Addition of oncolytic virotherapy to clinical isolated limb perfusion for melanoma and sarcoma activates antitumor immunity

Abstract

Background

We previously showed that oncolytic virotherapy delivered by isolated limb perfusion (ILP), combined with immune checkpoint inhibition, prevents both local tumor progression and systemic metastases in an animal sarcoma model.

Methods

We describe a first-in-human phase I/II trial combining oncolytic herpes simplex virus, talimogene laherparepvec (T-VEC), with melphalan and tumor necrosis factor-alpha delivered by ILP, in patients with locally advanced sarcoma or melanoma.

Results

T-VEC/ILP is well tolerated, with an overall response rate of 53% in all patients and 44% in sarcoma. Importantly, we report durable complete responses in sarcoma subtypes usually unresponsive to ILP. Translational analysis of longitudinal tumor and blood samples showed that T-VEC induced an inflammatory gene expression profile within injected tumors, which was more sustained in sarcoma than in melanoma. In relation to clinical outcome, responding patients with sarcoma showed a greater increase in gene expression for interferon response after virus treatment than non-responding patients. Analysis of the T-cell repertoire (TCR) in tumor and blood showed that clonality was higher in the tumor, but lower in the blood, in responders following virotherapy, suggesting that virus treatment may expand intratumoral T-cell clones that recognize tumor and/or viral antigens. Increased TCR diversity in the blood was suggestive of a systemic immune response.

Conclusions

These clinical and translational findings support the further development of oncolytic virotherapy/ILP combinations to activate both systemic and local antitumor immunity, including in tumor types such as sarcoma, which are largely refractory to current treatment with immunotherapy.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Isolated limb perfusion (ILP) is an effective limb salvage technique for patients with advanced extremity melanoma or sarcoma who are facing amputation but currently has no antitumor effects outside of the limb. Talimogene laherparepvec (T-VEC) is an oncolytic herpes simplex virus that is tumoricidal and immune-activating for melanoma, although without established efficacy in sarcoma.

WHAT THIS STUDY ADDS

• We have shown that combining ILP with T-VEC can result in durable complete responses in sarcoma subtypes that are usually unresponsive to ILP. Analyses of tumor and blood samples after T-VEC injection and ILP showed an expansion of intratumoral T-cell clones that potentially recognize tumor and/or viral antigens. We also found increased diversity of the T-cell repertoire in blood, which is suggestive of a systemic immune response.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The demonstration of enhanced local tumoricidal effects of this combination therapy in treatment-resistant sarcoma, coupled with evidence of generation of a systemic immune response, builds on our preclinical data and provides further data strongly supporting a future trial of high-dose intraperfusional delivery of virus and bio-chemotherapy by ILP, coupled with checkpoint inhibitors, in patients with locally advanced oligometastatic sarcoma.

Introduction

Isolated limb perfusion (ILP) is a surgical technique that allows the delivery of high-dose bio-chemotherapy to a limb that is affected by an advanced malignancy. It is most commonly used for soft tissue sarcomas or advanced in-transit malignant melanomas that are untreatable by surgical techniques other than amputation.¹ ILP involves the surgical isolation of the limb from the systemic circulation, by a proximal limb pneumatic tourniquet and large-vessel vascular clamps, and its perfusion using an oxygenated hyperthermic extracorporeal bypass circuit. In this way, concentrations of cytotoxic chemotherapy that are far greater than could otherwise be tolerated with systemic delivery can be delivered safely. In addition, the vasoactive cytokine, tumor necrosis factor-alpha (TNF-α), can be added to the perfusion circuit to increase the distribution of cytotoxic agents within the tumor.² After a 60 min perfusion, the chemotherapeutic agents are washed out from the limb and continuity between the limb and systemic circulation is re-established prior to surgical closure.³

ILP can be effective for localized disease. The overall local response rates for patients with either sarcoma or melanoma after a single treatment can be as high as 70%, and most patients with otherwise unresectable disease can avoid amputation.⁴ However, despite anecdotal evidence of abscopal (ie, distant) antitumor effects in small numbers of patients, ILP must be considered to be a purely local treatment. Despite good local disease control, most patients ultimately develop fatal distant metastatic disease.⁵

The development of immunotherapy, specifically immune checkpoint inhibitors, has markedly improved the outcome for patients with immunogenic malignancies such as malignant melanoma in the metastatic, adjuvant and, most recently, neoadjuvant setting.6,8 However, this benefit has not translated into meaningful responses in malignancies which are mostly immunologically inert, such as soft tissue sarcoma.9,11 Furthermore, even in immunogenic malignancies, response to immunotherapies varies, with little activity seen in a considerable proportion of patients.¹² Combining immune checkpoint inhibitors (ICI) with other immunostimulatory treatments is a promising strategy that could improve response rates in these patient groups.¹³

Oncolytic virotherapy (OV) is a prime candidate for such an approach. OV leads to local tumoricidal effects through direct tumor cell lysis following viral replication specifically within tumor cells, leading to activation of an innate and adaptive antitumor immune response targeting tumor-associated antigens.¹⁴ Furthermore, any local immune stimulation within the tumor (often targeted by direct injection) can invoke systemic immune responses against distant metastatic disease which has not been directly targeted.¹⁵ OV has proven clinical value as a monotherapy, as exemplified by the worldwide approval of the modified herpes simplex type I virus talimogene laherparepvec (T-VEC), which is licensed for the intratumoral treatment of patients with Stage IIIB-IV melanoma.¹⁶

