Oncolytic viruses (OVs) are a promising emerging class of anticancer immunotherapies that exploit the innate ability of certain replication-competent viruses to infect and preferentially lyse tumour cells while leaving non-neoplastic cells intact. OVs can be selected from native viral species on the basis of their innate ability to induce immunogenic cell death (ICD) in cancer cells, although they can also be genetically engineered to enhance tumour selectivity, promote replication competence, limit pathogenicity and increase immunogenicity1. Engineered viruses can be manipulated by the deletion or modification of viral genes or, in viruses with larger genomes, eukaryotic transgenes can be included as an additional ‘payload’ usually for the purpose of increasing the extent of tumour cell death or promoting antitumour immunity. Both DNA and RNA viruses capable of oncolytic activity in mammalian cells are available, although most clinical studies have used DNA viruses because their molecular biology and life cycle are currently better understood2. The generally larger genomes of DNA viruses have the additional advantage of facilitating recombinant gene expression. Further information on the molecular basis of OV development is published elsewhere3.

Talimogene laherparepvec (T-VEC) is an engineered oncolytic herpes simplex virus type 1 (HSV1) designed for preferential replication in tumour cells and induction of antitumour immune responses. Intratumoural injection of T-VEC was evaluated in a prospective randomized trial that met the primary end point of improved durable response rate (DRR, with a durable response defined as an objective response based on modified World Health Organization (WHO) criteria, lasting ≥6 months) in patients with accessible and unresectable melanomas4. In addition to DRR, T-VEC also demonstrated improvements in objective response rate (ORR), progression-free survival (PFS) and overall survival (OS), leading to full FDA approval in 2015. T-VEC has since also been approved for use in Europe, Australia and Israel.

Since its initial approval, T-VEC has been tested in numerous other clinical trials, including in combination with other therapies for patients with melanoma and as a monotherapy in patients with a variety of other cancers, and in real-world (or clinical practice) studies from both single and multiple institutions5. These studies have provided new insights into how best to integrate T-VEC into the expanding clinical landscape of therapeutic agents available for patients with melanoma. The clinical experience with T-VEC has also highlighted some of the challenges associated with the development of OVs and their clinical use as intratumoural agents. Intratumoural immunotherapy is currently in the early stages of clinical development and although T-VEC is the only FDA-approved OV, in December 2022 the FDA approved a non-oncolytic adenovirus encoding IFNα-2b for the treatment of bacillus Calmette–Guerin (BCG)-unresponsive, non-muscle invasive bladder cancer (NMIBC); and three other OVs have been approved globally (although one was discontinued in 2019; Table 1), suggesting that clinicians should seek to familiarize themselves with T-VEC and other emerging OV technologies.

Table 1 Currently approved oncolytic and non-oncolytic viruses worldwide

In this Review, we provide an overview of what we have learned from the clinical experience with T-VEC and define how best to include this OV in the clinical management of patients with melanoma. We also describe some of the preclinical, clinical and regulatory challenges associated with other OVs and intratumoural agents currently in development for the treatment of various advanced-stage cancers. A better understanding of how to optimize the clinical integration of T-VEC and an awareness of other promising emerging agents is expected to enhance the clinical management of patients with melanoma and perhaps also of those with other cancers.

Clinical development

T-VEC is indicated for the local treatment of unresectable cutaneous, subcutaneous and nodal lesions in patients with recurrent melanoma after initial surgery. Based on an attenuated HSV1 strain derived from herpetic stomatitis, T-VEC was engineered to replicate in cancer cells and induce an antitumour immune response6. In T-VEC, ɣ34.5, the HSV gene that encodes the neurovirulence factor ICP34.5 in the unmodified virus, is deleted. This modification diminishes pathogenicity and promotes the early transcription of another HSV gene, US11, which promotes selective tumour cell replication7. US12, the viral gene encoding ICP47, which normally blocks peptide entry into the antigen-processing pathway, is also deleted, allowing the virus to avoid immune detection. Because ICP47 is deleted in T-VEC, antigen presentation of tumour cell-derived peptides is uninhibited, enabling tumour-associated peptides to be loaded onto MHC class I molecules. T-VEC is also engineered to express granulocyte–macrophage colony-stimulating factor (GM-CSF) under the control of a CMV promoter, which provides another mechanism of enhancing local and systemic antitumour immunity by inducing the recruitment and maturation of dendritic cells5. In contrast to tumour cells, in which the intracellular antiviral machinery is often defective, non-malignant cells can readily detect infection and rapidly clear T-VEC, providing another mechanism of tumour-selective replication8. T-VEC has been shown to induce the release of HMGB1 and ecto-calreticulin in human melanoma cells and to promote melanoma tumour antigen spreading in mouse models of established melanomas9. Thus, T-VEC injection is hypothesized to induce selective ICD of cancer cells and promote T cell-mediated immunity against tumour-associated antigens released during tumour cell lysis in conjunction with soluble ‘danger’ signals’10.

Several early preclinical investigations studying T-VEC demonstrated in vitro lytic activity in a variety of human cancer cell lines, including those originating from patients with melanomas as well as breast, prostate and colorectal adenocarcinoma-derived cell lines11. Studies involving mouse models validated the in vivo anticancer activity of T-VEC at doses comparable to those administered in the clinic today. Preclinical studies were also successful in establishing that local production of GM-CSF expressed by the vector is essential for the promotion of antitumour immunity, especially in inducing the rejection of uninjected contralateral lesions, a so-called ‘abscopal’ response12.

The first-in-human phase I trial of T-VEC was designed to evaluate the safety profile and identify an optimal dosing schedule for intratumoural injections in patients with tumours deemed readily accessible for intralesional administration, including cutaneous, subcutaneous and lymph node tumours6. Thirty patients diagnosed with breast cancer, melanoma, head and neck cancer, or colorectal cancer (CRC) with disease progression on standard-of-care therapies were enrolled. T-VEC was well tolerated overall. Fatigue, nausea, fever, chills and injection site pain, all predominantly of grade 1–2 severity, were the most frequently reported adverse reactions. The trial was specifically designed to allow for seroconversion in patients who were HSV naive. No notable differences in response characteristics were observed between patients who were HSV seronegative at baseline versus those who were HSV seropositive; however, adverse events were less severe in patients who were HSV naive and received lower starting doses followed by a higher dose 3 weeks later, especially in those who were HSV seronegative6. Selective tumour cell lysis with viral replication and GM-CSF expression was confirmed via analysis of tumour biopsy samples. Six patients had some evidence of regression of injected and uninjected lesions, including responses in distant metastatic lesions in four. These phase I data suggested an effective dosing scheme comprising an initial priming dose of 106 plaque-forming units (PFU)/ml followed 3 weeks later with a dose of 108 PFU/ml every 2 weeks, independent of baseline HSV antibody status. This study also established a combined maximum injection volume per treatment visit for all lesions of 4 ml, with injection volume titrated to tumour volume based on the longest tumour diameter. While the initial phase I study stipulated the injection of only one lesion to allow early exploration of response in both injected and uninjected lesions, current guidelines permit the injection of multiple lesions as long as the total volume does not exceed 4 ml on any given treatment day. Although largely empirical, this dosing volume enables sufficient dosing without the risk of leakage from the injection site. This regimen was used in subsequent clinical trials and has been widely adopted in other trials testing OVs although studies designed specifically to evaluate the optimal dosing concentration and volumes are a high priority for the field.

