Research ArticleCancer

Therapeutic targeting of the RB1 pathway in retinoblastoma with the oncolytic adenovirus VCN-01

See allHide authors and affiliations

Science Translational Medicine  23 Jan 2019:
Vol. 11, Issue 476, eaat9321
DOI: 10.1126/scitranslmed.aat9321

An oncolytic virus to blast retinoblastoma

Eyes afflicted by chemoresistant retinoblastoma may need to be surgically removed to prevent life-threatening metastasis. To develop alternative treatments for retinoblastoma, Pascual-Pasto et al. investigated the oncolytic adenovirus VCN-01, which targets cells with dysregulated RB1. They found that VCN-01 can lyse even chemoresistant patient samples in vitro and was effective in mouse xenograft models. VCN-01 was safe in mice and juvenile rabbits. Initial results from the first pediatric retinoblastoma patients treated with intravitreous VCN-01 showed viral replication in tumor cells and no systemic inflammation. These encouraging results support the development of this oncolytic virus to be used for retinoblastoma therapy.

Abstract

Retinoblastoma is a pediatric solid tumor of the retina activated upon homozygous inactivation of the tumor suppressor RB1. VCN-01 is an oncolytic adenovirus designed to replicate selectively in tumor cells with high abundance of free E2F-1, a consequence of a dysfunctional RB1 pathway. Thus, we reasoned that VCN-01 could provide targeted therapeutic activity against even chemoresistant retinoblastoma. In vitro, VCN-01 effectively killed patient-derived retinoblastoma models. In mice, intravitreous administration of VCN-01 in retinoblastoma xenografts induced tumor necrosis, improved ocular survival compared with standard-of-care chemotherapy, and prevented micrometastatic dissemination into the brain. In juvenile immunocompetent rabbits, VCN-01 did not replicate in retinas, induced minor local side effects, and only leaked slightly and for a short time into the blood. Initial phase 1 data in patients showed the feasibility of the administration of intravitreous VCN-01 and resulted in antitumor activity in retinoblastoma vitreous seeds and evidence of viral replication markers in tumor cells. The treatment caused local vitreous inflammation but no systemic complications. Thus, oncolytic adenoviruses targeting RB1 might provide a tumor-selective and chemotherapy-independent treatment option for retinoblastoma.

INTRODUCTION

Each year, it is estimated that 8000 cases of retinoblastoma—a tumor of the developing retina—are diagnosed worldwide, representing 11% of cancers in children less than 1 year of age (1). Systemic and novel drug delivery approaches such as intravitreous and intra-ophthalmic artery administration routes have led to improved ocular preservation (2). However, even after receiving maximum feasible dosages of drugs locally, some tumors progress to a chemoresistant phenotype (3). In such cases, enucleation (surgical removal of the eye) is then performed to prevent extraocular metastases, which may be fatal (4). Otherwise, intensive local chemotherapy might preserve eyes; however, there may be an associated cost of compromised vision due to long-term retinal toxicity (5, 6). Thus, current research in retinoblastoma is focused on identifying new targeted therapies with improved antitumor activity and retinal safety profiles.

New knowledge about the biology and genetics of retinoblastoma has provided therapeutic leads (7, 8). The disease results in most cases from biallelic inactivation of the tumor suppressor retinoblastoma 1 gene (RB1), located on chromosome 13 (13q14.2) (911). About 40% of retinoblastomas are hereditary with a monoallelic germline RB1 mutation and present clinically with bilateral multifocal disease upon the second hit (inactivation of the second allele) in retinal cells (12). The remaining 60% of cases are nonheritable, with biallelic RB1 inactivation arising locally within the developing retina and involving one eye (unilateral disease). In both cases, retinal cells reveal dysfunctional status of the retinoblastoma protein (RB1), which promotes uncontrolled cell division and initiates retinoblastoma oncogenesis (13, 14).

In normal cells, functional RB1 binds to free E2F transcription factors, forming a complex that inhibits cell proliferation (15). E2Fs are a broad family of transcription factors with conserved DNA binding domains that bind target promoters controlling the expression of key cell cycle regulators (16, 17). Upon receiving a mitogenic signal, RB1 is phosphorylated by cyclin-dependent kinases, causing inability to bind E2F and promoting cell cycle progression from G1 to S (18). In retinoblastoma, the inactivation of RB1 removes the constraint on cell cycle control and constitutively increases the expression of free E2F (19). Heterozygous deletion of the E2F1 gene prevents retinoblastoma activation in murine models (20). Targeting the events associated with RB1 dysfunction and increased E2F expression represents a potential strategy in the treatment of retinoblastoma.

VCN-01 is a clinical-grade oncolytic adenovirus genetically modified from adenovirus type 5 (Ad5) designed to replicate in cancer cells with an altered RB1-E2F pathway (21). Modifications include a 24–base pair deletion in the adenovirus E1A gene that impedes viral replication in RB1-functional cells and the insertion of one E2F1 promoter under the E1A gene that permits transcriptional control in cells with high expression of free E2F-1 (22, 23). Cell receptors involved in adenovirus entry to host cells include the coxsackie adenovirus receptor (CAR; CXADR gene), a glycoprotein expressed in tumors including retinoblastoma cell lines Y79 and WERI-Rb1 (24, 25), and αvβ3 and αvβ5 integrins, which are receptors expressed by several cancer types (26). CAR and αv integrins mediate adenovirus binding to the cell surface and internalization into clathrin-coated vesicles, respectively (27). To enhance interaction with integrins α5 (gene ITGA5) and αv (gene ITGAV), the fiber heparan sulfate-glycosaminoglycans (HSG)–binding domain Lys-Lys-Thr-Lys (KKTK) was replaced with an Arg-Gly-Asp-Lys (RGDK) domain in VCN-01 (21). In addition, a hyaluronidase expression cassette included in VCN-01 enhances intratumor spread of the virus by degrading the extracellular matrix (21). Through these mechanisms, VCN-01 selectively infects and replicates in cancer cells, leading to antitumor activity in a variety of cancer models (21, 28).

