1INSERM U768 and Université Paris Descartes, Paris, France. 2Departement de Biothérapies and 3Unité d’Immunologie et Hématologie Pédiatrique, AP-HP Hôpital Necker-Enfants Malades, Paris, France.
Address correspondence to: Marina Cavazzana-Calvo, Département de Biothérapie, AP-HP Hopital Necker-Enfants Malades F-75015, Paris, France. Phone: 33-144-49-50-68; Fax: 33-144-49-25-05; E-mail: firstname.lastname@example.org.
1INSERM U768 and Université Paris Descartes, Paris, France. 2Departement de Biothérapies and 3Unité d’Immunologie et Hématologie Pédiatrique, AP-HP Hôpital Necker-Enfants Malades, Paris, France.
Address correspondence to: Marina Cavazzana-Calvo, Département de Biothérapie, AP-HP Hopital Necker-Enfants Malades F-75015, Paris, France. Phone: 33-144-49-50-68; Fax: 33-144-49-25-05; E-mail: email@example.com.
First published June 1, 2007 - More info
Inherited and acquired diseases of the hematopoietic system can be cured by allogeneic hematopoietic stem cell transplantation. This treatment strategy is highly successful when an HLA-matched sibling donor is available, but if not, few therapeutic options exist. Gene-modified, autologous bone marrow transplantation can circumvent the severe immunological complications that occur when a related HLA-mismatched donor is used and thus represents an attractive alternative. In this review, we summarize the advantages and limitations associated with the use of gene therapy to cure SCID. Insertional mutagenesis and technological improvements aimed at increasing the safety of this strategy are also discussed.
Gene therapy is a form of molecular medicine based on the addition of a corrected copy of a gene to the somatic cells of an individual in order to cure or to alleviate his/her disease. This strategy may provide new treatments for a large number of inherited diseases in the near future.
The practical use of gene therapy is currently limited to 2 cell types: accessible stem cells and terminally differentiated, postmitotic, long-lived cells. The characteristics of the target cells greatly influence the choice of viral vector for gene delivery; vectors able to integrate into the host genome are necessary when dividing cells are targeted, in order to ensure the replication of the transgene, while nonintegrative vectors can be used when gene transfer is performed in postmitotic cells. RNA viruses (retro- and lentiviruses) are used to mediate integrative gene transfer (1, 2). Conversely, DNA viruses provide nonintegrative means of transferring therapeutic genes. Two viruses have been abundantly used in clinical trials of gene therapy (3–5): adenovirus (Ad) and adeno-associated virus (AAV). Given that their genetic material does not integrate into cells and is not replicated at cell division, usage of these 2 viruses is limited to the infection of postmitotic cells.
Of the various stem cell compartments accessible to date, 2 display optimal characteristics for gene therapy purposes: stem cells in the hematopoietic system and those in the skin. These 2 stem cell compartments are well characterized by their capacity to self renew and maintain specific functions over an individual’s lifetime. The cell types’ accessibility, ability to survive in ex vivo cell culture, and transplantability are advantageous for their use in gene therapy; in fact, over 20 years of research on ex vivo manipulation of these stem cell compartments (in HSCT and burn repairs, respectively) has enabled rapid advances in the gene therapy field (6–10).
Terminally differentiated, long-lived cells (including postmitotic cells such as neurons and photoreceptor cells) also represent potential targets for gene therapy and are currently being evaluated in clinical trials (11–13). Cells in the retina are particularly good examples of such a target; they are readily accessible to gene therapy, fundus imaging, and electrophysiological techniques. Considerable progress has been made in the development of gene therapies for retinal degeneration occurring as a result of gene defects in photoreceptor and retinal pigment epithelial cells; recombinant AAV (rAAV) vectors have shown sustained (>3 years) efficacy in animal models (14, 15). The demonstration of long-term functional improvement following transfer of the gene encoding RPE65 (an RPE-specific visual cycle isomerase) has underpinned proposals for clinical trials of rAAV-mediated gene therapy for patients with the inherited retinal degenerative condition Leber congenital amaurosis (reviewed in ref. 15). Ethical approval has been granted by the NIH Recombinant DNA Advisory Committee for 2 proposed trials in the USA and by the Gene Therapy Advisory Committee for a trial in the United Kingdom.
Another condition that lends itself to gene therapy involves the in vivo production of proteins that must be secreted in order to exert their function. In this area, many approaches have been developed for the treatment of the hereditary blood coagulation disorder hemophilia B, with the knowledge that a low but sustained level of factor IX production would provide clinical benefit. This disorder could be used as a model for other diseases involving defects in secreted proteins (16, 17).