While concomitant treatment with T-VEC and ICI was unsuccessful,¹⁷ prior treatment with T-VEC has been shown to improve response rates to subsequent immune checkpoint blockade in patients with malignant melanoma.^{18 19} Responses were seen even in patients with immunologically-inert tumors at baseline, who would be unlikely to respond to ICI alone. T-VEC has also shown efficacy against both injected and non-injected lesions, demonstrating the ability of the virus to generate abscopal antitumor effects.²⁰ This ability of OV to remodel an immunologically inert “cold” into an inflamed “hot” tumor microenvironment, thus improving responses to subsequent immune checkpoint blockade, may not only be of benefit to patients with melanoma, but also for those with inherently more ICI-resistant tumors, such as advanced extremity sarcomas.

We previously hypothesized that combining OV with ILP may increase the intratumoral distribution and efficacy of the virus, both due to direct physical delivery of the virus into the afferent arterial circulation of the limb bearing the tumor and the vasoactive effects of TNF-α. To test this hypothesis, we developed an animal model of advanced soft tissue sarcoma suitable for ILP.²¹ Using this model, we demonstrated that combining intravascular OV with ILP led to improved distribution of OV within the tumor, which improved local responses compared with standard therapy.²² Further studies demonstrated that this combination also increased responses to subsequent ICI therapy, both improving local disease control and preventing distant metastases.²³ These systemic effects were associated with a marked remodeling of the local tumor microenvironment, characterized by increased CD8+T cell infiltration, and expansion and activation of antigen-presenting cells.

Building on these promising results, we conducted a phase I/II first-in-human clinical trial combining the oncolytic virus, T-VEC, delivered intratumorally (consistent with current clinical practice), with ILP in patients with extremity soft tissue sarcoma or in-transit melanoma. Here, we report the safety of this combination, alongside data on clinical responses and translational exploratory analyses of both local and systemic antitumor immune effects.

Results

Clinical trial design and patient characteristics

Between August 2018 and July 2021, a total of 15 patients were enrolled, comprising 6 in the initial phase I safety cohort and 9 in the phase II expansion. The study schedule is shown in figure 1. The demographics of the enrolled patients are summarized in table 1. The cohort included 8 (53%) women, with a median age at the time of ILP of 67 years (range 26–82). All patients had a performance status of 0 (87%) or 1 (13%). The pathology was melanoma in 6 (40%) patients and sarcoma in 9 (60%) patients. All patients had disease in the limb that was too extensive for limb-conserving surgery; that is, the only oncological surgical operation was an amputation. Histological subtypes in patients with sarcoma are provided in table 1.

Figure 1. Clinical trial design. The seroconversion dose of T-VEC was given at 4–6 weeks before ILP, the first treatment T-VEC dose was given 2–3 weeks before ILP and the final treatment dose of T-VEC was given on the day of surgery after induction of anesthesia but before the ILP. Tumor biopsies and blood samples for translational analyses were taken before the seroconversion T-VEC dose (baseline), on the day of surgery after the second T-VEC dose had been administered but before the ILP (pre-ILP) and at 2 weeks after the ILP (post-ILP). A final blood sample only was taken 12 weeks after ILP. ILP, isolated limb perfusion; IT. Intra-tumoural; TNF, tumor necrosis factor; T-VEC, talimogene laherparepvec.

Table 1. Patient demographics and baseline characteristics.

Patient characteristics		Frequency
		n (%)
Number of patients		15
Age at ILP (years) median (range), (IQR)		67 (26–82), (60–79)
Gender	Female	8 (53)
	Male	7 (47)
Ethnicity	White British	13 (87)
	White and black Caribbean	1 (7)
	Other	1 (7)
ECOG status at enrollment	ECOG-0	13 (87)
	ECOG-1	2 (13)
	ECOG-2	0 (0)
Histology subtype	Melanoma	6 (40)
	Sarcoma	9 (60)
Sarcoma subtype (n=9)	Angiosarcoma	3 (33)
	Myxofibrosarcoma	2 (22)
	UPS	2 (22)
	Epithelioid sarcoma	1 (11)
	Myxoid liposarcoma	1 (11)

ECOG, Eastern Cooperative Oncology Group; ILP, isolated limb perfusion; UPS, Undifferentiated Pleomorphic Sarcoma.

All six patients with melanoma had advanced, multifocal, limb-confined in-transit disease but no distant metastases (Stage IIIC). Of these six patients, three had undergone prior systemic treatment for locally advanced disease. One patient had received pembrolizumab alone, one patient had received combination nivolumab and ipilimumab and one patient had undergone a prior ILP, combination dabrafenib and trametinib, combination nivolumab and ipilimumab and treatment in a phase I clinical trial before entry into this study.

