Tumor-Localized Interleukin-2 and Interleukin-12 Combine with Radiation Therapy to Safely Potentiate Regression of Advanced Malignant Melanoma in Pet Dogs

The recent success of immune checkpoint inhibitors has ushered in a new era to treat advanced cancers through rational engagement of the immune system (1–3). Remarkable objective responses have been observed in primary tumors across a multitude of cancer immunotherapy strategies, although achievement of objective responses at metastatic sites remains an elusive clinical outcome for the majority of patients (4–6). As such, combinations of checkpoint inhibitors with immune agonists have been explored to enhance systemic antitumor responses by overcoming immune-suppressive barriers operative at these metastatic sites (7, 8). In particular, the cytokines IL2 and IL12 have garnered significant interest owing to their ability to proliferate, activate, and differentiate critical effector immune cell populations unleashed by checkpoint inhibitors (9, 10). Encouraging synergy has been observed with these IL/checkpoint inhibitor combinations in early clinical trials, although adverse side effects have been encountered in patients (11–13). As key signaling molecules between immune cells, endogenous immune-stimulating cytokines like IL2 and IL12 exhibit tightly controlled spatial distributions and diffusional kinetics to prevent aberrant and pathologic activation. However, in the therapeutic setting, systemically dosed cytokines can elicit on-target, off-tumor activation of immune cells and subsequently possess an extremely narrow therapeutic window constrained by dose-limiting toxicities (14–16). These clinical limitations resulting from systemically administered cytokines have driven the recent interest in protein engineering strategies to mitigate systemic toxicities through tumor-targeting immunocytokines (17–21), conditionally active/masked cytokines (22–26), and receptor-biased cytokine agonists (27–30) to enable their inclusivity alongside checkpoint inhibitors and other first-line cancer treatments such as radiation, chemotherapy, and surgery.

These elegant protein engineering efforts converge on the same objective for cytokine therapies that promote their accumulation within the tumor and constrain their signaling to the immediate tumor microenvironment. With advances in image-guided injection techniques, intratumoral dosing of therapies is now possible for the majority of solid tumor indications. As such, we and others have begun to explore strategies to physically retain cytokines like IL2 and IL12 within the tumor microenvironment after intratumoral injection through binding to co-dosed biomaterials (31–34) or extracellular matrix components like collagen (25, 35–37). These approaches minimize the systemic biodistribution, tumor accumulation, and toxicity challenges associated with systemic dosing of engineered cytokines and have led to marked improvements in both safety and efficacy profiles versus non-retained cytokines in mouse tumor models (31, 35, 36). However, mouse syngeneic transplant tumor models lack the long-term immune selection pressures that sculpt human tumor genetics, and thus they incompletely recapitulate critical evolutionary features of the complex human tumor microenvironment (38, 39). As a result, the achievement of treatment efficacy in mouse preclinical models with investigational immunotherapies is not sufficient for predicting their success when translated to human clinical trials (40–42). For this reason, naturally occurring tumors in larger companion animals complement these conventional model systems by illuminating the nuanced and complex tumor–immune interactions otherwise undetectable in mouse tumors, aiding translational investigation of novel anticancer strategies.

