Thymidine Kinase Suicide Gene-mediated Ganciclovir Ablation of Autologous Gene-modified Rhesus Hematopoiesis

Abstract

Despite the genotoxic complications encountered in clinical gene therapy trials for primary immunodeficiency diseases targeting hematopoietic cells with integrating vectors; this strategy holds promise for the cure of several monogenic blood, metabolic and neurodegenerative diseases. In this study, we asked whether the inclusion of a suicide gene in a standard retrovirus vector would allow elimination of vector-containing stem and progenitor cells and their progeny in vivo following transplantation, using our rhesus macaque transplantation model. Following stable engraftment with autologous CD34⁺ cells transduced with a retrovirus vector encoding a highly sensitive modified Herpes simplex virus thymidine kinase SR39, the administration of the antiviral prodrug ganciclovir (GCV) was effective in completely eliminating vector-containing cells in all hematopoietic lineages in vivo. The sustained absence of vector-containing cells over time, without additional GCV administration, suggests that the ablation of TkSR39 GCV-sensitive cells occurred in the most primitive hematopoietic long-term repopulating stem or progenitor cell compartment. These results are a proof-of-concept that the inclusion of a suicide gene in integrating vectors, in addition to a therapeutic gene, can provide a mechanism for later elimination of vector-containing cells, thereby increasing the safety of gene transfer.

Introduction

The body of evidence for autologous hematopoietic stem cell (HSC) targeted gene therapy as a potential effective treatment for a variety of human inherited immunodeficiency, neurodegenerative and blood disorders is growing and has been extensively reported.^{1,2,3,4,5,6,7} However, the serious genotoxic complications that have occurred in a subset of patients treated with HSC gene therapy for X-linked severe combined immunodeficiency, X-linked chronic granulomatous disease, and most recently Wiscott–Aldrich syndrome, together with preclinical studies that document nonrandom integration patterns for retroviruses and lentiviruses in animal models and insertional activation of adjacent proto-oncogenes have revealed the risks associated with the use of integrating vectors for gene transfer to HSC.^{8,9,10,11,12,13,14} Investigators have made significant progress in developing vectors less likely to activate adjacent genes by deleting long-terminal repeat enhancers in both retroviral and lentiviral vectors, and utilizing lineage-specific or weaker constitutive cellular promoters such as PGK or EF-1, however, the risk of insertional oncogenesis can never be eliminated.¹⁵ The recent report of a thus far clinically benign clonal hematopoietic expansion resulting from integration of an long-terminal repeat enhancer-deleted lentivirus vector expressing a β-hemoglobin gene and perturbing expression of the adjacent DNA-binding protein gene HMGA2 (high-mobility group AT-hook²) via aberrant splicing in a thalassemia patient treated with gene therapy highlights the need for caution and continued investigation of insertional mutagenesis.⁵ Additional approaches being investigated include utilization of chromatin insulators within vectors,^16,17 development of vectors with more random integration patterns based on human foamy virus¹⁸ or avian sarcoma leucosis virus,¹⁹ and targeted integration via specific zinc-finger nucleases.²⁰ However, in spite of these resources, any interruption of the genome by exogenous genetic material can perturb gene expression and result in adverse genotoxicity.²¹

The “suicide gene” ablation approach to improve gene therapy safety requires inclusion of a suicide gene in the gene transfer vector, encoding a protein that results in cell killing only when activated by a specific compound. A suicide gene generally encodes an enzyme that selectively converts a nontoxic prodrug into highly toxic metabolites, specifically eliminating cells expressing the enzyme.²² The first and thus far most commonly used combination of activator and suicide gene is the drug ganciclovir (GCV) in combination with Herpes simplex virus thymidine kinase (HSVtk) gene, which has been tested in a number of experimental and clinical settings.^23,24 The HSVtk gene product phosphorylates nontoxic GCV into GCV triphosphate, which is incorporated into DNA in replicating cells, inhibiting DNA synthesis and resulting in cell death.

The most successful application of suicide gene gene transfer technology to date has been in the field of allogeneic HSC transplantation. Donor T cells can exert powerful graft-versus-tumor and antiviral effects, however, these immunoreactive cells can also cause difficult to control life-threatening graft-versus-host disease. HSVtk suicide gene-modified donor T cells have been shown to be functional, accelerating immune reconstitution in the recipient and able to contribute to the graft-versus-leukemia effect. Administration of GCV was shown to efficiently ablate gene-modified donor T cells and mitigate active graft-versus-host disease.^25,26,27 Vectors expressing HSVtk have also been used in brain tumor treatment strategies, based on the concept that dividing tumor cells would specifically be transduced in vivo following vector injection, in contrast to normal quiescent neurons, and be rendered susceptible to GCV-mediated killing. Despite promising preclinical data, human randomized clinical trials did not show a benefit for this approach in patients with brain tumors.²⁸ The utility of suicide genes as biosafety mechanisms in gene therapy and regenerative medicine has begun to be explored in additional preclinical models, including ablation of vector-containing transplanted hepatocytes, teratomas derived from embryonic stem cells, and mesenchymal stromal cells and murine embryonic stem cell -derived hematopoietic progenitors in vitro.^29,30,31

Whether this approach can be effective for eliminating transduced long-term multilineage repopulating hematopoietic stem and progenitor cells (HSPC) and their progeny in vivo has never been explored. Given the particularly high risk of adverse outcomes linked to insertional mutagenesis in gene therapy targeting HSPCs, the development of a suicide gene approach for HSPC gene therapy seems warranted.³² The quiescent nature of the most primitive HSPCs and other tissue stem cells could hinder effective killing with suicide gene-activator approaches such as Tk thought to be dependent on DNA replication for toxicity. Using the rhesus macaque nonhuman primate autologous transplantation model, we evaluated the safety, efficacy, and kinetics of GCV ablation of HSVtk gammaretroviral vector-transduced long-term repopulating HSPC cells and their progeny in vitro and in vivo. Our results suggest that sustained and complete ablation of vector-positive hematopoietic cells can be achieved with tolerable doses of GCV.

