Thymidine Phosphorylase Is Both a Therapeutic and a Suicide Gene in a Murine Model of Mitochondrial Neurogastrointestinal Encephalomyopathy

ORIGINAL ARTICLE Thymidine phosphorylase is both a therapeutic and a suicide gene in a murine model of mitochondrial neurogastrointestinal encephalomyopathy

S López-Estévez1,5, G Ferrer1,5, J Torres-Torronteras2,3, MJ Mansilla1, S Casacuberta-Serra1, L Martorell1, M Hirano4, R Martí2,3 and J Barquinero1

Suicide gene therapy (SGT) is a promising strategy for treating cancer. In this work, we show that thymidine phosphorylase (TP) deficiency, the underlying genetic defect in mitochondrial neurogastrointestinal encephalomyopathy (MNGIE), presents an opportunity to apply SGT using capecitabine, a commonly used prodrug that is converted into 5-fluorouracil by TP. Using an immortalised B-lymphoblastoid cell line from a patient with MNGIE, the tumourigenic EL-4 cell line, lentiviral vectors encoding TP and a double knockout (TympÀ/ÀUpp1À/À) murine model, we found that EL-4 cell-derived TP+ tumours were exquisitely sensitive to capecitabine and generated a significant local bystander effect. In addition, we detected a spontaneous cytolytic immune response in a significant fraction of the animals surviving more than 20 days after termination of the therapy. These data indicate that, in individuals lacking TP expression, TP is a highly specific suicide gene, which can be used to treat tumours that could hypothetically arise in MNGIE patients undergoing gene therapy, as these tumours will likely originate from the gene-modified cells and will be selectively targeted by capecitabine. These observations have important implications for gene therapy for MNGIE.

Gene Therapy (2014) 21, 673–681; doi:10.1038/gt.2014.41; published online 8 May 2014

INTRODUCTION haematopoietic stem cell gene therapy is an effective long-term Mitochondrial neurogastrointestinal encephalomyopathy (MNGIE) treatment even at levels of correction below 10% of the bone 3 is a rare metabolic disease caused by autosomal recessive marrow cells and we are currently investigating novel preclinical mutations in TYMP, which encodes the enzyme thymidine gene therapy strategies that can be easily translated into humans. phosphorylase (TP). This enzyme catalyses the first step in the Depending on the target, long-term correction of genetic catabolism of thymidine (dThd) and deoxyuridine (dUrd); there- defects typically requires vectors that integrate into the genome fore, in patients with TP deficiency, these two nucleosides of target cells therapeutic genes that can be transmitted to accumulate to reach abnormally high systemic and plasma levels, progeny cells. Retroviral vectors were initially used in clinical gene which can be used diagnostically.1 This imbalance selectively therapy trials and, two decades later, are still the most widely used impairs mitochondrial DNA synthesis, which results in mitochon- vectors in stem cell-based applications and have been successfully drial DNA depletion, multiple deletions and site-specific point applied to treat inherited diseases including several primary mutations, without detectable changes in the nuclear DNA. In immunodeficiencies.4 Although these successes have brought healthy individuals and in heterozygous carriers of pathogenic renewed hope to the discipline, the development of T-cell TYMP mutations (who are otherwise asymptomatic), dThd and leukaemias in some of the patients treated unveiled the risk of dUrd produced during normal metabolism cross cell membranes insertional oncogenesis by integrative vectors. Although those via specific transporters, diffuse into the extracellular fluid and are first-generation vectors have been progressively replaced by more cleared from the circulation by tissues expressing the enzyme; efficient and safer vectors, like self-inactivating lentiviral vectors, thus, their systemic concentrations are maintained below a non- the potential oncogenic risks still represent a major concern to the 1 toxic threshold (approximately o0.05 μM in plasma). Over the clinical application of gene therapy. last years, we have studied different gene therapy strategies to Suicide gene therapy (SGT; also known as Gene Directed restore TP activity and to correct the biochemical alterations in Enzyme/Prodrug therapy or Gene Prodrug Activation Therapy) can TympÀ/ÀUpp1À/À double knockout (dKO) mice (the only be defined as the therapeutic introduction, into target cells or available animal model of MNGIE). These mice have no apparent tissues, of heterologous genes encoding enzymes able to trans- clinical abnormalities but display a biochemical phenotype form an inactive prodrug into an active drug. Most suicide genes (with increased systemic levels of dThd and dUrd) and it has encode enzymes involved in the nucleotide metabolism, for been reported that elderly animals also have mDNA depletion in example, herpes virus thymidine kinase or cytosine deaminase, the brain.2 We previously reported that sub-myeloablative which confer sensitivity in the transduced cells to ganciclovir or