Here, we build upon our prior work in murine tumor models by examining the safety and efficacy of intratumorally delivered, collagen-retained IL2 and IL12 cytokines in advanced malignant melanomas that spontaneously develop in outbred pet dogs. Dogs develop cancer at similar rates to humans and yet are an underutilized model to bridge the gaps between mouse and human studies of novel immunotherapies or treatment combinations (43–45). Canine tumors feature many of the same biological immune escape mechanisms and intratumor genetic heterogeneity that define human cancers, while also possessing more human-relevant body characteristics that enable the prediction of drug biodistribution and pharmacokinetics/pharmacodynamics (46–49). Canine malignant melanomas typically feature mutational burden at >1 mutation/Mb, with mutational enrichment across the p53 and RAS signaling pathways (50). Moreover, a significant fraction of pet dogs with cancer presents with metastatic disease, enabling the evaluation of locoregional response to intratumoral therapy, which has been far more difficult to model and test in murine tumors or genetically engineered mouse models. We previously evaluated the safety and mechanism of action of an intratumoral collagen-binding cytokine approach in canine soft-tissue sarcomas but did not have the opportunity to investigate long-term antitumor responses due to the medical ethical obligation to resect such tumors shortly after treatment (51). Guided by palliative regimens for malignant melanoma using hypofractionated radiotherapy (RT), we here report our studies of the safety and efficacy of a single RT dose with repeat dosing of tumor-localized IL2 and IL12 cytokines against malignant melanoma, a canine cancer that metastasizes in more than 70% of cases (52). Through a dose-escalation trial inclusive of key immunobiological endpoints, we observed provocative activity engendered at both primary and metastatic tumors in a defined cohort of pet dogs. Profiling of canine patients that progress after therapy inform hypotheses regarding new therapeutic combinations predicted to improve tumor response rates, and we intend to deploy these strategies in both mouse models and pet dogs with naturally occurring cancers. Collectively, these efforts underscore the potential utility of comparative oncology inclusive of canine tumors to build, test, and optimize treatment regimens prior to commencing human clinical studies.

This study complies with all relevant ethical norms and principles. This research study protocol was approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Illinois Urbana-Champaign (UIUC), and this study was conducted in accordance with the ARRIVE guidelines.

Client-owned pet dogs with cytologically or histologically confirmed oral malignant melanoma (OMM) were included in the study (Table 1). Eligibility criteria required the dogs to have (i) primary tumor measure between 0.5 and 7.5 cm in diameter, (ii) adequate organ function determined by laboratory evaluations (complete blood count, serum biochemical profile, and urinalysis), and (iii) a minimum 3-week washout period for RT, systemic chemotherapy, or any additional immunosuppressive/homeopathic/alternative therapy. No exclusion criteria for tumor stage or metastatic burden, age, weight, sex, or neuter status were applied for this trial. Tumor staging at enrollment was determined on the basis of the World Health Organization (WHO) staging scheme for dogs with oral melanoma (Supplementary Table S1; ref. 53). All patient owners provided written informed consent before enrollment, and all procedures were performed in accordance with the study protocol approved by the UIUC IACUC.

Canine cytokines (cLAIR-CSA-cIL2 and cIL12-CSA-cLAIR) were cloned and recombinantly expressed as previously described (51). Briefly, HEK293-F (Gibco; RRID: CVCL_D603) stable cell lines for each cytokine were prepared through cloning into the expression cassette of PiggyBac (System Biosciences) transposon vector, followed by dual transfection of the transposon vector and the Super PiggyBac transposase plasmid. Stable integration was confirmed after sorting enhanced green fluorescent protein–positive cells 3 to 4 days after transfection (BD FACSAria), and transformed HEK293-F cell lines were confirmed to be mycoplasma free (IDEXX RADIL). Protein was produced from IL2- and IL12-expressing stable lines during 1-week culture in serum-free media (FreeStyle 293, Invitrogen) and purified with HisPur Ni-NTA affinity resin (Thermo Fisher Scientific). Protein was analyzed by using size exclusion chromatography (Superdex 200 Increase 10/300 GL column, Cytiva Life Sciences, on ÄKTA FPLC system) for size and aggregation and validated to meet low endotoxin levels (<5 EU/kg) by using the Endosafe nexgen-PTS system (Charles River Laboratories). Activity of cytokines was confirmed through CTLL2 and HEK Blue IL12 activation assays, while collagen-binding was confirmed through ELISA. Both CTLL2 (ATCC, Cat# TIB-214, RRID: CVCL_0227) and HEK-Blue IL12 (InvivoGen; RRID: CVCL_UF31) cells were cultured according to the vendor's instructions in 5% CO₂ at 37°C. These cells also tested negative for mycoplasma. Aliquots of cytokines were snap-frozen in liquid nitrogen and thawed immediately prior to dilution in sterile saline for dosing intratumorally to dogs.