Results

In vitro effects of GCV on the survival and function of hematopoietic progenitor cells transduced with vectors expressing herpes thymidine kinase suicide genes

We evaluated the effect of GCV on the dose-dependent survival and function of human CD34⁺ (huCD34) cells transduced with a standard retrovirus vector (MND.TkSR39.LNGFR) expressing a highly active mutated Herpes virus thymidine kinase gene (SR39) and a truncated nonsignaling low-affinity nerve growth factor receptor (LNGFR) marker gene (Figure 1a).³³ In initial experiments, huCD34 cells were first transduced with viral supernatants, sorted for LNGFR expression by flow cytometry following staining with anti-CD271 antibody, and then subjected to escalating GCV doses in culture for 96 hours. The percentage of apoptotic (Ann-V⁺ 7-AAD⁺) CD271⁺ transduced cells was 95.8% ± 0.85 in the presence of 5–10 µmol/l GCV, levels achieved in human serum at therapeutic doses of GCV, and 96.7% ± 0.5 at higher 50–100 µmol/l concentrations (Figure 2a). Only 4.72% ± 1.57 of nontransduced huCD34⁺ cells were apoptotic in the therapeutic dose range of GCV under the same conditions (P = 0.005). Apoptosis of cells transduced with the same vector backbone, but lacking the HSVtk suicide gene was 1.52% ± 0.37; not different from apoptosis of nontransduced huCD34 (P = 0.22), confirming that the killing mechanism induced by GCV is highly specific for cells transduced with the suicide gene vector and not a consequence of the drug's reported hematologic toxicity.

Figure 1.

Vectors used for cell and animal studies. (a) MND.TkSR39.LNGFR: The MND backbone vector (5.2 kb) used in the studies contains the myeloproliferative sarcoma virus enhancer (MPSV). The truncated form of the low-affinity nerve growth factor (LNGFR) is the gene marker that is incorporated to the sequence through a viral internal ribosome entry site (IRES) from the encephalomyocarditis virus IRES (600 bp).⁵¹ By having an IRES rather than a separate promoter for LNGFR, the expression of LNGFR should be equal than that of the tk gene. The monoclonal antibody CD271 recognizes the tLNGFR on transduced cells. The Tk gene is the SR39 high-affinity mutant of wild-type HSVtk.³⁷ (b) MLV-Tkmut2S-GFP: The viral backbone is based on the MLV virus (7.1 Kb). The Tk gene has a silent T-C transition to remove a cryptic splice donor site. Green fluorescent protein (GFP) as gene marker is driven through the SV40 virus promoter. (c) Amino acid sequence of SR39Tk and Tkmut2S mutants compared to the wild-type HSVtk. The critical sites for GCV interaction are marked as GCV site in gray boxes.

Figure 2.

Effect of GCV administration on huCD34⁺ cells with or without MND.SR39Tk.LNGFR transduction. (a) After transduction and FACS selection for cells expressing surface CD271 (truncated NGFR), huCD34⁺ cells (closed square) were cultured with different concentrations of GCV for 96 hours, and the % of cell death was determined by flow cytometry via staining for apoptosis markers (Ann-V⁺ and 7-AAD⁺). Nontransduced huCD34⁺ cells (open square) and cells transduced with an MND-GFP vector lacking the HSVtk gene (in gray) were also cultured with GCV for 4 days. Values are mean ± SD (n = 2). *P ≤ 0.005. (b) After in vitro incubation GCV-transduced (closed square) and nontransduced (open square) huCD34⁺ were plated in CFU-C assays. Data were collected at two cell concentrations, and assays were plated in triplicate. Values are the mean ± SD. **P ≤ 0.001. (c) Coculture of mixtures of 10% vector transduced (CD271⁺) and nontransduced huCD34⁺ (closed triangle) under escalating GCV doses, compared with transduced (closed square) and nontransduced (open square) huCD34⁺(n = 1). 7-AAD, 7-amino-actinomycin-D; Ann-V, annexin V; CFU-C, colony-forming units cells; GCV, ganciclovir; GFP, green fluorescent protein; ns, not significant.

We next examined the extent to which 96 hours of GCV exposure influenced the differentiation and proliferation of hematopoietic progenitors. Transduced CD271⁺ huCD34⁺ cells were cultured for 96 hours with or without GCV and then plated in a standard methylcellulose colony-forming unit (CFU) assay. In the absence of GCV, the plating efficiency and characteristics of CFUs derived from CD271⁺ transduced huCD34 cells were similar to results of plating nontransduced CD34⁺ cells (P = 0.64) (Figure 2b), demonstrating that the presence of the suicide transgene itself did not interfere with the in vitro growth and differentiation of huCD34⁺ progenitors. However, when the CD271⁺ transduced huCD34 pre-incubated with GCV were plated, there was no CFU formation, indicating that suicide of vector-containing CFU with GCV exposure was very potent, with complete inhibition of survival and differentiation (Figure 2b). There was no impact of 5–10 µmol/l GCV preincubation on the plating efficiency or characteristics of nontransduced CFU (Figure 2b).

A “bystander” effect has been described in the literature, with some GCV-mediated killing of adjacent nontransduced adherent as well as hematopoietic tumor cells in the presence of Tk-expressing tumor cells.³⁴ Because of concern regarding the impact of any bystander effect on normal marrow function, we investigated for this phenomenon in vitro, realizing that modeling the complex interactions of hematopoietic cells and their environment in the bone marrow in vitro is difficult. HuCD34⁺ cells transduced with MND.TkSR39.LNGFR and sorted for CD271 expression were mixed in a 10% ratio with nontransduced CD34⁺ cells and grown in coculture at high density, and compared to nontransduced and 100% CD271⁺ transduced cells. At the target GCV concentration of 5–10 µmol/l, there was no evidence for a bystander effect, however, at very high concentrations of GCV, there was some evidence of bystander killing, with more cell death observed in the 10% transduced cell mixture than in the untransduced cells, as confirmed by trypan blue viability staining (Figure 2c). This result suggests that, while a bystander killing effect of nontransduced CD34⁺ cells can exist when high doses of GCV are administrated; it was not observed with the obtainable therapeutic doses of the drug (5–10 µmol/l).³⁵

In vivo GCV ablation of herpes thymidine kinase-expressing hematopoietic cells following rhesus macaque autologous transplantation of transduced CD34⁺ cells

Next, we used our nonhuman primate autologous transplantation model to test whether therapeutic doses of GCV could effectively eliminate cells containing the HSVtk gene. Table 1 summarizes the approaches used in a total of five individual animals. The total number of CD34⁺ cells/kg returned to the macaques ranged from 10 × 10⁶ to 14 × 10⁶ /kg. All animals had complete hematopoietic recovery and were stably engrafted for at least 4-6 months prior to initiating the therapy with GCV. Our previous extensive experience with this model indicates that vector-containing hematopoiesis obtained at this time point is generated by clones with very long-term repopulating activity, stable for many years, and consisting at least in part of clones with both myeloid and lymphoid potential.