However, a number of obstacles exist: (a) poor gene transfer, resulting in low protein expression, as observed in clinical trials (3); (b) risk of insertional mutagenesis when using integrating (i.e., retro- and lentivirus) vectors, which is still a safety concern; and (c) immunogenicity of the vector and sometimes of the transgene itself (18, 19), with the important exception perhaps being in the HSC compartment; in this latter setting, the transgene may be expressed by the hematopoietic cells that seed the thymus and may thus be able to induce specific tolerance.
Given the plethora of significant advances that have been made in a number of fields related to gene therapy, it is impossible to provide an exhaustive review. We therefore focus our discussion on gene therapy for SCID, i.e., what has been achieved to date and what we can expect to achieve in the future. Gene therapy is discussed in light of the clinical results obtained following allogeneic stem cell transplantation, which represents in some circumstances a partially satisfying treatment and a benchmark for gene therapy.
SCIDs comprise a number of rare, monogenic disorders whose common characteristic is the occurrence of a block in T cell differentiation together with a direct or indirect impairment of B cell immunity. The overall frequency of SCIDs is estimated at between 1:50,000 and 1:100,000 live births (20, 21). This could be a slight underestimate, as early death may prevent correct diagnosis in some cases. Studies of patterns of inheritance, immunological characteristics, and, more recently, genotypes have led to the identification of at least 11 distinct SCID conditions (Table 1).
Four main mechanisms of these diseases have been described: (a) Premature cell death caused by the accumulation of purine metabolites, as seen in adenosine deaminase (ADA) deficiency. (b) Defective cytokine-dependent survival signaling in T cell precursors (and sometimes NK cell precursors). This mechanism accounts for more than 50% of cases of SCID. Deficiency in expression or function of the γ common (γc) cytokine receptor subunit shared by the receptors for IL-2, IL-4, IL-7, IL-9, IL-15, and IL-21 causes the X-linked form of SCID (SCID-X1), characterized by the complete absence of both T and NK lymphocytes (22). Deficiency in JAK3, which is normally associated with the cytoplasmic region of γc, results in an identical phenotype. (c) Defective V(D)J rearrangements of the TCR and B cell receptor genes. In our experience, this group accounts for 30% of SCID cases. Deficiency in either RAG1 or RAG2 (the lymphoid-specific recombination–initiating elements) or Artemis (a factor involved in the nonhomologous end-joining repair pathway) leads to defective V(D)J rearrangements and thereby thymocyte and pre–B cell death. (d) Defective pre-TCR and TCR signaling. Pure T cell deficiencies are caused by defects in either a CD3 subunit (such as CD3δ, CD3ε, or CD3ζ) (23–25) or in the CD45 tyrosine phosphatase, key proteins involved in pre-TCR and/or TCR signaling at the positive selection stage (26).
Some researchers include other T cell immunodeficiencies in the SCID group, such as ZAP-70 deficiency (27), CD3γ deficiencies (28), HLA class II expression deficiency (29), purine nucleoside phosphorylase deficiency (30), ligase IV or Cernunnos deficiency (31, 32), and Omenn syndrome (33). However, since these conditions are characterized by the presence of mature (though functionally defective) T cells, they raise very distinct issues as far as therapy is concerned (see below) and so are not considered in this review.
The clinical presentation of the different SCID conditions is fairly uniform and is characterized by the early onset of infections (usually in the respiratory tract and the gut). Common opportunistic organisms such as Pneumocystis carinii and Aspergillus and intracellular organisms such as Cytomegalovirus can cause recurrent infections and a failure to thrive. The severity of these clinical manifestations makes SCID a medical emergency that, in the absence of treatment, leads to death within the first year of life.
It is important to note that the first clear-cut report of successful allogeneic HSCT (in 1968) dealt with the treatment of patients with primary immunodeficiencies, i.e., SCID and Wiskott-Aldrich syndrome (34, 35). In a striking parallel, the first successful gene therapy occurred in the former disorder, demonstrating the value of primary immunodeficiency in general and SCIDs in particular as models for new therapeutic strategies.
Starting with this pioneering experience in 1968, hundreds of SCID patients and hundreds of patients affected by other life-threatening forms of primary immunodeficiencies throughout the world have benefited from HSCT (6, 36–39). Presently, HSCT from an HLA-matched sibling donor confers at least an 80% chance of cure for children affected by primary immunodeficiencies (6), and there is about a 70% chance of cure when a fully HLA-matched unrelated donor is available (6, 38, 40, 41). This high success rate is the consequence of better management of the nutritional and infectious problems affecting these patients at the time of the treatment (Figure 1).
Cumulative probability of survival in SCID patients after HSCT according to donor-recipient compatibility. Figure reproduced with permission from The Lancet (6).