Here, we used a battery of preclinical retinoblastoma models derived from treatment-naïve or chemoresistant patient tumors to study the oncolytic activity of VCN-01 in cell cultures, three-dimensional tumorspheres, and orthotopic xenografts. After performing comprehensive biodistribution and toxicity studies in mice and rabbits, we conducted a clinical study to evaluate the safety and activity profile of the treatment in pediatric patients with chemorefractory retinoblastoma.

RESULTS

Targets for infection and replication of VCN-01 are expressed in retinoblastomas

Gene expression in human retinoblastomas of virus receptors CXADR, ITGA5, and ITGAV and viral molecular targets E2F1 and RB1 are shown in Fig. 1A and table S1. Compared to fetal retinas, mRNA expression of CXADR and E2F1 was increased in retinoblastomas (P = 0.0081 and P = 0.0000018, respectively; Fig. 1B and table S2). Compared to astrocytes, a cell type reported to express adenovirus receptors (29), retinoblastoma cultures showed similar expression of CXADR (fig. S1A). According to flow cytometry results, all retinoblastoma cultures expressed CAR and half expressed αv integrin (fig. S1B). Formalin-fixed, paraffin-embedded (FFPE) tumors from a cohort of 33 enucleated eyes (from 18 males and 15 females; mean age at diagnosis, 24 months) showed high expression of CAR and E2F-1 proteins, weak expression of αv integrin, and no expression of α5 integrin and RB1 proteins (Fig. 1C).

Fig. 1 Expression of targets for infection and replication of VCN-01 in retinoblastomas.

(A) Gene expression values (represented in log2) of CXADR, ITGA5, ITGAV, E2F1, and RB1 in retinoblastoma patient tumors, cell lines, and fetal retinas are represented in colors, from lower (blue) to higher (red) expression. (B) Fold-change expression (relative to fetal retinas) of the five genes in patient tumors and cell lines. *P = 0.0081 and **P = 0.0000018 compared to fetal retinas (limma t test). (C) CAR, α5 integrin, αv integrin, E2F-1, and RB1 protein expression in all evaluable cases (n = 32, 23, 29, 32, and 27 cases, respectively) included in the IHC study. One section per sample was assessed. (D) Representative cases of CAR, αv integrin, E2F-1, and RB1 staining in human samples. Control (−) slides were treated with no primary antibody. Pictures show selected areas of the conserved retina and retinoblastoma. PS, photoreceptor segments; ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer; NFL, nerve fiber layer.

No correlation between the immunohistochemistry (IHC) score and clinical data of the enucleated eyes [sex, laterality, age, tumor differentiation, International Classification of Retinoblastoma (ICRB) group (30), or treatment] was observed (tables S3 and S4). Representative images of staining in human samples are shown in Fig. 1D. CAR was focally expressed in the cytoplasmic membrane of tumor cells and in several layers of the conserved retina (ganglion cells, outer and inner plexiform layers, inner nuclear layer, and inner photoreceptor segments). α5 and αv integrins were stained on the luminal side of endothelial cells of retinal and tumor vessels, and αv integrin was expressed in several areas of the conserved retina (ganglion cells, inner nuclear layer, and inner photoreceptor segments). E2F-1 staining was highly positive in the nucleus of tumor cells and negative in retinas and blood vessels, even in those patients with contrasted RB1 germline mutation. RB1 protein was absent in tumors but highly positive in the retinal cells, even in patients with germline mutations, and in the endothelial cells of retinal and tumor vessels. In orthotopic xenografts implanted into mice, E2F-1 protein expression was highly conserved in human tumors, whereas the normal mouse tissues (retina and brain) were negative (fig. S1C). These results suggest that retinoblastoma is a well-suited disease for local treatment with VCN-01, because tumors show higher expression of CAR and E2F-1 than surrounding normal tissues.

VCN-01 is cytotoxic against multiple patient-derived retinoblastomas and RB1-knockout cells

VCN-01 induced cytotoxicity in 11 of 12 patient early-passage retinoblastoma tumorsphere cultures established from freshly excised tumors in serum-free, growth factor–supplemented medium (Fig. 2; clinical details of the cultures in table S5) and in Y79 cells (fig. S2A) (31). Cytotoxicity was concentration dependent in seven of eight patient-derived models established from heavily pretreated tumors (“chemorefractory”; Fig. 2A) and also in four of four models established from tumors that did not receive treatment before eye enucleation (“naïve”; Fig. 2B). In the chemorefractory models, median viral concentration inhibiting 50% of tumor cell proliferation (IC50), expressed as multiplicity of infection (MOI; i.e., transducing units of virus per cell), was 6.93 (range, 0.086 to 28.7) MOI. In the naïve group, the median IC50 was 4.35 (0.87 to 10.9) MOI. There was no significant difference between both groups (P > 0.05, Mann-Whitney test) (Fig. 2C).

Fig. 2 Oncolytic activity of VCN-01 in retinoblastoma and RB1−/− cells.

(A) Concentration-dependent cytotoxicity of VCN-01 in patient-derived cultures established from chemorefractory tumors at Hospital Sant Joan de Deu (HSJD; Barcelona, Spain). Dots are means of six replicates, and best-fit curves were built in GraphPad Prism. (B) Cytotoxicity of VCN-01 in patient-derived cultures derived from naïve tumors. Dots are means of six replicates, and best-fit curves were built in GraphPad Prism. (C) Comparison of viral concentration inhibiting 50% of tumor cell proliferation (IC50 values) between cultures derived from chemorefractory and naïve tumors. Individual data (labeled dots) are represented. VCN-01–resistant HSJD-RBVS-1 cells were excluded from this analysis. (D) Cytotoxic effects of VCN-01 (100 MOI) in a model of large vitreous seeding of HSJD-RBT-1 cells. Scale bars, 100 μm. (E) Cell viability at day 12 in large vitreous seeding treated with 1.5, 6.25, and 100 MOI of VCN-01 relative to untreated controls. Mean ± SD from six replicates is shown. *P = 0.002 compared to control, ANOVA. (F) Antiproliferative activity of VCN-01 in RB1+/− and RB1−/− iPS cells at day 3 after virus infection. Mean ± SD of six replicates and best-fit curves is shown.

