Assessing the Risk of T-cell Malignancies in Mouse Models of SCID-X1

It has long been recognized that severe combined immunodeficiency disorders (SCIDs) are favorable settings for the application of stem cell gene therapy. This genetically heterogeneous group of disorders includes adenosine deaminase (ADA) deficiency and X-linked SCID (SCID-X1),¹ the latter caused by mutations in the common gamma chain (γ_c) gene, which is required for a broad range of cell signaling events in lymphocytes and their precursors. The strong selective advantage for genetically corrected cells during development circumvents the requirement for high levels of transduction and engraftment, which remains a significant barrier for other blood diseases that lack an in vivo selection mechanism for transduced cells. Pioneering gene therapy trials for SCID-X1 have taken place in France,^2,3 the United Kingdom,⁴ and—for ADA deficiency—Italy⁵; these have clearly demonstrated the curative potential of this approach. Retroviral vectors derived from the murine leukemia virus (MLV) were used to transfer a normal copy of the diseased gene to autologous CD34⁺ cells from the infant patient's bone marrow. After reinfusion of these modified cells, most of the more than 30 treated patients have had full reconstitution of T-cell–mediated immunity, with restoration of B-cell function in fewer, but significant numbers of, cases. However, in both SCID-X1 trials, some patients developed T-cell malignancies that were clearly related to insertional mutagenesis from the integrated vector.^6,7,8

Molecular analyses of these five cases have shown that the tumors were caused by activation of known proto-oncogenes due to the action of the MLV enhancer element on those adjacent cellular promoters. Strikingly, four of the five cases showed activation of the LMO2 gene, a known T-cell oncogene involved in a number of T-cell acute lymphoblastic leukemia (T-ALL) chromosomal translocations.⁹ This unexpected adverse event has been a major limitation to this approach and has led to a new field within gene therapy focusing on the mechanisms of genotoxicity associated with integrating vectors.¹⁰ Much has been learned from these studies, yet important questions remain.

One question is why the leukemia cases have been seen only in the SCID-X1 trials and not in the trials involving ADA deficiency. Intriguingly, insertion-site analyses in ADA-deficient patients treated with MLV-derived retroviral vectors have shown minor clones with insertions into the LMO2 locus, but no cases of T-cell malignancies or other hematologic abnormalities have been noted.¹¹ This and other observations raise the question of whether the disease background influences the vector insertion patterns and risk of oncogenesis. Two patients in a German trial for chronic granulomatous disease (CGD) developed a severe form of early leukemia, called myelodysplasia, due to insertions into the MDS1/EVI1 gene,^12,13,14 a locus known to be preferentially targeted by MLV vectors in other contexts unrelated to SCID gene therapy.^15,16 These observation suggest that the genetic profile for vector toxicity may differ with the underlying disease.

There are many potential reasons for these differences; however, we still do not understand this phenomenon in great detail. In this context, the new report from Ginn et al. in this issue of Molecular Therapy is of particular interest.¹⁷ These investigators used a SCID-X1 mouse model to show that the SCID-X1 background itself is an important risk factor for T-cell malignancies, independent of vector-related perturbations of proto-oncogene expression. These results suggest that the increased susceptibility of SCID-X1 to transformational events will be an important consideration in future gene therapy efforts; they also underscore the need for safer approaches that retain the proven efficacy of such treatment but eliminate or reduce this serious complication.

This report describes a series of self-inactivating lentiviral vectors that were tested in a mouse model of SCID-X1 gene therapy. Various human cellular promoters were tested, including the now widely utilized elongation factor (EF)-1α promoter, along with a phosphoglycerate kinase (PGK) promoter and the promoter from the WASP gene. The EF-1α promoter was most effective and restored full immune function in transplanted mice, the PGK promoter gave intermediate results, and the WASP endogenous promoter had little activity with regard to reconstitution. These results are important in considering promoter choice not only for SCID-X1 but also potentially for other lymphodeficiency disorders that require expression of the therapeutic gene throughout T-, B-, and natural killer–cell compartments.

The mice in these studies were followed for more than 1 year, and 4 of 14 mice treated with the EF-1α-hγ_c vector developed malignancy. No tumors were seen in a smaller cohort of mice transplanted with untransduced SCID-X1 cells or in mice transplanted with wild-type cells transduced with an EF-1α-GFP vector. At first impression, it seemed that the EF-1α-hγ_c vector was inducing oncogenesis specifically on the SCID-X1 background. This was a surprising result, because the EF-1α promoter contains little to no enhancer activity and was considered to be unlikely to activate cellular proto-oncogenes adjacent to the insertion site.¹⁸ Consistent with this prediction, further molecular analyses of the integrated provirus in all four malignant clones showed no evidence of vector-induced oncogene activation. In two cases, the integration sites were not located within 150 kb of any cellular genes. In the other two cases, microarray expression analyses showed no evidence of vector-induced gene expression within a 300-kb window of the insertion site. These results suggest that the requirement for the EF-1α-hγ_c vector for inducing T-cell malignancies was not related to insertional mutagenesis but instead reflects the requirement for restoration of T-cell development via gene replacement.

