Observed Incidence of Tumorigenesis in Long-Term Rodent Studies of Raav Vectors

Gene Therapy (2001) 8, 1343–1346  2001 Nature Publishing Group All rights reserved 0969-7128/01 $15.00 www.nature.com/gt BRIEF COMMUNICATION Observed incidence of tumorigenesis in long-term rodent studies of rAAV vectors

A Donsante1, C Vogler2, N Muzyczka3, JM Crawford4, J Barker5, T Flotte3, M Campbell-Thompson4, T Daly1,6 and MS Sands1 1Department of Internal Medicine, and 2Department of Pathology, St Louis University School of Medicine, St Louis, MO; 3Powell Gene Therapy Center, 4Department of Pathology, University of Florida, Gainesville, FL; and 5Jackson Laboratory, Bar Harbor, ME, USA

Gene therapy using recombinant adeno-associated virus normal mice overexpressing human GUSB for the presence vectors (rAAV) is generally considered safe. During the of tumors and increased hepatocyte replication. The results course of a study designed to determine the long-term effi- of these studies do not indicate that MPSVII mice or mice cacy of rAAV-mediated gene therapy initiated in newborn overexpressing human GUSB are susceptible to tumor for- mice with the lysosomal storage disease, mucopolysacchar- mation; however, the number of animals examined is too idosis type VII (MPSVII), a significant incidence of hepato- small to draw definitive conclusions. Results from quantitat- cellular carcinomas and angiosarcomas was discovered. A ive PCR performed on the tumor samples suggest that the hepatocellular carcinoma was first detected in a 35-week- tumors are probably not caused by an insertional old mouse and by 72 weeks of age, three out of five rAAV- mutagenesis event followed by the clonal expansion of a treated MPSVII mice had similar lesions. These types of transformed cell. In a separate study, a relatively large group tumors had not been seen previously in long-term studies of of mice injected with varying doses and types of rAAV vec- MPSVII mice using recombinant enzyme or bone marrow tors had no evidence of hepatic or vascular tumors. Although transplantation. In an attempt to ascertain whether mouse the mechanism of tumor formation is currently unknown, the strain or GUSB expression confers susceptibility to tumor tumorigenic potential of rAAV vectors must be rigorously formation, we histopathologically examined untreated nor- determined in long-term in vivo studies. Gene Therapy mal mice of the same strain, untreated MPSVII mice, and (2001) 8, 1343–1346.

Keywords: adeno-associated virus; MPSVII; gene therapy; lysosomal storage disease; tumorigenesis