Table 1. Autologous bone marrow transplantation of rhesus macaques with gammaretrovirus transduced CD34⁺ and regimen of GCV administration long-term after transplant.

Two animals received cells transduced with the MND.TkSR39.LNGFR vector, containing the high-affinity thymidine kinase mutant. By 3-4 months following transplantation, the level of transduced cells as accessed via flow cytometry for LNGFR (CD271) was stable at 2–10% (Figure 3a, c). Expression of the LNGFR marker in different PB lineages was analyzed using flow cytometry for CD271. Representative data are shown in Supplementary Table S1. T lymphocytes (CD3⁺) showed the highest percentage of cells expressing CD271, whereas granulocytes (CD18⁺) had somewhat lower percentages of CD271⁺ cells. Ten months following transplantation, animal RQ7413 received GCV at 10 mg/kg/day, two times a day, for 21 consecutive days. This regimen of GCV did not result in any detectable toxicity or in significant suppression of peripheral blood counts (Supplementary Figure S1a–c). As shown in Figure 3a, with just one treatment course of GCV, the level of CD271⁺ cells steadily decreased, and became completely undetectable in the peripheral blood by the last day of GCV treatment. We performed genomic DNA PCR amplification to ask whether nonexpressing vector-containing cells persisted following GCV treatment. Neither conventional PCR (Supplementary Figure S2a, b and Supplementary Table S2) nor two independent blood sample determinations using real time-quantitative PCR (RT-qPCR) (Figure 3d and Supplementary Figure S3a, b) revealed residual vector-containing cells.

Figure 3.

Marking levels in macaques receiving cells transduced with the MND.TkSR39.LNGFR vector, before and after GCV treatment. Fate of Tk⁺PBL in macaques RQ7413 and DC2C before and after GCV treatment. (a) Level of CD271⁺ PB cells as assessed by flow cytometry is represented before and after the administration of 10 mg/kg/day GCV for 21 days in RQ7413. Dashed black panel represents the GCV infusions. (b) Level of CD271⁺ PB assessed by flow cytometry in animal DC2C pre and 6 months post-GCV therapy. Two cycles of 15 mg/kg/day GCV for 5 days (solid gray arrow), and 20 mg/kg/day for 5 days (dashed gray arrow) were given 1 month apart. (c) Flow cytometric dot plots of peripheral blood and bone marrow samples stained with the anti-CD271 antibody pre- and post-GCV. Left hand panels: animal RQ7413, right hand panels, animal DC2C. (d) HSVtk copy number as assessed by real time-quantitative PCR (RT-qPCR) analysis for vector sequences before and after GCV therapy. Data shown represents means ± SD of C_t values performed in sample duplicates. Dashed line shows limit of detection 10 copies per 200 ng genomic DNA. GCV, ganciclovir; LNGFR, low-affinity nerve growth factor receptor; PBL, peripheral blood leukocytes; UD, undetectable.

In order to mobilize quiescent progenitors and their progeny from the marrow, to determine whether a reservoir of vector-positive cells still persisted in the animal but could not be detected in the peripheral blood, we treated the animal with 5 days of subcutaneous granulocyte colony-stimulating factor 10 µg/kg/day, 12 months following transplantation and 2 months following GCV treatment. Following granulocyte colony-stimulating factor mobilization, we still could not detect vector-containing cells in the blood by flow cytometry or PCR. Both flow cytometry and PCR performed in parallel on PB and bone marrow samples of the animal have remained negative for vector marking now 18 months following the single cycle of GCV treatment (Figure 3b, left panels, and Supplementary Figure S3e, f). These data suggest that cells transduced with the MND.TkSR39.LNGFR vector were completely eliminated from animal RQ7413 following 21 days of GCV treatment.

We confirmed these results in a second rhesus macaque, DC2C, receiving autologous CD34⁺ cells transduced with the MND.TkSR39.tNGFR vector. Four months following transplantation, after stable marking at a level of 2% had been established, the animal was treated with a simpler GCV regimen consisting of 15 mg/kg/day GCV daily for 5 consecutive days, followed by a second cycle of GCV at a dose of 20 mg/kg/day for 5 consecutive days 1-month later (Figure 3c). Following this GCV regimen, the level of vector-containing cells in the blood dropped to undetectable levels by both flow cytometry for CD271 (Figure 3b, right panels) and by quantitative PCR (Figure 3d, and Supplementary Figure S3c–f). This regimen of GCV also did not adversely affect the animal's hematopoiesis or overall health (Supplementary Figure S1d, e).

Efficiency of in vivo ablation depends on the relative sensitivity of the herpes thymidine kinase to GCV

We investigated in vivo ablation utilizing an alternative standard retrovirus vector expressing a wild-type herpes thymidine kinase protein from a splice-corrected cDNA and a green fluorescent protein marker gene (GFP). Three macaques were transplanted with autologous CD34⁺ cells transduced with this MLV.Tkmut2S.GFP vector. All three animals engrafted promptly and had stable marking levels of 1–2.5% as assessed by flow cytometry (Figure 4a, b), somewhat lower than achieved with the MND.TkSR39.LNGFR vector. The GFP expression was analyzed in lymphoid and myeloid lineages separately, and in spite of the variation between animals and overall lower levels of marking with this vector, there was multilineage stable marking in all three animals before GCV therapy (Supplementary Table S1).

Figure 4.

