The Jeremiah Metzger Lecture: Gene Therapy for Inherited Disorders: From Christmas Disease to Leber's Amaurosis

Abstract

This paper will focus on recent developments in the field of gene therapy for inherited disorders. From a historical perspective, this Metzger lecture is a follow-on to one presented by Dr. William Kelley in 1987, entitled “Current Status of Human Gene Therapy” (Transactions Am Clin. Climatol. Assoc. 99:152–169) (¹). In 1987, gene transfer studies in human subjects were yet to be undertaken; the first clinical studies, infusion of genetically modified autologous T cells into two young girls with ADA-SCID, would not take place until 1990 (²). Today's lecture will summarize progress since that time in one area, that of in vivo gene transfer for genetic disease. I will describe progress in two areas, gene therapy for the bleeding disorder hemophilia B, and for a subset of retinal degenerative disorders termed Leber's congenital amaurosis, due to mutations in the gene encoding retinal pigment epithelium-specific 65 kilodalton protein (RPE65). This lecture will demonstrate the interconnected nature of progress in these two areas, as careful delineation of the obstacles in hemophilia led to the realization that success could be achieved in Leber's.

Genetic disease accounts for as many as 50% of hospitalizations in childrens' hospitals, and up to 10% of admissions in adult hospitals. Despite dramatic improvements over the last three decades in the ability to diagnose genetic disorders, progress in therapeutics has been considerably more limited, and options for most conditions are few. For some metabolic disorders such as phenylketonuria or homocystinuria, disease can be managed or ameliorated with dietary restrictions or supplements, and for diseases like cystic fibrosis, chest physiotherapy and the availability of a range of antibiotics have slowed the progression of pulmonary disease. For a few diseases, such as hemophilia, Gaucher's disease, and Fabry's disease, enzyme replacement therapy is available as a recombinant protein for periodic infusion. And in some cases, when end organ damage becomes life-threatening, organ transplantation can cure the disease, although one can argue that the accompanying need for chronic immunosuppression merely exchanges one chronic condition for another.

A longstanding goal of those who study genetic disease has been the development of genetic therapies for these disorders, so that patients would have access to a long-term cure rather than simply amelioration or management of symptoms. Toward this goal, both stem cell therapies and gene transfer therapeutics are now under development. Within the broad category of gene therapies, multiple strategies are under investigation (Figure 1). The most well-known strategy and that with which there is the most clinical experience is what is commonly referred to as gene addition therapy, in which a normal copy of a defective gene is delivered to a physiologically relevant target cell, using a gene delivery vehicle or vector (Figure 1A). From this brief description two important facts can be immediately deduced—first that this type of gene therapy will be no better than the vectors, and second that efficiency of delivery of the donated gene to the target cell will be a key determinant of outcome. Both of these conclusions have been abundantly borne out in clinical studies. A second strategy that is also under investigation is in situ gene correction therapy. One method for this is the use of vector-delivered zinc finger nucleases, which are engineered to cleave DNA at a specific site near a genetic mutation (^{3, 4}). Once a double-strand break has been introduced by the nucleases, cellular repair mechanisms including homology-directed repair are activated. Vector-mediated introduction of a wild-type sequence, which can serve as a template for homology-directed repair, results in in situ correction of the mutant gene, which is now under the control of the normal regulatory sequences (Figure 1B). Early applications of this technology to effect sequence changes in human lymphocytes have been described (^{5, 6}). In yet another strategy, small molecule drugs that facilitate read-through of stop codons have been used to treat genetic diseases caused by nonsense mutations (⁷) (Fig. 1C). These oral drugs are now in testing for cystic fibrosis and Duchenne muscular dystrophy (⁸); although they can only correct disease in a subset of patients, orally available therapies would be a welcome improvement over intravenous delivery of missing enzymes, or as is the case for cystic fibrosis, no treatment.

Fig. 1.

Novel genetic therapies under investigation for monogenic disorders. Panel A: In gene addition therapy, a normal copy of a defective gene is delivered to an appropriate target cell. Panel B: In gene correction therapy, a site-specific double-stranded break is introduced by zinc finger nucleases (X's), and the cellular mechanisms for homology-directed repair effect repair of the break using a wild-type template delivered by gene transfer. Panel C: Bypass therapy promotes read-through of premature stop codons and thus allows synthesis of full-length protein.

Gene addition therapy for genetic disorders requires long-term expression of the donated gene. There are two basic strategies for achieving this: the first is to use an integrating vector in a stem cell. In this setting, since the donated gene integrates into the host cell chromosomal material, it is passed to each daughter cell, and long-term expression ensues. The risk of this strategy, demonstrated in the trials of retroviral transduction of hematopoietic stem cells for X-linked SCID, is that the donated DNA will integrate at an untoward site and promote the development of malignancy (⁹). The second strategy, which we have pursued, is to introduce the donated DNA into a long-lived post-mitotic cell, such as cells in the central nervous system, hepatocytes, cardiac muscle, or skeletal muscle (The latter, in particular, are extremely long-lived and may, in fact, last the lifetime of the animal). In this case, there is no need for an integrating vector, as long as the donated DNA can be stabilized episomally, and thus problems related to vector integration are avoided. One final distinction that needs to be highlighted is the difference between ex vivo and in vivo gene transfer. In ex vivo gene transfer, commonly used when hematopoietic stem cells are the targets, cells are harvested from the patient, genetically modified in the laboratory, then returned to the patient. Although a few cell types lend themselves well to this procedure, many cell types of interest do not, and for these, in vivo gene transfer, in which the vector is directly introduced into the patient, is the only feasible method of accessing the cells. Ex vivo gene transfer is a labor-intensive procedure that employs many of the methodologies developed for bone marrow transplantation. In vivo gene transfer is a long-term goal of the field of gene therapy, since it achieves the goal of “gene medicine in a bottle” (i.e. an off-the-shelf pharmacologic agent that could theoretically treat any affected patient).

For in vivo gene transfer, development of gene therapies has followed a paradigm in which efficient gene transfer in tissue culture is followed by attempts to scale up to murine models of disease. If these are successful, the next step is scale-up into large animal models of disease. These have proven important in in vivo gene transfer for several reasons: first, in the early days of gene transfer, production of adequate quantities of vector to treat an organism larger than a mouse was frequently a challenge, and a large animal ranging in size from 5–20 kg was a reasonably stringent test. Second, large animals more nearly recapitulate many aspects of human anatomy and physiology, so that they constitute more accurate models for gene delivery methods and for vector biodistribution than do mice. Finally, because these animals tend to be outbred, their immune responses to both vector and transgene product may more accurately reflect the range of responses seen in human subjects, compared to inbred strains of laboratory mice. As will become clear, however, even extensive testing in large animal models may fail to reflect every aspect of testing in human subjects, which explains why drugs cannot be licensed without clinical trials.