Of the nine patients with sarcoma, all had disease confined to the limb and no evidence of distant metastases. Five patients presented with locally advanced primary disease and none of these patients had undergone any previous treatment. Four patients presented with multifocal locally recurrent sarcoma after previous unsuccessful limb-conserving surgery. One patient had previously received two lines of cytotoxic chemotherapy (single-agent doxorubicin, then single-agent paclitaxel) and a previous ILP, and one other patient with a locally metastatic epithelioid sarcoma in the arm had received tazemetostat in a clinical trial.

Clinical response, safety and tolerability

The same study treatment was given to all enrolled patients, comprising a single dose of up to 4 mL of T-VEC at 1×10⁶ PFU/mL (the “seroconversion” dose) and two therapy doses of up to 4 mL of T-VEC at 1×10⁸ PFU/mL. The third dose was given during general anesthesia, immediately prior to the ILP procedure. The total number of injected lesions ranged from 1 to 8 per patient. Multilesion injections were more common in patients with melanoma, as they had a greater incidence of multifocal disease within affected limbs.

The investigational T-VEC treatment did not impact the perioperative outcomes of the ILP procedure (table 2). ILP was technically successful in all patients. Early discontinuation was required in one patient due to a leak from the perfusion circuit, but leak rates were in keeping with previous series.²⁴ The median maximum perfusion temperature was 39°C (range 37.7–39.2) and the median length of hospital stay following ILP was 5 days. No evidence was noted of increased locoregional toxicity following ILP compared with historical series.

Table 2. Pre-ILP and post-ILP variables.

Variable		Frequency n (%)
Site	Lower limb	14
	Upper limb	1
ILP approach	Femoral	13
	Inguinal	1
	Brachial	1
Median leak rate (range)		0 (0–16)
Median max perfusion temperature (range), (IQR)		39 (36–39.4)
Median length of stay (range), (IQR)		5 (4–7)
Median Wieberdink toxicity (range), (IQR)		2 (1–3)

ILP, isolated limb perfusion.

Adverse events (AEs) were assessable in all enrolled patients. A total of 124 AEs were recorded, although the majority were grade 1–2 (88%). A total of six patients experienced grade 3 reactions. One grade 4 and one grade 5 AE were noted, neither of which was thought to be related to the study treatment. The grade 5 AE occurred outside the 30-day dose-limiting toxicity assessment window. One patient developed bradycardia and a very brief period of asystole in the immediate postoperative period following the ILP. This was assessed as a consequence of anesthesia rather than a reaction to the study treatment. The patient was immediately resuscitated, and no long-term sequelae have been observed. A second patient developed a malignant pleural effusion and a subsequent fatal pulmonary embolism due to the development of pulmonary metastatic disease 9 months after the ILP procedure. Further details of grade ≥3 AEs are shown in table 3, and a full list of AEs is provided in online supplemental table S1.

Table 3. Summary of ≥grade 3 adverse events (AEs) number (%).

AEs	Grade 3	Grade 4	Grade 5
Asystole		1 (7)
Cardiac arrest			1 (7)
Sinus bradycardia	1 (7)
Pain	1 (7)
Pleural effusion	1 (7)
Bullous dermatitis	1 (7)
Hypertension	1 (7)
Hypotension	1 (7)
Other	3 (20)
Total	9	1	1

Overall and progression-free survivals for all patients are shown in figure 2A and B, respectively. A swimmer plot demonstrating individual patient trajectories over time is shown in figure 2C. Eight patients developed progressive disease (PD) and the median progression-free survival for all patients was 12 months (figure 2B). Of those who progressed, three developed locally PD, three distant metastases and two both local and distant progression.

Figure 2. Clinical outcomes of patients undergoing talimogene laherparepvec isolated limb perfusion treatment. (A) Kaplan-Meier curve of overall survival (B) Kaplan-Meier curve of progression-free survival. (C) Swimmers plot displaying individual patient trajectories over time. (D) Waterfall plot showing maximal tumor size fold change from baseline. The dotted lines represent iRECIST criteria for disease progression (>20% increase in size) and partial response (>30% decrease in size); *indicates patients who developed metastatic disease. CR denotes complete response, PR denotes partial response, SD denotes stable disease, PD denotes progressive disease. (E) Clinical photographs of two patients with complete response to treatment at 3 years compared with baseline, arrows indicate lesions before treatment and absence of lesions at 3 years. (F) Kaplan-Meier curves categorized by response to treatment showing overall survival (median survival in non-responder 30 months, not reached in responders). (G) Progression-free survival (median PFS survival in non-responder 7 months, 26 months in responders). iRECIST, immune Response Evaluation Criteria in Solid Tumours.

The overall response rate was 53%, with RECIST-defined best responses of complete response (CR) in five patients (two melanoma, three sarcoma) and partial response (PR) in three patients (two melanoma, one sarcoma) (table 4). Six patients underwent surgical resection after ILP. One patient who presented with a locally advanced primary sarcoma abutting the neurovascular bundle in the lower leg underwent an ILP as a neoadjuvant strategy, which allowed for a planned marginal resection of the sarcoma at 6 weeks after the ILP, followed by postoperative radiotherapy. This patient did not relapse in the leg after this treatment. A further five patients (three sarcoma and two melanoma) with multifocal disease underwent local excisions of either progressing or persisting tumors at varying time points after the ILP because of local symptoms. Prior to these surgical resections, two patients had achieved PR, three had stable disease (SD) and one had progressed by RECIST.