Following the phase I study, an open-label, single-arm, multicentre clinical trial was conducted to determine the efficacy of T-VEC in 50 patients with unresectable, advanced-stage melanomas with a primary end point of ORR13. Ten patients had a complete response (CR) and three patients had a partial response (PR; ORR 26%) following a median of six injections of T-VEC. Based on these results, the efficacy of T-VEC was assessed in the open-label, randomized phase III OPTiM trial, which accrued 436 patients with stage IIIB/C–IV melanomas and compared the efficacy of T-VEC with that of intratumourally administered recombinant GM-CSF14. DRR was the primary end point. Modification of the WHO criteria allowed for treatment beyond progression as data from the phase II trial demonstrated ‘pseudoprogression’ (or delayed responses) in >50% of patients.

In the preliminary analysis, T-VEC met the primary end point with a significant improvement in DRR compared with GM-CSF (16.3% versus 2.1%, OR 8.9; P < 0.001)4. The ORR was 26.4% for patients receiving T-VEC versus 5.7% for those receiving GM-CSF. T-VEC was also associated with an improvement in modified PFS, which was defined as time to treatment failure (TTF) to account for potential pseudoprogression. TTF was defined as time from baseline to clinically relevant disease progression for which no objective response was subsequently achieved or until death. Median TTF was 8.2 months in the T-VEC arm versus 2.9 months in the GM-CSF arm (HR 0.42; 95% CI 0.32–0.54)4. Median OS was 23.3 months for T-VEC versus 18.9 months for GM-CSF (HR 0.79, 95% CI 0.62–1.00; P = 0.051 on initial preplanned analysis and P = 0.049 on final analysis at 4 years of follow-up monitoring)14. In the final analysis, DRR also increased to 19% for patients receiving T-VEC versus 1.4% for those receiving GM-CSF (unadjusted OR 16.6, 95% CI 4.0–69.2; P < 0.0001). At this point, ORR was 31.5% in the T-VEC group, including a 16.9% CR rate, probably reflecting delayed responses to treatment in a subset of patients. T-VEC resulted in an ORR of 15% in patients with uninjected visceral metastases. Similar to previous studies, T-VEC demonstrated a tolerable safety profile with low-grade fatigue, fever and chills as the most prevalent adverse events. These data led to FDA approval in 2015 followed by similar regulatory approvals in Australia and Israel (Table 1). T-VEC was also approved in Europe for patients with stage III–IVM1a disease based on a subgroup analysis indicating more substantial improvements in DRR and ORR with T-VEC compared to GM-CSF in these subgroups (DRRs 33% versus 0% for stage III and 24% versus 0% for stage IVM1a). The difference in OS was especially pronounced among patients with stage III–IVM1a disease who received T-VEC (46.8 months versus 21.5 months, HR 0.57, 95% CI 0.40–0.80; P < 0.001).

The OPTiM trial was one of the first randomized studies to assess the efficacy of any OV and was initiated at a time when immune-mediated pseudoprogression was suggested to occur in patients receiving immune-checkpoint inhibitors (ICIs)14. In particular, selecting an appropriate control for this trial was challenging given that no other intralesional agents were approved by the FDA for patients with melanoma. Intratumoural injections of non-recombinant HSV1 or recombinant GM-CSF could both have been relevant controls, yet the FDA did not permit these options owing to a lack of prior safety data on the local injection of these agents. At the time of study development, data suggested that recombinant GM-CSF given subcutaneously might have some activity in patients with stage III–IV melanomas15. Thus, recombinant GM-CSF was chosen for the comparator arm to provide a relevant control and avoid the use of placebo. Owing to the possibility of pseudoprogression, DRR was selected as the primary end point. DRR is a more rigorous end point than ORR as it requires an objective response to be maintained for at least 6 months. This end point has subsequently been validated as being strongly associated with clinical benefit although it remains underutilized16. Finally, the initial preplanned analysis suggested a lack of a significant improvement in median OS, yet this became statistically significant in the final analysis17. These limitations made data analysis problematic although this experience is perhaps not uncommon for a new class of agents with a distinctly different mechanism of action to that of other systemic therapies.

Over the past 5 years, clinical reports assessing the activity of T-VEC in patients with advanced-stage melanoma have confirmed both the safety and efficacy results reported in the pivotal trials18,19,20,21,22,23,24,25. For example, a multi-institutional observational study assessed the real-world safety and efficacy of T-VEC in 80 patients18. The safety profile of T-VEC was comparable to that described in the pivotal trial reports and a 57% ORR was observed, with 39% having a CR. The higher response rates observed probably reflect the fact that investigators generally selected patients with unresectable stage III tumours, including in-transit metastases. In another multicentre retrospective study, 76 patients with advanced-stage melanomas received T-VEC, with a 12-month OS of 77% reported among 42 patients with stage IIIB–IVM1a disease and 64.6% for the 30 patients in the IVM1b–IVM1c subgroup19. Data from a multi-institutional study involving 66 patients at various European sites indicate that 55.3% of patients receiving T-VEC discontinued treatment because they had no further lesions. There were some differences in patient selection by country and the authors concluded that patients with earlier stage III disease receiving first-line therapy are the optimal responders23. Another multi-institutional study evaluated retrospective data from patients receiving T-VEC either concurrently or sequentially with an anti-PD-1 antibody and found no notable differences in clinical outcomes or safety profile, suggesting that T-VEC can be safely administered in combination with or after an ICI21. In an international multicentre study involving 112 patients with stage IIIB–IV melanomas who received T-VEC following disease progression on an ICI, 37% had a CR and 14% had a PR based on assessments of all directly injected sites24. Collectively, these data suggest that the tolerability of T-VEC in routine clinical practice is consistent with that observed in prospective studies. Despite some theoretical concerns, no reports of transmission between household contacts or environmental transmission have emerged. Furthermore, ORRs were generally higher than those reported in clinical trials, which might reflect T-VEC being administered to patients with earlier-stage lesions and perhaps more prevalent use in the first-line setting.

Other studies have focused on identifying specific populations who are most likely to derive benefit from T-VEC. In a subgroup analysis of data from OPTiM, those with isolated head and neck lesions were evaluated for response26. Similar to analyses of the entire cohort, the DRR of patients receiving T-VEC was significantly improved compared to those receiving GM-CSF (36.1% versus 3.8%; P = 0.001); however, in this study, the majority of patients with durable responses also had a CR (CR rate 29.5%). Median OS was not reached for patients treated with T-VEC in this subgroup compared with 25.2 months for those receiving GM-CSF. In a retrospective single-institution study, the experience with T-VEC in older patients was reported22. In this study, 12 patients with advanced-stage melanomas (median 83 years of age; range 75–89 years) received T-VEC. The ORR was 58.3% and durable responses were seen in 41.7% of patients with no grade ≥3 treatment-related adverse events. The authors concluded that T-VEC might be a useful option for such patients who might not be able to tolerate other therapeutic interventions. Finally, a common theme emerging from the post-approval literature on T-VEC highlights a need for better training on the administration of intratumoural injections. Logistical modifications are needed for the clinic to ensure timely and safe drug administration. For example, each lesion must be measured on the day of T-VEC administration and a system for communication between the clinician and pharmacist needs to be in place to individualize drug administration at each visit. As oncology outpatient clinics become more familiar with the modified logistics, most can safely deliver T-VEC to eligible patients as demonstrated by emerging real-world data18.