Long-term cures of both hemophilia B and Leber's congenital amaurosis due to RPE65 mutations have been achieved using gene transfer in canine models of these diseases (^10–13). In both cases the vector used was a recombinant adeno-associated viral (AAV) vector (Figure 2). Engineered from a naturally occurring parvovirus with a small (4.7 kb) single- stranded DNA genome, wild-type adeno-associated virus (AAV) is naturally replication-defective (it requires a helper virus such as adenovirus to replicate) and non-pathogenic (i.e. not associated with any known illness). The recombinant vector is fully deleted of viral coding sequences, retaining only the viral inverted terminal repeats (ITRs); the therapeutic gene of interest is cloned between the two viral ITRs. Work by a number of groups in the mid-1990s demonstrated that the recombinant vector had a tropism for certain long-lived post-mitotic cell types, including cells in the central nervous system, hepatocytes, and skeletal and cardiac muscle; that the vector DNA was stabilized in a predominantly non-integrated form; and that, in contrast to results with adenoviral vectors, long-term expression could be achieved in immunocompetent animals, suggesting that the vector triggered very little immune response (^14–17).

Fig. 2.

Wild-type and recombinant adeno-associated virus. Panel A: The wild-type virus is a member of the parvovirus family, and contains a single-stranded DNA genome of ∼4.7 kb, encoding genes required for replication (rep genes) and those encoding the viral capsid (cap genes). Panel B: The genome of a recombinant virion contains the therapeutic gene of interest under the control of a promoter, usually tissue-specific, and a polyadenylation signal. It is fully deleted of viral coding sequences; the rep and cap genes, required for replication during the manufacturing process, are supplied in trans by an additional plasmid.

For a number of reasons, hemophilia has been a longstanding target of interest in the development of gene transfer therapeutics. The bleeding disorders, hemophilia A and B, are due to mutations in the genes encoding Factor VIII and Factor IX respectively. These factors are the cofactor and the enzyme that catalyze the conversion of Factor X to activated Factor X. A mutation in the F8 gene occurs in ∼1 in 5000 births, while mutations in the F9 gene occur in ∼1 in 35,000 births, making hemophilia one of the more common inherited disorders. The largest category of patients are those with severe disease characterized by circulating factor levels <1% normal. These individuals suffer spontaneous bleeding episodes, mostly in joints and soft tissues, but bleeds into other critical closed spaces, such as the intracranial space, can be rapidly fatal. Individuals with levels in the range of 1–5% have a more moderate disease phenotype, and those with levels >5% are mildly affected, generally do not suffer bleeds unless confronted with a hemostatic challenge such as surgery or trauma, and are free of the life-threatening bleeds that characterize the severe phenotype (^{18, 19}). This observation, based on the natural history of the disease, makes it clear that even a modest increase in circulating levels of clotting factor, into the range of a few percent, can greatly improve the clinical course of the disease. The disease is well-characterized at the molecular level, and both the natural history of the disease and experience with plasma-derived and recombinant protein concentrates demonstrate that raising circulating levels of clotting factor into the range of a few percent of normal can greatly improve the symptoms of the disease (²⁰). Moreover, since raising levels to 100% would still be within the therapeutic range, the therapeutic window is wide in this disease (i.e. there is no requirement for precise regulation of expression). Another advantage as a model for gene transfer is that biologically active clotting factors, although normally synthesized in the liver, can be made in a wide variety of tissues; for example the currently marketed recombinant clotting factors are made in baby hamster kidney cells or Chinese hamster ovary cells. This affords latitude in the choice of target tissue, an advantage for gene therapy. The existence of genetically engineered mice (^{21, 22}) and naturally occurring dog models of the disease (^23–25) facilitates testing of novel gene therapeutic strategies before moving to clinical studies; and measurement of therapeutic endpoints is a straightforward laboratory test that can be done in almost any hospital coagulation laboratory. This stands in stark contrast to diseases like cystic fibrosis, where determination of therapeutic endpoints relies on surrogate measures like transepithelial nasal potential differences, or requires years of follow-up to assess long-term effects on disease outcome.

Building on studies demonstrating long-term expression of a β-galactosidase gene following introduction of an AAV vector into murine skeletal muscle, we constructed rAAV vectors expressing Factor IX; Factor IX, rather than Factor VIII, was the logical starting point, since its shorter cDNA was easily accommodated within the size constraints of an AAV vector. (Subsequent work has extended these findings to Factor VIII, albeit with a truncated promoter and poly A signal (²⁶)). We initially demonstrated long-term expression of Factor IX after intramuscular (IM) injection of an AAV-F.IX vector in mice (²⁷), and subsequently showed we could achieve the same result by introducing the vector into the liver via portal vein injection (²⁸). In follow-on studies, we and others demonstrated that this result could be extended to the canine model of hemophilia B (^11–13), and that therapeutic levels could be achieved using either skeletal muscle or liver as the target tissue (Figure 3). Long-term follow-up of injected dogs failed to disclose any evidence of serious adverse events (^{29, 30}). These studies formed the basis for subsequent clinical trials in subjects with severe hemophilia B.

Fig. 3.

Time course of Factor IX levels following vector injection in dogs with severe hemophilia B. Injection of a recombinant AAV vector expressing canine Factor IX results in long-term expression (note time scale in weeks) of F.IX, whether vector is introduced into liver (dotted lines) or into skeletal muscle (solid line). Levels of F.IX expression are higher following delivery to liver, even at 10-fold lower vector doses, but levels in muscle (1–2%) are adequate to improve disease phenotype.