Conversely, when a related but HLA-mismatched donor is used (e.g., when one of the parents donates to a child), the survival rate is significantly lower than that of patients receiving HSCs from either an HLA-matched sibling or a fully HLA-matched, unrelated donor — despite the availability of different ex vivo methods for eliminating the contaminating, mature donor T cells (responsible for the occurrence of acute and severe forms of the graft-versus-host reaction) from the harvested bone marrow (6, 37, 38).
The widespread use of these related HLA-mismatched donors is limited by the following: (a) the age of the recipient at the time of transplantation — mortality is higher in older children, which correlates with that in concurrent infectious diseases; (b) immunological complications (such as graft-versus-host reaction) due to the high HLA disparity between donor and recipient; (c) dramatically slow and often partial immunological reconstitution, responsible for late complications including infections and inflammatory/autoimmune manifestations (42–44); and (d) long-term decline in T cell function, related to the absence of donor stem cell engraftment and (possibly) a premature decline in thymus function when SCID patients are transplanted late (45).
In the future, advances in the control of the allogeneic reaction and in speeding up immune reconstitution are expected to alleviate these obstacles, which prevent the application of related, HLA-mismatched HSCT to diseases other than immunodeficiencies (such as hemoglobinopathies and malignant disorders). Meanwhile, a gene therapy approach (based on gene-modified autologous HSCT) could circumvent the significant limitations described above and thus represent an attractive therapeutic alternative.
The vectors most commonly used to introduce genetic information into HSCs and/or progenitor cells are RNA viruses. They were actually the first viral delivery systems to be developed for gene therapy applications. Retroviruses are a large family of enveloped RNA viruses found in all vertebrates and can be classified into γ-retroviruses, lentiviruses, and spumaviruses. The γ-retroviruses are “simple” viruses (encoding only the structural genes gag,pol, and env) whereas lentiviruses and spumaviruses have a more complex organization and encode additional viral proteins (1, 2, 46).
The complete sequencing of the oncoretrovirus genome and detailed knowledge of the retroviral replication cycle (with integration of the provirus in the target cell genome) prompted researchers in the 1980s to envisage the use of retroviral vehicles for correcting monogenic disorders. Replication-defective murine leukemia virus (MLV) vectors are generated by replacing the viral protein–encoding sequences with the desired therapeutic gene (Figure 2A). In addition to the packaging signal, the vectors retain the viral long-term repeats (LTRs) and adjacent sequences that are essential for reverse transcription and integration. Vector RNA production is driven by either the U3 region of the LTR or by an internal promoter. In this latter setting, the vector’s U3 region can be inactivated to prevent the potent expression of the retroviral LTR. Packaging of retroviral vectors is achieved by providing the structural proteins in trans in the packaging cells. To prevent recombination between vector and packaging constructs, packaging cells have been developed that express gag, pol, and env from separate constructs.
Schematic representation of retroviral vectors and their modifications to improve safety. (A) The transcription of the therapeutic gene is driven by the enhancer-promoter activity of the U3 region of the retroviral LTR. (B) The transcription of the therapeutic gene is driven by the addition of an internal promoter. The U3 region of the retroviral LTR has been almost completely deleted. (C) The provirus contains the cHS4 element (i.e., insulator) in order to protect the transcriptional cassette against position effects. (D) This provirus contains 2 cassettes: (a) the therapeutic gene driven by a first internal promoter and (b) a suicide gene (e.g., thymidine kinase, TK) that could allow the elimination of gene-corrected cells if an adverse event such as a monoclonal proliferation occurs. EF-1a, elongation factor–1a; IRES, internal ribosome entry site; PGK, phosphoglycerate kinase; R, repeats; SA, site acceptor; SD, site donor.
In addition to gag, pol, and env, lentiviruses encode 3–6 other viral proteins, which contribute to virus replication and the persistence of infection (47). Although vectors based on simian, equine, and feline lentiviruses have been developed, the most extensively studied for clinical applications are the HIV type 1–based (HIV-1–based) vectors. HIV-1 encodes 6 accessory proteins (tat, rev, vif, vpr, nef, and vpu). The accessory genes are nonessential for lentiviral vector production and transduction and so are deleted from the packaging construct (48). Whereas γ-retroviral vectors can only be effective in dividing cells (because the preintegration complex cannot cross the nuclear membrane), lentiviral vectors do cross the nuclear membrane, particularly when the endogenous central polypurine tract (49, 50) element is added. The addition of this element in cis enhances the transduction rate and thus extends the field of application of these vectors.
The biosafety of these HIV-1–derived vectors has been increased by the development of self-inactivating vectors. Self-inactivating vectors do not contain the U3 enhancer-promoter in the LTRs (51) and are therefore likely to be less potent in activating cellular genes at the integration site (52) and less likely to be mobilized following infection with HIV (53) (Figure 2B).