Infection was confirmed by the expression of the early viral replication protein E1A in all cultures except in resistant HSJD-RBVS-1 cells, which were also resistant to infection with the nonreplicative adenovirus AdTLRGDK that expresses green fluorescent protein and luciferase (fig. S2, B and C) (32). Upon long-term exposure to VCN-01 (100 MOI), cell lysis was complete by day 12 (fig. S2D). Lysis was also detected by optical microscopy in a model of vitreous seeding (>250 μm tumorspheres) after 6 and 12 days of infection at 100 MOI, with loss of spherical shape and cell detachment from the spheroid (Fig. 2D). Such morphological changes correlated with significantly lower viability of the seedings at 100 MOI compared to untreated spheroids [P < 0.05, analysis of variance (ANOVA)] (Fig. 2E). Knockout of RB1 using CRISPR-Cas9 technology in fibroblasts derived from induced pluripotent stem (iPS) cells produced homozygous RB1−/− and heterozygous RB1+/− cells. RB1−/− cells lacked expression of RB1 protein and showed twofold higher E2F-1 expression than RB1+/− cells (fig. S2, E and F). VCN-01 activity was fivefold higher in RB1−/− cells than in RB1+/− cells [IC50 = 12.7 MOI for RB1−/− and 54.5 for RB1+/− cells; P < 0.0001, EC50 (half maximal effective concentration) shift analysis; Fig. 2F]. Thus, VCN-01 produces an oncolytic effect in most retinoblastomas, and such effect is mediated by the dysfunctionality of RB1.

Biodistribution of intravitreous VCN-01 is confined in the injected eye even in the presence of intraocular tumors

Tumor-specific distribution is a desirable safety property of oncolytic adenoviruses (33). First, we addressed whether the presence of intraocular tumors resulted in increased systemic distribution of an intravitreous dose of adenovirus because of the abnormalities of the tumor vasculature. We used HSJD-RBT-2 cells as a well-characterized patient-derived retinoblastoma with metastatic properties (31, 34). Orthotopic xenografts injected intraocularly with the tracer virus AdTLRGDK at the maximum feasible dose in the mouse (MFDmouse; determined by the maximum volume of injection of 2 μl in the mouse vitreous) achieved a similar number of viral genomes in the injected eyes (1.4 × 108 ± 0.96 × 108; median and SD) compared with mice without tumors (0.57 × 108 ± 0.24 × 108; P = 0.112, t test of log-transformed data; Fig. 3A). Viral genomes remained predominantly confined to the injected eyes, independently of the tumor content, being at least 4 logs lower or below the limit of quantification in contralateral eyes and collected organs at 48 hours (Fig. 3A). Only 2 of 12 mice had low but detectable viral genomes in plasma 4 hours after injection (fig. S3). Viral expression of luciferase was detected only in the injected eyes, whereas extraocular tissues and contralateral eyes were all negative for infection (Fig. 3B).

Fig. 3 Biodistribution of intravitreous adenoviruses in tumor-bearing mice and juvenile rabbits.

(A) AdTLRGDK genomes in mouse tissues 48 hours after intravitreous injection of the maximum feasible dose (MFDmouse; 3 × 109 vp in 2 μl) in mice with or without intraocular HSJD-RBT-2 tumors (n = 6 mice per group). Individual data from injected and contralateral (Contral.) eyes, brain regions, and liver are represented. Optic nerves (O. Nerve) were pooled, obtaining one data per group. *P < 0.0001 compared to all tissues (ANOVA). (B) Luciferase signal (normalized to the background of untreated mice) quantified in mouse samples (n = 6 mice per group). *P < 0.0006 and **P = 0.0001 compared to all tissues (ANOVA). (C) VCN-01 genomes in rabbit blood after intravitreous injection of one MFDrabbit (6 × 1010 vp in 40 μl of undiluted good manufacturing practice (GMP)–grade product containing VCN-01; n = 24 rabbits). (D) VCN-01 genomes in intraocular content (rabbit vitreous humor, aqueous humor, and retinas) of injected and contralateral eyes (n = 6 samples at each time point). (E) VCN-01 genomes in rabbit ocular surface tissues and fluids contiguous to the injection site (tear film, cornea, and optic nerve) (n = 6 to 24 samples at each time point). (F) VCN-01 genomes in rabbit liver and brain regions (n = 6 samples at each time point). LOQ, limit of quantification; BLD, below limit of detection (values BLD are represented as 1). Dots represent data from individual animals.

Next, we performed extensive biodistribution studies in 17- to 18-day-old rabbits, a model with a developing retina and a similar proportion of volume between aqueous and vitreous ocular compartments compared to the human eye (35). In most of the animals, we observed minimal reflux of fluid towards the ocular surface upon injection of the intravitreous dose. Viral shedding in blood at the intravitreous maximum feasible dose (MFDrabbit; determined by the maximum volume of injection of 40 μl in the juvenile rabbit vitreous) was detectable at 1 and 4 hours after administration in half of the injected animals [viral genomes were below the limit of detection (BLD) in the remaining rabbits], achieving a maximum median concentration of 1.8 × 105 viral particles per milliliter (vp/ml) (0.64 × 105 to 6.2 × 105; median and range of six animals with quantifiable viral genomes at 3 hours; Fig. 3C). At the first tissue sampling point (3 days), viral genomes remained in the injected eye (vitreous humor, aqueous humor, and retina; Fig. 3D) and in ocular surface tissues and fluids contiguous to the injection site and potentially exposed to the reflux upon injection (tear film, cornea, and optic nerve; Fig. 3E). Low systemic dissemination was confirmed at day 3 with detectable viral genomes in the liver (Fig. 3F) and in contralateral eyes, which contained 4-log lower viral genome amounts than injected eyes. Brain samples contained low or undetectable amounts of viral genomes at day 3 (Fig. 3F). Overall, mouse and rabbit experiments demonstrate that (i) functional VCN-01 remained confined to the injected eyes, (ii) the presence of intraocular tumors did not increase systemic dissemination, and (iii) slight and short-term leakage occurred at the local injection site and into the blood.