1Gene and Cell Therapy Laboratory, Vall d'Hebron Research Institute (VHIR)/Universitat Autònoma de Barcelona, Barcelona, Spain; 2Mitochondrial Pathology Unit, VHIR/Universitat Autònoma de Barcelona, Barcelona, Spain; 3Biomedical Network Research, Centre on Rare Diseases (CIBERER) ISCIII, Madrid, Spain and 4Department of Neurology, Columbia University, New York, NY, USA. Correspondence: Dr J Barquinero, Gene and Cell Therapy Laboratory, Vall d'Hebron Research Institute (VHIR)/Universitat Autònoma de Barcelona, Passeig Vall d'Hebron 119-129, 08035 Barcelona, Spain. E-mail: [email protected] 5These authors contributed equally to this work. Received 6 September 2013; revised 7 March 2014; accepted 26 March 2014; published online 8 May 2014 TP is both a therapeutic and a suicide gene in MNGIE S López-Estévez et al 674 5-fluorocytosine, respectively. This strategy has been used in combination of exogenous TP with 5-FU prodrugs has been used clinical trials to treat cancers and other conditions such as graft- experimentally as a form of SGT,13–15 with a documented versus-host disease.5–7 Capecitabine is an orally administered bystander effect.16 With this background, we hypothesised that À À chemotherapeutic agent used in the treatment of colorectal and the absence of TP activity in MNGIE patients (or in Tymp / À À metastatic breast cancers, as well as other cancer types. It is a Upp1 / dKO mice) would render virtually all their cells resistant TP-activated oral fluoropyrimidine that was rationally designed to to capecitabine because of inability to convert the prodrug into generate 5-fluorouracil (5-FU) preferentially within tumours. Upon 5-FU. One implication of this phenomenon is that capecitabine oral administration, this prodrug is rapidly absorbed in the will not be effective in treating tumours in MNGIE patients. On gastrointestinal tract as an inactive molecule. In the liver, it is the other hand, if a MNGIE patient is eventually treated with gene converted to 5′-deoxy-5-fluorocytidine by the action of a hepatic therapy and develops a tumour associated with the procedure carboxyl-esterase, which is subsequently metabolised to 5′-deoxy- (because of insertional mutagenesis), the therapeutic gene will 5-fluorouridine (doxifluridine, 5′-DFUR) by cytidine deaminase, also behave as a suicide gene, as the tumour will most likely derive and then to 5-FU by the TP expressed in the tumour, the liver or from transduced cells expressing TP (as opposed to the rest of the other normal tissues (Figure 1).8,9 When present, intracellular tissues in the organism), a unique scenario that will make tumour TP can convert 5-FU to 2'-deoxy-5-fluorouridine (2'-DFUR), which is cells selectively sensitive to the prodrug. Herein, we provide further phosphorylated by thymidine kinase 1 to the active evidence that both in vitro and in the murine model of MNGIE, the molecule fluorodeoxyuridine monophosphate, which inhibits therapeutic gene can also behave as a suicide gene. These results thymidylate synthase and disrupts DNA synthesis.10 On the other can most likely be extrapolated to MNGIE patients undergoing hand, 5-FU can undergo a series of enzymatic steps that result in gene therapy in the hypothetical scenario that they develop the generation of fluorouridine triphosphate, a false metabolite for tumours in association with the therapy. Thus, this form of SGT RNA synthesis. 5-FU catabolism is mainly due to the action of may represent a safeguard in cases of oncogenicity, although at dihydropyrimidine dehydrogenase, which is abundant in the liver. the potential expense of elimination of all gene-corrected cells, The relative selectivity of capecitabine in cancer partially derives thereby neutralising the therapeutic effect. However, this would from the fact that most tumours naturally overexpress TP, which probably be an acceptable side effect if the priority is to save the acts as an angiogenic factor, primarily through the production of patient's life, as gene therapy can be performed again later on. 2-deoxy-D-ribose.11 In most tumours, TP has a dual role. Although Moreover, we also show that the use of this murine model its angiogenic activity (mainly due to the release of 2-deoxy-D- provides an optimal tool to investigate various aspects of SGT with capecitabine, such as local and distant bystander effects or ribose) favours tumour growth, its overexpression sensitises the potential synergies between chemotherapy and the anti-tumour cells to the toxic effects of 5-FU and 5-FU prodrugs.12 This benefit immune response. is due to the same principle that supports SGT. In fact, the

RESULTS Cells expressing TP generate 5-FU upon exposure to 5′-DFUR Levels of 5-FU, 5′-DFUR and dThd in culture supernatants of both TP+ (transduced with lentiviral vectors encoding TP) and sham- transduced B-lymphoblastoid cells (B-LCs) from a MNGIE patient3 exposed to either 5-FU or 5′-DFUR (a one-step prodrug of 5-FU) for 48 h were assessed by liquid chromatography coupled to tandem mass spectrometry. As expected, 5-FU was detected in the supernatants of all cell cultures exposed to this drug and in cultures of TP+ B-LCs exposed to 5′-DFUR but not in their sham- transduced counterparts (their levels were negligible), indicating that these cells could not transform 5′-DFUR into 5-FU (two-tailed paired t-test; Po0.001, for the lower concentration; Po0.01 for the higher concentration). In addition, in the supernatants of TP+ cell cultures, 5-FU concentration levels after 48 h were propor- tional to the dose of 5′-DFUR used (Figure 2a). In contrast, 5′-DFUR was virtually absent in the supernatants of cultures exposed to 5- FU (5-FU is physiologically converted to 5-fluoro-2′-deoxyuridine (5-FDUR), but not to 5′-DFUR; data not shown). However, in TP+ cell cultures exposed to the lower concentration of the prodrug, Figure 1. Metabolic conversion of capecitabine to 5-fluorouracil the level of 5′-DFUR was 1.9-fold lower than in their sham- (5-FU). Upon oral administration and absorption in the intestine, transduced counterparts (Po0.01), most likely indicating its ability capecitabine (Xeloda) is first metabolised to 5′-deoxy-5-fluorocyti- to transform it to 5-FU (see Figure 2b). In the cultures exposed to dine (5′-DFCR) by carboxyl-esterases (CE) present in the liver and the higher 5′-DFUR concentration, the ratio was lower (1.2-fold, then to 5′-deoxy-5-fluorouridine (5′-DFUR) by cytidine deaminase Po0.01), which may be due to the saturation of the enzyme. This ′ (CyD). 5 -DFUR conversion to 5-FU is catalyserd by thymidine result is also in agreement with the results observed in Figure 2a. phosphorylase (TP), which is present in many tissues and highly expressed in the majority of tumours. 5-FU can be further metabolised to 2′-DFUR by TP and phosphorylated to fluorodeox- 5′-DFUR has selective toxicity for TP+ B-LCs, and exerts a bystander yuridine monophosphate (FdUMP) by thymidine kinase (TK) or effect in vitro fl either converted to uorodeoxyuridine diphosphate (FdUDP) To evaluate the sensitivity of the two B-LC lines to 5-FU and through the orotate pyroribosyltransferase (OPRT) or ribonuceotide 5 5′-DFUR, 2 × 10 cells were exposed to 50 or 500 μM of the two reductase (RR) pathway. FdUMP is an active compound that inhibits fl thymidylate synthase (TS), resulting in dTTP depletion that interferes substrates and cell viability was assessed 48 h later by ow with DNA replication and repair. FdUDP can be further depho- cytometry using APC-annexin V and 7-aminoactinomycin D sphorylated to FdUMP or either phosphorylated to FdUTP that can (7-AAD) labelling. As expected, TP+ B-LCs were sensitive to both be incorporated into DNA. 5-FU and 5′-DFUR, but sham-transduced B-LCs were less sensitive