The emerging clinical data support the use of T-VEC earlier in the course of treatment, for example, in patients with advanced-stage III or IVM1a melanomas. Indeed, patients with ‘early metastatic’ stage III or IVM1a disease in the OPTiM trial derived greater levels of benefit, with objective responses in 33% of patients with stage IIIB–IIIC disease and 16% in stage IVM1a compared with only 3.1% and 7.5% at later disease stages4. DRR also improved to 23.9% among patients receiving T-VEC as first-line therapy compared with 9.6% with second-line or later-line use4. As described previously, improvements in median OS were also more pronounced in patients with stage IIIB–IVM1a disease. T-VEC is specifically approved for use in Europe only in patients with stage III–IVM1a melanomas. In clinical practice, such patients might be identified based on the inability to technically resect all lesions or the consideration that surgery might cause substantial morbidities and/or functional compromise. Patients with lesions that might not be curable with surgery based on the pattern of recurrence, such as those with multiple recurrent cutaneous and/or soft-tissue lesions and/or rapid disease recurrence in whom additional surgical management is unlikely to be curative, might also be good candidates for T-VEC. Other patient-related factors that could also favour the use of T-VEC include older age (for example, in patients that might not tolerate general anaesthesia), complex comorbidities and inability to tolerate other systemic therapies, and/or tumours in certain favourable anatomical locations (such as head and neck lesions). Whenever possible, the management of individual patients should be discussed by a multidisciplinary team including surgeons who can assess the potential for cure with surgery and medical oncologists familiar with melanoma therapy (Boxes 1 and 2).

Finally, while T-VEC is already approved for patients with melanoma, new indications beyond melanoma are being actively explored. For example, T-VEC is currently under investigation in patients with head and neck cancer, peritoneal malignancies, bladder cancer, Merkel cell carcinoma, cutaneous squamous cell carcinoma, soft-tissue sarcoma, triple-negative breast cancer, pancreatic cancer, rectal cancer, and hepatic lesions from breast and colorectal cancers27,28. In addition, neoadjuvant studies with T-VEC and other OVs are especially interesting to consider given that promising clinical responses have been seen in this setting with ICIs and that the mechanism of action and safety profile of OVs are especially well suited to this setting29,30.

Neoadjuvant therapy

Recent clinical data has demonstrated significant clinical responses when immune-checkpoint blockade is used in the neoadjuvant setting across a variety of cancers31,32,33. In the Southwest Oncology Group (SWOG) S1801 study, 313 patients with resectable stage III–IV melanoma were randomized to neoadjuvant or adjuvant pembrolizumab34. The investigators reported a 2-year event-free survival of 72% for patients receiving three doses of neoadjuvant pembrolizumab followed by adjuvant pembrolizumab compared with 49% for patients receiving adjuvant pembrolizumab only. The reasons for this apparent superiority of neoadjuvant pembrolizumab are somewhat controversial but might include preferential expansion of pre-existing tumour-reactive T cells, induction of de novo antigen presentation by PD-1/PD-L1+ dendritic cells, and/or administration of treatment at a time of lower tumour volume and prior to extensive immunoediting29. Nonetheless, delivery of immunotherapy in the neoadjuvant setting, while promising, can also be clinically challenging to evaluate and implement because treatment requires a delay in definitive surgical management, which might also affect patient outcomes depending on the type of cancer and the underlying immune characteristics.

OVs might be appropriate for use as neoadjuvant therapy owing to the ability to induce an ‘in situ’ vaccination effect allowing for personalized induction of tumour-specific immune responses and might also induce T cell infiltration into tumours with an immune-excluded and/or deserted microenvironment. Neoadjuvant T-VEC has been evaluated in an open-label, randomized phase II trial involving 150 patients with resectable stage IIIB–IVM1a melanomas with a primary end point of 2-year recurrence-free survival (RFS)35. In this study, 76 patients were randomly allocated to receive 6 doses of T-VEC followed by surgery and 74 patients were allocated to surgery alone. The 2-year RFS was 29.5% in the T-VEC group compared to 16.5% for those undergoing surgery alone (HR 0.75, 80% CI 0.58–0.96). OS, a secondary end point, was also improved in the T-VEC arm at 2 years of follow-up (88.9% versus 77.4%; HR 0.49, 80% CI 0.3–0.79). These differences in both RFS and OS persisted at 3 years of follow-up. Patients receiving T-VEC had a 17% pathological complete response rate with associated increases in CD8+ T cell density at the tumour site. This trial resulted in an estimated 25% reduction in the risk of disease recurrence in patients receiving neoadjuvant T-VEC, thus meeting the primary end point. Based on the results of this study, a short course of neoadjuvant T-VEC could be considered for patients with borderline-resectable tumours.

Combination therapies

The mechanisms by which OVs eliminate tumour cells are distinctly different to those of other anticancer therapies; therefore, OVs are a rational candidate for combination with most other treatment modalities, including systemic chemotherapies, immunotherapies, targeted therapies and/or radiotherapy (Fig. 1). Furthermore, the toxicity profiles of OVs are limited and generally do not overlap with those associated with other therapeutic approaches6,13. Indeed, data from established preclinical models demonstrate the improved antitumour activity of OVs when combined with other systemic therapies, including chemotherapies and cellular therapies36,37,38. These promising data are supported by early clinical studies suggesting improved therapeutic activity with OV combination strategies39. For example, ICIs are typically less effective in tumours with limited immune cell infiltration, of the so-called ‘immune-desert’ phenotype, whereas OVs are known to promote the recruitment of immune cells into the tumour microenvironment31,32,33. In addition, the effectiveness of ICIs might be dependent on local PD-L1 expression, which can vary both between tumours and patients. Many OVs have also been shown to induce expression of PD-1 and PD-L1, probably through IFNγ production following viral infection9. Thus, OVs might be able to reverse at least some aspects of ICI resistance. The potential synergy of OVs and ICIs has been demonstrated in animal models40.

Fig. 1: Potential OV combination strategies in clinical development.
figure 1

Most oncolytic viruses (OVs) directly kill tumour cells in an immunogenic manner resulting in the release of soluble tumour-associated antigens and danger-associated molecular patterns (DAMPs), an effect known as immunogenic cell death. OVs can utilize established tumours as a source of antigens for individualized in situ vaccination without the need for prior antigen identification. Furthermore, OVs are generally well tolerated and can thus be administered repeatedly, if required. These features make OVs ideal agents for promoting personalized immune responses that can be safely combined with other cancer treatment strategies designed to mediate tumour regression through alternative mechanisms. These strategies can include cytotoxic chemotherapies, molecularly targeted therapies, immune-checkpoint inhibitors, chimeric antigen receptor (CAR) T cell and adoptive T cell therapies, and radiotherapy. Many preclinical and early phase clinical studies have provided data on the antitumour activity of combinations of OVs with one other modality. Additional benefit is now anticipated from approaches combining OVs with multiple modalities, either concurrently or sequentially. Ac, acetyl; HDAC, histone deacetylase; HDACi, HDAC inhibitor.

The response to T-VEC in combination with ipilimumab has been evaluated in a phase I trial involving 19 patients with advanced-stage melanomas, with a promising ORR of 50%, which is a higher response rate than that reported for ipilimumab or T-VEC alone39. This study was followed by a randomized phase II trial comparing T-VEC plus ipilimumab with ipilimumab monotherapy in 198 patients with metastatic melanomas41. Patients in the T-VEC plus ipilimumab group had an ORR of 39% versus 18% in the ipilimumab monotherapy group (OR 2.9, 95% CI 1.5–5.5; P = 0.002). The toxicity profile of this combination was consistent with that of both agents as monotherapies, with no notable additional adverse events arising from combination therapy. Furthermore, regression of visceral lesions that were not injected with T-VEC was reported in 52% of patients in the combination therapy group versus only 23% of those receiving ipilimumab monotherapy, suggesting the development of so-called abscopal responses. These studies, although promising, were initiated prior to the approval of anti-PD-1-antibodies.