VCN-01 improves ocular survival and inhibits metastasis in retinoblastoma xenografts

In survival studies of orthotopic retinoblastoma xenografts, we examined whether VCN-01 increased the time to ocular endpoint status, i.e., eye proptosis due to tumor growth leading to enucleation. Use of a single intravitreous dose (MFDmouse; 3 × 109 vp per eye) of VCN-01 increased median survival of mice with Y79 xenografts (P = 0.0002; Fig. 4A) and in mice bearing tumors identified as potent responders in vitro (HSJD-RBT-5 and HSJD-RBT-7), but not in resistant (HSJD-RBVS-1) and semiresponders (HSJD-RBT-2; fig. S4A). Antitumor activity (necrosis and infection) was confined to the injected eyes (Fig. 4, B and C). Retinoblastoma cells surrounding the necrotic areas showed Ad5 in the cytoplasm, colocalizing in a small percentage of cells with E1A nuclear staining (Fig. 4D). Lack of viral replication in the mouse retina was confirmed by immunoblotting (fig. S4B). Potency of the adenovirus treatment was boosted by a second intravitreous dose, improving survival in all treated models, including the semiresponder HSJD-RBT-2 (Fig. 4E).

Fig. 4 Efficacy of intravitreous VCN-01 in orthotopic retinoblastoma xenografts.

(A) Ocular survival of eyes with Y79 xenografts treated with VCN-01 at the MFDmouse (3 × 109 vp in 2 μl; n = 16) or not treated (n = 18). (B) Representative image of the local activity of the virus in the treated (right) eye of a mouse with bilateral HSJD-RBT-7 intraocular xenografts. (C) Representative case of hematoxylin and eosin (H&E), human nuclei (hNu), E1A, and hyaluronic acid (HA) staining in bilateral HSJD-RBT-2 xenografts treated with one dose (MFDmouse) of VCN-01 in the right eye (Injected) and not treated in the left eye (Control). Samples were obtained 30 days after treatment. Control (−) slide was treated with no primary antibody. Scale bars, 100 μm. One section was assessed per sample. (D) Immunofluorescence of E1A and Ad5 staining in an HSJD-RBT-2 xenograft. (E) Ocular survival of eyes with Y79, HSJD-RBT-2, or HSJD-RBT-5 xenografts treated with two doses of VCN-01 at the MFDmouse and compared to controls (n = 6 each). (F) Ocular survival of eyes with HSJD-RBT-2 xenografts treated with standard-of-care chemotherapy (SoC chemo), intravitreous melphalan, or VCN-01 at three clinically feasible human-equivalent dosages. Data were pooled from three independent studies, in which median survival of controls was similar. (G) Representative H&E staining in a VCN-01–treated (at the MFDmouse dose level) survivor eye at day 120. (H) Frequency of brain metastases in animals bearing orthotopic HSJD-RBT-2 tumors after enucleation of the second eye or after surviving 120 days. Numbers at top of the columns indicate the percent positive samples for metastasis. *P = 0.0094, χ2 test and Bonferroni correction; all groups were compared to vehicle controls.

To facilitate clinical translation, our next studies addressed the clinically feasible dosages of VCN-01. Human-equivalent VCN-01 doses were defined as those achieving the same concentration of VCN-01 particles in the vitreous of humans and animals. For translation from mice to humans, we considered that the volume of the vitreous cavity is 5.3 μl in mice and 4 ml in children (35, 36). The maximum feasible dose in children (MFDchildren) is limited by the maximum volume that can be injected safely into the vitreous (37), set to 120 μl of the GMP product in this study. Following these considerations, MFDmouse = 10 × MFDchildren, in terms of vp concentration achieved in the vitreous. The following studies in mice were performed starting at the MFDchildren and decreasing such dosage in logs of 10 to 1/10MFDchildren and 1/100MFDchildren. Upon application of Bonferroni correction for comparison of median ocular survivals from multiple treatment groups, P < 0.0033 was considered significant. We observed that the antitumor activity of VCN-01 at MFDchildren was superior to standard-of-care chemotherapy and intravitreous melphalan in HSJD-RBT-2 xenografts (P < 0.0001; Fig. 4F). Median ocular survival in groups receiving two doses of VCN-01 at 3 × 107 (1/10MFDchildren) and 3 × 106 vp per eye (1/100MFDchildren) was 82 and 62 days, respectively, significantly longer than that in vehicle-treated controls (48 days; P < 0.0001). The group receiving 3 × 108 vp per eye (MFDchildren) extended median ocular survival to a value greater than the evaluation period (>120 days; P < 0.0001). Median ocular survival of mice treated with standard-of-care chemotherapy and two intravitreous injections of melphalan was 52 and 51 days, respectively, with no significant differences compared with controls (P = 0.2491 and P = 0.0142). Eyes surviving longer than 120 days (survivors) showed few tumor cells in the ocular structures, but protein detritus in the vitreous cavity, corneal edema, and corneal neovascularization were present (Fig. 4G). Viable tumor load was assessed by quantification of human CRX mRNA expression in enucleated eyes (31). The limit of detection of this method was one HSJD-RBT-2 retinoblastoma cell in 104 mouse cells (fig. S4C). Eyes enucleated at ocular endpoint were fully loaded with viable tumor, whereas survivor eyes enucleated at day 120 had a lower tumor load (fig. S4D).