Figure 2. Quantification of 5-FU (a) and 5′-DFUR (b). Cell cultures of sham-transduced B-LCs (dashed bars) or TP+ B-LCs (solid bars) were exposed to either 5-FU or 5′-DFUR and the concentrations of both substances were measured 48 h later in the culture supernatants (*Po0.05; **Po0.01; ***Po0.001). 5-FU and 5′-DFUR concentrations were lower than those initially added to both TP+ and TPÀ cell cultures, which could be due to TP-independent metabolic transformations of both compounds. to 5′-DFUR and 5-FU, but only at high doses (Figure 3a). treated with capecitabine, TP+ tumours were absent or barely To investigate a potential bystander effect in vitro, each cell type detectable. In contrast, tumours from animals treated with vehicle was exposed to the culture supernatant of the other cell type and and sham-transduced cell-derived tumours treated with capecita- incubated for 48 h in the presence of the prodrug. Both cell types bine in dKO animals were clearly palpable and larger than TP+ were sensitive to the supernatants of their counterpart cultures tumours from the animals treated with the prodrug. 5-FU was exposed to 5′-DFUR, suggesting that indeed there is a bystander detectable in two TP+ tumours (out of two evaluable) and in one effect in vitro (see Figure 3b). In addition, to investigate the sham-transduced tumour out of three wt mice treated with potency of the bystander effect in vitro, mixtures of TP+ (stained capecitabine. To confirm the bioavailability of the prodrug, we with PKH26) and TPÀ B-LCs at different ratios were exposed to also analysed the 5′-DFUR concentration in all tumours. It was 50 μM of the prodrug for 48 h and cell viability was measured in detectable in all tumours from mice treated with capecitabine. both cell populations by flow cytometry. As expected, TP+ cells Because the conversion of 5′-DFUR to 5-FU is catalysed by TP; to showed homogeneously low rates of survival, whereas in their assess whether 5-FU was only present in TP-expressing tumours, TPÀ counterparts, survival rates were ratio dependent (Figure 3c). we measured TP enzyme activity in all tumours. Sham-transduced The reduced survival of TP+ cells at decreasing TP+/TPÀ ratios tumour cells did not show detectable TP activity in the four dKO could be attributed to the fact that, because the cell mixtures were mice, whereas in their wt counterparts it ranged from 17 to 35 set with a minimum fixed amount of TP+ cells, cell densities and nmole Thy per h per mg protein, probably due to stromal cells nutrient consumption was higher in the culture wells with the present in the tumour, from host origin, that can express TP. lower cell ratios, especially after 48 h. TP activity in TP+ tumours ranged between 824 and 840 nmol Thy per h per mg protein in animals treated with vehicle. However, fi + TP activity in the tumours is required to convert 5′-DFUR into 5-FU this activity was signi cantly reduced in the smallest TP tumours and metabolises dThd in dKO mice that were treated with capecitabine, probably because the cells with the highest TP expression levels had already been killed by Having demonstrated that enforced TP expression conferred + sensitivity to 5′-DFUR and that there was a bystander effect the therapy. Finally, dThd content was reduced in the TP tumours in vitro, we evaluated whether capecitabine, an oral 5-FU prodrug analysed, reaching undetectable levels in the tumours with the and a 5′-DFUR precursor that is commonly used in oncology, had highest TP activity (Table 1). a selective therapeutic effect for TP+ tumours in an in vivo tumour model. To this end, we used EL-4 murine cells, a tumorigenic Capecitabine has a TP-dependent therapeutic effect on an thymoma-derived cell line previously shown not to express TP. experimental EL-4 tumour model and reveals a local bystander Upon confirming that TP activity was absent in these cells, we effect transduced them with lentiviral vectors encoding either human TP In another set of experiments, upon demonstration that the TP+ and eGFP or only eGFP.3 Highly enriched populations (>99%) of EL-4 cells were sensitive and their sham-transduced counterparts transduced cells were generated by cell sorting based on eGFP were resistant to 5′-DFUR in vitro (data not shown), the expression. therapeutic effect of capecitabine was evaluated in an in vivo To determine the bioavailability of 5-FU in tumours of animals tumour model. To this end, 105 transduced EL-4 cells were treated with capecitabine, we first analysed the content of the injected s.c. into the flanks of groups of six to seven mice that prodrug and its metabolites in the tumours. TP+ and sham- were treated and followed up according to the experimental transduced EL-4 cells were injected into each flank of four wild- setting in Figures 4a and b. When the tumours started to be type (wt) and four dKO mice.2 When the tumours were palpable in palpable in the first set animals, treatment was started with either the first animals, 7 days after injection, mice began treatment with oral capecitabine (1.5 mmol kg − 1 per day) or vehicle, 5 days a either oral capecitabine (1.5 mmol kg − 1 per day) or vehicle. week until day 30 post-injection. Animals were weighed and Treatment was administered daily from day 7 until day 11, and tumour size was monitored daily using a caliper. For statistical the tumours were then measured, dissected and frozen at − 80 °C comparisons of tumour volumes between groups, we used the until needed. Although the number of samples analysed is small two-way repeated-measures analysis of variance test. The vast and no statistical comparisons can be made, we believe that the majority of animals receiving vehicle had to be killed by day 20 results shown in Table 1 can be representative of what occurs in because of tumour growth. In T1 mice (injected with TP+ and the tumours and are worth being shown. In wt and dKO mice sham-transduced EL-4 cells in opposite flanks, see Figure 4a),