In a phase Ib trial testing the combination of T-VEC and the anti-PD-1 antibody pembrolizumab, an ORR of 62% was reported in 21 patients with advanced-stage melanomas42. In this study, T-VEC was given twice prior to the administration of pembrolizumab, resulting in increased local PD-L1 expression and CD8+ T cell infiltration in the tumour microenvironment. Notably, certain patients with tumours lacking detectable T cell infiltration at baseline and/or an inflammatory gene signature still had complete responses, suggesting that T-VEC promotes responsiveness to ICIs. No unexpected safety signals emerged from this trial. These data provided the rationale for MASTERKEY-265, a prospective, randomized clinical trial comparing T-VEC plus pembrolizumab versus pembrolizumab monotherapy43. In this study, 692 patients with stage IIIB–IVM1c melanomas were randomly allocated 1:1 to receive T-VEC plus pembrolizumab or placebo plus pembrolizumab. Response rates were numerically higher in the combination group (ORR 48.6% versus 41.3%); nonetheless, this study failed to meet its dual primary end points of PFS (median 14.3 months versus 8.5 months, HR 0.86, 95% CI 0.71–1.04; P = 0.13) and OS (median not reached versus 49.2 months, HR 0.96, 95% CI 0.76–1.22; P = 0.74). The reasons for this failure are not entirely clear but might be related to inadequate statistical power owing to the higher response rates in the pembrolizumab-only group in a population skewed towards stage III patients considering that all patients needed to have clinically accessible lesions. Another difference seen in MASTERKEY-265 was that T-VEC and pembrolizumab were administered concurrently, whereas, in the phase I study, pembrolizumab exposure was delayed until after two doses of T-VEC. Early exposure to an ICI could theoretically promote antiviral immunity and therefore result in more rapid viral clearance. The activity of OVs and ICIs might be dependent on many factors, including the underlying tumour histology and patient-specific immune factors. For example, in a phase I trial assessing the combination of perioperative nivolumab plus intralesional IL-12 gene therapy for patients with recurrent glioblastoma, a median OS of 16.9 months was reported, leading to an active phase II study of this combination (NCT04006119)44.

In a related strategy, an oncolytic HSV was engineered to directly express a single-chain variable antibody against PD-1 (ref. 45). This approach delayed tumour growth in two mouse models of glioblastoma and might offer an improved immune-related adverse event profile while providing direct immune-checkpoint inhibition locally within the tumour microenvironment. In addition to T-VEC, other OVs, including HSVs, reoviruses, poxviruses, coxsackieviruses and adenoviruses, are also currently under investigation in combination with ICIs, either mediated via transgenes or as systemic therapies in combination studies2.

OVs are also being evaluated in combination with chimeric antigen receptor (CAR) T cell therapies owing to the ability of OVs to promote the recruitment of T cells to tumours upon release of cytokines, such as TNF, IL-2 and IL-15, as well as of chemokines such as CXCL9 and CXCL10. Improved antitumour activity has been reported in preclinical studies exploring this approach46. Preclinical investigations with an IL-7-encoded oncolytic adenovirus in combination with CAR T cells targeting B7H3 were conducted in a mouse model of glioblastoma47. Both in vitro and in vivo synergistic anticancer activity was observed with this approach. Similarly, mesothelin-specific CAR T cells combined with an oncolytic adenovirus therapy promoting the expression of a soluble TGFβ receptor II fused with a human IgG Fc fragment were investigated in a mouse xenograft model of triple-negative breast cancer48. Rapid antitumour activity and cancer cell death were observed with this combination. These studies suggest that OVs could provide an interesting strategy for extending the benefits of CAR T cell therapy to patients with solid tumours, and one study combining an oncolytic adenovirus with HER2-specific CAR T cells is currently under way (NCT03740256). Furthermore, the combination of an oncolytic poxvirus with adoptive cell therapy using tumour-infiltrating lymphocytes (TILs) demonstrated substantial tumour regression and tumour-specific accumulation of TILs in a mouse model49.

Evidence suggests that OVs might promote the development of abscopal effects when combined with radiotherapy. Radiotherapy might increase viral replication and intercellular spread with increased antigen presentation, while data from other studies suggest that overlapping mechanisms involving inhibition of DNA repair following viral infection might be important50. Combination therapy with a genetically modified oncolytic vaccinia virus and radiotherapy has been tested in mouse models of BRAFV600E/D-mutant melanoma51. In this study, combination therapy promoted tumour cell death, including induction of caspase activity, leading to tumour regression in vivo. In another study, irradiation resulted in dose-dependent tumour regression in a mouse xenograft model of CRC following infection with a TRAIL-expressing oncolytic adenovirus52. Clinical experience with this approach is currently limited, although an oncolytic adenovirus expressing the human telomerase reverse transcriptase promoter OBP-301 demonstrated a favourable safety profile with promising activity in 13 patients when delivered via endoscopic intratumoural injection with concurrent radiotherapy in patients with oesophageal cancer deemed unable to tolerate standard therapies53. Furthermore, T-VEC in combination with radiotherapy is currently being tested in patients with soft-tissue sarcomas amenable to surgical excision (NCT02453191) as well as in patients with melanomas and other solid tumours (NCT02819843).

Traditional cytotoxic agents, such as DNA intercalators, nucleotide analogues or alkylating agents, are also being considered for combination with OVs. Chemotherapies are thought to directly target rapidly dividing cells; therefore, combining chemotherapy and an OV might lead to improved therapeutic responses. Preliminary data supporting this hypothesis have been reported. For example, in vitro and in vivo studies combining docetaxel with the IL-24-encoding oncolytic adenovirus ZD55-IL-24 demonstrated a significant increase in caspase 3 and caspase 8 expression and a dramatic increase in the extent of apoptosis in a mouse prostate cancer xenograft model compared with tumour tissues exposed to either ZD55-IL-24 or radiotherapy alone54. Furthermore, data from several preclinical studies involving mouse models demonstrate that previous exposure to cyclophosphamide enhances the potency of several OVs; therefore, initiating therapy prior to administration of an OV might be the optimal approach55,56,57. Clinical data on combinations of OVs and cytotoxic agents are currently limited, although emerging data are providing the impetus to reconsider the role of such combinations in chemotherapy58. In a phase II study involving 17 patients with head and neck squamous cell cancers, the combination of T-VEC and chemoradiotherapy resulted in a RECIST-defined response in 14 (82.3%) patients and a pathological CR was observed in 14 of 15 patients (93.3%) who underwent surgery27.

In contrast to systemic chemotherapy, which inhibits all rapidly dividing cells, molecularly targeted therapies inhibit tumour cells harbouring highly specific mutations, providing a highly personalized treatment strategy. For example, in patients with BRAFV600E-mutant melanoma, response rates to BRAF/MEK targeted therapies are typically quite high (~60–70%) although acquired resistance often follows initial treatment59,60. Therefore, the strategy of combining targeted therapies with OVs is being explored. Several groups have now reported promising data on the combination of T-VEC and BRAF/MEK inhibitors in mouse models of melanoma and thyroid carcinoma9,61,62. Similarly, preclinical studies have demonstrated the persistence and antitumour activity of the oncolytic HSV1 HF10 in combination with the EGFR tyrosine kinase inhibitor erlotinib in mouse xenograft models of pancreatic cancer63. Cetuximab is a monoclonal anti-EGFR antibody with synergistic activity against human CRC cells in combination with oncolytic HSV1 (ref. 64). Preclinical studies involving mouse models of colon cancer featuring high levels of matrix metalloproteinase 3 (MMP3) expression have shown no response to targeted MMP3 inhibitors alone although delayed tumour growth was observed when MMP3 inhibitors were given in combination with an oncolytic vesicular stomatitis virus (VSV)65.