Marking levels in macaques receiving cells transduced with the MLV-Tkmut2S-GFP vector, before and after GCV treatment. (a) Level of GFP⁺ cells in the peripheral blood of three animals transplanted with MLV-Tkmut2S-GFP-transduced autologous CD34⁺ cells. Sustained long-term engraftment demonstrated by GFP⁺ expression in all the animals prior to GCV treatment, and the effect of the GCV treatment (arrows) is shown. Gray arrows represent 10 mg/kg/day GCV administrated intravenous (i.v.) for 21 days at month 10 (macaque RQ7476) or month 8 (macaques 05E119 and RQ7364) after transplant, respectively. Black arrows represent dosing the animals at 15 and 20 mg/kg/day GCV for 5 days, respectively. (b) FACS dot plots illustrating GFP⁺ PBL before and after GCV therapy. GCV, ganciclovir; GFP, green fluorescent protein; LNGFR, low-affinity nerve growth factor receptor.

All three animals were treated with 10 mg/kg GCV twice a day for 21 days, beginning 8–10 months following transplantation. By flow cytometry for GFP⁺ cells, all had a marked decrease but not complete elimination of GFP⁺ cells from the blood (Figure 4a, b). Next we assessed whether escalating the dose of GCV to first 15 mg/kg/day as a single daily dose for 5 consecutive days, then to 20 mg/kg/day for 5 days could eliminate the residual GFP⁺ cells. There was a further drop in marking in two of three animals, but very low levels of residual GFP⁺ cells persisted (Figure 4a). The resistance of residual Tk⁺ PBL to GCV killing was also demonstrated in vitro using sorted GFP⁺ cells from the peripheral blood of macaque 05E119. The GFP⁺ PBL were sorted and plated in 1–10 µmol/l GCV (Supplementary Figure S4a). Only 50% killing was observed after 96 hours of in vitro culture with the drug (Supplementary Figure S4b).

In previous clinical trials utilizing a similar vector for T cell ablation following allogeneic transplantation, GCV-resistant cells were documented to have acquired mutations in the Tk transgene, explaining loss of suicide function.³⁶ Therefore, we examined the sequence of the Tkmut2 gene in DNA extracted from PBL in those animals with persistent GFP⁺ PB cells obtained in two independent experiments. The nucleotide sequence was found to be 100% identical to the one present in the original vector, ruling out this mechanism for resistance at least in the majority of the pool of persistent GFP⁺ cells (Supplementary Figure S4c). The immunophenotype of PBL was evaluated in all animals; no evidence of abnormalities was found, and the animals remain completely well with normal hematopoiesis, suggesting that an abnormal clonal outgrowth was not responsible for persistence of vector-containing, GCV-resistant cells.

Differences in the affinity of the enzyme Tk for its substrate GCV have been documented to impact on GCV sensitivity.^37,38 Thus, we next compared whether a difference existed between the GCV sensitivity among the two Tk versions we used for the animal experiments. K562 cells transduced at an efficiency resulting in on average one copy of vector per transduced cell were challenged with increasing concentrations of GCV for 96 hours. With this regimen, the GCV sensitivity of K562-Tkmut2S-transduced cells in vitro was 1.6–2.5-fold lower than that of SR39Tk-transduced cells (24.1% ± 3.4% versus 51% ± 18%; P = 0.045) (Figure 5) at concentrations of 5 and 10 µmol/l GCV, in two independent experiments. Note that both concentrations of GCV are within the range reached during in vivo drug administration. These results suggest that differential Tk GCV sensitivity could account for the lower efficiency and incomplete ablation observed in vivo in the animals that received autologous transplantation with the vector-driven Tkmut2S gene as compared to the SR39Tk gene. It is also possible that the different vector backbones resulted in less uniform or lower overall level of expression of the Tk gene in Tkmut2S vector, and that in vivo selection resulted in residual clones with insufficient Tk expression to result in effective killing.

Figure 5.

Comparison of the killing efficiency of TkSR39 and wild-type Tk proteins in K562 cells. After transduction, cells were exposed to 5 µmol/l GCV in vitro for 96 hours and assayed for cell death markers Ann-V and 7-AAD. *P = 0.045. Ann-V, annexin V. 7-AAD, 7-amino-actinomycin-D.

Immune response against HSVtk did not contribute to transduced cell ablation

Interferon-γ (IFN-γ) enzyme-linked immunospot assay analysis was performed in the macaques after autologous transplantation to evaluate whether Tk (in RQ7413; RQ7476; 05E119; and RQ7364 macaques) or GFP (in RQ7476; 05E119; and RQ7364 macaques) may have served as immunogenic foreign proteins, thereby contributing to elimination of transduced cells, as has been previously described even in myeloablative settings. Peripheral blood mononuclear cell from control and experimental macaques responded strongly to the incubation with the polyclonal stimuli Con-A and Phorbol 12-Myristate13-Acetate (PMA) (P = 0.10 and P = 0.06, respectively, compared to the response of seven control macaques) (Figure 6a–e and Supplementary Figure S5a–b). In contrast, the incubation of peripheral blood mononuclear cell from experimental macaques with several concentrations of Tk and recombinant GFP peptides did not induce specific IFN-γ–producing T cells (Figure 6a–d); by comparison, control macaques had low but statistically significant immune responses to each of the Tk peptides at each concentration tested (Figure 6e). We also assessed the T cells subsets in the animals, and although the CD4⁺/CD8⁺ ratio was found inverted in three of four experimental macaques, the absolute number of T lymphocytes was within the normal range (Figure 7a,b).³⁹ In summary, we observed no evidence that T-cell–mediated immune responses were generated against the HSVtk transgene in our experimental setting.

Figure 6.

Detection of herpes Tk and green fluorescent protein (GFP)-specific interferon-γ (IFN-γ)–producing T cells in transplanted animals. Number of spots in Elispot assays carried out on peripheral blood mononuclear cells in the presence of priming with the specified stimulators. Each bar represents the mean number of spots in duplicate (RQ7413) or triplicate wells ± SD. (a) Macaque RQ7413 transplanted with autologous cells transduced with the MND. TKSR39.LNGFR vector thus cryopreserved peripheral blood mononuclear cell (PBMC) were only primed at increasing concentrations of Tk but not GFP peptides. (b) Macaque RQ7476 transplanted with autologous cells transduced with MLV-Tkmut2S-GFP, primed with both tk and rGFP. (c) Macaque 05E119 transplanted with autologous cells transduced with MLV-Tkmut2S-GFP primed with both Tk and rGFP. (d) Macaque RQ7364 transplanted with autologous cells transduced with MLV-Tkmut2S-GFP, primed with both Tk and rGFP. (e) Detection of antigen-specific IFN-γ–producing T cells in control macaques to polyclonal stimulus (PMA and Con-A), to Tk, and rGFP peptides. Con-A, concanavalin-A, IFN-γ, interferon-γ, PMA, Phorbol 12-Myristate13-Acetate, rGFP, recombinant green fluorescence protein; Tk, thymidine kinase.