An initial question in the translation to human subjects was the choice of target tissue. When these studies were initiated, recombinant AAV vectors had been introduced into the sinuses and upper airways of patients with cystic fibrosis (³¹), but had not been injected parenterally in human subjects. For a number of reasons, we felt that intramuscular injection was a more conservative initial approach: intramuscular injections are a safe, familiar and relatively non-invasive approach compared to catheter-based delivery of vector into the hepatic artery; hepatitis B and C, which affect many adults with severe hemophilia B, would not be a contraindication to a muscle-directed approach, but might be for liver delivery; and the risk of germline transmission of vector was defined and appeared to be low for the IM delivery route based on studies in animal models (³²). However, it was also clear from animal studies that the risk of immune response to the transgene product Factor IX was higher with delivery to skeletal muscle, whereas delivery to liver appeared to promote tolerance to the transgene product (^{13, 33–35}). And based on studies in hemophilic dogs, there was clearly a dose advantage in favor of liver, on the order of 10–100 fold, probably owing to higher efficiency-secretion into the circulation from the hepatocyte. The first clinical study of AAV-F.IX in hemophilia B, which was also the first study of AAV in human skeletal muscle, began in 1999. The initial results showed that parenteral injection was safe, but that circulating levels of F.IX did not reach a therapeutic level (>1%) at the doses tested (^{36, 37}). Moreover, it proved difficult to reach the predicted therapeutic dose in humans, because of the large number of injections required to transduce adequate amounts of skeletal muscle. However, given the strong safety record of parenteral injection of AAV in these first eight subjects, it seemed reasonable to proceed to liver delivery, where it was anticipated that there would be a substantial dose advantage. Studies in hemophilic dogs had demonstrated long-term expression of Factor IX in the range of 6–8% of normal levels following injection of a dose of 1 × 10¹² vector genomes (vg)/kg into either the portal vein or the hepatic artery (¹³). The previous study of AAV-F.IX delivered by IM injection had established the safety of doses as high as 2 × 10¹² vg/kg in human subjects (³⁷); thus it was anticipated that the liver trial could reach a predicted therapeutic dose.

Before discussing the results of the AAV-F.IX liver trial, it is perhaps worthwhile to note that there had been substantial strides in AAV vector manufacture since the first studies were undertaken in large animal models, and the ability to manufacture sufficient amounts of vector of high purity and potency for human studies was well in hand. Most preparation methods still rely on transfection of HEK293 cells with 3 plasmids, one containing the therapeutic gene of interest cloned between the viral ITRs, another encoding the viral genes required for replication and encoding the viral capsid, and a third encoding the adenoviral helper genes required for replication (³⁸). Forty-eight hours after transduction, the cells are lysed and recombinant virions are purified from the cell lysate. It is worth noting that AAV is a rather inefficient virus, and the major product from the cell lysate is empty capsid (i.e. the viral capsid with no encapsulated DNA). A smaller fraction consists of vector capsid containing the therapeutic gene of interest; the other plasmids, lacking the viral ITRs, fail to package efficiently. Our purification method differs from many others in that we intentionally separate out the empty capsid, since we view it as a contaminant which is not inert in terms of the immune response (^39–41). The final product then is a recombinant virion consisting of a vector capsid composed of 3 distinct capsid proteins, containing an ∼4.5 kb sequence encoding Factor IX under the control of a liver-specific promoter (the α₁-antitrypsin promoter and the Apolipoprotein B enhancer). The amount of empty capsid in these preparations is <2%.

The liver study began in 2001, and was also the first study in which AAV was introduced into the liver; the trial was designed as a dose-escalation study, with a starting dose of 8 × 10¹⁰ vg/kg. This initiation at a low dose was required because of the lack of experience with administration of AAV to the liver in human subjects. Subjects were adult males with severe hemophilia B and no evidence of advanced liver disease (Metavir score ≤2) (⁴²) based on liver biopsy if they were HCV RNA viral load-positive. The first two doses were sub-therapeutic based on resulting Factor IX levels, but appeared to be safe with no evidence of side effects or adverse events. The first subject infused in the third dose cohort (2 × 10¹² vg/kg) showed therapeutic levels of Factor IX for the first 4 weeks after vector infusion, with circulating levels of 10–12% in a subject with a baseline level of <1%. At week 5, however, the F.IX level was measured at 6%, at week 6, 2.7%, and by week 10 was down to the baseline of <1%. This inexorable decline in Factor IX levels was accompanied by another finding that had never been seen in animals, namely a rise in liver transaminases that first appeared 4 weeks after vector infusion, then peaked at 5 weeks with an ALT of 532 IU/L and an AST of 254, then slowly returned to normal, without medical intervention, and with a time course similar to the decline in Factor IX levels (Figure 4). During this time, the subject felt well and none of the other laboratory studies were abnormal. He was evaluated for other causes of transaminase elevation, but all studies were negative. A second subject infused at this dose (before any abnormalities had been observed in the first subject) failed to show any evidence of Factor IX expression or of transaminase elevation. Of note, the second subject had neutralizing antibodies to AAV present at baseline in a titer of 1:17 (³⁹). The likelihood is that the antibodies neutralized vector before transduction occurred, underscoring the need to choose subjects without pre-existing neutralizing antibodies to AAV.

Fig. 4.

Time course of F.IX expression following vector injection in a human subject. The dotted line shows Factor IX levels (left axis) in a subject injected at a dose of 2 × 10¹² vg/kg. Note rapid decline in circulating levels between 4 and 10 weeks post vector infusion. The solid lines denote AST and ALT levels (right axis) over the same time course. Redrawn based on Manno CS, Pierce GF, Arruda VR, et al. Successful transduction of liver in hemophilia by AAV-Factor IX and limitations imposed by the host immune response. Nat Med 2006;12(3):342–7.

This third dose cohort was extremely instructive since it demonstrated that studies in dogs had accurately predicted the required dose of vector in human subjects, and it also uncovered a phenomenon that had not been observed in pre-clinical studies in animal models, namely a decrease in Factor IX levels and an increase in transaminases that started 4 weeks after vector infusion and resolved without medical intervention. In reviewing factors that differed between humans and other animals in which AAV had been tested, we noted two: 1) most adults with severe hemophilia B have been infected with hepatitis C, so that the hepatic microenvironment might differ from that of animals with normal livers, and 2) humans are a natural host for wild-type AAV-2, and thus might harbor memory T cells to the vector capsid, not present in other species. We also considered the differential diagnosis for the transient asymptomatic transaminase elevation. Routine virologic studies had excluded causes unrelated to the gene transfer. We reasoned that direct hepatocellular injury from the vector should have resulted in transaminase elevation immediately after vector infusion, rather than 4 weeks later. Indeed, the kinetics of the response, with delayed onset of a decline in Factor IX levels and simultaneous onset of transaminase elevation, suggested an immune response targeted to hepatocytes harboring vector (Figure 5).

Fig. 5.

Working hypothesis explaining results after AAV injection in liver of human subject. Vector enters the hepatocyte through clathrin-coated pits, and is trafficked to the perinuclear region. The sequence of events surrounding escape from the endosome, vector uncoating, and translocation to the nucleus, is poorly understood. Capsid that escapes into the cytosol undergoes proteasomal processing; capsid-derived peptides may be presented via MHC Class I molecules to circulating CD8+ T cells, resulting in destruction of transduced hepatocytes, loss of F.IX expression, and release of transaminases into the circulation.