Preclinical data on the activity of OV-containing combination therapies provide considerable support for broader clinical testing; however, clinical translation has thus far been challenging. Further mechanistic studies are needed to fully understand how the various therapeutic combinations mediate antitumour activity. Additionally, predictive biomarkers are needed to guide clinical development. Combination approaches, including potential triplet combinations, might be anticipated in the near future.

Challenges in clinical development

Preclinical challenges

The currently available mouse models have several important limitations that might impair the evaluation of OVs. Many mammalian viruses have a highly restrictive cell tropism, which hinders the use of mouse models for OVs that are designed to be active in humans. Immunocompromised mouse strains enable the use of cancer cells of human origin and might overcome the limitations of syngeneic mouse models; however, such immunocompromised hosts are generally not appropriate models for studies involving live replicating viruses. For such studies, an immunocompetent host with an intact adaptive immune system is generally preferred. Certain humanized mouse models might be useful in this regard but even these might lack certain specific cell types and/or the molecular elements needed to fully understand how OVs interact with the human immune system. For example, vaccinia virus is generally tropic for most mouse-derived and human-derived cancer cell lines, making this an ideal OV across several mouse models, whereas most mouse cells are resistant to HSV1 (the source of T-VEC) with a few notable exceptions, including A20 lymphoma and D4M melanoma cell lines9. Alternatively, humanized mouse models can be used for viruses that are only permissive in human tumour cells. However, humanized mouse models are expensive, labour-intensive and time-consuming and the need for patient blood matching creates further challenges. In particular, many humanized mouse models require HLA-matching and human immune cell engraftment, which can be challenging, especially when considering that such mice might remain vulnerable to systemic viraemia. Another limitation is that most preclinical studies involving OVs use subcutaneous tumours to simplify the administration of the OV under investigation. Limitations aside, these models do not provide an accurate model of the tumour microenvironment and generally do not reflect the disease characteristics of patients with multiple metastatic lesions. Orthotopic mouse models might provide more appropriate tumour sites and comparable microenvironments, although introducing the virus to the tumour site can be more challenging in these models. Patient-derived xenograft mouse models, despite providing a more accurate representation of the genetic landscape with adaptation to the mouse microenvironment, often also have long latency periods (typically around 75 days)66. Regardless of the model used, a lack of standardized assessment criteria for the quantification of any abscopal effects further hinders accurate in vivo preclinical testing. Developing more relevant and sophisticated mouse models that enable detailed investigations of both the mechanisms of resistance and sensitivity to OVs would ameliorate the development of these therapies67.

OVs can act as an in situ cancer vaccine by inducing ICD, which leads to the spatiotemporal release of soluble tumour-associated antigens, danger-associated molecular patterns and pathogen-associated molecular patterns. The success of such in situ vaccinations depends on optimal activation of the antiviral immune response; if this response is too strong, it might result in rapid viral clearance, which could limit the induction of antitumour immunity68,69,70,71. Thus, achieving optimal tumour regression with an OV depends on finding the optimal balance between the antiviral immune response and the antitumour immune response (Fig. 2). Humoral immunity-mediated rapid neutralization of OVs, for example, by virus-specific antibodies as well as by components of the complement system, is a considerable obstacle for the activity of nearly all OVs. This challenge might be especially relevant for OVs derived from endemic viruses, such as HSVs or adenoviruses, because patients previously exposed to viruses from the same family might have pre-existing cross-reactive antibodies that can impair effective viral replication72,73. Furthermore, patients with advanced-stage cancers are likely to require multiple OV injections and the emergence of neutralizing antibodies might preclude this.

Fig. 2: The balance between antiviral and antitumour immunity.
figure 2

Immune-mediated rejection of established cancers depends on the balance between antiviral immune responses and antitumour immunity. Antitumour immunity can potentially be ameliorated by the viral-induced immune response although, if the antiviral response is too strong, the virus will be rapidly cleared, resulting in insufficient induction of antitumour immunity. For example, certain viral infections, such as those involving herpes simplex viruses and poxviruses, result in rapid induction of neutralizing antiviral antibody titres and viral-specific T cells, non-specific complement activation, and rapid induction of pro-inflammatory signals. By contrast, antitumour immunity mediated by oncolytic viruses (OVs) often takes much longer to emerge and requires additional cross-presentation of tumour antigens in the context of the ongoing viral infection. Thus, an ideal OV will result in persistent viral infection with lower levels of neutralizing antiviral immunity, enabling prolonged antigen presentation. Ultimately, the emergence of antiviral and antitumour immunity can influence both tumour cells and their microenvironment. Tumours exposed to an effective OV are likely to undergo immunogenic cell death (ICD), resulting in increased exposure to recruited lymphocytes, which might be virus-specific or tumour-specific depending on the influence of the viral to tumour immune balance and most probably also involving cells of the tumour microenvironment (TME). The local expression of transgenes by OVs can further influence the immune balance within injected tumours. Thus, the optimal OV is sufficiently potent enough to induce a pro-inflammatory immune response but does not induce an antiviral immune response capable of premature elimination of the virus prior to the establishment of effective antitumour activity, including the release of relevant danger-associated molecular patterns (DAMPs) and/or pathogen-associated molecular patterns (PAMPs), and the induction of tumour-specific T cell responses.

Several strategies are being either tested or used in an attempt to reduce the incidence of antiviral immunity in patients receiving OVs. For example, protective coatings using cell-derived nanovesicles, liposomes or chemical polymers designed to physically protect OVs from passive or active immune components are all being explored74,75,76. Preclinical studies using ionic polymers or graphene oxide to shield the measles virus from pre-established neutralizing antibodies have led to increased antitumour activity in mouse models, probably owing to protection of the virus from rapid clearance74,75. Protective coatings can also be designed to incorporate a second function such as a specific cancer-targeting ligand that enables the OV to traffic specifically to tumour cells. For example, a cell membrane-derived adenovirus containing a nanovesicle coated with preS1 (a ligand targeting the taurocholate co-transporting polypeptide, which is expressed on the surface of hepatocellular carcinoma cells) protected the virus from neutralizing antibody-mediated clearance and promoted trafficking to cancer cells in a transgenic mouse model76. This approach seems promising although developing such protectively coated OVs has several complications, including difficulties regarding storage and/or stability, higher manufacturing costs, and scalability challenges.

Another strategy designed to limit early viral clearance is the use of ex vivo OV-loaded cells that can then be administered to the patient. Cellular carriers for OVs are often derived from tumour-infiltrating immune cells, such as dendritic cells, T cells, mesenchymal stem cells and macrophages, as these cells can all naturally traffic to sites of tumour growth77,78. Interest in using cancer cell-derived carriers loaded with OVs as delivery vehicles is also emerging, although concerns regarding tumorigenicity could limit the future clinical development of this approach78,79. The ability to deliver OVs to brain metastases might also be aided by the use of certain cellular carriers. For example, tumour-tropic mesenchymal stem cells carrying the adenovirus-based OV ICOVIR17 demonstrated extended survival in mouse models of glioblastoma with substantial accumulation of the OV observed in brain lesions80. The safety and tolerability of the adenovirus-based OV δ24-RGD, delivered via infusion into carefully selected central nervous system (CNS) arteries, is currently being assessed in a phase I trial involving patients with recurrent high-grade gliomas (NCT03896568)81. A phase I/II trial investigating an intraperitoneally administered oncolytic measles virus in women with recurrent gynaecological cancers is also currently under way (NCT02068794). The OVs administered in both studies use mesenchymal stem cells as delivery vehicles.