© 2014 Macmillan Publishers Limited Gene Therapy (2014) 673 – 681 TP is both a therapeutic and a suicide gene in MNGIE S López-Estévez et al 676 capecitabine, as opposed to dKO mice (Po0.001), in which the only cells expressing TP were the surviving TP+ tumour cells (Figure 4d). To investigate the potential bystander effect of capecitabine metabolism by TP+ tumour cells on the growth of sham-transduced cell-derived tumours, additional groups of animals were injected via s.c. with a 1:1 mixture of TP+ and sham- transduced EL-4 cells (T2 group). In these mice, capecitabine was very effective in reducing tumour growth, indicating a strong local bystander effect both in wt and in dKO animals (Po0.001 and Po0.05, respectively; Figure 4e). Interestingly, in dKO mice treated with vehicle, tumour growth was reduced in comparison with their wt counterparts (Figures 4c–e), and these differences were more striking in the T2 groups (Figure 4e), although these differences were not statistically significant, perhaps due to the high heterogeneity of tumour growth in this group, as spontaneous tumour regressions were observed in two out of four animals. The possibility of a distant bystander effect was investigated by comparing groups of mice injected with only sham-transduced cells in one flank (T3 group) with groups of animals injected with both TP+ and sham-transduced EL-4 cells in opposite flanks (T1 group). In T1 mice (carrying two different tumours in their contralateral flanks), sham-transduced tumours were only partially sensitive to capecitabine in wt animals, in a manner similar to their counterparts in T3 (mice carrying only a sham-transduced tumour; Figures 4c and f). However, in dKO animals receiving the prodrug, the sham-transduced tumours grew despite the presence of TP+ cells in the contralateral flank, suggesting the absence of a significant distant bystander effect in this model. The potential contribution of TP expressed by tumour or stromal cells to tumour growth was evaluated by comparing TP+ versus sham-transduced tumour growth in dKO T1 mice receiving vehicle, and in dKO versus wt T3 mice treated with vehicle. No statistically significant differences were observed between these groups, although the effect might have been masked by the rapid tumour growth. The analysis of survival also reflected the efficacy of capecitabine in TP+ tumours or sham-transduced tumours in wt animals (Supplementary Figure 1). However, survival data have to be interpreted with caution, especially in the T1 group, because these animals carried two different tumours that had different responses to therapy, so that their mortality could not necessarily be attributable to the growth of a specific tumour type. In the other two groups, the effect of capecitabine was highly specific. We observed a longer survival in the animals with some expression of TP (endogenous or by the TP+ tumours) and treated with the prodrug, especially in dKO animals, where the TP expression was restricted to the tumour, and consequently, was also the Figure 3. Toxicity of 5-FU and 5′-DFUR in cultured TP+ and sham- conversion of capecitabine to 5-FU. Wt animals treated with transduced B-LC lines. (a) Survival of B-LCs was measured by flow vehicle had lower survival rates than their dKO counterparts, cytometry using 7-AAD and Annexin V labelling after exposure to which could suggest an angiogenic effect of TP. 5-FU or 5′-DFUR for 48 h. Toxicity of both drugs was significantly higher for TP+ cells, but TPÀ cells were partially sensitive to a high concentration (500 μM) of both 5-FU and 5′-DFUR (see text). Long-term survival was associated with immune responses to the (b) Survival rates of TP+ and sham-tranduced B-LC lines 48 h after tumour cells exchanging the supernatants between the two cell lines previously At day 30 post-injection, after the third weekly cycle of treatment exposed to 5′-DFUR for 48 h. The higher toxicity of supernatants of was finished, a total of 21 (of 65) animals were still alive: 17 had TP+ cells on their counterparts suggests an indirect bystander effect + been treated with capecitabine and 4 had been treated with in vitro. Sham-transduced B-LCs exposed to the supernatants of TP vehicle. We monitored tumour growth and the clinical parameters B-LC cultures (dashed bars); TP+ B-LCs exposed to the supernatants of sham-transduced B-LC cultures (solid bars). (c) Survival of TP+ and for 20 additional days in the absence of any further intervention. TPÀ B-LCs mixed at different ratios after exposure to 5′-DFUR for At day 50, 11 animals had survived, 7 of which had no detectable 48 h. Bars represent mean ± s.d. (n = 3). TP+ B-LCs (solid bars) sham- tumour. Interestingly, 8 of these 11 survivors were from the T2 transduced B-LCs (dashed bars). *Po0.05; **Po0.01; ***Po0.001. group, most likely because this was the only group that was not injected with sham-transduced cells (that were resistant to the prodrug). To evaluate whether these mice had developed an TP+ EL-4 cell-derived tumours were highly sensitive to capecita- immune response to the tumour cells, they were killed and bine in both wt and dKO mice, as opposed to their sham- cytotoxicity (CTL) assays were performed using splenocytes and transduced counterparts (Po0.001; Figures 4c and d). In wt irradiated transduced EL-4 cells as target cells. Significant CTL animals, sham-transduced tumours were partially responsive to activity could be demonstrated in six of the eight evaluable

Table 1. TP activity, Thd and capecitabine metabolites content in tumours derived from EL-4 cells