Recombinant adeno-associated virus (rAAV) vectors are ingly, the average GUSB-specific activities in the liver, generally considered safe. No side-effects have been spleen, kidney, heart, lung, brain, and serum were vir- reported in the many in vivo studies designed to assess tually identical to those measured in other treated the efficacy of rAAV vectors.1 Wild-type AAV (wtAAV) MPSVII animals analyzed at 1 year.6 At necropsy, two of is not associated with any disease and the virus integrates the three animals had obvious circumscribed 1 to 2 cm in with high frequency into a single site on human chromo- diameter tan-brown tumor nodules beneath the capsular some 19. However, the site specificity is lost in rAAV vec- surface of the liver. Following this unexpected discovery, tors and they have been shown to either persist as epi- the remaining two rAAV-treated MPSVII mice and eight somes or to integrate randomly into the host genome.2–5 age-matched normal controls were killed and examined. Therefore, the potential for rAAV to have adverse side- In addition, tissue from animals from the same study that effects due to insertional mutagenesis may exist. had died or been killed earlier in the experiment were In this issue of Gene Therapy, Daly et al6 report on the retrospectively examined for abnormalities. In all, six long-term (1 year) efficacy of intravenous neonatal rAAV treated MPSVII animals with tumors were discovered treatment for murine mucopolysaccharidosis type VII (see Table 1). Five animals had hepatocellular carcinomas (MPSVII). Following submission of the paper, three of (HCC) (Figure 1a–c), one had a metastatic angiosarcoma five rAAV-treated MPSVII mice that were part of a lon- (AS) (Figure 2a–c), and one of the five that had HCC also gevity study were killed at approximately 18 months of had an angiosarcoma. None of the eight age-matched age for histochemical, biochemical, and histopathological controls had gross lesions at 18 months of age. analyses. All of the animals appeared healthy and the Several hypotheses could explain the formation of the three animals killed were chosen at random. Interest- tumors. The first hypothesis is that MPSVII mice are pre- disposed to tumors. However, since untreated MPSVII mice have a reduced lifespan (6 to 10 months),6–8 they Correspondence: MS Sands, Box 8007, 660 S Euclid Ave, Washington may not live long enough to develop neoplasms. To University School of Medicine, St Louis, MO 63110, USA 6Current address: Department of Pathology, University of Alabama, address the possibility that the tumors might be strain- Birmingham, AL, USA specific or specific to older MPSVII animals that have sur- Received 29 March 2001; accepted 5 July 2001 vived because of treatment, four groups of mice were Recombinant AAV toxicity A Donsante et al 1344 Table 1 Tumor identification in AAV-injected newborn mutants that received non-ablative bone marrow trans- mps/mps mice plantation at birth. We also examined transgenic mice that harbor the same expression cassette encoded by the Animal Noa Age Cause of death Diagnosisb Metastasesc rAAV vector9,10 to evaluate the possibility that over- (weeks) expression of human GUSB is oncogenic. All of the experiments described above were performed in mice on 8374 35 killed HCC NE the same genetic background (C57Bl/6), except for a 6669 52 spontaneous HCC NE 7648 52 spontaneous AS (ut) spl, bm mutation in the minor histocompatibility locus bm1 in the 6654 78 killed HCC colon animals not receiving bone marrow transplantation. Only 7645 78 killed HCC, AS NE the rAAV-treated MPSVII mice exhibited HCC or angio- (br) sarcomas, though one of the bone marrow transplanted 6477 78 killed HCC NE animals that received radiation had a pulmonary adeno- carcinoma. These data are consistent with previous bone a41 treated animals killed between one and 35 weeks of age had marrow transplantation and enzyme replacement studies no visible lesions, but have not been examined microscopically. where the tumor incidence is low.8,11–14 The only example Twelve treated mice either died spontaneously or were killed between 35 and 78 weeks of age and had no gross lesions. of HCC described in these studies occurred in a 494-day- bHCC and AS indicate hepatocellular carcinoma and angiosarcoma, old normal animal receiving irradiation and bone mar- respectively. The primary angiosarcomas were in the uterus (ut) row transplantation.8 None of the treated MPSVII mice and at the base of the brain (br). developed HCC. These results suggest that the tumors cThe metastatic lesions were visible on gross examination and con- found in the current study are unique to animals receiv- firmed histologically. Mice without visible lesions in other organs ing rAAV. However, we have not evaluated the tumori- were not thoroughly examined (NE) for metastases. spl and bm indicate spleen and bone marrow, respectively. genic potential of rAAV in normal mice. Studies to evaluate the toxicity of rAAV in nomal animals is currently underway. examined histologically for tumors (Table 2). They It is also possible that MPSVII mice could be at include: (1) the rAAV-treated MPSVII mice; (2) the nor- increased risk for liver tumors if they have an increased mal untreated controls from the efficacy study; (3) hepatocyte replication rate. To determine the fraction of mutants that received bone marrow transplantation at hepatocytes undergoing replication, young adult animals birth following 100 rads of gamma radiation; and (4) (MPSVII mice, normal mice, and transgenic mice overex-

Figure 1 Morphology of hepatocellular carcinomas observed in rAAV-treated MPSVII mice. (a) A normal mouse liver has hepatocytes aligned in regular cords and portal spaces (arrowhead). Kupffer cells line the normal sinusoids and there are occasional Ito cells. Hepatocytes are regular in size and have small nuclei with inconspicuous nucleoli. Mitotic figures are not seen in the normal liver (toluidine blue, 1 cm = 31 microns). (b) The interface (arrowhead) between a tumor nodule (bottom) and normal liver (top) is distinct, although there is no tumor capsule. The neoplastic tumor cells in the nodule have marked pleomorphism with variation in cell size and large bizarre nuclei (arrow) in contrast to the regularly sized hepatocytes in the adjacent normal liver (toluidine blue, 1 cm = 31 microns). (c) Although there are areas of preserved liver cell cords, the neoplastic cells in the grossly identified tumor nodules are larger and more pleomorphic than normal hepatocytes. The nuclei are variable in size and have prominent nucleoli. There were occasional multinucleated cells and occasional large bizarre mitotic figures were seen, as in this figure with a tripolar mitosis (arrow) (toluidine blue, 1 cm = 31 microns).