Figure 7.

T cells subsets in experimental macaques subjected to interferon-γ (IFN-γ) Elispot assay between 4 and 6 months after autologous transplantation with MND-TkSR39-LNGFR (RQ7413), and MLV-Tkmut2S-GFP (RQ7476, 05E119, and RQ7364). (a) Frequency of CD3⁺ T cells and the CD4⁺ and CD8⁺ subpopulations. (b) CD4⁺/CD8⁺ ratio of experimental macaques. GFP, green fluorescent protein; LNGFR, low-affinity nerve growth factor receptor.

Discussion

In this study, we show that the administration of well-tolerated treatment courses of GCV to rhesus macaques was effective in eliminating detectable retrovirus vector-containing hematopoietic cells from animals that engrafted long-term with autologous CD34⁺ cells, which were transduced with the highly sensitive herpes thymidine kinase mutant HSVtkSR39 transgene. Even 18 months after the last dose of GCV, flow cytometric and molecular studies remained persistently negative for vector expressing or containing cells in PB and in the bone marrow, suggesting that ablation was effective at both the hematopoietic progenitor and even more primitive long-term repopulating cell level of the animals. We were unable to detect residual vector-containing cells even following cytokine stimulation and mobilization of marrow cells into the peripheral blood. We found no evidence for a T-cell mediated immune response against the HSVtk transgene as a mechanism for cell clearance, suggesting that HSVtk transgene transfer into primitive repopulating cells induced central tolerance, as previously well-described by our group and others^40,41 thereby provided an approach to overcome immune rejection of gene-modified cells, in contrast to studies using Tk in a mature T-cell target population.⁴² No evidence of killing via a bystander effect on nontransduced cells was observed in our in vitro experiments using a mixture of 10% Tk⁺CD34 and 90% nontransduced CD34 cells at clinically obtainable GCV concentrations, and we saw no evidence for unexpected hematologic toxicity of GCV in vivo, suggesting that only specific depletion of sensitive Tk⁺ transduced cells occurred in the bone marrow in the rhesus experiments during normal hematopoiesis, although ruling out some small component of a bystander effect in vivo is impossible.

A marked decrease, although not complete ablation, of vector-containing cells was observed after the administration of GCV to three macaques whose CD34⁺ cells were transduced with the vector containing the splice site-corrected but wild-type protein sequence thymidine kinase transgene Tkmut2S. In patients treated with GCV to ameliorate GvHD after adoptive T cells therapy, resistant cells arose due to the presence of cryptic splice donor and acceptor sites within the endogenous HSVtk gene sequence, resulting in production of mRNA with a 227-bp deletion and thus loss of sensitivity to GCV, with selection for cells containing this deletion in vivo after GCV therapy.³⁶ This issue was resolved by the deletion of the splicing signal in the Tkmut2S transgene, but the protein coding sequence itself remains wild type. In the animals receiving cells transduced with the Tkmut2S vector, there were residual vector-containing cells detected after GCV administration. As far as we could detect by the nucleotide consensus sequences obtained of the Tkmut2S transgene in residual GFP⁺ cells in these animals at two different time points after treatment, resistance did not result from loss of activity mutations in the transgene, but we would not be able to detect partial or full deletions in the very low percentage of persistent vector-containing cells. It seems more likely that the residual GCV-resistant cells resulted from borderline drug sensitivity of primitive hematopoietic cells to killing using GCV. Cells with vector insertions resulting in relatively low expression of Tk might fall below the threshold of sensitivity to GCV, due to low resultant concentrations of GCV-triphophate in very slowly cycling HSPCs. The contrasting complete ablation achieved in the animals receiving cells transduced with the HSVtkSR39 transgene supports this hypothesis, and is supported by our experiments documenting higher activity of the TkSR39 variant compared to the wild-type Tk in killing of hematopoietic K562 cells with GCV. These results confirm the report showing that the mutant TkSR39 has a GCV K_m of 14.3-fold lower than that wild-type Tk.^37,38

Importantly, ablation of Tk⁺ hematopoiesis with GCV was achieved with good clinical tolerance and without significant myelotoxicity. Even dose escalation to 20 mg/kg/day did not produce cytopenias in the three macaques that received these higher doses in an attempt to ablate residual 10% vector-positive Tkmut2S cells. The proliferation-dependent killing mediated by the wild-type Tkmut2 protein makes this suicide protein particularly useful for protocols (i.e., GvHD control after adoptive T cells therapy) requiring preferential elimination of proliferating effector cells.²⁷

The incorporation of toxic GCV triphosphates into the nuclear DNA in dividing cells inhibits new strand synthesis and results in cell death. In addition, the incorporation of these toxic triphosphates into mitochondria DNA has been demonstrated to be an additional potential mechanism of drug killing in cells which are not actively dividing.⁴³ Although these studies used transient expression of much higher levels of Tk using adenoviral gene transfer, a similar effect was found using a liver-specific lentivirus vector in cynomolgous monkeys for ablation of transduced hepatocytes.³⁰ Thus, it is at least conceivable that the GCV ablation of Tk⁺ primitive hematopoietic cells, which are primarily in a quiescent state, may be secondary to damaged mitochondria, abnormal reactive oxygen species production, and thereby genomic instability leading to cell death, rather than incorporation into genomic DNA. However, we cannot exclude the possibility that killing of sensitive Tk⁺ progenitor cells with GCV or more generalized myelotoxicity that is associated with this drug could have recruited quiescent HSCs into cell cycle and thus then made them susceptible to GCV killing via incorporation of the GCV metabolite nucleosides into genomic DNA. Finally, it is possible that the most primitive, deeply-quiescent HSCs were not transduced with these vectors, however the long-term stable, multilineage engraftment with retroviral-vector-containing cells in nonhuman primate models as well as in human clinical trials strongly suggests that functional HSCs are transduced with these vectors, albeit at low efficiency.^44,45