Treatment Tumour transduction Volumea TP activityb 5-FUc 5′-DFURc Thdc

dKO-1 Capecitabine TP+ NT –––– Sham 204 Und Und 1635 1305 dKO-2 Capecitabine TP+ 20 68 195 6895 395 Sham 454 Und Und 2525 1060 dKO-3 Vehicle TP+ 196 840 Und Und Und Sham 234 Und Und Und 1255 dKO-4 Capecitabine Sham 274 Und Und 2690 1215 wt-1 Capecitabine TP+ 20 199 670 2710 Und Sham 270 35 305 1400 1270 wt-2 Capecitabine TP+ NT –––– Sham 333 28 Und 885 920 wt-3 Vehicle TP+ 168 824 Und Und Und Sham 384 35 Und Und 755 wt-4 Capecitabine Sham 182 17 Und 1095 530 Abbreviations: 5-FU, 5-fluorouracil; 5′-DFUR, 5′-deoxy-5-fluorouridine; dKO, double knockout; NT, no tumour detected; TP, thymidine phosphorylase; Und, undetectable; wt, wild type. aTumour volume (mm3) measured at day 11 after EL-4 cell injection. bTumour TP activity (nmol per h per mg protien). cTumour 5- FU, 5′-DFUR and dThd concentration (pmol per mg protien). animals. Interestingly, two of these eight surviving mice had been 2′-deoxyinosine (a deoxyribose 1-phosphate precursor)23 or treated with vehicle; one of them displayed the strongest anti- antineoplastic agents such as paclitaxel or cyclophosphamide.24,25 tumour CTL activity we measured (Figure 5). TP activity was also a predictive marker of sensitivity to 5-FU.15,25,26 Interestingly, some studies reported that TP gene transfer increased the sensitivity to 5′-DFUR but not to 5-FU, which DISCUSSION could be explained by a basal level of TP activity in the cell line In the present study, we show that the defective gene in MNGIE, studied. The high sensitivity of our TP-transduced cell lines to 5-FU which encodes the enzyme TP, can also function as a suicide gene and especially to 5′-DFUR reveals that TYMP is not only the using capecitabine in a syngeneic tumour murine model. In a defective and the therapeutic gene in MNGIE but can also behave previous work, we showed that TP activity can be restored using as a suicide gene in tumours expressing TP, such as those that haematopoietic stem cell gene therapy in a cellular model and in could develop due to insertional oncogenesis in the context of sublethally myeloablated dKO mice, and we suggested that MNGIE gene therapy for this disease. Thus, we chose the EL-4 thymoma is an excellent candidate for gene therapy.3 Because one of the model because of its clinical relevance, as the leukaemias reported major concerns for the clinical application of gene therapy is the in haematopoietic stem cell gene therapy clinical trials have been potential genotoxicity and oncogenicity of the integrative vectors,17 related by retroviral vector insertions causing oncogene transacti- we hypothesised that some 5-FU prodrugs such as capecitabine, vation in T-cell precursors.17 whose metabolites are converted into 5-FU by TP, would provide a The high sensitivity of mixed TP+/sham-transduced tumours window of opportunity in case of eventual insertional oncogenesis to capecitabine suggests that there is a significant local bystander associated with gene therapy in this specific disease. Here, we show effect, which is more striking (although not significantly different) that TP gene transfer into cells that do not normally express the in dKO mice (Figure 4e), most likely because in these animals, enzyme, such as the human B-LC lines from MNGIE patients or the 5-FU concentration is higher within the tumour, as TP+ tumour murine thymoma EL-4 cells, confers the ability to convert the cells are the only cells that are able to convert the prodrug into prodrugs into 5-FU and sensitivity to its toxic effects. In the B-LC the active drug. In agreement with previous studies,16 our cultures, 5-FU was detectable upon addition of 5′-DFUR in TP- results suggest that 5-FU is released from TP+ cells into the transduced cell cultures but not in their controls, which suggests local environment, most likely by facilitated transport, thus that the conversion of 5′-DFUR to 5-FU is mediated by TP. resulting in bystander toxicity in the neighbouring cells. Some Previous studies had already demonstrated that an elevated or researchers have suggested that this effect is actually caused by enforced expression of TP in cancer cell lines increased their 2'-DFUR, one of 5-FU metabolites.27 Another potential cytotoxic sensitivity to these prodrugs.13,15 In our human B-LCs and the mechanism we cannot rule out in this model is Fas/FasL-mediated murine EL-4 cellular models, TP gene transfer and 5′-DFUR apoptosis. Indeed, Fas and FasL-induced expression was reported exposure also resulted in sensitisation to this prodrug, allowing to mediate apoptosis in TP-transfected human colorectal the formation of 5-FU. At high concentrations of 5′-DFUR, we cancer cells upon exposure to capecitabine in the presence of observed some degree of sensitisation in human sham- HepG2 cells.28 transduced B-LCs, perhaps due to a constitutive expression of The demonstration of specific anti-tumour cytolytic immune uridine phosphorylase, which would allow some conversion of the responses in a significant fraction of the mice surviving 20 days prodrug to 5-FU.18,19 In addition, in our cellular model, after therapy withdrawal, and even in some animals that did not we also observed that TP expression increased the sensitivity receive capecitabine, is intriguing. Because the EL-4 cell line was to 5-FU, as this enzyme converts 5-FU to 2'-DFUR (floxuridine), derived from a thymoma of a male C57BL6/J mouse, which is which is further converted by thymidine kinase 1 into syngeneic with the mouse strain we used in our tumour model, 5-fluorodeoxyuridine monophosphate, an inbititor of thymidylate some potential antigens that could have contributed to the anti- synthase, an enzyme required for DNA synthesis. Our results are in tumour immune responses include minor histocompatibility agreement with previous studies reporting an association antigens; male H-Y antigens (in the case of female recipients); between TP overexpression and the sensitivity of human tumours eGFP, a modified protein from a fluorescent jellyfish; and the TP to capecitabine20 or the increased sensitivity to 5′-DFUR, for expressed by the transduced EL-4 cells (in this case only for the example, by inducing TP overexpression using cytokines such as dKO mice), among others. Although it was not the purpose of the tumor necrosis factor, interleukin-1α or interferon-γ,21 interferon-α,22 study to determine which antigens elicited the immune response,

Figure 4. Experimental setting of the in vivo studies and results on tumour growth in the experimental mice. (a) Scheme of the experimental setting. The three categories (T1, T2 and T3) shown do not correspond to experimental groups based on the therapy given or the type of animals, but solely on the type of tumour cells that were injected (see Materials and methods). The experiment was designed in a way that each of these three groups contained equivalent proportions of wt and dKO and of mice of both sexes treated with capecitabine and vehicle (a minimum of six animals per experimental condition). (b) Timeline of the experimental procedure (d = days). Lower panel: growth of the different types of tumours in the different experimental groups of mice. (c)TP+ tumour growth in T1 mice; (d) sham-transduced tumour growth in T1 mice; (e) mixed tumour growth (T2 mice); (f) sham-transduced tumour growth (T3 mice). Mean ± s.e.m. (n = 6). Arrows indicate the days in which treatment (capecitabine or vehicle) was administered.

the overall higher CTL activity towards TP+ EL-4 cells observed in showed that this drug enhanced anti-tumour immune response comparison to their sham-transduced counterparts (n =8, by eliminating myeloid-derived suppressor cells.29 However, Po0.001) suggest that TP expression could have contributed to although most animals, including virtually all treated with vehicle, the immunogenicity of some tumours. In addition, anti-tumour could not avoid tumour growth and had to be killed, it is CTL activity was more frequently found in females than in males intriguing that some animals treated with vehicle, such as two (5/3), which also suggests that male antigens migh have played a females injected with the mixture of TP+:sham-transduced cells, role, although these differences were not statistically signiﬁcant.5 developed tumours that vanished spontaneously. In such cases, As the vast majority of animals surviving for at least 20 days after an immune response is the most plausible explanation for their the therapy was withdrawn, and those in which we documented a long-term survival, although we could only detect it in a fraction cytolytic anti-tumour immune response had received capecita- of the surviving animals. It is also of interest that the vast bine, it is tempting to speculate that these immune responses majority of animals surviving at day 50 had been injected with the actually synergised with chemotherapy in our model. Overall, the mixture of TP+ and sham-transduced cells, which were targets of data suggest that the the anti-tumour CTL responses detected the local bystander effect (Supplementary Figure 1). Finally, were not targeting a single antigen but several, which may also although we cannot generalise due to the limited number of differ between different animals. In the case of 5-FU, a report samples analysed, we think that the facts that 5-FU levels were