Figure 2 Morphology of angiosarcomas and metastases observed in rAAV-treated MPSVII mice. (a) One animal had an angiosarcoma that involved the uterus, spleen and bone marrow. In the spleen, the lesions were distinct from the adjacent spleen, multifocal and had large blood-ﬁlled vascular spaces along with more solid areas (hematoxylin and eosin, 1 cm = 322 microns). (b) The bone marrow from the same animal had foci of similar vascular neoplasia (hematoxylin and eosin, 1 cm = 322 microns). (c) The more solid areas consisted of variably sized vessels with erythrocytes admixed with small, closely packed endothelial cells (hematoxylin and eosin, 1 cm = 50 microns).

Gene Therapy Recombinant AAV toxicity A Donsante et al 1345 Table 2 Clinical ﬁndings in aged treated and untreated mice

Mouse genotypea Treatment No. of animals Age (months) Abnormalitiesb mps/mps 8 × 107 i.u. rAAV i.v. at birth 12 12–18 HCC 33% angiosarcoma 17% +/? none 8 18 none mps/mps BMT 107 cells at birth 6 17 simple ovarian cyst 17% mps/mps BMT 107 cells, 100 rads at birth 18 17 simple ovary cyst 28% lung tumor 6% transgenic none 6 12–15 none aMice receiving bone marrow were on the C57BL/6J (B6) background. All other mice were on a B6-H-2bm1− background. mps/mps designates an MPSVII mutant. +/? indicates the non-mutant controls for the MPSVII mutants that received rAAV. Transgenic mice have an unknown number of copies of the viral transgene inserted into their genome. bThese abnormalities were identified under gross examination of the animals. HCC indicates a hepatocellular carcinoma. pressing GUSB) were given BrdU intraperitoneally (i.p.) and killed 5 to 6 h later. The livers were harvested, sec- Table 3 Quantitative analysis of tumor and normal tissues tioned, and stained for BrdU. The frequency of BrdU- positive hepatocytes in all groups of mice were similar 2 Animal No. Normal tissue rAAV Tumor tissue AAV (approximately 20 per cm ). However, MPSVII mice had genomes/cella genomes/cella more total positive cells in the liver (about three-fold more). The increased number of replicating cells reflected 6654 0.025 0.113 infrequent aggregates of small BrdU-positive cells mor- 6477 0.018 0.000 phologically consistent with lymphocytes. By H&E stain- 7645 0.008 0.088 ing, normal, MPSVII, and transgenic animals had similar 8374 0.000 0.000 numbers of lymphocyte aggregates. Therefore, the only 7515 0.005 — difference we detected between the three groups was an 7229 0.010 — increase in DNA replication and presumably cell division a in the lymphocyte population in MPSVII mice. The sig- The number of rAAV genomes per diploid genome found in each sample was determined by a Taqman PCR assay detecting the nificance of this finding is not clear. Although this type chicken beta-actin promoter and the CMV enhancer used in our of finding could be due to an infectious agent, none were rAAV construct and using a GAPDH primer-probe set to determine identified during the course of this study in sentinel ani- diploid genome number. The assay is sensitive to about 0.001 copies mals housed in the same room as the rAAV-treated per cell. Normal tissue represents samples taken from the normal- MPSVII animals. Importantly, no rAAV-treated MPSVII appearing portion of the liver of the treated animal. Tumor tissue mice were similarly analyzed to determine if the lympho- represents tissue taken from the liver tumor. Two animals did not present liver tumors. The average copies per genome for the normal cyte proliferation is corrected or worsened by rAAV tissue and the tumor tissue are not significantly different. treatment. To evaluate the possibility that the tumors in rAAV- treated MPSVII mice were induced by rAAV, we quantit- ated the rAAV genome in the tumor cells. DNA was extracted from the tumors, from normal-appearing liver from animals with tumors, and from pieces of liver taken from rAAV-treated animals without tumors. The rAAV Table 4 Lack of toxicity in other rAAV injected animals genome was quantified by real-time PCR on a PE/ABI 7700 instrument (Table 3). A FAM-labeled fluorescent Treatment Total Number Number Number probe specific for the enhancer/promoter cassette was number у11 moa у12 mob у13 moc used to assess the vector copy number. The diploid genome number was assessed with a VIC-labeled fluorescent Neonatal IV 43d 22 22 22 probe specific for the murine glyceraldehyde-3-phos- Adult portal 79 22 16 2 phate dehydrogenase gene (GAPDH). None of the tissue Adult IV 15 9 3 0 e f g samples analyzed had more than 0.113 copies of our Total animals 137 53 41 24 rAAV cassette per cell, and several of the samples had less than one copy per 1000 diploid genomes, the sensi- These data represent animals from various studies. None of these tivity of this assay. These data suggest that the tumors animals were reported to have hepatocellular carcinomas. a,b,cThese are the total number of animals that received each treat- were not caused by an integration event resulting in ment and were killed after 11, 12, and 13 months, respectively. clonal expansion of a transformed cell. In that case the dThis data set (0 of 43 with tumors) is different from the MPSVII quantitative PCR results would theoretically yield a num- data set (six of 59 with tumors), P Ͻ 0.001. ber greater than or equal to one for the number of rAAV eThis data set (0 of 137 with tumors) is different from the MPSVII data set (six of 59 with tumors), P Ͻ 0.05. genomes per host genome. However, these data do not f exclude rAAV as the causative agent for the tumors in This data set (0 of 53 with tumors) is different from the MPSVII data set (six of 18 with tumors, у35 weeks of age), P Ͻ 0.01. the rAAV-treated mice. gThis data set (0 of 41 with tumors) is different from the MPSVII As stated previously, there was no evidence before this data set (five of 12 with tumors), P Ͻ 0.05. study of rAAV-associated toxicity. A relatively large All meta-analysis comparisons were performed using the ␹2-test.