The possibility of treating and even curing a number of congenital blood, immunologic and neurologic diseases with HSC-targeted gene therapy now seems closer than ever, by combining the approaches shown to be effective in the landmark clinical trials with intense efforts to understand genotoxicity and improve vector safety. But even with safer vectors, the risks of insertional mutagenesis can be mitigated but not completely avoided by any gene transfer strategy requiring permanent integration of exogenous DNA sequences into the genome. Therefore, it seems prudent to develop safety approaches, such as suicide gene strategies, to permit ablation of vector-containing cells if needed. In contrast to the addition of insulator sequences in the viral construct, we found that the inclusion of a suicide gene sequence did not hamper viral titers and performed well driven by the proviral promoters of MND and MLV to achieve efficient transduction of rhesus CD34⁺cells, with sustained expression of the GFP or LNGFR marker genes also encoded by the vectors. In addition to the vector safety modifications ensuring that integrated vector proviruses are less likely to activate adjacent proto-oncogenes, there may be a role for inclusion of suicide genes into vectors,^46,47 further decreasing risk. As other stem cell populations with potentially even higher risks for tumorigenesis are developed for therapeutic applications, such as induced pluripotent stem cells, suicide gene strategies to enhance safety in these settings may also merit consideration.

Our study in a clinically relevant large animal model presents the first in vitro and in vivo proof-of-concept that including a suicide gene sequence in integrating viral vectors can permit controlled and nontoxic ablation of vector-modified primitive hematopoietic cells and their progeny. It is encouraging that even highly quiescent long-term repopulating cells appear to be efficiency deleted via this strategy. Of course there are major hurdles to overcome before more generally applying this strategy. Although we did not find evidence of transgene deletion or silencing in our experimental model, a previous study reported that Tk-expressing lymphoma cells (EL-4) injected in mice resulted in GCV-resistant tumor relapses due to genetic and epigenetic modifications of the retroviral insert resulting in gene loss or silencing, suggesting that such events could present obstacles, however, chromosomal instability and silencing may be more common in fully transformed cell lines than primary hematopoietic cells.⁴⁸ In this context, it will be critical to test whether cells fully transformed by insertional mutagenesis or via overexpression of an oncogenic transgene such as HoxB4 can be ablated in animal models. It would be ideal if only transformed or abnormal preleukemic cells could be ablated, retaining engraftment with gene-corrected cells not impacted by insertional mutagenesis. Strategies utilizing suicide gene promoters active only in transformed cells could be considered, for instance, a promoter only active in cells overexpressing Lmo2 given the frequency of dysregulation of this particular gene in the cases of T-ALL resulting from insertional genotoxicity. Or constitutive promoters could be utilized, with ablation only carried out as a last resort should progression towards leukemia be detected, and other treatment options exhausted.

Because both GCV and the related drug acyclovir are valuable resources for controlling cytomegalovirus, herpes simplex, and herpes zoster infections in the clinical setting, particularly following stem cell transplanation and in other patients with acquired or congenital immune deficiencies, development of suicide gene strategies employing agents not commonly used in clinical practice would be desirable. We are currently testing an inducible caspase 9 suicide gene in combination with a nontoxic and specific chemical dimerizer (AP1903) (Bellicum, Houston, TX) to initiate apoptosis in vector-containing cells. This strategy recently has shown to be effective and safe in ablation of allogeneic T cells in a human clinical trial.⁴⁹

Materials and Methods

HSVtk vectors. The MND.TkSR39.IRES.LNGFR vector PG13 producer line (Figure 1a) was provided by Dr Frank Marini (Wake Forest Institute for Regenerative Medicine, Winston-Salem, NC).³³ Figure 1c shows the deduced amino acids substitutions in the HSVtk mutant SR39. The MLV-Tkmut2S-GFP vector PG13 producer line, provided by Dr Claudio Bordignon (Molmed S.p.A., Milano, Italy) (Figure 1b), drives Tkmut2 expression from the viral long-terminal repeat and the transgene coding amino acid sequence is identical to wild-type HSVtk (Figure 1c). The viral titers collected from producer cell lines were 0.5–2 × 10⁶ viral particles/ml for both vectors.

Collection and transduction of human and rhesus CD34⁺ cells. Young rhesus macaques weighing 3–5 kg were handled in accordance with the Institute of Laboratory Animal Research guidelines. Protocols were approved by the Animal Care and Use Committee of the National Heart, Lung, and Blood Institute (National Institutes of Health, Bethesda, MD). Peripheral blood mobilization with granulocyte colony-stimulating factor and stem cell factor (SCF), CD34 immune-selection and ex vivo transduction procedures were performed as described.⁵⁰ A 24-hour prestimulation of 2.5–5 × 10⁵/ml rhesus CD34⁺ cells in the presence of rhuSCF (100 ng/ml), rhuTPO (100 ng/ml), and rhuFlt-3L (100 ng/ml), was performed in CH-296 fibronectin-coated T162 flasks (RetroNectin; Takara, Bio, Otsu, Japan). Freshly collected retrovirus supernatants at calculated multiplicity of infection of 2, and 4 µg/ml protamine sulfate were added to the cultures at 24, 48, and 72 hours. After 96 hours total culture, the CD34⁺ cells were collected and resuspended for in vitro studies or autologous transplantation. Human granulocyte colony-stimulating factor mobilized peripheral blood CD34⁺ cells were obtained from normal volunteers entered on an institutional review board-approved clinical protocol.

In vitro GCV sensitivity. MND.SR39Tk.LNGFR-transduced human CD34⁺cells were sorted for the truncated LNGFR expression after staining with a phycoerythrin-CD271 monoclonal antibody (BD Pharmigen, San Diego, CA). CD271-positive cells and nontransduced control CD34⁺ cells were plated at a concentration of 10⁵ cells/ml in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Hyclone, Logan, UT), 1% penicillin-streptomycin, and rhuSCF (100 ng/ml), TPO (100 ng/ml), and rhuFlt-3L (100 ng/ml). After 24 hours of culture, 0, 5, 10, 50, or 100 µmol/l GCV (Cytovene-IV; Roche Laboratories; Nutley, NJ) was added to the wells, and the cultures were continued for a total of 96 hours. Cells were harvested and stained with an annexin V phycoerythrin- conjugated monoclonal antibody and 7-amino-actinomycin D (7-AAD) (BD Pharmigen) to assess cell death via flow cytometry. Data were normalized using the formula: percentage of killing induced by the GCV = 100% – (% viability in cells treated with GCV / % viability of nontreated cells).