Gene Therapy (2014) 673 – 681 © 2014 Macmillan Publishers Limited TP is both a therapeutic and a suicide gene in MNGIE S López-Estévez et al 679 EL-4 is a tumourigenic cell line established from thymocytes of a lymphoma induced by 9,10-dimethyl-1,2-benzanthracene in a male C57BL6 mouse (American Type Culture Collection, Manassas, VA, USA; TIB-39). EL-4 cells were cultured in DMEM high glucose 4.5 g l − 1 (PAA, Cölbe, Germany) medium supplemented with 10% FBS Gold. All culture − 1 media were also supplemented with 2 mM L-glutamine, 100 U ml penicillin and 0.1 mg ml − 1 streptomycin and cell cultures were incubated at 37 °C and 5% CO2. These cells were transduced using either the human TYMP or the sham (eGFP)-encoding lentiviral vectors,3 using a multiplicity of infection of 10. Plates were centrifuged at 508 × g for 1 h and incubated + for 48 h at 37 °C and 5% CO2. To obtain highly enriched cultures of TP and sham-transduced EL-4 cells, two sequential cell sortings based on eGFP expression were performed using a FACSAria cytometer (BD, San Jose, CA, USA).

B-LCs in vitro CTL assay The toxicities of 5-FU and its prodrug 5′-DFUR were determined by analysing cell viability by ﬂow cytometry using 7-AAD and APC-Annexin V labelling. In preliminary experiments, we had observed that toxicity was more evident at 48 h than at 24 h of exposure, so we used this time period in all further experiments. TP+ and sham-transduced B-LCs were seeded at a density of 2 × 105 cells per well in six-well plates and were incubated for 48 h in the presence of 50 and 500 μM 5-FU or 5′-DFUR (both from Sigma, St Louis, MO, USA) at 37 °C and 5% CO2 (the concentrations achieved in patients receiving these drugs are in this range). Then, cells were washed with PBS and counted using a Neubauer chamber. A minimum of 4 × 105 cells were resuspended in 500 μl of cold 1X Binding buffer (BD Pharmingen, San Jose, CA, USA) and 7-AAD and 5 μl APC-Annexin V (BD Pharmingen) were added before they were incubated for 15 min at room temperature in the dark. Then, 300 μl of cold 1X binding buffer was added, and cell viability was analysed using a FacsCalibur ﬂow cytometer (BD).

Analysis of the bystander effect in vitro To evaluate the cytotoxic effect of the 5-FU generated by TP+ B-LCs on Figure 5. Anti-tumour CTL activity of splenocytes of mice surviving their sham-transduced counterparts, both cell lines were seeded at a cell density of 2 × 105 cells per well in six-well plates; they were exposed to 50 at day 50 after the injection of transduced EL-4 cells. Mean values of μ + a and 500 M of the prodrug and incubated for 48 h at 37 °C and 5% CO2. relative killing are represented for TP EL-4 cells ( ) and sham- fi μ fi transduced EL-4 cells (b)(n = 8). Wt mice (solid squares), dKO mice Then, the supernatants were collected and ltered through a 0.45- m ltre o (Whatman, Piscataway, NJ, USA) to remove all cells, as they grow in (solid circles), non-injected control mice (empty squares; *P 0.05; fi **Po0.01; ***Po0.001). E:T, effector/target cell ratio. suspension. Sham-transduced B-LCs were then resuspended in ltered supernatants from TP+ B-LCs at a density of 2 × 105 cells per well and, reciprocally, TP+ B-LCs were resuspended in filtered supernatants of their undetectable in the extracts of sham-transduced tumours in dKO sham-transduced counterparts. After 48 h of incubation at 37 °C and 5% mice treated with capecitabine, that TP activity was present in all CO2, cell viability was measured as described above. To analyse the tumours generated in wt mice, in two (out of two) TP+ tumours in potency of the bystander effect in vitro,TP+ B-LCs were stained with PKH26 (Sigma-Aldrich, St Louis, MO, USA) following the manufacturer's recom- dKO mice, and that some production of 5-FU was detected in the À two tumours of a wt mice treated with the prodrug (most likely mendations, mixed with non-stained TP B-LCs at different ratios, exposed to 50 μM of 5′-DFUR for 48 h, and cell viability was measured in both due to the TP expressed by the tumour stroma, which is PKH26+ and PKH26À cell populations by 7-AAD staining and flow host derived), are relevant, and suggest that TP is actually cytometry. All culture wells contained a minimum of TP+ cells (3 × 105) mediating 5-FU generation in these tumours. to properly analyse their viability independently in both cell types. Altogether, these results suggests that, in MNGIE patients, TP is a highly specific suicide gene in combination with a commonly 5-FU, 5′-DFUR and dThd quantification used 5-FU prodrug, and this represents a valuable safeguard in 5-FU, 5′-DFUR and dThd levels were measured in cell culture media and case these patients develop tumours as a consequence of tissue homogenates by liquid chromatography coupled to tandem mass insertional oncogenesis. Finally, we describe a novel murine SGT spectrometry. The cultured cells were collected and centrifuged at 300 × g model in which the angiogenic contribution of TP activity by the for 5 min, and the supernatant was recovered and deproteinised through fi tumour stroma or the bystander effect of capecitabine or other ultra ltration (Amicon Ultra 0.5 ml, 3 kDa, EMD Millipore, Billerica, MA, USA). For the determination of 5-FU, 5′-DFUR and dThd levels in tumours, 5-FU prodrugs can be dissected in a much cleaner way than with an aliquot of tumour tissue homogenate obtained for TP activity the use of wt animals. assessment containing 300 μg of protein was deproteinised by the addition of perchloric acid (0.5 M, final concentration) and the supernatant was used for quantification. Then, 7.5 μl of deproteinised medium or MATERIALS AND METHODS homogenate was injected into an Acquity UPLC-MS/MS apparatus (Acquity Vectors and cell lines UPLC-Xevo TQ Mass Spectrometer, Waters, Milford, MA, USA), using an Acquity UPLC BEH C18 column (100 × 2.1 mm, 130 Å pore, 1.7 μm particle, – B-LCs from MNGIE patients transformed with the Epstein Barr virus and Waters). The components of the sample were resolved at 0.5 ml/min transduced with lentiviral vectors encoding TP and/or eGFP as a reporter through a binary gradient-elution using a saline buffer (20 mM ammonium + À gene (TP and sham-transduced (TP ) B-LCs) have been reported acetate, pH 5.6) and acetonitrile as follows: 0–1.1 min, isocratic 100% saline elsewhere.3 Both cell lines were cultured in RPMI 1640 medium (Gibco, buffer; 1.1–5 min, gradient from 0 to 13.6% acetonitrile; 5–5.1 min, gradient Paisley, UK) supplemented with 15% FBS Gold (PAA Laboratories GmbH, from 13.6 to 100% acetonitrile; 5.1–6.1 min, isocratic 100% acetonitrile; Cölbe, Germany). 6.1–7.2, isocratic 100% saline buffer. The detection of the eluate