The HIV-1 env glycoprotein has a highly restricted host range, in that it only infects cells expressing CD4 and its coreceptors. To broaden the host range of lentiviral vectors, the latter can be pseudotyped, for example, with the vesicular stomatitis virus glycoprotein (VSV-G) or the gibbon ape leukemia virus (GALV) env, which is provided in trans and imparts a broad tropism; VSV-G or GALV pseudotyped vectors can also be concentrated to produce high-titer supernatants. Stable modifications of the viral envelope have been introduced in order to target a desired cell type or to increase the transduction rate (54–57).
SCID conditions provide several circumstances that favor successful outcomes following gene therapy. Since gene mutations lead to loss of protein function with a block in T cell development, a selective growth advantage is conferred to transduced cells if the corrective transgene is expressed. The effect is sustained because the gene-corrected differentiated T cells are long-lived.
Two additional considerations led to the initiation of a gene therapy trial in 1999 for SCID-X1: the γc subunit receptor is ubiquitously expressed by cells of all hematopoietic lineages from a very immature stage (i.e., CD34+ lineage–negative cells) (personal communication from Emmanuelle Six, INSERM U768, Paris, France) up to and including mature cells. Hence, there is no risk of inducing unwanted ectopic expression. Secondly, potential γc overexpression (in the absence of the other specific cytokine receptor subunits with which γc associates) (58) cannot induce signal transduction and thus cell activation. Membrane γc expression is actively regulated by the available quantity of the other cytokine receptor subunits to which γc binds (such as IL-7 receptor-α). In these studies, the γc cDNA was placed under the control of the viral LTR, and the defective MLV was produced using an amphotropic ψ CRIP packaging cell line. Following extensive preclinical studies, 10 children under the age of 1 were enrolled between 1999 and 2002 into a SCID-X1 gene therapy clinical trial (59). In 9 out of the 10 patients, γc gene transfer into CD34+ cells resulted in the emergence of gene-corrected T and NK lymphocytes. In 7 of the 9 patients who developed T cells, T cell counts reached normal levels within 3 months and have remained so ever since, according to the last follow-up. The other 2 patients only experienced partial T cell reconstitution, presumably due to the lower number of CD34+γc+ cells/kg they received (59), indicating that a threshold dose of transduced precursor cells was required to achieve full T cell reconstitution. Now, 7.5 years later, these 7 patients retain a functional immune system, enabling them to live normally. The diversity of the T cell repertoire and T cell functions has been observed over this period. Meanwhile, in the United Kingdom, a second gene therapy trial aimed at treating the same form of SCID has been initiated (60). To date, 17 out of the 20 enrolled SCID-X1 patients have been successfully treated in the London and Paris clinical trials. Likewise, ADA deficiency has also been successfully treated using a similar approach that also required mild chemotherapy prior to the gene therapy. The immunodeficiency was partially or fully corrected in all 9 patients (61, 62).
Within the past 10 years, spontaneous, partial corrections of the phenotype of severe T cell immunodeficiencies (e.g., ADA deficiency, SCID-X1, Wiskott-Aldrich syndrome, RAG1 deficiency, CD3 deficiency, and NF-κb essential modulator (NEMO) deficiency syndrome) have been reported (63–67). In each of these conditions, it has been found that 1 or several T cell precursors carrying a wild-type sequence of the disease-causing gene (or a mutation with a less deleterious effect) can differentiate into mature, functional T cells that can provide significant immunity for several years (64–66, 68). Detailed analysis performed on an individual with a partial SCID-X1 phenotype (resulting from reversion to a wild-type γc sequence in a T cell precursor) (64) emphasized the expansion capacity of T cell precursors and the longevity of mature T cells. These results constitute a strong rationale for the development of a gene therapy approach that recapitulates these rare spontaneous events.
The selective advantage conferred by the expression of either γc or ADA in lymphocyte progenitors was confirmed in 3 gene therapy clinical trials (60, 62, 69). In the γc trials, no more than between 0.1% and 1% of the myeloid cells were transduced, indicating that the presence of even a small number of transduced hematopoietic progenitor cells was enough to replenish the T cell pool (70). Furthermore, an additional, mild, myelosuppressive treatment (for patients with ADA deficiency) led to a higher proportion of transduced progenitors being detectable in vivo due to better engraftment of transduced HSCs. In fact, analysis of the retrovirus integration sites indicated that the number of transduced clones that effectively led to T cell development was only around 100. It is also striking to note that gene therapy for ADA deficiency was only successful in patients who did not concomitantly receive polyethylene glycol–ADA (PEG-ADA) enzymatic substitution. Gene therapy failed in patients receiving PEG-ADA — a setting likely to abrogate the selective advantage conferred on transduced cells (71). Despite the absence of any clear clinical benefit, a marked difference was reported in the percentage of transgene-containing cells upon comparison of T lymphocytes (1%–10%) and cells of the myeloid lineage (0.01%–0.1%), validating the concept of selective advantage. In 1 of the 2 treated patients, a single provirus integration site was found to be predominant and was present at a stable level for 8 years. This T cell progenitor was able to generate a restricted pattern of TCR rearrangements (71).