Table 4. Responders and non-responders as per RECIST criteria.

All patients	Response n=15 (%)
CR	5 (33)
PR	3 (20)
SD	5 (33)
PD	2 (13)
Responder (CR+PR)	8 (53)
Non-responder (SD+PD)	7 (47)

CR, complete response; PD, progressive disease; PFS, Progression free survival; PR, partial response; RECIST, Response Evauation Criteria in Solid Tumours; SD, stable disease.

The maximum changes in tumor size in response to treatment (excluding tumors that were surgically resected) are shown in the waterfall plot of figure 2D. Although overall response rates are compatible with those seen previously with ILP alone, interesting responses were seen, particularly in patients with sarcomas. These included CRs in patients with undifferentiated pleomorphic sarcoma and myxofibrosarcoma, subtypes that typically have poor responses to ILP (figure 2E). These responses were maintained up to 3 years after the ILP procedure. Overall (figure 2F) and progression-free survival (figure 2G) curves are presented in terms of response to treatment, with evidence of significant benefit in those with CR or PR to ILP.

T-VEC injection induces an inflammatory gene expression profile, which is more maintained in sarcoma than in melanoma

Tumor biopsies and blood samples were taken as shown in figure 1. To address the hypothesis that T-VEC injection increases the immunological heat within tumors,¹⁸ we analyzed RNA sequencing (RNA-seq) data from biopsies taken before T-VEC (“Baseline” in figure 1), immediately before ILP after T-VEC (“Pre-ILP” in figure 1), and 2 weeks after ILP (“Post-ILP” in figure 1). Analyzing all patients together demonstrated clear evidence of T-VEC inducing a transcriptional signature reflecting significantly higher levels of immune cell infiltration in the tumor immune microenvironment. T-VEC increased expression profiles consistent with cytotoxic T cells, B cells and dendritic cells (figure 3A). This positive change had largely reversed by 2 weeks post-ILP when an increase in neutrophils and fibroblasts within tumor biopsies was seen. When patients were divided into responders and non-responders (table 4), the increased inflammatory gene expression profile after T-VEC was more marked in responders (figure 3B); the changes in expression levels of some exemplar immune genes are shown in figure 3C, including CD8A and PDCD1 (the gene for programmed cell death protein-1 (PD-1)). The pattern of increased (with T-VEC), followed by decreased (after ILP), inflammatory gene expression seen in tumor tissue was also reflected systemically in the blood, on similar analysis of RNA-seq from bulk peripheral blood mononuclear cells (PBMC, figure 3D and E). While the patterns seen in blood samples generally matched the changes in tissue when all patients were analyzed together (figure 3D), there were no clear differences observed in the blood (figure 3E), when responders and non-responders were compared (in contrast to the patterns seen in tumor tissue). This suggests that the time course in the tumor compartment of increased inflammation induced by T-VEC, followed by suppression with ILP, is related to clinical response, but that this is not mirrored in the systemic immune response. Interestingly, there were also blood samples available at the later time point, 12 weeks after ILP, by which time the systemic immune cold profile seen at week 2 post-ILP, had largely reversed (figure 3E).

Figure 3. Oncolytic virotherapy and ILP induce transcriptional changes in the tumor and systemic circulation. Tumor biopsy and blood specimens were taken as illustrated in figure 1. (A) Tumor immune cell populations estimated from RNA sequencing transcriptional profiling of tumor biopsies in all patients using human microenvironment cell population counter. (B) Data from (A), subdivided into responders and non-responders. (C) Changes in expression levels of exemplar immune genes over time in responders and non-responders. (D) Transcriptomic deconvolution to demonstrate immune cell composition in peripheral blood in all patients, and (E) comparing responders and non-responders. Tumor immune cell composition over time in melanoma (F) and sarcoma (G) patients. ILP, isolated limb perfusion; NK, natural killer; T-VEC, talimogene laherparepvec.

Next, we analyzed the RNA-seq profile of patients with melanoma and sarcoma separately because the transcriptional profile of the different tumor types differed markedly (online supplemental figure S1). This analysis showed that T-VEC-induced immune changes were more prolonged (ie, between pre-ILP and post-ILP time points), in sarcoma than in melanoma (figure 3F and G). This is particularly interesting, given the general perception of sarcoma as an ICI-non-responsive tumor. The evidence that T-VEC can invoke a sustained, intratumoral inflammatory response, together with the clinical responses seen in figure 2E, is encouraging.