© 2014 Macmillan Publishers Limited Gene Therapy (2014) 673 – 681 TP is both a therapeutic and a suicide gene in MNGIE S López-Estévez et al 680 components was performed using Multiple Reaction Monitoring, with REFERENCES negative electrospray for 5-FU (transition mass: 128.6–42.0; cone voltage: 1 Wang GP, Garrigue A, Ciuffi A, Ronen K, Leipzig J, Berry C et al. DNA bar coding 24 V; collision energy: 14 V), and positive electrospray for 5′-DFUR and pyrosequencing to analyze adverse events in therapeutic gene transfer. (transition mass: 246.8–117.0; cone voltage: 12 V; collision energy: 8 V) Nucleic Acids Res 2008; 36: e49. and dThd (transition mass: 242.8–127.1; cone voltage: 10V; collision energy: 2 Lopez LC, Akman HO, Garcia-Cazorla A, Dorado B, Marti R, Nishino I et al. 12 V). Calibration curves made with aqueous standards were processed in Unbalanced deoxynucleotide pools cause mitochondrial DNA instability in parallel, and the concentrations were obtained from interpolation of the thymidine phosphorylase-deficient mice. Hum Mol Genet 2009; 18:714–722. peak areas. 3 Torres-Torronteras J, Gomez A, Eixarch H, Palenzuela L, Pizzorno G, Hirano M et al. Hematopoietic gene therapy restores thymidine phosphorylase activity in a cell – TP activity determination culture and a murine model of MNGIE. Gene Therapy 2011; 18:795 806. 4 Rivat C, Santilli G, Gaspar HB, Thrasher AJ. Gene therapy for primary immuno- 3 TP activity was determined in tumours as detailed elsewhere. deficiencies. Hum Gene Ther 2012; 23: 668–675. Briefly, 74 μl of the tumour homogenate supernatant containing 5 Simpson E, Scott D, James E, Lombardi G, Cwynarski K, Dazzi F et al. Minor H 100 μg of protein was incubated at 37 °C for 1 h in the presence of antigens: genes and peptides. Eur J Immunogenet 2001; 28: 505–513. 10 mM dThd in the reaction buffer, and the thymine formed was 6 Lu M, Freytag SO, Stricker H, Kim JH, Barton K, Movsas B. Adaptive seamless measured using ultra performance liquid chromatography coupled with design for an efficacy trial of replication-competent adenovirus-mediated suicide ultra violet detection (Waters). The protein content was determined using gene therapy and radiation in newly-diagnosed prostate cancer (ReCAP Trial). 30 the Bradford method. Contemp Clin Trials 2011; 32:453–460. 7 Bonini C, Ferrari G, Verzeletti S, Servida P, Zappone E, Ruggieri L et al. HSV-TK gene Mice and tumour model with transduced EL-4 cells transfer into donor lymphocytes for control of allogeneic graft-versus-leukemia. Science 1997; 276: 1719–1724. dKO mice, a model of MNGIE generated in a C57BL/6J background by the 8 Walko CM, Lindley C. Capecitabine: a review. Clin Ther 2005; 27:23–44. group of M Hirano (University of Columbia, NY, USA), have been 9 Ajani J. Review of capecitabine as oral treatment of gastric, gastroesophageal, and 2 described elsewhere. dKO mice and normal inbred C57BL/6J were bred esophageal cancers. Cancer 2006; 107:221–231. and housed in the institutional animal facility. All procedures 10 Longley DB, Harkin DP, Johnston PG. 5-fluorouracil: mechanisms of action and were approved by our institutional Ethical Committee on Animal clinical strategies. Nat Rev Cancer 2003; 3:330–338. Experimentation. Tumours were induced in dKO and parental C57BL/6J 11 Haraguchi M, Miyadera K, Uemura K, Sumizawa T, Furukawa T, Yamada K et al. 5 + mice by s.c. injection of 10 TP , sham-transduced EL-4 cells or a 1:1 Angiogenic activity of enzymes. Nature 1994; 368:198. mixture of both cell types. Ten- to twelve-week-old male and female mice 12 Ciccolini J, Evrard A, Cuq P. Thymidine phosphorylase and fluoropyrimidines were used for the experiments. Once the tumours became mea- efficacy: a Jekyll and Hyde story. Curr Med Chem Anticancer Agents 2004; 4: surable (approximately 0.02 cm3), they were treated with 200 μlof 71–81. either capecitabine (5′-deoxy-5-fluoro-N-[(pentyloxy)carbonyl]cytidine; 1.5 mmol 13 Schwartz EL, Baptiste N, Wadler S, Makower D. Thymidine phosphorylase med- − kg 1 per day) or vehicle (citrate buffer) by gavage, five times per week. iates the sensitivity of human colon carcinoma cells to 5-fluorouracil. J Biol Chem Tumour growth was monitored daily until fulfilment of the end point 1995; 270: 19073–19077. criteria (more than 2 cm3 of tumour volume, weight loss of more than 20%, 14 Evrard A, Cuq P, Robert B, Vian L, Pelegrin A, Cano JP. Enhancement of presence of skin ulcers or clinical signs of pain or suffering), in agreement 5-fluorouracil cytotoxicity by human thymidine-phosphorylase expression in with the recommendations of our institutional Ethical Committee on cancer cells: in vitro and in vivo study. Int J Cancer 1999; 80: 465–470. Animal Experimentation. Tumour volumes were estimated using the 15 Morita T, Matsuzaki A, Tokue A. Enhancement of sensitivity to capecitabine in formula V = AB2 × 0.4, with A and B as the shorter and longer diameters of human renal carcinoma cells transfected with thymidine phosphorylase cDNA. the tumour, respectively. Int J Cancer 2001; 92: 451–456. 16 Evrard A, Cuq P, Ciccolini J, Vian L, Cano JP. Increased cytotoxicity and bystander effect of 5-fluorouracil and 5-deoxy-5-fluorouridine in human colorectal Evaluation of anti-tumour CTL immune responses cancer cells transfected with thymidine phosphorylase. Br J Cancer 1999; 80: CTL assays were performed in the cohort of mice surviving at least 19 days 1726–1733. after the therapy, which was stopped at day +30. Splenocytes of tumour- 17 Hacein-Bey-Abina S, Garrigue A, Wang GP, Soulier J, Lim A, Morillon E et al. bearing mice were used as effector cells, and transduced EL-4 cells were Insertional oncogenesis in 4 patients after retrovirus-mediated gene therapy of used as targets. A total of 3 × 105 splenocytes were co-cultured with SCID-X1. J Clin Invest 2008; 118: 3132–3142. irradiated (80 Gy) TP+ or sham-transduced EL-4 cells at different ratios (1:1, 18 Ishitsuka H, Miwa M, Takemoto K, Fukuoka K, Itoga A, Maruyama HB. Role of 10:1, 25:1, 50:1 and 100:1) for 18 h in 96-well plates and the specific killing uridine phosphorylase for antitumor activity of 5′-deoxy-5-fluorouridine. Gann was calculated by measuring the percentage of viable eGFP+ EL-4 cells 1980; 71: 112–123. present in the cultures by flow cytometry using a Fortessa flow cytometer 19 Cao D, Ziemba A, McCabe J, Yan R, Wan L, Kim B et al. Differential expression of (BD), as previously described.31 Specific killing was normalised using uridine phosphorylase in tumors contributes to an improved fluoropyrimidine – negative control data, without effector cells. therapeutic activity. Mol Cancer Ther 2011; 10:2330 2339. 20 Andreetta C, Puppin C, Minisini A, Valent F, Pegolo E, Damante G et al. Thymidine phosphorylase expression and benefit from capecitabine in patients with Statistical analysis advanced breast cancer. Ann Oncol 2009; 20:265–271. All analyses were made using the GraphPad Prism software (GraphPad 21 Eda H, Fujimoto K, Watanabe S, Ura M, Hino A, Tanaka Y et al. Cytokines induce Software, Inc. La Jolla, CA, USA). To compare the means of quantitative thymidine phosphorylase expression in tumor cells and make them more sus- variables, we used Student's t-tests or the two-way repeated-measures ceptible to 5′-deoxy-5-fluorouridine. Cancer Chemother Pharmacol 1993; 32: analysis of variance. The values represented are means ± standard error 333–338. of the mean (s.e.m.). P-values o0.05 were considered statistically 22 Schwartz EL, Hoffman M, O'Connor CJ, Wadler S. Stimulation of 5-fluorouracil significant. metabolic activation by interferon-alpha in human colon carcinoma cells. Biochem Biophys Res Commun 1992; 182:1232–1239. 23 Ciccolini J, Peillard L, Evrard A, Cuq P, Aubert C, Pelegrin A et al. Enhanced CONFLICT OF INTEREST antitumor activity of 5-fluorouracil in combination with 2'-deoxyinosine in human colorectal cell lines and human colon tumor xenografts. Clin Cancer Res 2000; 6: The authors declare no conflict of interest. 1529–1535. 24 Endo M, Shinbori N, Fukase Y, Sawada N, Ishikawa T, Ishitsuka H et al. Induction of thymidine phosphorylase expression and enhancement of efficacy of capecita- ACKNOWLEDGEMENTS bine or 5′-deoxy-5-fluorouridine by cyclophosphamide in mammary This work was supported by the Instituto de Salud Carlos III (grant PI12/00322 to RM), tumor models. Int J Cancer 1999; 83: 127–134. and the United Mitochondrial Disease Foundation (postdoctoral grant 12-029 to JT). 25 Ma T, Zhu ZG, Ji YB, Zhang Y, Yu YY, Liu BY et al. Correlation of thymidylate We thank Alex Rojo and Marta Rosal (VHIR animal facility) for their help with the synthase, thymidine phosphorylase and dihydropyrimidine dehydrogenase with in vivo experiments. We also thank María José Carreras for providing the capecitabine sensitivity of gastrointestinal cancer cells to 5-fluorouracil and 5-fluoro-2'-deox- for the in vivo experiments. yuridine. World J Gastroenterol 2004; 10:172–176.