The absence of insertional mutagenesis in this model stands in marked contrast to the clinical results from the SCID-X1 and CGD trials, which invariably show that vector insertions lead to transcriptional activation of cellular proto-oncogenes. Furthermore, the clinical trials have shown a remarkable preference for—and/or selection of—integration into a few particular proto-oncogenes, a feature not seen in this mouse model. In this sense, the mouse model does not recapitulate several key findings seen in the clinical trials, namely, preferred integration into the LMO2 locus and transcriptional activation due to the vector-encoded enhancer. This discrepancy has also been noted in other mouse model studies, in which no LMO2 insertions were seen in tumor cells.^19,20 These considerations argue that this mouse model will not be helpful in testing the relative safety of newer vectors designed to avoid the activation of proto-oncogenes. Other good alternatives now exist for vector safety testing that do not rely on mouse transplantation assays.^21,22 Another problem with mouse transplant assays is that they are generally associated with a high background of spontaneous tumors due to irradiation and other factors.²³ Therefore an important feature of this study is that it shows how mouse models and human trials differ with regard to retroviral vector–related insertional mutagenesis and that it illustrates the limitations of using mouse models for making safety determinations.

Newer retroviral vectors that are less prone to the activation of flanking cellular genes are just now becoming available. One important design feature is the use of self-inactivating vectors that are completely devoid of viral transcriptional elements, particularly enhancers (see the Editorial in this issue²⁴). These vectors, instead, contain cellular promoters that allow for adequate expression of the therapeutic gene but lack the ability to activate proto-oncogenes dominantly. The EF-1α-hγ_c vector used in this study is an example of such a design. Another safety feature being employed in the use of genetic enhancer blocking elements is the cHS4 insulator fragment from the chicken β-globin locus,^25,26 which has been shown to attenuate the effects of viral enhancers on cellular oncogenes.²² Other recent studies suggest that a lentiviral vector may be less prone than MLV vectors to induce transformation, even when relative enhancer strength is controlled for.²⁷

It is still a challenge to define the best safety measures for evaluating these newer lentiviral vectors. At this time, there are no preclinical assays that can “guarantee” the safety of any retroviral vector for use in a clinical trial. Instead, investigators must use an appropriate battery of preclinical assays to make the case for a potentially increased safety profile and then design a definitive safety evaluation in an appropriately designed clinical trial where the potential risks are fully disclosed. This position is supported by reflecting on the unique toxicity data that have arisen from the SCID-X1, ADA deficiency, and CGD trials and the fact that none of these risks were prospectively identified in numerous preclinical studies.

The potential mechanisms of the increased susceptibility in this mouse model are unknown. Further genetic analyses of the four tumors have shown that these clones had acquired activating mutations in Notch1. This feature has also been seen in the SCID-X1 T-cell malignancies^8,28 and is thought to be a secondary acquired mutation that collaborates with vector-induced LMO2 activation. Other acquired mutations have been recurrently noted in human SCID-X1 transformation and have been assumed to occur during the 24- to 68-month latency period associated with these cases. For instance, many of the human SCID-X1 cases have displayed biallelic deletions of the tumor suppressor gene CDKN2a. These deletions can occur when aberrant recombinase activity results in gene deletion by engaging cryptic recombinase recognition sites, which are known to be present in the CDKN2a locus. This genetic mutation occurs frequently in “naturally” occurring cases of human T-ALL^29,30 that are not associated with gene therapy.

The authors discuss several possible mechanisms for these genetic events and the generally increased susceptibility. One possibility could be increased signaling due to deregulated expression of the vector-encoded γ_c gene, as others have argued,³¹ but this study found no evidence for autonomous signaling in vector-transduced cells. Another possibility is that the “stalled” pool of γ_c-deficient lymphoid progenitors may have increased susceptibility to mutations during the rapid clonal expansion induced by the sudden acquisition of cytokine signaling. Associated high levels of proliferation could lead to subsequent DNA damage and mutations during a period of “proliferative stress.” It would be interesting to determine whether the mutations in Notch1 and/or CDKN2a were actually acquired and present prior to vector transduction and subsequent proliferation. Another possibility is that the lack of tumor immune surveillance in the SCID-X1 transplant setting could predispose to T-cell malignancies. In this regard, it is interesting to note that control animals receiving wild-type donor cells did not develop leukemia, most likely because the adoptive transfer of mature immune cells, most particularly T cells, would be predicted to result in rapid immune reconstitution. This is known to occur both in mouse transplants using wild-type bone marrow cells and in human allogeneic transplants using matched sibling donors. In contrast, immune reconstitution is often delayed for 3 to 4 months in the gene therapy setting using corrected autologous cells and in the haploidentical allogeneic transplants that are commonly used for patients lacking a matched sibling donor. Such patients require high-grade T-cell depletion of the graft, thereby potentially leading to prolonged periods of inadequate surveillance.

There are certainly other possibilities to be considered; this paper provides new information and systems for evaluating this susceptibility process. It remains important to keep in mind that, unless future human cases provide new information, all cases of retroviral vector-induced leukemia and myelodysplasia have demonstrated clear evidence of vector-induced oncogene activation due to enhancer elements within the vector as a “smoking gun” feature. We think that this is a solvable problem, and it will be interesting to see whether elimination of this specific vector property will eliminate or reduce the oncogenic features of SCID-X1 gene therapy. Less likely, but still not ruled out, is the possibility that new mechanisms of transformation will emerge as limiting features.