Fifteen eligible dogs were enrolled into a modified Fibonacci 3 + 3 dose escalation trial design of four different cohorts. The trial consisted of a regimen involving treatment with a single 9-Gy dose of RT followed by six doses of cLAIR-CSA-cIL2 (IL2) and cLAIR-CSA-cIL12 (IL12) every 2 weeks. Radiation was delivered using a Varian TrueBeam linear accelerator with 6 MV photons at standard dose rate of 6 Gy/minute (Varian Medical Systems, Palo Alto, CA). Depending on the location and proximate organs at risk, the dose was delivered either using manual calculations for parallel opposed portals or with three-dimensional conformal radiation plan using CT guidance and a treatment planning system (Varian Eclipse v.15). The dose was calculated to the central axis for parallel opposed portals and with the goal of 100% of dose to 95% of the planning target volume (gross tumor volume plus a 3 to 5 mm expansion) for computer plans. The initial doses of IL2 and IL12 cytokines used in this work were determined from prior evaluation in both healthy beagles and pet dogs with soft-tissue sarcomas (51). The reported values for single-agent maximum tolerated doses (MTD) of IL2 in humans (54–56) and IL12 immunocytokine in dogs (57) were used to select the maximum cytokine dose in our prior work, with the “1×” dose level described here representing 0.1× the MTD (Supplementary Table S2). Doses of cLAIR-CSA-cIL2 (17.4 μg/kg at “1×” dose level) and cIL12-CSA-cLAIR (2.08 µg/kg at “1× dose level”) were prepared from frozen protein aliquots and combined in a total volume not exceeding 0.5 mL in sterile saline. A 29-gauge, 0.5-inch insulin syringe was used to slowly inject the full dose volume via a single insertion point using a fanning pattern into the tumor. No additional measures were used to avoid any internal necrotic areas within the tumor. Radiation therapy was performed using the Varian TrueBeam system. Adverse events were classified and graded in accordance with the Veterinary Cooperative Oncology Group’s Common Terminology Criteria for Adverse Events (v2; ref. 58). An additional cohort of three dogs was recruited to examine the response to IL2/IL12 cytokine-only dosing. Three other dogs undergoing RT for their oral melanoma consented to blood draws following administration of radiation to examine circulating cytokine/chemokine biomarkers as the control cohort. These patients undergoing RT-only were not followed for survival analyses.

Clinical and vital evaluations were conducted on all patients at baseline and preceding each treatment administration at the UIUC Veterinary Teaching Hospital. In addition, after intratumoral cytokine administration, a 48-hour monitoring period was initiated to assess the presence of any toxicity-related symptoms, coupled with blood sampling for complete blood count, serum biochemical profiling, and urinalysis. In addition, after each treatment, blood draws by jugular venipuncture were performed for cytokine/chemokine analysis before treatment and at 2, 4, 8, 24, and 48 hours after treatment. Patients were followed up until their death or removal from the trial.