This hypothesis, the most parsimonious to account for the series of findings in the subject, was potentially testable, since the vector itself has a simple structure, consisting only of the vector capsid, and of a short segment of DNA encoding wild-type Factor IX. The capsid protein, not encoded in the vector, should be present only transiently and eventually degraded and cleared from the cell, while the DNA, after vector uncoating, should be stabilized episomally and should direct expression of wild-type Factor IX for a prolonged period. It should be underscored here that either of these, vector capsid or wild-type Factor IX, or both, could appear as a foreign protein to the hemophilic recipient. To test this hypothesis, we proposed to analyze PBMCs from the next recipient, using an IFN-γ ELISPOT assay to determine whether the immune response was directed to vector capsid, wild-type F.IX, or both. Peptide libraries spanning the AAV capsid sequence and the F.IX coding sequence were prepared and grouped into pools to be used to stimulate PBMCs collected at baseline and at a series of timepoints after vector infusion. Pools eliciting IFN-γ synthesis could be further broken down to identify reactive epitopes. Concern about the level of the transaminase elevation in the previous subject resulted in a directive from the regulatory agencies to study an additional subject at a dose of 4 × 10¹¹ vg/kg. Enthusiasm for this maneuver was limited on the part of the investigators, since the previous subjects infused at this dose had not achieved therapeutic levels of F.IX expression. Nonetheless the infusion of the next subject at this dose proved highly informative; the subject was negative for any previous exposure to hepatitis B or C, so that this could be excluded as a possible factor accounting for transaminase elevation. This subject also had non-detectable neutralizing antibodies to AAV, so that there was likely to be adequate transduction after hepatic artery infusion of vector. The results showed, as predicted, no evidence of F.IX levels >1%, but the development of modest transaminase elevation on a time course consistent with that seen in the earlier subject. Analysis of PBMCs from this subject showed that he was consistently negative for reactivity to any vector capsid or F.IX peptides at baseline, but developed reactivity to two AAV capsid epitopes after vector infusion, with a peak CD4⁺ T cell response at 2 weeks and a peak CD8⁺ T cell response at 5 weeks after vector infusion. No response to F.IX epitopes was detected (³⁹). High resolution HLA-typing on the infused subject allowed use of bioinformatics programs to predict peptides within the AAV capsid sequence most likely to bind to the subject's Class I MHC molecules; these predicted the same epitope, VPQYGYLTL, that had been identified experimentally by IFN-γ ELISPOT analysis of patient lymphocytes. These data also permitted the generation of a pentamer reagent, consisting of 5 molecules of the HLA-B*0702 molecule loaded with the peptide VPQYGYLTL, which can be used for direct determination of the number of antigen (capsid)-specific CD8⁺ T cells in the patient's circulation (⁴³). This analysis demonstrated that the population of capsid-specific CD8⁺ T cells expands after vector infusion, reaching a maximum 5 weeks later, and contracts thereafter in a time course that closely matches the rise and fall of transaminases. Further analysis of CD8⁺ T cells from this subject demonstrated that they were able to lyse peptide-loaded, HLA-matched target cells, and that they cross-reacted with the analogous peptides derived from other AAV serotypes (there are >100 AAV serotypes but sequence conservation of capsid is high, in the range of 60–100% (⁴⁴)).

This analysis raised two important questions, the answers to which will determine the outline of AAV-mediated, liver-directed gene transfer going forward. The first unsolved question is why this series of events was observed in humans but not in other species. An early hypothesis was that, since humans are natural hosts for AAV, and since natural infection can only occur in the context of a helper virus such as adenovirus which has a powerful adjuvant effect, humans likely harbor memory T cells to AAV which on vector infusion are activated. Since memory T cells require only a very low level of peptide- MHC complexes to be activated, this would account for the loss of expression in this species, but long-term expression in species without memory T cells to AAV. However, since non-human primates (NHP) are also natural hosts for some AAV serotypes, they might be expected to exhibit a similar phenomenon, but to date have not. The evidence for this is less than perfect, since most investigators screen NHPs for absence of neutralizing antibodies to AAV, to insure that transduction will occur. This may also insure that only naïve animals are tested. It is a fair criticism to note that a rigorous study of NHPs that have been previously exposed to AAV has not been conducted.

Another possible explanation may be the greater responsiveness of human T cells to T cell receptor stimulation due to the evolutionary loss of sialic acid-recognizing Ig-superfamily lectins on human T cells (⁴⁵). It is worth noting that the events described here would not be the first instance in which the human immune response was remarkably underestimated by studies in experimental animals (⁴⁶); the development of cytokine storm in six human volunteers infused with the CD28 superagonist TGN1412 (⁴⁷), which had proven safe in NHPs at 500-fold higher doses (⁴⁸), illustrates clearly the degree to which pre-clinical studies in animals may fall short of the mark in accurately modeling the human immune response.

The second critical question raised by these results is whether there is any maneuver that would allow us to avoid or block this response, so that the therapeutic result observed initially, with circulating F.IX levels of 10–12%, could persist in humans as it has in other species. One can imagine a number of potential solutions (⁴⁹). These might include altering the vector capsid to remove the immunodominant epitope(s); however, the immune system may simply target a subdominant epitope, and in addition the existence of well over 100 different MHC Class I molecules, each potentially binding a different epitope, would make this a daunting task. Some investigators have proposed that switching AAV serotypes may avoid this response, due to capsid-dependent differences in intracellular trafficking or transduction of antigen presenting cells; this will be tested in a study to be initiated in 2009, in which an AAV serotype 8 vector expressing human Factor IX will be infused into subjects with severe hemophilia B (⁵⁰). Rather than changing the capsid, one could envision a solution in which one changes the T cell response to capsid, either by tolerizing the recipient to the AAV capsid, or by using pharmacologic suppression of the immune response. A third possibility would be to introduce the vector into an immunoprivileged site; while this may be an option for some transgenes (vide infra), it is not likely to be feasible for Factor IX. If the memory T cell response is critical, it may be possible to avoid this by infusing young subjects with no previous exposure to wild-type AAV, but this is not likely to be acceptable in hemophilia, where the existence of an established treatment has resulted in a paradigm in which the safety of new treatments is evaluated first in adults. Finally it is likely to be beneficial to decrease the total antigenic capsid load entering the cell, so maneuvers such as removal of empty capsid and/or codon optimization, that increase the total amount of Factor IX synthesized per virion entering the cell, may at least diminish the immune response to the transduced hepatocyte. The use of self-complementary constructs (⁵¹) is likely to be helpful for the same reason.