Responding patients with sarcoma express more interferon pathway genes at baseline than non-responders, and the number of these differentially expressed genes increases on oncolytic virus treatment

Therefore, we analyzed the sarcoma data alone in more detail, including a comparison between the four responders and five non-responders. At baseline before treatment, there were some differentially expressed genes between the two groups (figure 4A). However, pre-ILP, after two injections of T-VEC, a greater number of genes showed significantly different expression levels (figure 4B). On Gene-Set Enrichment Analysis of these differentially expressed genes, both at baseline and pre-ILP, there was clear enrichment in clinical responders of the interferon alpha (IFN-α) and gamma (IFN-γ) responses (online supplemental table S2). The activation of such pathways is consistent with an anti-viral response, and also potentially with activation of antitumor immunity.²⁵ While there was some enrichment in IFN-α/IFN-γ RNA in responders even at baseline, more IFN-related gene expression levels changed after viral injection, demonstrating an increase in the strength of this signal on treatment. A baseline IFN signature may reflect a predisposition to response to T-VEC followed by ILP, as has been reported for ICI therapy.²⁶ However, a further enhancement in this signature following treatment with T-VEC (at the pre-ILP time point) also appears to be associated with treatment response.

Figure 4. Differentially expressed genes between sarcoma responders and non-responders. (A) Volcano plot of differentially expressed genes between responders and non-responders in patients with sarcoma at baseline (A) and pre-isolated limb perfusion (B). Positive log2 fold change indicates overexpression in responding patients. Genes marked in red pass significance thresholds and the most significant are labeled. FC, fold change; NS, non significant.

T-VEC increases infiltration of CD8+ T cells into injected tumors

We also performed immunofluorescence staining, focusing on just one sarcoma responder (patient T07, figure 2E, CR at 3 years), and a sarcoma non-responder (patient T04). A 6-plex immunofluorescent panel, consisting of CD8, CD4, FoxP3, CD68, CD206 and DAPI, was deployed on both baseline and pre-ILP (directly following the final T-VEC injection) time points. As shown in figure 5, CD8+T cell infiltration (as measured by CD8⁺ cell density per mm²) was both higher at baseline, and increased after T-VEC injection, in patient T007 but not T004. While we also stained samples from other patients and for other markers (eg, M1 CD68⁺CD206⁻ and M2 CD68⁺CD206⁺ macrophage subsets), no significant differences were seen beyond CD8⁺ T cells, between sarcoma responder and non-responder groups.

Figure 5. CD8⁺ T cells higher in selected sarcoma responder versus selected sarcoma non-responder. Patient T007 (responder) and T004 (non-responder) were selected for analysis by mIF. (A) Multiplex immunofluorescence staining of formalin fixed paraffin embedded whole slides for baseline (TP1) and pre-ILP directly following talimogene laherparepvec treatment (TP2). Full slide (left panel) with a selected region (white box) for magnified view (right panel). CD8 (yellow) and DAPI (blue). (B) Quantification of CD8⁺ T-cell mIF staining. CD8⁺ T-cell infiltration is shown as normalized cell density (CD8 T-cell count per mm²). ILP, isolated limb perfusion.

T-VEC therapy with isolated limb perfusion reshapes the T-cell receptor repertoire in the tumor and blood

To probe further the immune mechanisms underlying response to treatment, we performed T-cell repertoire (TCR) sequencing at all time points in tumor and blood. The number of clones detected was, as expected, higher in the blood than in the tumor across all patients and time points (online supplemental figure S2). In terms of clonality, we found that this was higher in the tumor in responders than non-responders (all patients) after T-VEC pre-ILP. There was a similar, though non-statistically significant, trend at baseline pretreatment (figure 6A). This suggests that within the tumor, particularly after virotherapy, there may be an expansion of T-cell clones recognizing tumor and/or virus (post-therapy) antigens that is more evident in responders over non-responders. In contrast, the clonality was significantly lower in the blood of responders after T-VEC with, again, a non-significant trend pretreatment. This higher diversity of the peripheral T-cell repertoire in responders may reflect a more widespread systemic activation of the immune response following exposure to the virus. The data in figure 6A shows that responding patients appear to demonstrate an on-treatment, T-VEC-induced increased T-cell antigenic focus within the tumor, alongside a more diverse clonal population in the blood; within these circulating T cells may lie antitumor T cells primed to expand and infiltrate the tumor for therapy, particularly on further intervention with ICI, as we have demonstrated in a preclinical model of ILP.²³ At later time points, 2 weeks post-ILP (in tissue and blood) and 12 weeks post-ILP (blood only), although there were fewer samples, it is noteworthy that the clonality differences between responders and non-responders were no longer apparent. This is consistent with the reversion towards baseline characteristics over time shown in the blood transcriptomic data of figure 3D and E.

Figure 6. TCR sequencing showing changes in tumor microenvironment and systemic antitumor immunity in responders compared with non-responders. (A) TCR sequencing—clonality (1-normalized Shannon Index) of the peripheral and systemic TCR repertoire across time points in responders and non-responders. (B) Triplet CDR3 amino acid kernel clustering of the intratumoral TCR repertoire. (C) Quantification of cluster sizes from (B); green responders, pink non-responders. ILP, isolated limb perfusion; TCR, T-cell repertoire.