Most OVs require intratumoural administration, which poses another major challenge to clinical development. Pharmacological studies designed to clearly define the optimal biodistribution and pharmacokinetics/pharmacodynamics of OVs following intratumoural delivery are currently a priority in the field. Furthermore, standardized metrics for the assessment and quantification of several parameters, including viral replication within tumour cells, the absence of replication in non-malignant cells, the detection of viral shedding in patient bodily fluids, the presence of latent infection, antiviral immunity and antitumour immune responses, are still needed. Thoroughly assessing these metrics will require the development of standardized assays and end points for use in clinical trials. To date, OV dosing strategies have largely been empirical and data from studies designed to compare the performance of weight-based based dosing versus that of tumour volume-based dosing are unavailable. Similarly, the optimal dose and volume of OV to administer is not clear and might depend on several factors including the virus, the histology of the target tumour and other patient-related factors. The timing of viral delivery, especially in relation to other systemic therapies, remains largely unexplored and is critical for appropriate clinical study design.

Finally, knowledge of the mechanisms of innate and acquired resistance to OVs is limited and these might differ between directly injected tumours versus uninjected lesions. Furthermore, predictive biomarkers to enable better patient selection for OV therapy are at a very early stage of development. In a study report published in 2021, tumours harbouring loss-of-function mutations in JAK1/2 were resistant to ICIs although cells derived from these lesions had enhanced viral replication when exposed to both HSV1 and VSV compared to JAK1/2 wild-type cells82. Deficient JAK signalling probably interferes with the antiviral immune response, which might explain why OVs appear to be more permissive in JAK-deficient tumour cells. Similarly, studies in preclinical models of melanoma indicate that STING deficiency is associated with improved OV replication and lysis by T-VEC83. Data from these studies suggest that factors related to intracellular antiviral activity might be potential biomarkers of OV response although further clinical validation is needed (Fig. 3). Other potential biomarkers of OV activity that merit further exploration include tumour-intrinsic factors such as tumour mutational burden, inflammatory/antiviral gene expression and tumour cell PD-L1 expression, all of which have already been investigated as biomarkers for use with other immunotherapies in selected cancers84. Immune system-related factors, such as TILs, and systemic factors, such as circulating tumour DNA and exosomes, should also be explored as potential biomarkers in prospective clinical trials (Fig. 3).

Fig. 3: Biomarker discovery for OVs.
figure 3

Thus far, no predictive biomarkers of clinical benefit from an oncolytic virus (OV) have been established although several potential molecular and/or cellular characteristics are under consideration. These include tumour-intrinsic characteristics, immune-related characteristics and systemic characteristics. Tumour-intrinsic characteristics include tumour mutational burden, inflammatory and antiviral gene expression, tumour cell PD-L1 expression, and other factors related to the tumour microenvironment. Immune-related characteristics include the induction of innate immunity and the accumulation of infiltrating tumour-reactive lymphocyte populations. Systemic characteristics refer to elements measurable in the peripheral circulation such as circulating tumour DNA and exosomes. Further research is needed to identify potential biomarkers and these will need to be validated in prospective clinical trials.

Clinical challenges

OVs generally have a tolerable safety profile; nonetheless, OVs are still live replicating viruses that require special attention when used in the clinic (Box 2). The replicative potential of OVs necessitates special logistical and biological safety considerations related to the risks of viral shedding and of unintentional transmission from patients to health-care workers, close household contacts and the environment. Minimizing these risks requires careful attention to storage, preparation and handling, and administration of the OV as well as careful management of injected sites and the correct disposal of soiled bandages and any other potentially infected materials. Guidelines and a developmental plan outlining protocols for the safe storage, handling and administration of OVs as well as policies on how to handle accidental spills, overdoses and the proper sterilization of treatment areas in which the OV is administered are all required, and several guidance documents are now available85,86,87,88. Safe use of OVs also depends on health-care provider education and training as well as education of patients on injection site management. Regarding the latter, limited data from studies evaluating differences in technical approaches, for example, comparing the fan technique versus central injection or intralesional versus intradermal delivery, are currently available89. Guidance on selection of the most appropriate lesions for injection and whether a wider virus distribution might be preferable compared to a higher dose at one location is also limited.

To date, no reports of confirmed transmission to close household contacts or environmental exposures have emerged although a few health-care workers have reportedly been infected while handling T-VEC and, in one study, 8.4% of household contacts reported cold sores, although these were not confirmed as infections from T-VEC and none were considered clinically severe90. T-VEC is sensitive to antiviral therapy such as acyclovir, and this agent can be used following accidental exposures such as from needlestick injuries. Certain safety warnings might be required regarding the use of specific viruses in patients who are immunosuppressed. Concerns also exist regarding the use of genetically engineered OVs containing recombinant DNA elements and the potential for such elements to recombine with naturally occurring wild-type viruses.

The need for clinical studies assessing the role of dosage, volume, timing, route of administration and combinations remains a priority and such data are expected to make OVs easier to handle, manage and administer. For example, the current requirement for storage at –80 °C presents an immediate barrier to implementation in pharmacies owing to a lack of freezer units at the appropriate temperature. The preparation and handling of OVs also require a sterile biosafety cabinet although such dedicated spaces are often unrealistic in pharmacies, which are also required to prepare a wide range of other agents. Furthermore, the requirement of different doses and time points for initial injections versus later injections as well as only being able to determine the final drug volume when the patient’s lesions are measured can create a further burden. These challenges might lead to delays in administering the OV to patients. Furthermore, challenges related to intratumoural injections arise from the need for tumours to be physically accessible. This poses issues for patients with occult lesions or those that regress to an undetectable size upon PR. The use of ultrasonography might help in navigating lesions for OV injections.

Clinicians should be aware that delayed responses might be common in patients receiving T-VEC and perhaps also with other OVs. This is a situation in which new tumours might arise or established lesions can increase in diameter on measurement owing to local inflammation and/or the accumulation of immune cells rather than true tumour-specific expansion. Indeed, in OPTiM, >50% of patients had measurable tumour growth prior to an objective response. Thus, data from calliper measurements and/or imaging might be misleading. If tumours remain stable or a delayed response is suspected, biopsy sampling might be indicated to delineate tumour growth from a true pro-inflammatory response. If viable tumour cells are detected, patients should be considered to have true tumour progression and consultation with a pathologist familiar with immunotherapy might be helpful.

The need for intratumoural injection is a major obstacle to the clinical adoption of OVs and a growing body of preclinical data supporting a role for intravenous delivery are currently emerging. This approach theoretically enables the exposure of all lesions, potentially enabling widespread OV infection and also obviates the need for complex localization equipment and expertise. The intravenous delivery of experimental OVs has been investigated in several early phase clinical trials. For example, enadenotucirev, a chimeric oncolytic adenovirus, was administered intravenously to patients with resectable solid tumours 1–2 weeks prior to surgery in a phase I study, with evidence of viral particles consistently detected on examination of resected tumour material91. In another study, the oncolytic vaccinia virus pexa-vec was given intravenously to three patients with melanoma and six with CRC prior to the surgical resection of metastatic lesions. This study demonstrated an acceptable safety profile and the presence of the OV was confirmed upon immunohistochemical examination of resected tumour specimens92. Intravenous delivery of OVs offers a clear logistical advantage and is being explored further; however, several drawbacks also need to be considered. Firstly, the optimal dosing for intravenous OV delivery is unclear as the virus will be diluted in the peripheral circulation, making bioavailable titres at any individual tumour site unpredictable. In addition, viral particles could be prematurely cleared by circulating neutralizing factors (such as by antiviral antibodies or complement) that might further limit the effective dose of available virus at any given lesion. Thus, further studies on intravenous delivery are needed to better understand the pharmacokinetics of specific OVs upon entering the systemic circulation. New strategies aimed at protecting intravenously administered viruses, such as nanodelivery vehicles or intracellular viral delivery, could provide novel and interesting methods of improving the local delivery of OVs via intravenous administration. Studies of intravenously administered OVs should include an assessment of viral titres at the tumour site in order to ensure the presence of a sufficiently high quantity of replicating virus; many studies rely on PCR or immunohistochemical analysis and this approach might misrepresent the bioequivalent dose of replicating virus. In a review of data from all OV clinical trials, intravenous delivery was generally safe and was used in many studies albeit with lower response rates than those reported following intratumoural delivery2. Thus, intravenous delivery should be considered experimental for now. Finally, several challenges exist regarding the administration of OVs in the outpatient setting, which might be amplified by combination therapies for which injected patients might need to travel to open infusion units.