From among these possibilities, we have elected to pursue a short-term course of immunosuppression (IS) administered in conjunction with AAV-F.IX vector. There is considerable experience with IS drugs and their safety record is well-established in the setting of organ transplantation (^52–54). We have shown the safety of co-admininstration of MMF and rapamycin with AAV-F.IX vector in non-human primates; these drugs do not alter the transduction characteristics or biodistribution of AAV following vector administration through the hepatic artery (⁵⁵). The critical decision in this setting is the duration of the IS regimen. Immunosuppression would be needed as long as capsid peptide-MHC complexes are displayed on the cell surface, but the kinetics of this are not established. Unpublished studies in hemophilic dogs demonstrate that vector capsid can be detected in hepatocytes as long as 11 weeks after vector infusion (KAH, unpublished studies). In a clinical study conducted in Canada, in which AAV vector was injected intramuscularly and subjects received MMF and cyclosporine simultaneously, the duration of immunosuppressive therapy was 12 weeks, but the outcomes in terms of transgene expression are not yet known. However, the co-admininstration of immunosuppressants and vector appeared to be safe in the 12 subjects injected (⁵⁶). The plan in the currently approved hemophilia trial is to begin MMF 7 days before vector infusion, and sirolimus on the day of vector infusion. The starting vector dose in this iteration of the study will be 1.2 × 10¹² vg/kg, 60% of the dose that was therapeutic in the highest dose cohort. Immunosuppression will be continued for a total of 8 weeks; if no F.IX expression is detected at that point, it will be stopped without a taper. If on the other hand F.IX is present in the circulation, the IS regimen will be continued for a total of 16 weeks, with a taper beginning at that point and continuing through week 20. If a decline in F.IX expression or elevation in liver enzymes is detected, then the taper would be stopped and IS continued for 2–4 more weeks. This trial is anticipated to begin in 2009.

As may be readily appreciated from the description of this study, the timeline stretched out as the investigators sought to understand the series of events that led to the unanticipated adverse events of transaminase elevation and loss of circulating F.IX (Figure 6). Eventually, the biotechnology company that had manufactured vector for the study decided that the timelines involved were not compatible with their survival, and they dropped active pursuit of gene therapies. This left the investigators facing the quandary of where to obtain clinical-grade AAV to pursue the studies. Fortunately, The Children's Hospital of Philadelphia, convinced of the long-term utility of gene therapy approaches for the treatment of genetic disease, agreed to fund and staff an in-house facility for the manufacture of clinical-grade recombinant AAV. Thus after a brief interval to renovate space and recruit personnel, needed project resources were again in place, with the added bonus that the investigators could pursue other gene therapy projects of interest without regard to their appeal to the marketplace, a recurring problem that has prevented the biotechnology sector from pursuing therapies for genetic diseases that affect small numbers of patients. The Hospital also provided funding for other critical resources required to mount clinical studies of novel cell and gene therapeutics, including clinical, regulatory and administrative infrastructure. The availability of resources to carry out early phase testing of novel therapeutics entirely in the academic medical center is likely to be crucial to the development of novel cell and gene therapies. The NIH, despite its interest in translational work, is still struggling to come to grips with the high costs of this type of investigation, and the biotech industry has had difficulties with the long timelines that typically characterize development of novel classes of therapeutics. The current economic downturn is likely to further exacerbate this inability of small biotechnology companies to support development of novel products through to licensing. Thus, the availability of expertise and funding support placed us in a unique position to advance the development of novel gene therapies for inherited disorders.

Fig. 6.

Timeline of AAV-Factor IX liver trial. Initial progress was slow due to transient contamination of semen by vector. These issues were solved but dose escalation revealed an unexpected transaminase elevation not encountered in animal models. The complexity of this problem resulted in withdrawal of support from the biotechnology company that had manufactured the clinical grade vector. This resulted in further delays while in-house facilities for vector manufacture were established.

Based on the data we had generated on immune responses to AAV capsid, we felt confident that delivery of vector to an immunoprivileged site would be unlikely to trigger the immune response that had led to the destruction of the transduced cells in the liver. We therefore approached investigators at the University of Pennsylvania, Drs. Albert Maguire and Jean Bennett, who had previously demonstrated that introduction of an AAV vector into the subretinal space could correct a retinal degenerative disorder termed Leber's congenital amaurosis (LCA) in dogs with this disorder (¹⁰). We hypothesized that introduction of small doses of vector into a relatively immunoprivileged site such as the subretinal space would avoid the immune responses seen in the hemophilia trial and would be likely to result in long-term expression.

LCA is an autosomal recessive disorder characterized by early onset retinal degeneration. Symptoms appear in infancy or early childhood, and the diagnosis is typically made when the parents notice that the baby fails to track visually. There is no treatment for this disease and patients are eventually totally blind, although there may be some low level of vision in childhood. Mutations in any one of a number of genes can result in LCA, but the existence of a large animal model of one of these (^{57, 58}), deficiency of retinal pigment epithelium-specific 65 kilodalton protein (RPE65), had focused attention on this disorder. From an anatomic standpoint, this mutation affects the retinal pigment epithelium, the layer of cells that forms the blood-retinal barrier, nourishes the photoreceptors and recycles vitamin A. From a biochemical standpoint, RPE65 is a critical component of the retinoid cycle. The visual cycle converts a photon of light to an action potential in the retina. This process is dependent on the photoisomerization of the chromophore 11-cis retinol to all-trans retinol; all-trans retinol is eventually released from opsin and travels back to the RPE layer, where it is recharged to regenerate 11-cis retinol. The conversion of all-trans retinol to 11-cis retinol is dependent on the isomerohydrolase RPE65; without functional RPE65, rhodopsin cannot be produced and the visual cycle is broken. An important characteristic of this disease is that despite the early loss of function, there is long-term preservation of the RPE and photoreceptor layers, in contrast to a number of other retinal degenerative disorders that exhibit early structural deterioration.

In 2001, a team of investigators led by Dr. Jean Bennett at the University of Pennsylvania had published the results of studies in the canine model of RPE65 deficiency. Their work demonstrated that vision could be restored following subretinal injection of an AAV vector expressing RPE65. Endpoints included pupillometry (video recording of pupillary responses to light), electroretinograms, and navigation through an obstacle course in dim light. These promising results were successfully repeated by a number of investigators, who established that treatment relatively early in life (before 14 mos of age in the dog) was critical for restoration of vision (^{59, 60}). Long-term follow-up of these animals has demonstrated maintenance of vision without adverse events for periods of up to 9 years, with observation ongoing.