Gene Therapy (2014) 673 – 681 © 2014 Macmillan Publishers Limited TP is both a therapeutic and a suicide gene in MNGIE S López-Estévez et al 681 26 Wang T, Wang L, Qian X, Yu L, Ding Y, Liu B. Relationship between gene 29 Vincent J, Mignot G, Chalmin F, Ladoire S, Bruchard M, Chevriaux A et al. expression of 5-fluorouracil metabolic enzymes and 5-fluorouracil sensitivity in 5-Fluorouracil selectively kills tumor-associated myeloid-derived suppressor cells primary cancer cells isolated from malignant ascites. Cancer Invest 2011; 29: resulting in enhanced T cell-dependent antitumor immunity. Cancer Res 2010; 70: 130–136. 3052–3061. 27 Zhuo LJ, Gao MQ, Huang AM. [Improved sensitivity of gastric carcinoma cells to 30 Bradford MM. A rapid and sensitive method for the quantitation of microgram fluorouracil-related drugs by transfection of thymidine phosphorylase gene]. quantities of protein utilizing the principle of protein-dye binding. Anal Biochem Zhonghua Bing Li Xue Za Zhi 2011; 40:330–333. 1976; 72:248–254. 28 Ciccolini J, Fina F, Bezulier K, Giacometti S, Roussel M, Evrard A et al. Transmission 31 Eixarch H, Gomez A, Kadar E, George M, Martinez N, Espejo C et al. Transgene of apoptosis in human colorectal tumor cells exposed to capecitabine, Xeloda, is expression levels determine the immunogenicity of transduced hematopoietic mediated via Fas. Mol Cancer Ther 2002; 1: 923–927. grafts in partially myeloablated mice. Mol Ther 2009; 17: 1904–1909.

Supplementary Information accompanies this paper on Gene Therapy website (http://www.nature.com/gt)