Subsequently, we explored clustering of the T-cell clones detected in the tumor. This analysis shows the closeness of the relationship between T-cell clones within the tumor which, in turn, reflects how likely they are to be recognizing a shared antigen. While sufficient data for this analysis was only available for a small number of patients—three non-responders (one melanoma and two sarcoma), and two responders (one melanoma and one sarcoma)—we found that there was a significant increase in cluster size between baseline and pre-ILP time points in responders, but not in non-responders (figure 6B and C). An increase in cluster size in non-responders between baseline and 2 weeks post-ILP was seen, which potentially reflects an expansion of anti-viral T-cell clones. Nevertheless, the significant increase in tumor TCR cluster size that was seen only in responding patients in samples obtained before and after T-VEC treatment, is consistent with an expansion of related T-cell clones within the tumor recognizing potentially functionally relevant antigens, consistent with the clonality data presented in figure 6A.

Peripheral blood immune cell profiling shows activation of a systemic immune response following localized treatment with T-VEC and ILP

To explore further the systemic immune response on treatment, we analyzed the phenotype of circulating immune cells. Routine clinical full blood count parameters showed a significant decrease in the neutrophil count and neutrophil-to-lymphocyte ratio across all patients, comparing pre-ILP with 2-week post-ILP (figure 7A). This is likely due to the recognized immune-suppressive effects of melphalan.^{1 27} No differences were seen between responders and non-responders in the entire cohort or in the sarcoma-only group. On Fluorescence -activated cell sorting (FACS) analysis of blood samples, we analyzed a range of surface and intracellular protein markers (online supplemental table S4) within PBMCs, including CD8, CD4, CD69, PD-1, CD25, IFN-γ, CD107a and IL-2, and compared responders and non-responders in all patients and separately in sarcoma patients. The majority of parameters were not significantly different, but some changes were seen in expression of the early activation marker CD69 and in PD-1. In all responding patients, the CD69 expression on conventional CD4+ (but not CD8+) T cells was higher 2 weeks post-ILP while, in sarcoma responders, CD69 was also significantly upregulated relative to non-responders on regulatory T cells, unconventional (CD3+CD4−/CD8−) T cells and natural killer cells at this time point (figure 7B). With regard to PD-1 expression, a notable finding was a trend towards increased expression on conventional CD4+cells (though not in CD8+cells), in responders at 12 weeks post-ILP (figure 7C). While the numbers of patients in the study is too few to draw definitive conclusions, this data shows that the lack of difference between responders and non-responders in the transcriptomic signature of activation in bulk PBMC (figure 3E), within which specifically CD69 expression was not different, may not detect potentially significant differences in activation markers such as CD69 measured by FACS at the protein level (figure 7B). Moreover, increased circulating PD-1+ conventional CD4+ cells, appearing relatively late in this analysis (3 months post-treatment), may signal activation of a memory effector population related to clinical antitumor response, as has previously been described.28,30 No differences were seen between all, or sarcoma-only, responders versus non-responders.

Figure 7. Immune cell activation. (A) Full blood count parameters over time in all patients. Peripheral blood mononuclear cells were surface-stained with anti-CD45, CD3, CD19, CD4, CD8, CD56, CD25, CD127, CD69 and PD-1 antibodies for 30 min at 4°C and analyzed by flow cytometry. Graphs show fold-change versus baseline in percent CD69 (B) and PD-1 (C)-positive cells in the indicated population. P values for responders versus non-responders patients shown for the indicated time point by two-way analysis of variance followed by Šídák multiple comparison test. ILP, isolated limb perfusion; PD-1, programmed cell death protein-1; R, Responder; NR, Non-responder ; WBC, white blood cell.

Discussion

We have conducted a first-in-human phase I/II clinical trial combining T-VEC OV with ILP, with associated translational studies. Combining these two separate oncological approaches was well tolerated and did not have a higher incidence of AEs than would be seen by delivering these treatments independently. A CR rate of 33% and a PR rate of 20%, giving an overall response rate of 53%, is comparable to previously reported historical series of ILP treatment alone.²⁴ However, in this study, 50% of the patients with melanoma had been previously treated with systemic immunotherapy, which is known to be associated with lower CR rates after ILP.³¹ Furthermore, 44% of patients with sarcoma had sarcoma subtypes known to be more resistant to ILP³² (myxofibrosarcoma, undifferentiated sarcoma). However, two such patients demonstrated sustained CRs (figure 2G), suggesting that the addition of T-VEC to ILP enhanced the clinical response. Analysis by immunofluorescence of the local tumor microenvironment in repeated tumor biopsy specimens, taken at baseline and pre-ILP after two T-VEC injections, indicates that these remarkable clinical responses may be immune-mediated, as reflected by an increased infiltration of CD8+T cells after injection of virus (figure 5).