The capacity to differentiate into CFU was evaluated in HSVTk-transduced and nontransduced huCD34⁺ cells. Sorted CD271⁺ CD34⁺ were first cultured with 0, 5, 10, 50, or 100 µmol/l GCV for 96 hours as described above, then seeded in dishes with methylcellulose medium (MethoCult H4034; Stem Cell Technologies; Vancouver, BC) at a concentration of ≈ 10³ cells/dish. The culture dishes were incubated at 37 °C with 5% CO₂ for 15 days. Colonies with >50 cells were scored as positive.

Rhesus autologous transplantation and in vivo GCV treatment. Each animal received two 500-cGy doses of total body irradiation at 24 and 48 hours preceding infusion of transduced autologous CD34⁺ cells, as described.⁵⁰ Autologous CD34⁺ cells were collected from transduction cultures, resuspended in autologous serum, and infused intravenously. Following engraftment, peripheral blood cells were collected and monitored for GFP expression, or for LNGFR expression by flow cytometry for CD271⁺ cells.

GCV (Cytovene-IV; Roche Laboratories) was administered intravenously at 10 mg/kg/day split twice daily for 21 days. In three animals (macaques ID# RQ7476, 5E0119, and RQ7364) demonstrating persistence of vector-containing cells following the initial GCV treatment, the dose was escalated to 15 and then to 20 mg/kg/day given once daily for 5 days. Complete blood counts as well as blood chemistries were monitored throughout.

Quantitative real-time PCR (q-RTPCR) for the HSVtk transgene. To measure copy numbers of the HSVtk transgene in all animals, TaqMan PCR analyses using forward primer 5′-AAACCACCACCACGCAACTG-3′, reverse primer 5′-CATCGGCTCGGGTACGTAGAC-3′ (IDT, Coralville, IA) and probe 5′ –FAM- TGGTGGCCCTGGGTTCGCG-TAMRA (IDT) were performed with DNA extracted from the macaques' blood cells using the DNeasy Blood Mini kit (Qiagen, Huntsville, AL). To quantify HSVtk gene copy number, a plasmid standard curve ranging from 10⁷ to 10 copies was generated using tenfold serial dilutions of plasmids known to contain the target viral sequences TkSR39 and Tkmut2, respectively. The plasmid DNA was suspended in normal genomic DNA of a control macaque to stabilize the molecule. The macaque's β-actin gene was used to normalize the amount of genomic DNA, primers and probes: Fw primer 5′-GGATCAGCAAGCAGGAGTATGA-3′, Rv primer 5′-GCGCAAGTTAGGTTTTGTCAAG-3′, and probe 5′ –FAM- TCCATGGTCCACCGCAAATGCT –TAMRA (IDT). Reactions were performed using Chromo 4 (BioRad, Hercules, CA) with universal mastermix (Applied Biosystems, Foster City, CA). The amplification conditions were 2 minutes at 50 °C and 10 minutes at 95°, followed by 45 cycles of 95 °C for 15 seconds, 60 °C for 1 minute. Data were analyzed with Opticon software (BioRad) using the following parameters: average over cycle range, manual threshold above the nontemplate control or the normal DNA control; and, for baseline and threshold determinations, a cycle range between 3 and 15 cycles. We established the acceptance criteria of the calibrator curve as follows: slope (–3.23 to –3.77), Y-intercept (37.61 to –44.52), and R² values (0.984–1.008). The calibrator curve could be edited by removing no more than three data points.

Interferon-γ enzyme-linked immunospot assay. Cryopreserved peripheral blood mononuclear cells from experimental macaques at 4 months (RQ7476, 05E119, and RQ7364) or 6 months (RQ7413) after transplantation and before GCV treatment (> 96% viable, and at 5 × 10⁵/well, along with peripheral blood mononuclear cell from seven control macaques, were incubated with a mix of two targets Tk antigens: the N-terminal (sc-28037) and the internal Tk region (sc-28038) between residues 50 and 100 and between residues 215–265, respectively (Santa Cruz Biotechnology, Santa Cruz, CA)] at concentrations of 1, 2.5, and 5 µg/ml. The same procedure was carried out using recombinant GFP (Clontech, Mountain View, CA). Additionally, the immune response to polyclonal mitogens was tested in triplicate using PMA (20 µg/ml) and Concanavalin A (5 µg/ml) (all from Sigma-Aldrich, St Louis, MO). The incubation was carried out in 96-well plates at 37 °C in 5% CO₂ incubator for 2 days. The cells were transferred to 96-well MultiScreen-IP polyvinylidene difluoride (Millipore, Billerica, MA) plates precoated with 10 mg/ml mAb GZ-4 (against monkey IFN-γ) (Mabtech, Nacka Strand, Sweden), and incubated at 37 °C, for 24 hours. Triplicate test and negative control wells (Dulbecco's modified Eagle's medium + 10% fetal bovine serum) were set up for all assays. Biotinylated secondary mAb to IFN-γ (7-B6-1) (Mabtech) at 1 µg/ml was added. Avidin–Peroxidase Complex (Vector Laboratories, Burlingame, CA), and AEC-substrate solution (Sigma-Aldrich) were used according to the manufacturers' instructions. Plates were dried at room temperature and then shipped to Zellnet Consulting (New York) for spot counting using an enzyme-linked immunospot assay reader system. Positive response was considered when the number of spots in the sample was more than five times the negative control response. The mean number of spots and standard deviation for triplicate wells was reported. For statistical analysis, the significance of differences between means was evaluated by the two-tailed t-test.