Plans were made by several groups to develop a clinical trial based on these results. These proposed trials, one in London and the other in the U.S., were planning to include only adults as the initial subjects, a trial design that was likely to limit the ability to assess efficacy if the dog data were predictive. In July 2005, our group at The Children's Hospital of Philadelphia began a collaboration with Dr. Bennett to develop a clinical trial that would include pediatric subjects with RPE65 deficiency (Figure 7). The Pediatrics Research Equity Act of 2003 underscored the importance of testing drugs and biologics intended for the pediatric population in pediatric subjects. In RPE65 deficiency, it is clear that the eventual goal is to inject children as soon as possible after diagnosis, to maximize the chances for restoration of vision before the photoreceptor layer degenerates. Thus the eventual target population, as well as the pathophysiology of the disease, supported the development of a pediatric trial.

Fig. 7.

Timeline of translational studies for AAV-RPE65. These studies were completed on a relatively brisk timeline. IR, immune response. TG, transgene.

In presentations to the regulatory authorities, we emphasized that the animal studies suggested that outcome was dependent on age, since both the viability and the amount of target tissue decreased with age. We also noted that, given our proposal that only subjects who were legally blind would participate, the risk to the subjects of diminution of visual function in the injected eye was relatively low. A critical consideration was the starting dose. The U.S. Code of Federal Regulations sets strict guidelines for the inclusion of children in research studies. If the research involves greater than minimal risk to the child, then it must present the prospect of direct benefit to the subject; if it does not present the prospect of direct benefit, then it must involve only a minor increase over minimal risk. The risks in this study included the risks associated with surgery and general anesthesia, the risk that vector would be distributed beyond the injection site, the risk of inflammation in the injected eye, and the risk that there would be loss of visual function in the injected eye. The risk of general anesthesia for a healthy child is extremely low; data from The Children's Hospital of Philadelphia over the past 20 years suggest that this risk, for a child classified as ASA 1 or 2, is well below the risk of mortality for an average healthy child, estimated to be ∼4/million for those age 15–19, and 0.6/million for younger children (⁶¹). The other three risks, related to vector distribution and potential for inflammation, were also judged to be low based on pre-clinical studies in dogs and non-human primates. The NIH Recombinant DNA Advisory Committee unanimously judged the study as one that involved greater than minimal risk but presented a prospect of direct benefit based on the pre-clinical studies, and therefore ethically appropriate. The Institutional Review Board of the Children's Hospital of Philadelphia concurred that the study carried more than minimal risk, but felt that the requirement for prospect of direct benefit could only be met if the starting dose was the dose that had shown efficacy in at least 90% of affected dogs (i.e. a dose of 1.5 × 10¹⁰ vg). All regulatory agencies eventually agreed to this point of view, and the trial was approved as a study with 3 dose cohorts, each to include 3 subjects, with doses of 1.5 × 10¹⁰, 4.8 × 10¹⁰, or 1.5 × 10¹¹, to be delivered to a single eye.

In parallel with the regulatory efforts, we also conducted additional studies aimed at achieving an optimal translation of the pre-clinical data to a clinical study. These included efforts to maximize expression from the expression cassette, by optimizing the Kozak sequence and by inserting a stuffer sequence in the therapeutic plasmid used to produce vector, to prevent “reverse packaging” of the vector backbone rather than the expression cassette (both are located between the ITRs and so could be packaged, but if the backbone exceeds the packaging limit, it will be inefficiently packaged). We carried out a series of experiments to determine whether surfactant should be included in the final formulation to prevent vector from adhering to the walls of the delivery devices approved for human use in the subretinal injection procedure (⁶²). These studies demonstrated that, at the doses planned for the clinical study, up to 80% of the dose was lost on the walls of the device if the vector was formulated without surfactant, but that virtually all the vector was recovered if the final formulation contained 0.001% Pluronic F68. The optimized vector, essentially identical to the material to be used in the clinical trial, was tested in affected dogs and in non-human primates and demonstrated efficacy and absence of toxicity, as had been previously observed with research-grade vector preparations.

The approved study included subjects ranging in age from 8–27. The lower limit of 8 years was selected for two reasons, first because the risk of amblyopia is lower in children >8 years than in younger children, and second because age 8 is used as a minimum age for a child to give assent to participate in a clinical trial at our hospital. The upper limit of age 27 was somewhat arbitrary but represented an age at which improvement might still occur in the judgment of the investigators. Subjects with mutations affecting both alleles of RPE65 (homozygotes or compound heterozygotes), confirmed on two independent samples, and with visual acuity ≤20/200 or visual fields <20°, underwent further examination to determine eligibility. If sufficient viable retinal cells were present as judged by ophthalmoscopy and optical coherence tomography (OCT), subjects underwent additional psychophysical and objective visual testing to identify the worse eye, which was selected for injection.

Under general anesthesia, the vector was injected into the subretinal space in a volume of 150 μl. This creates a dome-shaped bleb which in all subjects resolved within the first 24 hours after injection. To date there have been no serious adverse events in any of the subjects. One subject, subject 2, developed a macular hole first noted 14 days after vector injection on opthalmoscopy and OCT. The patient was unaware of the finding, and it has been stable over the ensuing months of observation. This was likely related to the procedure itself rather than the vector, and subsequent modifications to the procedure should further reduce the likelihood of recurrence of this event. In terms of immune responses, there has been no evidence of development of either T- or B-cell responses to RPE65 protein, and no evidence of IFN-γ ELISPOT responses to AAV capsid on testing of peripheral blood mononuclear cells from injected subjects. Thus administration of AAV-2-RPE65 seems safe in subjects tested thus far (⁶³).

All subjects have reported a similar clinical course after vector injection. Approximately 2 weeks after vector injection, they begin to perceive increased brightness in the injected eye; this is followed at 3–4 weeks by an ability to begin to recognize patterns (e.g. numbers on a cell phone). Most subjects have reported continuing improved resolution even beyond 6–8 weeks, the time frame in which AAV-mediated gene expression generally reaches a maximum and then plateaus. This may reflect the complex interplay of retinal function and neural pathways; subject age (because of the plasticity of the CNS) may affect outcome in this disease much more profoundly than in a setting like hemophilia.