Although increased local response rates are of clinical value, we wished to understand whether the combination of an oncolytic virus with local ILP could evoke significant local immune effects in immunologically inert tumor subtypes which could potentially translate into a systemic immune response that could prevent the development of metastatic disease. Data from our preclinical studies supported our hypothesis, since we have shown that oncolytic vaccinia virus delivered by ILP alongside TNF and melphalan with checkpoint inhibitor treatment prevented the development of metastases in a highly metastatic preclinical rat sarcoma model.²³ By necessity, this trial differed from that animal model in two aspects. First, intraperfusional delivery of virus has been shown to increase the concentration of virus within the substance of large bulky sarcomas while, at the same time, preventing both systemic sequestration of virus in the reticuloendothelial system and systemic side effects.²¹ However, the licensed indication for T-VEC for intratumoral injection mandated delivery only by this route, rather than via the perfusion circuit. Therefore, it is likely that the delivery to the tumor in our patients was less than would have been achieved had the virus been delivered to the tumor via the afferent artery during ILP. Second, data from our animal sarcoma model indicated that oncolytic vaccinia virotherapy, when combined with melphalan and TNF, was highly effective locally but, in the absence of checkpoint blockade, was unable to prevent the development of metastases.²² It was only with the addition of a checkpoint inhibitor that a systemic therapeutic effect was seen.²³ At the time of development of the current study protocol, that preclinical rodent data was not available. In addition, as a first step, clinically testing the addition of single-agent T-VEC to ILP was appropriate prior to consideration of further combinations incorporating ICIs. We plan to progress this work to studies that infuse the virus into the ILP circuit to increase tumor infection and co-treatment with an ICI in patients with locoregional and metastatic disease, to directly test the potential of activating systemic immunity against metastases.

From the translational perspective, the repeat tumor biopsies and blood samples taken in this study allowed us to explore the immunogenomic response to treatment, including differences between responders and non-responders, although in small numbers and over two tumor histotypes. Transcriptomic analysis, including deconvolution of RNA-seq, confirmed the inflammatory effects of virus injection within the tumor that were reported previously for T-VEC in advanced melanoma¹⁸ (figure 3A). This reaction had abated at 2 weeks post-ILP, potentially due to the suppressive effects of the ILP and/or the time course of viral infection followed by clearance; by this time, the tumor was more infiltrated by neutrophils and fibroblasts, consistent with a wound-healing response. Intriguingly, when responders and non-responders from the entire cohort were compared, the transcriptomic signature of recruitment of effector immune cells into the tumor by T-VEC was greater in responders (figure 3B). It is worth noting that RNA-seq did not show any Herpes simplex virus (HSV) transcripts at any of the time points of tumor biopsy. This observation is consistent with our previous experience with oncolytic reovirus treatment in patients³³ and, most likely, reflects effective viral clearance during the period between T-VEC injection(s) and the timing of the biopsy (2–3 weeks for pre-ILP and 4–5 weeks for post-ILP biopsies). RNA-seq analysis of bulk PBMC also showed increased expression of activation markers after virus administration, which decreased after ILP (figure 3D). In this case, there were no differences in readouts between responders and non-responders. Moreover, on FACS analysis of surface and intracellular protein markers, few significant changes were seen. In responding patients, CD69 expression was higher on some T-cell subsets in responders than non-responders (figure 7B), while there was a trend towards increased expression of PD-1 on conventional CD4+cells 3 months after treatment, which may reflect activation of a significant memory effector population. Taken together, our data show signals of systemic immune activation after localized treatment with ILP and intratumoral T-VEC.

An important indication of activation of a relevant immune response within tumor on T-VEC treatment was our finding that the clonality of the TCR repertoire was higher in responders than non-responders (figure 6A). A similar trend was seen even before treatment, suggesting that responders’ tumors have antitumor T cells which are poised to expand on virus treatment. This finding is consistent with the previously described significance of the TCR repertoire to clinical outcome and response to ICI.³⁴ This indication that the immune status of the tumor before treatment impacts subsequent response, is further supported by our finding of differential gene expression at baseline between sarcomas that did and did not respond to treatment (figure 4). In addition, these differences were amplified after virus treatment in responding patients, with a particular increase in activation of IFN-α/IFN-γ pathways (supp table 2). In contrast to the higher clonality seen in the tumors of responding patients, the lower clonality seen in their blood after T-VEC (figure 6A) suggests an increase in the diversity of the peripheral T-cell repertoire on treatment. Given the limited numbers of samples available in our analyses, a fuller understanding of the changes in TCR repertoire, and how they relate to response, requires more in-depth analysis, preferably at the single-cell level, to allow tracking of individual clones, their functionality, and whether they are shared between the tumor and blood.

It is important to recognize that the changes seen in the TCR repertoire in the current study are most apparent at early time points (figure 6A and C). The dynamic nature of the changing T-cell repertoire after T-VEC, is further illustrated by the analysis of the clustering of TCR clones shown in figure 6B and C. While the numbers of patients is small, the finding that the average intratumoral T-cell cluster size before and after virotherapy increases in responders, but not in non-responders, supports a central role for the T-cell clonal response triggered by T-VEC in therapy. Moving forward, understanding the function and reactivity of these clones which signal potential benefit, in terms of their antigen targets (tumor and/or viral antigens), will be critical in elucidating the mechanisms responsible for therapy using oncolytic viruses, ILP and potentially ICI as an additional intervention which has shown preclinical promise.²³

In conclusion, we have demonstrated that OV can be safely delivered in combination with melphalan and TNF-α via ILP. This combination appears to show favorable responses, particularly in some sarcoma subtypes that are less responsive to ILP alone. Analyses of longitudinal biopsy and blood samples show evidence of immune activation within the tumor after T-VEC treatment, and also changes in the PBMC phenotype/transcriptome and T-cell repertoire in the blood, consistent with potential systemic immune activation in response to local combination therapy. This preliminary data strongly supports a next stage of investigation of combination OV therapy with ILP using an initial immune priming intratumoral viral injection followed by a high-dose intraperfusional viral delivery coupled with additional ICI. We are planning for such a study to be performed in patients with locally advanced sarcoma who also have low volume metastatic disease, allowing for clinical assessment of potential systemic responses and local pathological response rates in this group of patients with very limited treatment options. Finally, the translational findings from the current study encourage more in-depth analysis in future trials of the systemic as well as local immune consequences of localized treatment, to increase our mechanistic understanding of how OV can lead to maximal antitumor immune activation.