Large-scale, sequence-based surveys of the integration sites have (a) confirmed a preference for provirus integration into genes and (b) further stressed the role of selective advantage in the successful outcomes reported in these 3 trials (60, 62, 69). For example, in our SCID-X1 gene therapy trial, the finding of recurrent integrations at specific gene loci suggests that these particular insertions might have provided the affected target cell clones with a nonrandom growth or survival advantage (72). Thus, vector integration in common integration sites may actively influence the fate of corrected cell clones in vivo. Alternatively, common integration sites could result from a preferential targeting of actively transcribed genes in progenitor cells. The influence of integration on outcome has recently received further confirmation (73) following the observation of clonal myeloid cell expansion in a retroviral vector–based gene therapy clinical trial for correcting chronic granulomatous disease, a disorder characterized by the absence of bactericidal function in neutrophils that otherwise appear to differentiate normally. In the 2 treated patients, extensive expansion of myeloid cells with either myelodysplasia syndrome 1 (MDS1)–, PR domain–containing 16 (PRDM16)–, or SET binding protein 1 (SETBP1)–selected integration sites has been detected, leading to a 3- to 4-fold expansion of the gene-corrected cell fraction (73).
Together, these data support the concept that the successful outcomes of the gene therapy trials for 2 different SCID conditions were primarily accounted for by the selective advantage provided by introduction of the therapeutic transgene (γc or ADA). Furthermore, in the absence of a selective advantage conferred by the transgene, insertion of the retrovirus near growth-promoting genes might also play a role (as observed in the chronic granulomatous disease trial). If the selective advantage provided by the retrovirus vector integration site is instrumental in the success of gene therapy because of an influence on expression of an adjacent gene, use of vectors with reduced enhancer activities might result in lower clinical efficacy. Additional in-depth analyses of the patients’ hematopoietic systems years after gene therapy are likely to provide a wealth of information on the eventual selection of cells that result from the specific interactions between the integrated vectors and the hosts’ genomes (74).
The clinical benefit of gene therapy has been tempered by observations of varying significance. First of all, restoration of a normal immunological phenotype depends on the number of gene-corrected HSCs administered to the patient. A direct correlation between this number and the robustness of the immunological reconstitution has been observed. In contrast, it is more difficult to define an upper threshold for reducing the risk of insertional mutagenesis. Ideally, methods capable of isolating the true stem cells and optimizing their bone marrow homing should significantly reduce the total number of gene-corrected cells injected into the patients and thus the associated oncogenic risk.
It appears that there is also an age limit for success: in 2 SCID-X1 patients (the first, over 15, in whom allogeneic HSCT failed, and the second, who had a partial SCID phenotype because of hypomorphic mutation) an otherwise successful procedure (i.e., gene therapy) failed, probably because thymic function was absent as a consequence of time and infection (75). In light of these results, the definition of a therapeutic window for the use of gene-modified autologous HSCs in SCID is an important matter, bearing in mind that different parameters can play a role in its definition. Residual thymic function, the number of available lymphoid precursors, and the latter’s homing capacity are likely to be relevant parameters but, at this point, are very difficult to measure. No evidence for transgene silencing has been observed in SCID–X1 and ADA clinical trials, although one cannot exclude a nondetectable loss of T cells in which the γc protein would no longer be synthesized. This phenomenon of transgene silencing is well known in mouse models of gene therapy (76).
Most importantly, insertional mutagenesis occurred in 3 patients with SCID-X1; γc gene transfer into CD34+ cells resulted in clonal T cell proliferation that became clinically overt after 3 years, leading to the death of 1 patient; it was controlled by chemotherapy in the 2 remaining patients. These serious complications clearly resulted from retroviral integration leading to the aberrant expression of the proto-oncogene LIM domain only 2 (LMO2) in the first 2 observed cases (Figure 3) (77). The occurrence of these severe adverse events obviously raises the question of how to prevent such incidents in the future.
LMO2 gene map of activating retrovirus vector insertions. The loci of retroviral insertion in the clones of the first 2 cases of monoclonal proliferation that occurred in P4 and P5 patient clones were characterized by LAM PCR sequencing of the 5′ insertion site fusion sequence. The enhancer activity of the proviral LTR allows the hyperexpression of LMO2 transcripts of the expected 3.3 kb size as well as the production of a high quantity of normal-size LMO2 protein. MFG-γc, retroviral vector.