Clinical and caliper measurements of the maximum tumor and lymph node dimensions were conducted by board-certified veterinary oncologists, and measurements were documented in millimeters during each examination. In addition, primary tumor or metastatic lesions were assessed by CT (SOMATOM Definition AS, Siemens Healthineers) at pretreatment, day 28, day 70, and day 84 (2 weeks after the last treatment). The tumor size and percentage of change were determined on the basis of CT measurements. Because determination of the longest dimension is challenging with these frequently irregularly marginated tumors, the tumor volume was used to measure the response to therapy. Volume was determined using the RT treatment planning software (Eclipse v15, Varian, Palo Alto, CA) by importing CT scan images (1.5-mm slices) before and after treatment. The gross tumor volume was delineated on the basis of distortion of normal tissues by the mass effect combined with changes in Hounsfield units, which reflect contrast enhancement due to changes in electron density. The software yields a three-dimensional volume based on the contours that are created. Standard criteria for volumetric assessment of tumor response were used. Furthermore, the assessment of tumor response was carried out during each visit and was determined in accordance with the guidelines established by the RECIST in Dogs (v1.0; Veterinary Cooperative Oncology Group; ref. 59). These guidelines similarly reflect the human RECIST v1.1 response criteria in the assessment of canine tumors: complete response (resolution of all target lesions and no new lesions), partial remission (at least 30% reduction in target lesions and no new lesions), stable disease (SD; no more than 20% increase in target lesions or a decrease in target lesions less than 30% and no new lesions), or progressive disease (PD; >20% increase in target lesions and new lesions). Patients presenting with SD or PD were allowed to remain in the study under the condition that no adverse events were observed or if such events could be mitigated through the implementation of a dose reduction protocol. Clients had the option to remove their dogs from the study if their pets’ conditions worsened, if they showed signs of declining health, or if the treatment caused unbearable side effects. The decision could be made by the investigator, the dog’s owner, or both.

Serum samples collected from patients following treatment were examined for concentrations of 13 cytokine and chemokine analytes, including GM-CSF, IFNγ, IL2, IL6, IL7, IL8/CXCL8, IL10, IL15, IL18, IP10/ CXCL10, KC-like, MCP1/CCL2, and TNFα (MILLIPLEX Canine Magnetic Bead Panel, Millipore Sigma) at Eve Technologies (Calgary, AB, Canada). Individual analyte concentrations were determined from panel standard curves for each cytokine or chemokine. Time course analysis of patient response to IL2/IL12 and RT was performed by determining the log₁₀ fold change of analyte concentrations relative to their pretreatment levels. IFNγ and IL10 serum concentrations following treatment with collagen-anchored cytokines and RT were further measured using the Canine IFNγ Quantikine ELISA Kit (R&D Systems) and the Canine IL10 Quantikine ELISA Kit (R&D Systems) according to the manufacturer’s instructions.

RNA was isolated from 10-μm formalin-fixed, paraffin-embedded (FFPE) samples from resected canine primary melanoma tumor or metastatic tumor lesions using an RNEasy FFPE Kit and deparaffinization solution (QIAGEN). Isolated RNA was examined by using Bioanalyzer (Agilent) for the assessment of fragment size prior to hybridization with nCounter probe sets (NanoString). Canine RNA samples were hybridized with the Canine IO nCounter Panel code set for 22 hours at 65°C per the manufacturer’s instructions. Following hybridization, the samples were loaded into the analysis cartridge and scanned at maximum resolution using NanoString Prep Station and Digital Analyzer.

Canine RCC count files were normalized using nSolver software (NanoString) after background thresholding using the mean of eight negative control probes and batch correction against a standard control panel. Normalized gene counts were processed using the nSolver Advanced Analysis module for differential expression and pathway enrichment analysis. P value adjustment was performed using the Benjamini–Hochberg method to estimate FDRs of differentially expressed genes.

Relative abundance of the tumor-infiltrating cell fraction was estimated from bulk NanoString profiling data by employing CIBERSORTx (60) algorithm using a validated leukocyte gene signature matrix (LM22). Bulk NanoString profiling data were assessed in relative mode, with 100 permutation runs and without quantile normalization.