Gene Therapy Recombinant AAV toxicity A Donsante et al 1346 number of adult and newborn animals, including those for assistance with BrdU labeling. This work was funded described by Song et al15 in this issue, were injected intra- by grants from the NIDDK DK58327 (NM, TF), the venously with a variety of vectors and were subsequently NHLBI HL59412 (NM, TF), NIDDK 27726 (JEB), and examined for the presence of toxicity, particularly hepatic NIDDK 53920 (MSS). Many thanks to Barry Byrne, Alfred toxicity (Table 4). These data indicate that 41 mice, receiv- Lewin, and Philip Laipis for contributing retrospective ing doses of 3.2 × 109 to 5.2 × 109 infectious units (i.u.) of data. rAAV and being killed 12 to 16 months later, had no evidence of liver pathology. Meta-analysis indicated that the References tumor incidence in these animals differed significantly Ͻ 1 Monahan PE, Samulski RJ. AAV vectors: is clinical success on (P 0.05) from the incidence observed in the rAAV- the horizon? Gene Therapy 2000; 7:24–30. treated MPSVII mice. Unfortunately, this study was not 2 Flotte TR, Afione SA, Zeitlin PL. Adeno-associated virus vector originally designed to test toxicity: several strains of gene expression occurs in nondividing cells in the absence of mice, many rAAV dosages, and several different vector vector DNA integration. Am J Resp Cell Mol Biol 1994; 11:517– constructs were used. However, animals injected as neo- 521. nates or injected via the portal vein (Table 4) did receive 3 Kearns WG et al. Recombinant adeno-associated virus (AAV- vectors containing the same rAAV backbone and pro- CFTR) vectors do not integrate in a site-specific fashion in an moter as that used in the MPSVII study6,9,10 and none immortalized epithelial cell line. Gene Therapy 1996; 3: 748–755. developed tumors. In addition, all of the rAAV prep- 4 Ponnazhagan S et al. Differential expression in human cells from arations used in the studies described in Table 4 and in the p6 promoter of human parvovirus B19 following plasmid transfection and recombinant adeno-associated virus 2 (AAV) the MPSVII study were packaged and purified using the infection: human megakaryocytic leukemia cells are non-per- same techniques at the University of Florida Vector Core missive for AAV infection. J Gen Virol 1996; 77: 1111–1122. 16 Facility. This suggests, but does not prove that there 5Afione SA et al. In vivo model of adeno-associated virus vector were no infections or chemical agents unique to the persistence and rescue. J Virol 1996; 70: 3235–3241. rAAV preparations used in the MPSVII study. If there 6 Daly TM et al. Prevention of systemic clinical disease in MPSVII was a common tumorigenic contaminant in the prep- mice following AAV-mediated neonatal gene transfer. Gene arations used for the studies described here, tumors Therapy 2001; 8: 1291–1298. 7 Birkenmeier EH et al. Murine mucopolysaccharidosis type VII: should have been observed in all of the long-term stud- ␤ ies. Although not all of these rAAV preparations were characterization of a mouse with -glucuronidase deficiency. J Clin Invest 1989; 83: 1258–1266. tested for adenovirus contamination, there was no infec- 8 Birkenmeier EH et al. Increased life span and correction of meta- tious adenovirus used in the production of these prep- bolic defects in murine mucopolysaccharidosis type VII after arations. Previous rAAV preparations produced at the syngeneic bone marrow transplantation. Blood 1991; 11: 1381– University of Florida Vector Core Facility by this method 1392. were free of adenovirus contamination. Taken as a whole, 9 Daly TM et al. Neonatal intramuscular injection with recombi- these data suggest that rAAV is not overtly toxic, but it nant AAV results in prolonged ␤-glucuronidase expression in does not exclude a relationship between our rAAV vector situ and correction of liver and spleen pathology in mucopoly- and the tumors observed in the rAAV-treated MPSVII saccharidosis type VII mice. Hum Gene Ther 1999; 10:85–94. mice. 10 Niwa H, Yamamura K, Miyazaki J. Efficient selection for high- Based on the above experiments, it is still impossible expression transfectants with a novel eukaryotic vector. Gene 1991; 108: 193–200. to determine whether or not the tumors identified in the 11 Sands MS et al. Treatment of murine mucopolysaccharidosis rAAV-treated mice were a consequence of the rAAV type VII by syngeneic bone marrow transplantation in neonates. treatment. Age-matched control mice and MPSVII mice Lab Invest 1993; 68: 676–686. from previous studies of other therapies have not 12 Sands MS et al. Syngeneic bone marrow transplantation reduces developed hepatocellular carcinomas or angiosarcomas. the hearing loss associated with murine mucopolysaccharidosis In addition, large numbers of mice have been exposed to type VII. Blood 1995; 86: 2033–2040. other rAAV vectors without developing these tumors, 13 Vogler CA et al. Enzyme replacement with recombinant ␤-gluc- and the PCR results are not consistent with a simple uronidase in murine mucopolysaccharidosis type VII-impact of model of insertional mutagenesis followed by clonal therapy during the first week of life on subsequent lysosomal storage, growth and survival. Ped Res 1996; 39: 1050–1054. expansion. However, the observation of tumors in rAAV- 14 Sands MS et al. Murine mucopolysaccharidosis type VII: long- treated MPSVII mice highlights the necessity of a detailed term therapeutic effects of enzyme replacement and enzyme long-term study designed to determine the tumorigenic replacement followed by bone marrow transplantation. J Clin potential of rAAV-based gene transfer vectors. Invest 1997; 99: 1596–1605. 15 Song S et al. Stable therapeutic serum levels of human alpha-1 antitrypsin (AAT) after portal vein injection of recombinant Acknowledgements adeno-associated virus (rAAV) vectors. Gene Therapy 2001; 8: 1299–1306. Many thanks to Dr Ronald Marks who assisted with the 16 Zolotukhin S et al. Recombinant adeno-associated virus purifi- statistical comparisons, Amy Pourier for technical sup- cation using novel methods improves infectious titer and yield. port for the real-time PCR assays and Dr Kathy Ponder Gene Therapy 1999; 6: 973–985.

Gene Therapy