Despite all the toxicological studies performed in various mammalian species (i.e., mice, dogs, and monkeys) and the absence of any observed toxic effects in the 40 different clinical protocols based on the use of retrovirus vectors, the occurrence of these 3 clonal lymphoproliferations in our SCID-X1 trial appeared to challenge all the biological data upon which the trial was based (78, 79). It is possible that differences exist between murine and human virus-induced oncogenetic processes. In replication-competent, retrovirus-induced oncogenesis, the replication capability of the virus led to the accumulation of several integrations, leading to transforming events that cooperated to trigger the tumorigenic process independently of their order of appearance. In contrast, in the first 2 cases of monoclonal proliferation observed in the SCID-X1 trial, a single integration site was detected. It led to the uncontrolled transcription of LMO2, a master gene in human and murine hematopoiesis, but one that lacks a known function in the T cell differentiation pathway. Despite the description of human translocations involving this transcription factor and the existence of LMO2 transgenic mice that develop lymphoma at 10 months of age, LMO2-mediated oncogenesis remains poorly understood (80). In addition, it is not clear whether the mechanism underlying LMO2-associated primary T cell acute lymphoblastic leukemia is the same as that at work in retrovirus-triggered LMO2 overexpression in the SCID-X1 patients (77, 81–83).
Some recently reported data suggest that cooperation between LMO2 activation and dysfunctional expression of the γc gene is involved in cell deregulation. By using neonatal mice infected with replication-competent retrovirus as a model of insertional mutagenesis, Dave et al. identified a neoplastic process in which both the LMO2 gene and the γc gene are targets for retroviral insertion within the same tumor (84). Simultaneous, random insertional mutagenesis of these 2 genes in a tumor is highly improbable, supporting the theory that a synergistic interaction may provide the affected clone with a significant proliferative advantage. In contrast, transduction of hematopoietic repopulating stem cells from LMO2-transgenic mice with the γc retrovirus–expressing vector did not increase the incidence or accelerate the development of lymphomas (85). Recently, a study in γc-deficient mice (in which γc complementation was performed with a lentiviral vector) provided evidence for the role of γc overexpression in lymphoma induction (86). However, these data were not supported by those from other studies performed in γc-transgenic mice and in other experimental models of γc gene transfer, in which no increase in tumor induction was observed (87). Thus, the potential interaction (if any) between LMO2 overexpression and retrovirus-mediated γc gene expression remains unclear. Consequently, in the absence of a complete understanding of the pathophysiology of this adverse event, it remains a challenge to design a reliable, preclinical predictive assay to test vector transduction protocols and identify changes that can improve protocol safety while preserving efficacy. It should be emphasized that safety assessment requires long-term follow-up and the performance of secondary transplantations. The latter can reveal clonal dominance, as observed with retroviral vectors expressing a neutral transgene with respect to cell growth potential (88).
An important remaining question is whether the risk of insertional mutagenesis is restricted solely to gene therapy for SCID-X1. The strong proliferative advantage of γc-transduced lymphoid progenitors could predispose these cells to transformation. Other factors increasing the oncogenic risk could be specific to SCID-X1 patients and might therefore be absent in the treatment of other diseases. One specific factor might be the age of the patient at the time of treatment. It is thought that below the age of 1, the bone marrow stem and progenitor compartments have a higher proliferative capacity. Alternatively, an expanded target population of lymphoid progenitors susceptible to vector-induced mutagenesis may be present because of the differentiation block (89). In an attempt to prove these hypotheses, experimental gene transfer was performed in cancer-prone murine models. Shou et al. recently generated a double-knockout murine model in which the Arf and γc genes were both ablated (89); more than 90% of these mice developed lymphomas within 1 year of gene therapy with γc–/– donor bone marrow cells transduced with an MLV γc–expressing vector. The transformation process required transduction with a γc vector, the γc–/– background of donor cells, selection of insertion sites near or within the cellular proto-oncogene, and biallelic deletion of Arf (89). Comparing the occurrence of lymphomas in Arf-deficient mice with either a γc or other gene defect (such as Rag1, as is presently being studied) treated by gene transfer should reveal whether γc itself plays a role. A similar model using cyclin-dependent kinase inhibitor 2A–deficient mice (Cdkn2a–/– mice) was recently reported by Montini et al. (90). Cdkn2a encodes p16 Ink4a (a regulator of Cdk4/6-mediated Rb1 phosphorylation) and p19 Arf (a modulator of Mdm2-mediated degradation of p53) (91). Loss of Cdkn2a results in a combined deficiency in the Rb1 and p53 pathways — 2 major regulators of cell proliferation, apoptosis, and senescence. Deficits in these tumor suppressor pathways enhance the survival and proliferative potential of the cells. Despite obvious limitations, these models should help us define the role of the genetic defect/transgene and assess vector modifications designed to increase safety.