The presence of positive CRX mRNA expression in brain tissue was considered metastasis (31). Brain samples were obtained either from mice sacrificed upon enucleation of the second eye or from mice sacrificed at day 120 (mice with one or two surviving eyes). Intravitreous VCN-01 at the MFDchildren significantly reduced the frequency of brain metastases compared to vehicle-treated mice (Fig. 4H).

Toxicity in juvenile rabbits is local, mild, and reversible

Local adverse events to adenoviruses might be related to the activation of the immune system against the viral coat proteins (38). Thus, to characterize the toxicity of VCN-01, we used immunocompetent juvenile rabbits in which we administered two intravitreous injections in 14 days. At human-equivalent dosages MFDchildren and 1/10MFDchildren, intravitreous VCN-01 did not induce systemic symptoms or alterations in hematologic, coagulation, or biochemistry parameters of juvenile rabbits. At necropsy, microscopic structure of the collected organs was normal and E1A staining was negative in all cases, indicating that there was no viral replication in off-target organs. Clinical signs were mild and confined to the injected eyes and included conjunctiva redness, edema, eye discharge, and swelling. Edema recovered completely 1 week after the second dose. Tonometry studies revealed that no animals developed glaucoma or anterior uveitis. According to slit-lamp imaging, cornea and anterior chamber were normal, and vitreous and lens showed cellular infiltration, more frequently in animals treated with the highest dose, although also present in vehicle-treated eyes. Similarly, treated eyes (VCN-01 or vehicle) showed small and superficial epithelial dystrophies, likely related to trauma during the intravitreal injection. No alterations in the retina were found in any animal, neither morphologic (fundoscopy) nor functional [electroretinography (ERG)] (table S6).

Histopathologic evaluation revealed the presence of macrophage infiltration and accumulation in uvea, ciliary bodies, anterior and posterior eye chambers, inner surface of the retina, and optic disc in VCN-01–treated rabbits. These findings were reversible by day 42 (fig. S5, A and B). The virus did not replicate in the rabbit ocular tissues, as shown by a lack of E1A staining.

Clinical translation of VCN-01 results in tumor-selective replication, antitumor activity, and immune activation

Upon completion of the preclinical studies with intravitreous VCN-01, we conducted phase 1 clinical trial for patients with intraocular retinoblastoma refractory to chemotherapy (systemic, intra-arterial, or intravitreal) or radiotherapy, in whom enucleation was the only recommended treatment (NCT03284268). Clinical details for the first two patients enrolled are shown in table S7. Patients received two doses of VCN-01 (2 × 109 vp per eye, 1/100MFDchildren, patient 1; 2 × 1010 vp per eye, 1/10MFDchildren, patient 2) injected 14 days apart, according to the trial protocol (clinical monitoring and sample collection schema are shown in fig. S6A). After virus injection, patients did not present with any systemic complications and viral genomes were not detected in patient blood (Fig. 5A). Systemic concentrations of neutralizing antibodies increased during treatment, reaching high titer (1:41,000) in patient 2 (Fig. 5B). Retinal fundoscopic photographs before and during VCN-01 treatment are shown in Fig. 5C. Both patients developed intravitreous inflammation recorded as vitritis after the first administration. In patient 1, vitritis-associated turbidity precluded monitoring the tumor activity and the eye was finally enucleated as per protocol guidelines. Intravitreous inflammation was controlled in patient 2 using corticosteroids [systemic methylprednisolone (2 mg/kg per day) and local prednisolone (10 mg/ml every 2 hours)]. There was evidence of antitumor response in patient 2 consisting of a reduction in size and number of tumoral vitreous seeds, which was evident after the first dose. Retinal tumors, however, remained unchanged in both patients after two doses (fig. S6B). No significant changes in retinal function measured by ERG (photopic b-wave amplitude) were observed after VCN-01 treatment (Fig. 5D).

Fig. 5 First-in-human intravitreous injection of VCN-01.

(A) Viral DNA in blood and aqueous humor. First injection was given at day 0. (B) Neutralizing antibodies in blood and aqueous humor during treatment. (C) Chronologic changes in fundoscopy of patients treated with VCN-01. Vitreous seeds in patient 2 (arrows). (D) Electroretinographic signals in treated eyes during treatment. Normal values of a healthy eye are shown for comparison. (E) H&E, CD4, CD8, EMA (plasma cells), and E1A staining in conserved fragments of the retina, peritumoral areas, viable tumor, and areas of necrosis and inflammation. Control (−) was treated without primary antibody. One section was assessed per sample.

Histopathology studies of the enucleated eye from patient 1 showed a calcified and ossified tumor with areas of viable solid tumor, inflammatory exudates, and extensive areas of necrosis (fig. S7, A and B). The tumor marker synaptophysin was positive in the viable tumor and in areas of tumor necrosis (fig. S7B). Inflammation was confirmed by the presence of CD4- and CD8-positive T cells in the retina and necrotic tumor areas in the vitreous (Fig. 5E). Plasma cells [epithelial membrane antigen (EMA)–positive cells] were also abundant (Fig. 5E), likely explaining the high titer of anti-adenovirus antibodies in the vitreous at day 42 (1:2560). Areas of viable tumor contained a low number of T cells and plasma cells and were negative for the viral replication marker E1A (Fig. 5E). E1A and Ad5 were identified in tumor necrotic areas, confirming VCN-01 replication and subsequent tumor lysis (Fig. 5E and fig. S6C). No signs of necrosis or viral replication were observed in the conserved retina (Fig. 5E).

