal of the field of gene therapy though is the ability to correct a mutation in situ rather than transfer in an additional gene. The advantages of gene correction are several, but one of the most appealing is that it reconstitutes the normal gene under the control of the endogenous regulatory signals. Gene correction is based on gene targeting, the therapeutic utility of which has traditionally been limited because of low homologous recombination rates in most cell types. However, recent work with a novel class of fusion proteins, zinc finger nucleases (ZFNs), has been shown to increase targeting efficiency to clinically meaningful levels in haematopoietic cells [88]. The zinc finger moieties in the molecule recognise a specific DNA sequence, and the endonuclease domains of the molecules dimerise to introduce a double-strand break (DSB) at a specific locus (Fig. 1). In the presence of a donor template with arms of homology to the regions flanking the DSB, the cellular mechanisms that mediate homology-directed repair (HDR) will effect repair of the break using the (wild-type) donor template, to restore a normal sequence at the site of the break. Thus one can envision a therapeutic strategy in which synthetic ZFNs that recognise a site in the 5′ portion of the F8 or F9 gene are introduced into the liver along with a donor template that carries the downstream cDNA sequences; at sites of template-driven HDR, a functional gene will be restored under the control of the endogenous promoter. We have recently described the use of this strategy to restore haemostasis in a mouse model of haemophilia B [89]. The major safety risk of this approach is off-target cleavage by the ZFNs, that is introduction of double-stranded breaks at other sites in the genome. Further studies will be required to assess this risk before clinical application in the liver, but ZFN-mediated modification of human target cells is already in testing in clinical studies [90].

Figure 1.

 (A) Zinc finger nucleases, designed so that the zinc finger domains recognize specific sites near a mutation, and introduce a double-strand break in the DNA, which is then repaired by the cellular process of homology-directed repair, using as template a wild-type sequence introduced into the cell (adapted from Nature Vol 435[2]:577). (B) Human F9 genomic map, indicating site of a potential zinc finger nuclease target. c. Level of target gene disruption in Hep 3B cells (human hepatoma cell line) transfected with plasmids encoding the indicated ZFNs targeting F9 at two different loci. Lower migrating products (arrows) are a direct measure of ZFN-mediated gene disruption, the substrate for homology-directed repair.

Disclosure of Conflict of Interest

Dr. High has received consulting income from Bristol-Myers-Squibb, Genzyme, Forrest Research Laboratories, Shire, Biogen-Idec, Tacere, Inc., bluebird bio, and Hoffmann-LaRoche, Inc. She has ongoing collaborations with Sangamo Biosciences, and PTC Therapeutics.