Another factor involved in the initiation of the observed serious adverse events could be related to the transduction protocol. Indeed, no serious adverse events have occurred so far in patients treated in the United Kingdom in accordance with the SCID-X1 protocol, although the difference is not statistically significant (60). Thrasher et al. used a GALV envelope instead of the amphotropic one used in our trial. They also used a 3-fold lower concentration of IL-3 and did not add a low level of fetal calf serum to the culture medium. The GALV envelope might target a slightly different (and possibly less differentiated) cell subset that may be less prone to transformation.
Despite access to improved experimental models for the assessment of protocol safety (see above), implementation of a clinical trial for the treatment of a given disease requires specific evaluation of the risk/benefit ratio as compared with any existing alternative therapies. In the case of SCID-X1 patients for whom an HLA-genoidentical sibling or an unrelated, fully HLA-matched donor does not exist, it appears that the efficacy of gene therapy is superior to that of a mismatched HSCT (as discussed above). However, the number of cases studied is low ( n = 20), and the follow-up time has been relatively short (median 4 years, maximum 7.5 years). This evaluation of the risk/benefit ratio justifies the efforts of several groups seeking to make existing vectors safer. In this rapidly moving field, 2 trends can be identified: (a) continuation of current clinical protocols with modification of the available integrative vectors (short-term modifications) and (b) identification and use of new technologies (long-term modifications).
The oncoretroviral vectors used in all the clinical trials performed to date (60, 62, 69, 73) possess strong enhancer and promoter elements within each integrated LTR. This strong enhancer drives the expression of distant genes up to 100 kb downstream or upstream and is known to be involved in the overexpression of LMO2 (77). Vectors with a self-inactivating LTR have thus been designed so that the promoter and enhancer elements from both the 5′ and 3′ LTRs are eliminated upon proviral integration. Expression of the therapeutic gene would subsequently occur via an internal promoter with little or no enhancer activity (Figure 2B). In addition, improved safety may be achieved by using tissue-specific promoters to drive transgene expression. By using more specific promoters, it should be possible to achieve more tightly regulated protein expression. Tissue- and gene-specific promoters may prevent oncogenesis in cells of the relevant lineages.
Insulators may also contribute to the improvement of vector safety by limiting the activation of genes surrounding the insertion sites. Insulators are small DNA elements that act as barriers, thereby preventing promoter-enhancer elements and/or chromatin modifications from influencing the expression of neighboring genes (92). The cHS4 insulator (derived from the chicken β-globin locus) has been most intensely studied for gene therapy purposes (93). When insulators are incorporated into integrating retroviral vectors, position-effect variegation is decreased while the degree of transgene expression is increased for each insertion (94–97) (Figure 2C). It has also been suggested that insulators could protect the surrounding genome from insertional activation by retroviral LTRs (61). However, introduction of these elements is only expected to decrease (and not eliminate) enhancer activity. Studies directly testing the degree to which insulators protect the surrounding genome from LTRs are ongoing.
A third potential safety measure consists of the introduction of a suicide gene (such as that encoding thymidine kinase from the type 1 herpes simplex virus) that could facilitate the killing of abnormally growing cells by use of a prodrug (e.g., ganciclovir) (13, 98, 99) (Figure 2D). Nevertheless, there are unsolved issues surrounding this approach, including expression and function of a second gene cassette, induction of in vivo resistance, immunogenicity of the thymidine kinase protein, and finally, inherent limitations in the use of antiviral drugs for the treatment of infections once gene-modified cells have been injected. Selection of more robust suicide genes might thus be desirable.
Major advances are expected from the development of new approaches that target specific genomic sites for integration or gene correction. The use of site-specific integrases (INs) could allow this goal to be attained. Bacteriophage IN phiC31 is able to perform site-specific integration via recombination between an attP recognition site encoded in the phage genome and an attB site in the bacterial chromosome. Hence, phiC31 IN can mediate integration of vectors bearing attB into mammalian chromosomes at genomic sequences that are similar to attP (termed pseudo-attP sites), since it was previously demonstrated that a few (between 100 and 1,000) integration sites do exist in the mammalian genome (100, 101). Observation that the phage IN phiC31 is able to induce site-specific integration in human cells (101, 102) has led to testing of its in vivo efficacy in experimental models (Figure 4). Recent studies have demonstrated stable genomic integration, prolonged expression of therapeutic genes, and therapeutic efficacy in animal models of tyrosinemia type 1 (103), epidermolysis bullosa (104), factor IX deficiency (100), and SCID-X1 (105). Unfortunately, concerns were raised because there was evidence that the phiC31 IN can induce recombination between pseudo- attP sites present in the human genome, thereby causing genomic instability and nonspecific integration. To minimize these possible toxic effects, custom, high-specificity INs may be required to restrict targeting to a more limited number of sites (101).