DISCUSSION

Therapeutic interventions for children with chemorefractory retinoblastoma are currently limited to the enucleation of the eye, which results in loss of vision in functional eyes and is perceived by many families as a mutilating procedure, although it efficiently controls the metastatic dissemination of the disease (39). In this study, we used a large preclinical platform of retinoblastoma models, gene-edited cells recapitulating the dysfunctional RB1 pathway, and adequate animal species for biodistribution and toxicity studies. We show that the oncolytic adenovirus VCN-01 targeting the canonical genomic alteration of human retinoblastomas was active, tumor selective, safe, and clinically translatable, providing a new therapeutic option to save vision in these children. Patients with retinoblastoma and RB1 inactivation would be suitable candidates for treatment with VCN-01 because their tumors express abundantly the molecular target E2F-1 for replication of the virus, as shown by our study and others (40, 41).

One of the main objectives of our work was to demonstrate tumor selectivity of the treatment in (i) very young patients whose retinas would be still in development (i.e., cells undergoing division or migration) and (ii) patients with germline RB1 mutations (i.e., children at risk of developing bilateral disease). In the first case, low expression of the E2F1 gene found in the developing retinas was consistent with sustained expression of RB1 in the normal mouse and human fetal retinas during development (42). VCN-01 did not replicate in the developing retinas of juvenile rabbits, and previous work has shown that Ad5 vectors either do not infect or induce short-term infections in the retina of mice (43, 44) and infect only the nerve fiber and ganglion cell layers, but not photoreceptors, in rabbits (45). Our first experience in a pediatric patient further supported the lack of viral replication in the retina. In the case of patients with germline mutations, our finding of the lack of E2F-1 protein staining in conserved retinal fragments of enucleated eyes supports that VCN-01 would not replicate in their normal retinal tissues, which was further demonstrated with our experiments using RB1-mutant iPS cells. Thus, VCN-01 replication should select tumors—and not retinas—universally in the population of retinoblastoma patients, including very young children and patients with germline mutations. The high and uniform expression of CAR in retinoblastomas compared to the developing retinas further supported adenovirus treatment, despite the fact that expression of αv integrin was moderate and α5 integrins were not expressed in tumors. Previous studies demonstrated that the interaction of the virus fiber with CAR is likely more relevant for viral transduction than the interaction with αv integrins (46). Differential expression of the virus receptors in cell cultures might explain in part the different susceptibility of each culture to the treatment. For instance, the selectivity for integrin receptors incorporated by gene editing of VCN-01 (32) could explain why cells derived from the vitreous seeding of the first patient, who showed no expression of αv integrin, were resistant to VCN-01 treatment, whereas cells derived from the retinal tumor of the same patient, showing positive αv integrin expression, were highly sensitive in our study.

The route by which a small fraction of the intravitreous dose of VCN-01 exited the eye and redistributed to blood and distant tissues is still not fully understood. For drugs dispersed in the vitreous, distribution to systemic blood takes place mainly through the blood-retinal barrier and choroidal blood route, and alternatively through the blood-aqueous barrier and ciliary body blood flow (47). Although in our preclinical studies viral genomes were transiently positive in blood, liver, and brain, this finding did not translate to gene expression or viral replication in extraocular locations, suggesting that efficient viral uptake by liver Kupffer cells impeded hepatocyte transduction by adenovirus (48) and/or that the low amount of circulating virus became inactivated during its systemic redistribution. A previous clinical study of an intravitreous adenovirus vector for gene therapy in pediatric patients with retinoblastoma did not detect adenovirus genomes in the blood of eight patients receiving up to 1011 vp per injection (49). In that clinical study and in our patients, immune activation in the injected eyes was evident and might have neutralized the viruses before moving from the eye to the systemic circulation. In addition, from our biodistribution analysis in mice with intraocular tumors, we infer that the presence of tumors in patients would likely not lead to a higher systemic exposure to the virus through clearance at the blood-tumor barrier. In addition, we have previously shown that this barrier remains intact in retinoblastoma (34).

We describe here improved antimetastatic efficacy of VCN-01 compared to animals treated with only enucleation (untreated controls), which might be explained by improved local control of tumors in VCN-01–treated animals, leading to prolonged ocular survival. A preclinical study in mice showed that an oncolytic Seneca Valley virus inhibited metastasis of the Y79 cell line to the brain compared with no treatment (50), although in this study the administration route was intravenous.

Limitations of our preclinical design for assessment of VCN-01 toxicity include the incompletely understood immunogenicity of the Ad5 serotype in rabbits, which might be different than the human immune response. In addition, the limited size of xenografts in the mouse eye might have amplified the activity of local virus treatment, and the limited tumor vasculature might have diminished the activity of systemic chemotherapy in mice (34). It is likely that with bulky intraocular human tumors, it is quite challenging to accomplish homogeneous infection with intravitreous treatment. Our ongoing clinical trial will elucidate whether clinical benefit differs between patients in which vitreous seeding predominates compared to patients with large retinal tumors. In this sense, antitumor activity in the first two patients enrolled to the clinical trial support further studies in patients. Local inflammation, controlled with corticosteroids, and immune activation surrounding the solid tumor were remarkable, but the question whether such response was mainly antitumoral or antiviral remains unclear. Our finding of patient seroconversion is in contrast to a previous study showing the absence of systemic immune response after intravitreous administration of a nonreplicative Ad5 vector in retinoblastoma patients, although sampling times were not identical in both studies (51). Production of immunogenic proteins, such as E1A, upon replication of VCN-01 in our patient tumors, might explain the appearance of antiviral systemic immune response in our patients (52). Because adenoviruses activate induction of systemic antitumor immunity (53), we speculate that immune activation upon oncolysis might be beneficial for patients with solid bulky tumors in the retina. In summary, the preclinical study described here supported successful clinical translation of intravitreous VCN-01 as a new candidate therapy for patients with refractory retinoblastoma.