Regulatory challenges

T-VEC is the first and thus far only FDA-approved OV although three others have been approved globally, including an oncolytic adenovirus in combination with chemotherapy for patients with nasopharyngeal cancer in China, an oncolytic HSV1 for the treatment of glioblastoma in Japan, and an oncolytic echovirus for melanoma approved in several countries in Eastern Europe (although this was withdrawn from the market in 2019; Table 1). Nonetheless, additional approvals have been elusive and might be hindered by several regulatory challenges (Box 3). In contrast to most anticancer therapeutics, OVs are unique in two main ways. Firstly, these are live, replicating viruses that are, in some cases, also engineered to express human transgenes. Secondly, most OVs are administered intratumourally. These factors make many of the traditional methods of establishing clinical trial eligibility end points, pharmacokinetics, dosing and scheduling inappropriate for the assessment of OVs. In the absence of preclinical data, much of the clinical development has been empirical, although prior studies provide a roadmap for regulatory considerations.

Obtaining evidence justifying a specific dosing strategy might be challenging and investigators should consider several factors, including the maximum concentration achievable based on current good manufacturing practices, manufacturing capabilities, the immunogenicity of the virus, the likelihood of pre-existing neutralizing antiviral antibodies, tumour histology and the effects of any transgene expression. Additional considerations should include whether dosing will be adjusted to tumour volume and what the maximum safe daily dose can be based on viral pathogenicity. Data from animal models might not be especially useful in this regard unless models that permit viral replication are available. Further, attention to the selection of lesions that are most suitable for injection is necessary and the inclusion of appropriate eligibility and exclusion criteria based on the underlying pathogenicity and immunogenicity of the specific virus should be considered.

Inclusion of the most appropriate clinical end points is another important issue. Current response criteria, such as RECIST, are designed largely to evaluate responses to cytotoxic anticancer agents. However, a response to most OVs involves a substantial immune component and delayed clinical responses have also been reported; therefore, alternative end points and/or an allowance for treatment beyond standard disease progression might be needed. Furthermore, the patterns of response with T-VEC are somewhat distinct from those seen with ICIs and can include the appearance of new lesions early in the course of treatment; active treatment of these lesions often leads to an objective response. Indeed, in OPTiM, just over 50% of patients had evidence of such delayed responses4. Thus, in the clinic, considering whether patients have ‘true’ disease progression warranting a change in therapy or if newly emergent lesions in fact reflect a delayed response that merits further injection is important. Typically, this phenomenon occurs early in the course of treatment with T-VEC and new lesions often appear in a regional pattern close to other pre-existing injected lesions. Furthermore, because certain lesions might not be amenable to direct injections, monitoring both injected and uninjected lesions is advisable. Additional response assessments might include CT or PET imaging, photography documentation, and/or biopsy sampling, although the timing of these interventions needs to be standardized prior to study implementation. Intratumoural RECIST criteria have been proposed in an attempt to standardize evaluations of response to OVs93. Intratumoural RECIST accounts for an aggregate response across all lesions, implements iRECIST for the assessment of potential pseudoprogression and delayed responses, and includes metrics for the assessment of both directly injected and uninjected lesions. However, these criteria are quite cumbersome as they stipulate direct measurements of all established lesions, therefore these criteria have not been widely adopted or validated relative to standard RECIST.

Other issues include selection of the most appropriate controls for randomized studies, which should include another intratumourally administered therapy whenever possible. Further discussions on whether abscopal responses are a necessary part of the evaluation and approval of an OV are also essential. Alternative strategies, such as neoadjuvant therapy followed by the assessment of pathological responses in resected tumour specimens, might also be helpful in the initial assessment of a potential OV. Similar to the experience with other immunotherapies, neoadjuvant administration of OVs is an especially interesting aspect given that the tumour burden is often much lower in this setting and tumours have probably undergone less immunoediting and are less exposed to immunosuppressive pressures. For example, serial gene expression profiling suggests a very rapid change in tumour cell gene expression within 24 h of intratumoural T-VEC injection in patients with cutaneous tumours94. This observation suggests that the immunological benefits occur rapidly after injection, including enabling the intact tumour to act as a source of neoantigens prior to surgery. Thus, OVs might be an ideal method of inducing an ‘in situ’ vaccination effect and can be administered rapidly, enabling early surgical management.

Promising novel OVs

Numerous OVs are currently in clinical development. While a full description of all OVs currently in clinical development is beyond the scope of this Review, this information is available elsewhere3,40,95. Here, we describe selected OVs in the context of the most promising clinical indications, in which several OVs are currently in the later stages of clinical development.


Malignant glioma is one of the most aggressive cancers, with few effective therapeutic options. Patients with gliomas usually undergo debulking surgery followed by chemoradiation; nonetheless, median OS remains dismal at around 15 months96. Treatment of patients with gliomas is further complicated by the need for any systemically administered agents to cross the blood–brain barrier and the need to avoid excessive inflammation as this might lead to severe neurological deficits. Nonetheless, glioma is generally restricted to the CNS; therefore, local approaches, such as OVs, are an intriguing treatment option given the localized nature of tumour development and that glioma cells are often quite permissive to viral infections, including by HSV1, adenoviruses and polioviruses.

Teserpaturev is a triple-mutated, third-generation HSV1-based OV approved in Japan for patients with malignant gliomas96,97. Results from a single-arm phase II trial demonstrated a 1-year OS of 84.2% in patients with recurrent and/or residual glioblastomas. Teserpaturev was generally well tolerated in this study, with fever, vomiting, nausea and leukopenia reported as the most common adverse events97,98. The oncolytic HSV1 strain, G207, is also under investigation for the treatment of gliomas99,100,101. This agent was evaluated in 12 paediatric patients with high-grade gliomas and resulted in only grade 1 adverse events. Responses were observed in almost all patients (11/12) with a median OS of 12.2 months101. Also of note, rQNestin34.5v.2, a genetically engineered HSV1 OV, is being tested in combination with cyclophosphamide in patients with recurrent malignant gliomas (NCT03152318). Thus far, IL-10 and CCL2 have been detected in the serum of patients with glioblastoma receiving this OV, probably indicating the development of an immunosuppressive microenvironment. However, pharmacological inhibition of NOTCH signalling has been shown to rescue rQNestin34.5-induced immunosuppression in mouse models102.

Tasadenoturev, an adenovirus type 5-based OV, has been assessed in 12 paediatric patients (median 9 years of age) with diffuse intrinsic pontine gliomas. Measurable reductions in tumour diameter were observed in 75% of patients. Three serious adverse events were reported, including grade 3 headaches and muscle weakness. Median OS was 17.8 months103. The safety and efficacy of tasadenoturev have been further assessed in a cohort of 37 adult patients with recurrent malignant gliomas103; 20% of patients included in the long-term efficacy analysis remained alive at 3 years post-treatment, including 3 with near-complete responses (loss of ~95% of tumour material)104. Only 2 patients had adverse events that were deemed to be treatment related and all events were grade 1–2 in severity. No dose-limiting toxicities were observed.