Materials and methods

Trial design and study population and safety analysis

The TITAN trial (NCT03555032) was an open-label, single-arm, single-center safety and phase II expansion trial of T-VEC in combination with ILP in extremity melanoma and sarcoma. The primary endpoint was safety and secondary endpoints included clinical efficacy and exploratory translational readouts. Because the trial design was a combination study using dosing schedules of approved agents, there was no dose escalation within the study. However, there was an initial review of AEs and dose-limiting toxicities, by the Safety Review Committee (SRC), 30 days following treatment completion in three patients, which confirmed the safety of the study to continue. Thereafter, the trial was open to recruitment of an expansion phase II cohort of 9 patients, to a total of 15 patients, with any individual dose-limiting toxicity (DLT) being reported to the SRC. Full informed consent was obtained from all patients participating in the study. DLT was defined as either systemic or local. Systemic DLT was defined as any of: severe neutropenia (<0.5×10⁹/L) lasting greater than 5 days or combined with sepsis; severe thrombocytopenia (<25×10⁹/L); or any other drug-related non-hematological grade 3 or 4 toxicity, with the exception of influenza-like symptoms, nausea and vomiting. Local DLT was defined using the Wieberdink criteria³⁵ for local toxicity in regional chemotherapy, with a local DLT classified as either a grade 4 toxicity (extensive epidermolysis; definite functional impairment; threatening or manifest compartment syndrome) or grade 5 toxicity (a local reaction requiring amputation).

Safety analysis

The primary endpoint was the incidence of DLT within 12 months of the ILP. Toxicities were graded and tabulated using the National Cancer Institute Common Terminology Criteria for Adverse Events (V.4.0). Inclusion criteria included males or females aged 18 years with a confirmed histological diagnosis of in-transit malignant melanoma with or without regional lymph node metastases or limited visceral metastatic disease (AJCC Stage IIIb/c and IVa/b), or locally advanced soft-tissue sarcoma with or without regional or distant metastases (T2a/b, N0/1, M0/1) not suitable for surgical resection by standard limb-conserving oncological surgery. Exclusion criteria included cerebral metastases, concurrent immunotherapy during, and for the number of days equal to the half-life of that agent before or during the study therapy, immunosuppression for any reason, open herpetic skin lesions or a history of hypersensitivity to T-VEC or its excipients.

T-VEC injections and isolated limb perfusion

The T-VEC dosing schedule was a “seroconversion” dose of up to 4 mL of 1×10⁶ PFU/mL of T-VEC delivered by direct intratumoral injection at week −4 to −6 (week 0 defined as the date of ILP), a treatment dose of up to 4 mL of 1×10⁸ PFU/mL of T-VEC delivered by direct intratumoral injection at week −2 to −3, and then a final treatment (third) dose of up to 4 mL of 1×10⁸ PFU/mL direct intratumoral injection, delivered under general anesthetic immediately before the ILP was commenced. The exact volume injected into each tumor was dependent on the size of the lesion as previously described.¹⁷ ILP was performed as has been described previously^{3 24} and surgical details are available in online supplemental materials.

Response assessment

Disease within the affected limb was assessed by direct ruler measurement and MRI. Distant disease was assessed by CT. Tumor biopsies were taken at baseline under local anesthesia prior to the first T-VEC injection (week −4 to −6), at week 0 in the operating room under general anesthesia prior to the third T-VEC injection, and at week 2 after ILP under local anesthesia. Peripheral blood samples for translational analyses were taken at these time points, and also at week 8. The secondary endpoint was to determine efficacy in accordance with RECIST V.1.1³⁶ criteria which was assessed at 4, 12, 24 and 52 weeks after ILP. Good responders were those with a CR+PR and non-responders were those with SD+PD. If a patient underwent a palliative surgical resection during follow-up for progressive symptomatic disease which resulted in a CR/PR by RECIST V.1.1, this would be classified as PD and the patient categorized as a non-responder for the subsequent translational analyses. Exploratory analysis included immunological and molecular correlates of response to therapy. Simple and descriptive statistical analysis was performed using Microsoft Excel (Microsoft, Washington, USA). Kaplan-Meier survival analysis and group analysis, including analysis of variance, was performed using GraphPad Prism V.9 software (GraphPad Software, California, USA). The test used is indicated in the legend of each figure. Without mention, differences are not statistically significant. Statistical significance was defined as a p value of <0.05.

Fluorescent immunohistochemistry: available in online supplemental materials

RNA sequencing: available in online supplemental materials

Peripheral blood immune profiling: available in online supplemental file materials

T cell receptor sequencing: available in online supplemental materials

Data availability statement

Data are available upon reasonable request.