MATERIALS AND METHODS

Study design

The goal of this research was to study the preclinical efficacy, pharmacokinetics, and safety of the oncolytic adenovirus VCN-01 administered intravitreously in retinoblastoma, with an emphasis on ascertaining the selectivity of VCN-01 for cells with a dysfunctional RB1 pathway. First, we collected human samples for gene expression and protein analyses to determine whether targets for viral infection and replication were significantly overexpressed in tumors compared to the developing human retina. In parallel, we established patient-derived tumor models cultured in serum-free medium as tumorspheres, and fibroblasts with the homozygous or heterozygous RB1 mutation inserted by CRISPR-Cas9. Cytotoxicity of VCN-01 in vitro was characterized with the compound [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS assay)]. Several patient-derived cells and the Y79 cell line were engrafted orthotopically in mice to assess the efficacy of intravitreous VCN-01 at maximum dosages. At lower, clinically relevant dosages, we compared the efficacy of two administrations of VCN-01 with standard-of-care chemotherapies (systemic vincristine and carboplatin or intravitreous melphalan). Endpoints of experiments were defined in advance for each experiment, and Kaplan-Meier curves were built for ocular survival analyses. At endpoint, the effect of the treatment in terms of the extent of extraocular tumor dissemination was also evaluated using a previously described molecular method (31). Biodistribution and toxicity studies were performed in mice and rabbits using at least six samples per time point. All animals were assigned randomly to treatment groups. Evaluation of pathology was blinded. Results of the preclinical study were used as the basis for the first-in-children phase 1 trial of VCN-01 for chemorefractory retinoblastoma. Primary data are reported in data file S1.

Retinoblastomas, cells, and adenoviruses

Twelve primary retinoblastoma tumorsphere cultures (established from 10 patients enucleated at HSJD; table S5) and the immortalized Y79 cell line (American Type Culture Collection) were cultured as previously described (31). Human astrocytes (K1884, Life Technologies) were cultured in Gibco Astrocyte Medium (Life Technologies). The human adenovirus VCN-01 and the nonreplicative AdTLRGDK vector were previously developed and characterized in detail (21, 32).

Efficacy in orthotopic xenografts

Mouse experiments were approved by the Animal Experimental Ethics Committee at the University of Barcelona (animal protocol number 542/15) and carried out in accordance with institutional and European guidelines (EU Directive 2010/63/EU) and the “Animal Research: Reporting of In Vivo Experiments” (ARRIVE) guidelines (54). Retinoblastoma cells (2 × 105) were inoculated into the posterior segment of the eye of 6-week-old athymic nude mice (Envigo) using a dull 33-gauge needle, as previously described (31, 34). Intravitreous treatments (2 μl) were performed after a similar procedure. Mice were monitored for local clinical signs (e.g., damage of the lens, corneal inflammation, and uveitis) and tumor growth. In survival studies, eyes were enucleated under ketamine-xylazine anesthesia upon achievement of the experimental endpoint (tumors loading the posterior and anterior chambers and causing eye proptosis).

A first set of experiments addressed the activity of VCN-01 at the dosage MFDmouse (3 × 109 vp per eye by intraocular injection of 2 μl of the undiluted GMP-grade product containing VCN-01). Treatments started at day 7 after tumor inoculation. In an initial experiment, to address the overall efficacy of VCN-01, Y79 cells were inoculated in both eyes of the mice. Treatment consisted of one single dose of VCN-01 in both eyes (n = 16 eyes); control mice did not receive treatment (n = 18 eyes). Eyes with no evidence of tumor growth at day 100 after tumor inoculation were considered survivors.

In a second experiment, to verify whether the antitumor action was confined to the VCN-01–inoculated eye, bilateral xenografts received one single dose of VCN-01 in the right eye, and contralateral eyes were not treated. Xenograft models showing dissimilar in vitro response to VCN-01 were used, either resistant HSJD-RBVS-1 cells, semiresponder HSJD-RBT-2, or potent responders HSJD-RBT-5 and HSJD-RBT-7 (n = 6 each). Experiments were considered completed upon achievement of the endpoint in both eyes, or after day 100 (survivor eyes).

In a third experiment, to study whether a second dose of VCN-01 boosted antitumor activity of VCN-01 treatment, 24 to 28 eyes inoculated with HSJD-RBT-2, HSJD-RBT-5, or Y79 cells received either two doses separated by 14 days or were not treated. Ocular survival was assessed as described.

The second set of in vivo experiments was designed to study the activity of clinically feasible dosages of VCN-01. Bilateral HSJD-RBT-2 xenografts received two intraocular injections of VCN-01 at dosages starting at the MFDchildren, and antitumor responses were compared with those of systemic and intraocular chemotherapy agents used in standard of care for retinoblastoma. Eighty-eight mice (176 eyes) were enrolled and randomized in six groups. Three groups were bilaterally treated with two injections of VCN-01 separated by 14 days at three different dosages: 3 × 108 vp per eye (equivalent to MFDchildren; 38 eyes), 3 × 107 vp per eye (1/10MFDchildren; 28 eyes), and 3 × 106 vp per eye (1/100MFDchildren; 16 eyes). One control group (42 eyes) was injected bilaterally with two injections of vehicle of VCN-01 [20 mM tris (pH 8.0), 25 mM NaCl, and 2.5% glycerol]. Systemic chemotherapy (three cycles of carboplatin and etoposide) was administered in a group of 28 eyes. Carboplatin (34 mg/kg) was injected intraperitoneally on days 1 and 21 of each cycle; etoposide (6 mg/kg) was injected intraperitoneally on days 1 to 3 and days 21 to 23. Cycles were repeated every 28 days. Intravitreous melphalan (0.03 μg per eye, two doses separated by 14 days; regimen selected following the same volume-based criteria to calculate human-equivalent dosages from a clinical dose of 20 μg) was injected in a group of 24 eyes. The study was finalized 120 days after tumor inoculation.

Mouse tissues including retinas, tumors, and brains were collected upon eye enucleation due to tumor growth or at day 120 (end of study for survivors) and analyzed for tumor load by quantification of human CRX mRNA (31) or processed in 4% paraformaldehyde and embedded in paraffin for IHC analysis.