In addition to the DNA viruses, several RNA viruses are currently also being tested in clinical trials involving patients with gliomas. These include OVs based on poliovirus, despite the considerable pathogenicity of unmodified strains of this virus. To overcome this obvious safety concern, the viral internal ribosome entry site of the widely used Sabin oral poliovirus vaccine, comprising live-attenuated strains of three different poliovirus serotypes, was swapped with the internal ribosome entry site of human rhinovirus type 2 for further attenuation105,106,107. This recombinant attenuated poliovirus, known as PVS-RIPO, enters tumour cells via the CD155 receptor, which can be found in many cancers, including gliomas and melanomas108. PVS-RIPO is currently being evaluated in early phase studies105,109. In a phase II study, patients with recurrent glioblastomas receiving intratumoural PVS-RIPO had a 1-year OS of 21% with grade 3–5 adverse events (including 1 treatment-related death) in 19% of patients105.

Bladder cancer

NMIBC provides another intriguing indication for OV therapy. Most of these tumours arise in the urothelium of the bladder and are amenable to both direct injections via cystoscopy or intravesical infusions. Indeed, intravesical infusions of BCG, a minor human bacterial pathogen, are the current standard-of-care adjuvant therapy for patients undergoing surgery for intermediate-risk and especially high-risk NMIBC.

Several adenoviruses are currently under clinical investigation for patients with BCG-refractory NMIBC. CG0070, an adenovirus serotype 5-based OV encoding GM-CSF, has been evaluated in a phase I/II trial110. A promising dose schedule-dependent complete response rate of 48.6% was reported among 35 patients receiving one or more intravesical infusions with a median duration of response of 10.4 months110. In a phase II trial exploring this approach in 45 patients with BCG-refractory NMIBC, CG0070 demonstrated a 47% 6-month CR rate111. CG0070 was well tolerated; the most frequently reported treatment-related adverse events included bladder spasms, haematuria and dysuria. Urinary tract infections were the most frequent treatment-related infectious events (in 16% of patients) and other likely OV-related events included flu-like symptoms and fatigue. The combination of CG0070 plus pembrolizumab is also currently being investigated in a similar patient population with promising results reported from an interim analysis (NCT04387461)112. Nadofaragene firadenovec is a replication-deficient, non-oncolytic adenovirus that enables the delivery of genes to urothelial cells, specifically cDNAs encoding the human IFNA2B gene. Data from a phase III multicentre study of the safety and efficacy of nadofaragene firadenovec in patients with BCG-refractory NMIBCs were reported in 2021 (ref. 113). In this study, 157 patients received a single intravesical dose of 3 × 1011 viral particles per millilitre of nadofaragene firadenovec in a total fluid volume of 75 ml. Repeat dosing was permitted at 3, 6 and 9 months in the absence of high-grade recurrence and the primary end point was CR in patients with carcinoma in situ (with or without a high-grade Ta or T1 tumour) at any point. Overall, 53.4% of patients had a CR within 3 months of the first dose and responses were maintained in 45.5% of patients at 12 months. The most frequent grade 3 adverse events were micturition urgency in two patients (1%). Based on these data, the FDA approved intravesical infusion of nadofaragene firadenovec for the treatment of BCG-unresponsive NMIBC in December 2022.

Other cancers

Seprehvir is an HSV1716-based OV that is currently being evaluated in a phase I/II trial involving children and young adults (11–30 years of age) with advanced-stage solid tumours114. Preliminary data indicate no dose-limiting toxicities, with the HSV1 genome detected in the peripheral blood of four of the nine patients, an observation consistent with de novo viral replication. Two patients had stable disease in response to seprehvir114. Another oncolytic HSV1-based OV, HF10, is being assessed in combination with ipilimumab in patients with metastatic or unresectable melanomas115 and as a monotherapy in patients with advanced-stage cutaneous solid tumours and/or those with superficial lesions116. Preliminary reports indicate that HF10 is well tolerated, with the majority of adverse events being of grade ≤2 in severity. Three patients had HF10-related grade ≥3 adverse events (embolism, lymphoedema and diarrhoea). ORR at 24 weeks was 41% and 68% of patients had stable disease. Median PFS was 19 months and median OS was 21.8 months115. In a phase I dose-escalation trial assessing HF10 in patients with advanced-stage cutaneous solid tumours and/or those with superficial lesions, multiple intralesional injections of HF10 were well tolerated. Six patients reportedly had HF10-related adverse events, including flu-like symptoms and injection site reactions. Despite rapid clearance from the blood, HF10 was shown to have some potential to provide antitumour activity114.

Among other HSV1-based OVs, RP1, which encodes GM-CSF and a fusogenic form of the envelope glycoprotein of gibbon ape leukaemia virus that is designed to promote local viral spread, is being evaluated in combination with ICIs in patients with cutaneous squamous cell carcinomas (NCT04050436) and melanoma and non-melanoma skin cancers117. Similarly, ONCR-177 is currently being tested in phase I trials involving patients with advanced-stage and/or metastatic solid tumours and is being assessed both alone and in combination with pembrolizumab (NCT04348916).

Reoviruses are ubiquitous double-stranded RNA viruses. Most people have antibodies against this family of viruses owing to prior exposure; however, cancer cells overexpressing EGFR and/or with activation of the RAS signalling pathway are highly susceptible to reovirus infection118. Reolysin, an oncolytic reovirus, has been evaluated in several clinical trials for various indications with promising antitumour activity and a tolerable safety profile119,120,121,122,123,124. Data are available from several clinical trials designed to assess reolysin in combination with radiotherapy, immunotherapy and chemotherapy125,126. For example, safety and tolerability of intravenous reolysin in combination with pembrolizumab plus chemotherapy (5-fluorouracil, gemcitabine or irinotecan) has been tested in a phase Ib study involving patients with advanced-stage pancreatic adenocarcinomas127. Disease control was observed in 3 of the 10 evaluable patients, with 1 patient having a PR lasting 17.4 months. Grade ≥3 adverse events were reported in 2 patients only127. In a phase II study, patients with pancreatic adenocarcinomas received reolysin in combination with gemcitabine: 1 patient had a PR and 23 had stable disease. Median OS was 10.2 months and combination therapy was well tolerated (grade 3–4 toxicities included anaemia, neutropenia and thrombocytopenia, in 27%, 27% and 6% of patients, respectively)126. Coxsackievirus-based OVs have also been evaluated alone and in combination with ICIs, including ipilimumab and pembrolizumab in patients with advanced-stage cutaneous and uveal melanomas, as well as in those with non-small-cell lung and bladder cancers. Intravenous administration of oncolytic reoviruses and coxsackieviruses have also been evaluated, although the research programmes for these agents have now been discontinued128.


OVs offer immense promise as anticancer therapies. Based on supporting data from clinical trials and real-world clinical studies with T-VEC, the selection of patients has become more focused and T-VEC provides an additional therapeutic option for selected patients with melanoma. Owing to the acceptable safety profile and a mechanism of action that largely does not overlap with that of other therapeutic modalities, T-VEC is being evaluated in combination with other anticancer treatment strategies, with promising early phase clinical data available. A number of other promising OVs are in clinical development, and ongoing research in this area includes identifying novel OV delivery methods such as intravenous administration. Considerable preclinical, clinical and regulatory challenges continue to impair OV development, which is further complicated by issues relating to the storage and administration of a live virus and the need for intratumoural injections. Renewed efforts to better understand the biology and immunology of OVs are leading to new OV strategies and the identification of potential predictive biomarkers. Furthermore, ongoing discussions between regulatory agencies, scientists, clinicians, industry representatives and professional societies are leading to better patient selection and study designs that should enable the full potential of OVs to be realized for patients with cancer.