SUPPLEMENTARY MATERIAL Figure S1. (a–c) Peripheral blood values in macaque RQ7413 over time after transplantation and after GCV treatment. (d), (e) Peripheral blood values in macaque DC2C over time after transplantation and after GCV treatment. WBC-white blood count, ANC-absolute neutrophil count. SEGS-segmented polymophonuclear granulocytes, LYMPHS-lymphocytes, MONOS-monocytes. Figure S2. (a) Agarose gels showing PCR amplification of a fragment of 139 bp size within the TKSR39 transgene, and (b) the β-actin gene (232 bp size) on samples from the peripheral blood of macaque RQ7413 at different time points after transplant and after 21-days of GCV treatment. As negative control, genomic DNA extracted from PB cells prior to transplantation is included. The positive control is TKSR39-LNGFR plasmid DNA. Figure S3. (a) Amplification plot of a real-time quantitative PCR for the TKSR39 transgene in macaque RQ7413 shows samples falling out of detection range (Ct35) after therapy with GCV. 10-fold diluted plasmid standards ranging from 1,000,000 down to 10 copies of TKSR39 plasmid DNA are shown. (b) The calibrator curve and acceptance criteria regarding slope, Y-intercept, and R^2 values are described in Methods section. (c), (d) Amplification plot and calibrator curve respectively of RT-qPCR of TKSR39 transgene in animal DC2C. (e), and (f). Amplification plot and calibrator curve of RT-qPCR of the TkSR39 transgene performed in cells of bone marrow samples in animals transplanted with autologous CD34 cells transduced with MND-TkSR39-LNGFR vector (RQ7413 and DC2C) at ~18 months after GCV administration. Figure S4. (a) Representative flow cytometry dot plot shows the persistance of GFP⁺ peripheral blood cells in macaque 05E119 one month following 20 mg/kg/day GCV. Mononuclear cells from this time point were sorted for GFP⁺ and these cells were used for the in vitro killing assay with GCV. (b) Viability of 05E119 GFP⁺ MNC and control animal MNC after in vitro treatment with a range of therapeutic plasma doses of GCV. (c) Nucleotide coding sequence of Tkmut2 transgene obtained from sequencing of residual in GFP⁺ cells in animals transplanted with autologous CD34 cells transduced with MLV-Tkmut2S-GFP (RQ7476, 05E119, and RQ7364) following GCV treatment compared to the coding sequence in the plasmid. Abbreviation: MNC: mononuclear cells. Figure S5. (a) Triplicate wells from an Elispot assay of a representative experimental macaque (RQ7476) are shown. (b) Wells in triplicate from an Elispot assay of a control macaque showing the negative control, positive control, and positive wells with response to the higher TK concentration used for the assay. Abbreviation: IFN-γ: interferon gamma, TK: thymidine kinase, Con-A: Concanavaliti-A, PMA: Phorbol 12-Myristatel3-Acetate, rGFP: recombinant green fluorescence protein, PHA-P: Phytohemagglutinine, DMEM: control culture media. Table S1. Lineage marking in macaques transplanted with autologous CD34 cells transduced with MND-TK-SR39 (RQ7413 and DC2C) and MLV-Tkmut2S-GFP (RQ7476, 05E119, and RQ7364) long-term after transplantation. Table S2. Sequences of primers used for PCR and sequencing reactions.

Supplementary Material

Figure S1.

(a–c) Peripheral blood values in macaque RQ7413 over time after transplantation and after GCV treatment. (d), (e) Peripheral blood values in macaque DC2C over time after transplantation and after GCV treatment. WBC-white blood count, ANC-absolute neutrophil count. SEGS-segmented polymophonuclear granulocytes, LYMPHS-lymphocytes, MONOS-monocytes.

Figure S2.

(a) Agarose gels showing PCR amplification of a fragment of 139 bp size within the TKSR39 transgene, and (b) the β-actin gene (232 bp size) on samples from the peripheral blood of macaque RQ7413 at different time points after transplant and after 21-days of GCV treatment. As negative control, genomic DNA extracted from PB cells prior to transplantation is included. The positive control is TKSR39-LNGFR plasmid DNA.

Figure S3.

(a) Amplification plot of a real-time quantitative PCR for the TKSR39 transgene in macaque RQ7413 shows samples falling out of detection range (Ct35) after therapy with GCV. 10-fold diluted plasmid standards ranging from 1,000,000 down to 10 copies of TKSR39 plasmid DNA are shown. (b) The calibrator curve and acceptance criteria regarding slope, Y-intercept, and R^2 values are described in Methods section. (c), (d) Amplification plot and calibrator curve respectively of RT-qPCR of TKSR39 transgene in animal DC2C. (e), and (f). Amplification plot and calibrator curve of RT-qPCR of the TkSR39 transgene performed in cells of bone marrow samples in animals transplanted with autologous CD34 cells transduced with MND-TkSR39-LNGFR vector (RQ7413 and DC2C) at ~18 months after GCV administration.

Figure S4.

(a) Representative flow cytometry dot plot shows the persistance of GFP⁺ peripheral blood cells in macaque 05E119 one month following 20 mg/kg/day GCV. Mononuclear cells from this time point were sorted for GFP⁺ and these cells were used for the in vitro killing assay with GCV. (b) Viability of 05E119 GFP⁺ MNC and control animal MNC after in vitro treatment with a range of therapeutic plasma doses of GCV. (c) Nucleotide coding sequence of Tkmut2 transgene obtained from sequencing of residual in GFP⁺ cells in animals transplanted with autologous CD34 cells transduced with MLV-Tkmut2S-GFP (RQ7476, 05E119, and RQ7364) following GCV treatment compared to the coding sequence in the plasmid. Abbreviation: MNC: mononuclear cells.

Figure S5.

(a) Triplicate wells from an Elispot assay of a representative experimental macaque (RQ7476) are shown. (b) Wells in triplicate from an Elispot assay of a control macaque showing the negative control, positive control, and positive wells with response to the higher TK concentration used for the assay. Abbreviation: IFN-γ: interferon gamma, TK: thymidine kinase, Con-A: Concanavaliti-A, PMA: Phorbol 12-Myristatel3-Acetate, rGFP: recombinant green fluorescence protein, PHA-P: Phytohemagglutinine, DMEM: control culture media.

Table S1.

Lineage marking in macaques transplanted with autologous CD34 cells transduced with MND-TK-SR39 (RQ7413 and DC2C) and MLV-Tkmut2S-GFP (RQ7476, 05E119, and RQ7364) long-term after transplantation.

Table S2.

Sequences of primers used for PCR and sequencing reactions.