Endpoints in this trial have included both psychophysical measures of vision such as visual acuity, visual field testing, and performance on a standardized obstacle course, and objective measures such as pupillometry and electroretinograms (Figure 8). Psychophysical measures have the disadvantage that they are dependent on subject participation, and can be affected by factors such as fatigue or learning and memory; for these reasons objective data are perhaps more compelling. Figure 8 demonstrates results of both types of testing in the subjects from the lowest dose cohort. Visual acuity was measured using the Early Treatment Diabetic Retinopathy Study (ETDRS) charts, and visual fields by Goldman visual field examination. In each case, the right eye, the injected eye, demonstrated statistically significant improvement in both visual acuity and visual fields when tested 6 weeks after vector injection. No changes were detected in the uninjected eye, except for improved visual acuity in the left eye of subject 2; this may reflect a reduction in nystagmus that allows better visual acuity. Panel B shows results of objective testing, specifically of pupillometry testing in a normal subject and in subject 3 before and after vector injection. The normal pupillary light response is consensual (i.e. light striking one pupil drives constriction of both pupils). This is easily appreciated in the tracing of the pupillary diameter of the normal subject in a protocol modeled after the “swinging flashlight” test; whether light strikes the left eye (blue vertical bar) or the right eye (pink vertical bar), both pupils constrict. For the subject with RPE65 deficiency, the pre-injection tracing (top tracings in right figure, offset from baseline 7 mm diameter to facilitate comparison with post-injection tracing) demonstrates essentially no response to a 0.04 lux intensity flash for either eye. However, one month after injection of vector into the right eye, there is a clearly detectable bilateral pupillary constriction after a 0.04 lux stimulus to the right eye. Note that a stimulus to the left eye has no effect and indeed the pupils continue to dilate until the right eye is again stimulated. Similar data were obtained for all subjects in this dose cohort. This establishes that vector delivery of the RPE65 gene seems to restore the retinoid cycle and the ability to perceive light as evidenced by the pupillary light response. Electroretinograms (ERGs), which are less sensitive than pupillometry, remained negative in the low dose cohort; the goal of dose escalation will be to identify a dose that restores the signal on ERG as well, although it is unclear if this is possible for subjects over a certain age limit.

Fig. 8.

Psychophysical and objective measures of visual function in human subjects injected with AAV-RPE65. Panel A: Results of mutation testing, pre- and post-visual acuity testing, and Goldman visual field testing in 3 subjects injected with AAV-RPE65 (*p < 0.0001; **p < 0.0001; ***p=0.0017). Panel B: Pupillary light responses in a normal subject and in subject 3 before and after injection with RPE65. Redrawn with permission from Maguire AM, Simonelli F, Pierce EA, et al. Safety and efficacy of gene transfer for Leber's congenital amaurosis. The New England Journal of Medicine 2008;358(21):2240–8. See text for details.

To date, all subjects have shown improvement in both psychophysical and objective measures of visual function; the improvements have been stable over the period of observation (i.e. since October 2007). This is perhaps not surprising given the finding that visual function in injected dogs has been stable for a period of 9 years with observation still ongoing, but it stands in marked contrast to the clinical result in hemophilia, where long-term expression in the dog model did not predict long-term expression in humans with hemophilia. This underscores the importance of the role of specific tissues in determining the immune response to a foreign protein, and also confirms the initial hypothesis that delivery of an AAV vector to a relatively immunoprivileged site would be unlikely to trigger a harmful immune response. No serious adverse events such as inflammation or loss of visual acuity have occurred, and there have been no adverse events attributed to the vector, although a small macular hole developed in one subject, related to the injection procedure itself.

Certain aspects of this disease are likely to present challenges in determining a dose response and an optimal dose. The age of the subject at injection, affecting both CNS plasticity and extent of retinal degeneration, the severity of the underlying mutation, and whether the subject had residual vision in childhood (and thus developed neural pathways for vision) are all likely to affect outcome. The disease is rare, and the number of genotyped subjects is even smaller, so it is not possible to control for these differences among subjects by trying to select a more uniform subset of patients. In addition to these considerations, outcome is also dependent on the sites of injection and the surgical procedure itself, so comparing results across trials and even within trials will not be as straightforward in this disease as it is for a disease like hemophilia. Fortunately, as is often the case for progress in novel therapies, there are multiple gene therapy trials ongoing for this disease (^{64, 65}), and the results seem generally to reinforce those presented here. This should accelerate the accumulation of generalizable knowledge on gene therapy for retinal disorders.

Key elements of success in the RPE65 trial include the fact that correction can be achieved by injection of low doses of vector into a focal immunoprivileged site; that the natural history of the disease is that photoreceptor cells are preserved even in the absence of vision; that the cDNA is easily accommodated in an AAV vector; that there seem to be no dominant negative effects from mutant transgene products; and that there are naturally occurring and genetically engineered animal models of the disease, allowing rational design of studies based on pre-clinical testing. A logical progression toward development of the eventual ideal therapy would include extending the safety studies to younger children, administering vector to both eyes rather than a single eye, and determining duration of expression and whether re-administration is possible. The relative preservation of photoreceptor cells in RPE65 deficiency is not observed in many other forms of retinal degeneration, so for those disorders treatment early in life will be the only opportunity for restoration of sight. This highlights the importance of studies in younger children.

Clearly, the optimal application of a gene transfer approach for retinal degenerative disorders will require changes in the practice of ophthalmology. For example, many ophthalmologists do not currently screen for mutations in patients with Leber's congenital amaurosis, because heretofore there has been no treatment for the disorder. Genotyping of patients diagnosed with the disorder will need to become commonplace. Another unanswered question is the upper age limit at which some improvement can occur; enrollment of additional subjects in clinical trials should help to answer this question. Perhaps one of the most difficult questions involves the large number of genes involved in vision. Over 180 genes have been identified as causing retinal disease. Clearly, the complexity and cost of mounting gene therapy trials for these disorders should be lessened as the field gains experience, but it is still daunting to consider the challenges involved in extending this approach to a large number of inherited retinal disorders.

In the last 20 years, gene therapy has moved from the pre-clinical realm to the clinical arena, and some successes have been achieved. The hard-won knowledge of the initially slow pace of clinical gene therapy studies is beginning to bear fruit. It is worthwhile to recall that other novel therapies such as bone marrow transplantation and monoclonal antibodies also had lengthy initial timelines, as investigators struggled to solve problems that had been uncovered in the first small-scale human studies. In the realm of in vivo gene therapy for genetic disease, our work has shown that delivery of small doses of an AAV vector to a relatively immunoprivileged site such as the retina can result in expression of the donated gene for prolonged periods in humans, much as had been shown years earlier in a large animal model (^{10, 63}). This was not the case, however, in the studies in hemophilia, and the next phase of the work in this disease is designed to test the hypothesis that it is possible to create a window of immunoprivilege by pharmacologic means, that will allow long-term expression of clotting factors in human liver, as has already been achieved in a large animal model of hemophilia.