Schematic representation of the integration process mediated by the bacteriophage phiC31 (INT). Two phiC31 molecules attach to the attB recognition site and the attP pseudo-sites in the mammalian genome (attP′). The recombination process is based on a 2-bp staggered cut of the DNAs and rotation of the DNA targets by 180°. The integration reaction results in integration of the transgene flanked by 2 hybrid att sites, attL and attR. Figure modified and reproduced with permission from Molecular Therapy (118).
In a similar manner, the process of integration site selection using retroviruses and transposons is primarily determined by the transposase itself. For both the Tn10 transposase (106) and retrovirus/lentivirus INs (107, 108), it has been shown that regions within the catalytic core domains of the transposase/IN proteins are involved in mediating interactions with the target DNA. Thus, several research groups are investigating target-directed integration into specific DNA sites by use of fusion proteins composed of HIV-1 or avian sarcoma virus IN and the Escherichia coli LexA repressor (109). To date, it has been shown that integration at specific sites can work well in vitro; however, a variety of obstacles (including the absence of physiological targets in human DNA) make in vivo applications of this approach quite problematic (110).
One class of DNA-binding protein that should offer specificity and flexibility in conferring integration site specificity is the synthetic zinc finger protein (ZFP). Work from the Chandrasegaran (111) laboratory has shown that a ZFP can be coupled to the nonspecific DNA cleavage domain of the Fok1 type IIS restriction enzyme to produce a zinc finger nuclease (ZFN) that can cut the ZFN-specified DNA sequence. An important mechanism for increasing DNA-binding affinity and specificity derives from the requirement for 2 ZFNs to bind the same locus in a precise orientation and spacing relative to each other so that the nuclease can create a double-strand break at the desired place, which can be repaired by homologous recombination (Figure 5). Work from Jasin and colleagues (112) followed by others (113, 114) demonstrated that the endonuclease I-Sce1 can potentiate homologous recombination into loci previously engineered to contain its own recognition site. Carroll et al. (115) and Baltimore et al. (116) have shown that a ZFN-induced double-strand break increases the rate of site-specific homologous recombination in model systems. More recently, Urnov et al. (117) showed that a ZFN can recognize an endogenous target gene (in this case the γc gene) and induce correction rates of up to 17% in the human K562 cell line and 5% in primary T cells. However, despite the optimism raised by this proof-of-concept experiment, it is now essential to evaluate the frequency of off-target, nonspecific DNA breaks and thus the potential for oncogenic mutations, especially since double-strand breaks have been shown to be mutagenic and can induce chromosomal translocations.
Usage of replication-defective γ-retroviruses has provided proof of principle for gene therapy as a powerful therapeutic approach to correcting monogenic diseases affecting the hematopoietic system and justifies all the ongoing efforts to improve these preliminary clinical results. Continuing technological progress in gene targeting and stem cell manipulation should improve safety and efficacy and thus prompt a significant extension of the application of gene therapy as a treatment for inherited diseases and notably those of the hematopoietic system and the skin.
Note added in proof. Since writing this paper, a fourth case of T-cell clonal lymphoproliferative disease occurred in our group of SCID-X1 patients treated by gene therapy.
The authors wish to thank the staff of the Immuno-Pediatric Service and the Department of Biotherapy at AP-IP Hôpital Necker-Enfants Malades for their work and support. This work is supported by grants from the Institut National de la Santé et de la Recherche Médicale (INSERM) and l’Assistance Publique des Hôpitaux de Paris (AP-HP), CEE grant CONSERT LSHB-CT-2004-005242, and grants from the Agence Nationale de la Recherche (ANR) and AFM-Telethon grant. We thank Debbie West and Brigitte Ikonomov for skillful secretarial assistance.
Nonstandard abbreviations used: AAV, adeno-associated virus; ADA, adenosine deaminase; γc, γ common; GALV, gibbon ape leukemia virus; HIV-1, HIV type 1; HSCT, HSC transplantation; IN, integrase; LMO2, LIM domain only 2; LTR, long-term repeat; MLV, murine leukemia virus; SCID-X1, X-linked SCID; ZFN, zinc finger nuclease; ZFP, zinc finger protein.
Conflict of interest: The authors have declared that no conflict of interest exists.
Citation for this article:J. Clin. Invest.117:1456–1465 (2007). doi:10.1172/JCI30953.