Human translation

We report the first two patients receiving two intravitreous injections of VCN-01 separated by 14 days at 2 × 109 vp per eye (i.e., 1/100MFDchildren) and 2 × 1010 vp per eye (i.e., 1/10MFDchildren) in clinical trial NCT03284268. Intravitreous injections were performed under general anesthesia, as previously described (37). Blood samples, aqueous humor, ocular surface, and nasal swabs were obtained to quantify viral DNA before and after the injection, as described in preclinical biodistribution studies. Neutralizing antibodies were also measured in serum and vitreous humor (the latter upon eye enucleation). Ophthalmological examinations included tonometry, ERG, and RetCam imaging before each injection (days 1 and 14) and on days 28 and 42 after the first dose. Tumor and tissue morphology was also evaluated by brain magnetic resonance imaging.

Statistical analysis

Statistical analyses were performed using GraphPad. Student’s t test and Mann-Whitney t test (for nonnormally distributed data) were used for nonpaired comparisons of two groups. One-way ANOVA and Kruskal-Wallis (for nonnormally distributed data) were applied to determine differences between multiple groups. χ2 test was used to compare proportions. The EC50 shift analysis was applied to compare IC50 values. Median survivals were calculated using the Kaplan-Meier method, and curves were compared using the log-rank test. For in vivo experiments including more than two groups, Bonferroni correction was applied. The threshold for significance (α) was set at 0.05 unless otherwise specified.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/11/476/eaat9321/DC1

Materials and Methods

Fig. S1. mRNA and protein expression for VCN-01 infection and replication targets.

Fig. S2. In vitro characterization of VCN-01 activity in retinoblastoma models.

Fig. S3. AdTLRGDK genomes in mouse plasma after one intravitreous administration.

Fig. S4. In vivo characterization of VCN-01 efficacy and replication.

Fig. S5. Macrophage accumulation in the uvea/ciliary body and the inner surface of the retina in VCN-01–treated rabbits.

Fig. S6. Human translation of VCN-01 for retinoblastoma treatment.

Fig. S7. Histopathology of a VCN-01–treated human eye.

Table S1. Gene expression in retinoblastoma primary tumors, retinoblastoma cell lines, and fetal retinas (provided as an Excel file).

Table S2. Differential gene expression for fetal retinas and retinoblastomas (limma t test) (provided as an Excel file).

Table S3. CAR, α5 integrin, and αv integrin expression scores (high, H; moderate, M; low, L; and negative, N) in tumors in relation to clinicopathologic characteristics of enucleated retinoblastoma patients.

Table S4. E2F-1 and RB1 expression score (high, H; moderate, M; low, L; and negative, N) in tumors in relation to clinicopathologic characteristics of enucleated retinoblastoma patients.

Table S5. Clinical details of patient-derived retinoblastoma cell models.

Table S6. Ocular toxicity monitoring (edema, intraocular pressure, and slit-lamp imaging) in juvenile rabbits treated with VCN-01 (provided as an Excel file).

Table S7. Clinical details for the first two patients of the clinical trial.

Data file S1. Primary data (provided as an Excel file).

References (5557)

REFERENCES AND NOTES

Acknowledgments: Dedicated to the memory of our colleague C. de Torres. We thank E. Rodríguez and M. J. Nagel for technical support, M. Ferrer and A. Raya for the work with CBiPS30-4F-5 cells, and J. Martinovic and A. Benachi for providing retinal samples. Funding: R.A., M.C., and A.M.C. were funded by Retos MINECO (Cure4RB project RTC-2015-4319-1). A.M.C. was funded by ISCIII-FEDER (CP13/00189), the AECC Scientific Foundation, and European Union Seventh Framework Programme (FP7/2007-2013) under a Marie Curie International Reintegration Grant (PIRG-08-GA-2010-276998). F.R. was funded by the Retinostop Association, the INCa, and the INSERM (in the framework of the Cancer Plan). We thank the Xarxa de Bancs de Tumors de Catalunya (XBTC; sponsored by Pla Director d’Oncologia de Catalunya). Author contributions: G.P.-P., M.B.-P., R.A., M.C., G.L.C., and A.M.C. conceived the studies; M.B.-P., M.C., and A.M.C. supervised experiments; G.P.-P., M.C.-V., and A.M.C. carried out biological experiments; A.K. and M.C. provided GMP-grade materials, methods, or facilities; G.P.-P., N.G.O., A.M.-B., D.O., S.P., M.V.-U., M.C.-V., H.C.-E., G.B., L.G.-G., M.G.-A., P.A.-L., M.F.-S., S.T.-M., D.R.-L., R.M., and M.T. processed biological samples; G.P.-P., K.W., S.P., H.M.-G., M.R., and M.S. developed and performed pathology assays; C.A.R.-P., K.W., and M.S. carried out pathology analysis; G.P.-P. and A.M.C. generated patient-derived retinoblastoma models; D.O., G.C., I.A., F.D., N.C., E.C., J.C., C.L., C.d.T., L.F., F.R., F.L.M., J.C.-M., J.M., G.L.C., and A.M.C. provided and analyzed patient data; G.P.-P., M.B.-P., and A.M.C. analyzed and interpreted biological data; G.P.-P. and A.M.C. wrote the manuscript; all authors revised and approved the manuscript. Competing interests: M.C., M.B.-P., G.L.C., and A.M.C. are coinventors of one patent application (WO2018091151A1) concerning the use of VCN-01 in retinoblastoma. M.B.-P., A.M.-B., M.G.-A., P.A.-L., M.F.-S., and M.C. are employees and R.A. is a consultant in VCN Biosciences. M.C. and R.A. have ownership interest in VCN Biosciences. Data and materials availability: VCN-01 is available from R.A. and M.C. under a material transfer agreement (MTA) with IDIBELL-ICO. Patient-derived retinoblastoma cultures can be obtained from A.M.C. through collaborative MTA with HSJD. All data related to this study are present in the paper or in the Supplementary Materials.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article