DISCUSSION

Mackowik, Baltimore: In terms of the problem with an immune response to your vector, would it make any difference if you used a vector like canary pox to which humans have not been exposed.

High, Philadelphia: Well actually, one strategy that we did take was to try to isolate parvoviruses from species like snakes or goats or something like that that we might not have immunity to; and they don't transduce human cells very well but obviously the goal of the field is to find a vector that seems to elicit very little immune response, and actually, at least in animals, that is certainly the case with AAV, but we haven't found one. Let me say another long-term goal of the field is to develop non-viral gene delivery procedures, but so far, they are much less efficient than most of the viral vectors.

Billings, Baton Rouge: Kathy, I thought that was phenomenal. What implications does this have for macular degeneration in the audience to which you speak?

High, Philadelphia: Well, I am sure you know that this is an area being very actively pursued by a number of biotech companies. So I think the ability to deliver genes to the sub-retinal space or the intravitreal space will be important for AMD. So that's not what we work on but don't worry, there are lots of people working on that, Fred.

Brenner, San Diego: Do you know what percentage of hepatocytes were infected by the AAV Factor IX, either in dogs or in your patients?

High, Philadelphia: Yes, We know that for AAV2 we are transducing probably 5 to 10% of the cells with the doses that were given there. Now with other serotypes, you can transduce a much larger percentage of the cells. If you are having an immune response, that is not going to be a good idea.

Wasserman, La Jolla: Just to make your village into a modest-sized city, the NIH is sponsoring a huge immune tolerance network where a number of strategies have been developed to do exactly what you are talking about here, mostly with transplant-for example using anti-CTLA4 or CTLA4 for IG constructs to permit islet cell transplants or to deal with type 1 diabetes or other transplants. So I think there is a large infrastructure to deal with the kind of immune responses that you are seeing, and I don't think you have to rediscover the wheel but to get involved with these large organizations that may have the solution already for you.

High, Philadelphia: Well, that's a very interesting point because we have presented our data to Jeff Bluestone and other people involved with the Immune Tolerance Network, but I have to say, I really felt that the best advice I got was from Hans-Dieter Volk. I don't know if you know him. He's a famous immunologist in Berlin and an expert in immunosuppressive reagents, and so we actually had the opportunity to present our data to Hans-Dieter Volk and, you know, I thought he gave me the best advice I have gotten; and I have presented these data lots of places and he said, “Well you have a very interesting problem here, and you have done a lot of work to work it up, and now you are presenting your data to many people because you desire to get the best advice for your patients. But this is the answer to your question, no one knows the answer.”

Wasserman, La Jolla: Well that's clear or it would have been solved already, but there are a lot of people who are attempting to answer your question with a variety of different strategies.

High, Philadelphia: Yes and what I would say about that too is that, we picked a strategy, and it's been vetted by lots of people, and we are going to start with this. Now it may not be the right strategy at the beginning, and it may need some tweaking. I would be very surprised if we were lucky enough to get the first one right, but we'll see.

Glassock, Laguna Niguel: This is kind of a follow up on the previous question or previous comment. There is an old phenomenon known as oral tolerance which has been reinvestigated recently by the group at Hammersmith with really remarkable results in a system with an antigen, peptide sequence understood. I wonder has anybody explored oral tolerance mechanisms as a means of avoiding this vector response?

High, Philadelphia: Well, I mentioned we are very interested in trying to identify a tolerizing regimen, and most of the work that has been done with those oral tolerizing regimens as far as I know is that they are trying to tolerize to a eukaryotic protein; and I am not sure it is going to be as easy to tolerize to a viral capsid protein, and that's been the hang up.

Czeisler, Boston: Beautiful presentation! I am interested particularly in the issue of the congenital Leber's Amaurosis and its treatment with this novel therapy. We had reported over a dozen years ago that a number of totally blind patients could still maintain circadian photoreception - that we could detect the maintenance of circadian photoreception in those patients, and I wondered whether first of all, at what age, if someone is in their mid-thirties, do you think that this would still be possible given that these intrinsically photosensitive ganglion cells are still maintaining their function, or do you think that the photoreceptors would be completely destroyed by that point. Secondly, in terms of pupillary light reflexes, given that there are two phases of the pupillary light reflexes, if you had maintained a more sustained light stimulus, you might see some preservation of pupillary light reflexes even in the untreated eyes if you kept the stimulus on long enough.

High, Philadelphia: Well actually, what you can show, these were low light stimulus-.04 LUX and if you give them a very high intensity flash, you know a 10 LUX, you can get a little bit of a response. So even in the uninjected eye and even pretreatment, so a very bright flash for some of these people showed a little bit of a response.

Czeisler, Boston: And that may be mediated by these intrinsically photosensitive ganglion cells?

High, Philadelphia: In answer to your other question, I think really from the mouse and dog data, we thought that these people who were 26-years-old might not get any sort of response, but we were wrong about that, and so I think we don't know the upper limit, the upper age limit where somebody might get some recovery of function; and that will be part of the work to be investigated in the next phase of the testing - to extend the upper age limit and see whether people get a response or not.

Davignon, Montreal: This is an outstanding presentation and very exciting. In the days of Clifford Steer, I was impressed by the notion of chimeroplasty and then there was an article in Science, and it seems to have disappeared. Is this still alive, or is this still a possibility to do chimeroplasty or transposon therapy?

High, Philadelphia: You know, I think that I don't know exactly what the difficulty is with chimeroplasty. I think it was very difficult to make the reagents in a reliable and reproducible way, and that was one of the issues. Certainly the field is interested in gene correction as opposed to simply gene addition, and I think the approach of zinc finger nucleases with a good donor template looks more promising at this point than chimeroplasty; and for some of these things really, the ability to make a reagent in a reliable way is just key, and I don't think they ever crossed that barrier in chimeroplasty.

Kelley, Philadelphia: Kathy I just can't tell you my immense pride in seeing how your work has progressed, and gee, what a wonderful clinical research this is and just the tremendous effort and perseverance you have had and staying with it and my warmest congratulations to you. What a terrific example, and we could have only imagined such a thing 20 years ago and gosh, I hardly know what to say; and it is so wonderful that you and Jean got together, and of course to see that long list of names also makes me feel wonderful about the tremendous effort and collegial experience that all of you have had there; and again, my warmest congratulations and thank you very much.

High, Philadelphia: Well I should have said and I want to underscore this that Bill Kelley, of course, was responsible for the large number of investigators there are at Penn who are interested in gene therapy, and you know it's been very important for the work to take place in an environment where people even through the darkest days in the field continued to support the work. So thank you.