Canine advanced malignant melanoma tumors were resected at specific indicated timepoints. Following resection, tumor tissues were fixed in 10% formalin and subjected to a paraffin processing and embedding protocol. IHC was used to determine the presence of inflammatory cells, specifically positive for CD3 (T lymphocyte; Biocare, CP215C), Iba-1 (macrophage; Biocare, Cat# CP 290 B, RRID: AB_10583150), and Melan-A (melanoma-specific antigen; Biocare, A103) for melanoma cells. All samples were histologically evaluated and classified by a single board-certified veterinary pathologist. Tumor tissues were classified based on CD3 T-cell infiltration status into an immune phenotype, defined as (i) inflamed—highly infiltrated by CD3⁺ T cells, (ii) immune desert/cold—devoid of CD3⁺ T cells, and (iii) immune excluded—bordered yet not infiltrated by CD3⁺ T cells (61). Cytology of tumors was performed at specified timepoints. Cellular specimens were collected using a 22-gauge needle attached to a 5-mL syringe. Following aspiration, the samples were smeared onto a glass slide for subsequent cytochemical staining. Cytology slides were then evaluated by a board-certified veterinary pathologist. IHC staining and cytology samples were assessed on an Olympus BX45 microscope using a high-power 10× microscope objective. Digital images were captured with an Olympus DP28 digital camera and processed using Olympus cellSens Imaging Software (v4.2).

Statistical analyses were conducted using Prism v10 (GraphPad, RRID: SCR_002798). Power calculations were not conducted to predetermine sample size. The details of statistical analysis have been provided in the descriptions for Figs. 2 and 4.

The data generated in this study are available within the article and its supplementary files (Supplementary Data S1), and the raw data generated are available from the corresponding author upon reasonable request. NanoString expression data for canine tumor expression in dogs progressing after completion of RT with the IL2 and IL12 therapy has been made publicly available in Gene Expression Omnibus (RRID:SCR_005012) at GSE253243.

For this study, clients whose dogs met the trial inclusion criteria provided written informed consent before enrollment, and all procedures were performed in accordance with the study protocol approved by the UIUC IACUC. The dogs were eligible after histologic or cytologic confirmation of OMM (n = 14) or malignant melanoma involving other facial structures (n = 1) and if their primary tumor was between 0.5 and 7.5 cm in diameter. Eligible dogs were also required to have adequate organ function as measured by standard laboratory tests and have had a minimum 3-week washout period if they had been recently treated with RT, systemic chemotherapy, immunotherapy, or any additional homeopathic/alternative therapy. There were no exclusion criteria for tumor stage or metastatic burden, age, weight, sex, breed, or neuter status for this study. Dogs were sequentially enrolled into a modified Fibonacci 3 + 3 dose escalation trial design, with the initial IL2 and IL12 cytokine dose chosen from prior identification of an MTD after evaluation in both healthy beagles and pet dogs with soft-tissue sarcomas (Table 1; ref. 51). The low dose level (“1×”) tested here represented one-tenth the MTD identified in a prior work (51), which was informed by a reported IL12 immunocytokine MTD in dogs (57) and allometric scaling of single-agent IL2 doses in humans (Supplementary Table S2; refs. 54–56). In total, 15 dogs with median age 11 years (min: 4, max: 16) were enrolled into the trial, with 10/15 (66%) dogs presenting with tumors of World Health Organization stage III or greater, indicating metastatic disease at lymph nodes or lung tissue sites (Supplementary Table S1).

The primary objective of this study was to examine the antitumor efficacy potentiated by the combination of intratumoral collagen-anchored IL2 and IL12 with a single dose of RT. As current veterinary practice patterns favor the use of hypofractionated RT protocols using 8 to 10 Gy fraction size for OMM (62–64), dogs treated in this study were provided a single RT dose of 9 Gy to stimulate tumor cell death and antigen generation. Local and regional lymph nodes were not irradiated, regardless of appearance or suspicion of possible metastatic disease. Dogs then received six doses of intratumoral collagen-anchored cytokines at the same 2-week cadence similar to an existing FDA-approved intratumoral immune strategy [e.g., Talimogene laherparepvec (T-VEC); Fig. 1A]. Pursuit of consecutive additional RT doses was not instituted because of concerns for detrimental lymphodepletion within the tumor and draining lymph node following preliminary experiments in the murine B16F10 model and other reports (Supplementary Fig. S1A–G; refs. 65–67). Moreover, the subsequent dosing of intratumoral cytokine alone enabled attribution of patient symptoms uniquely to cytokine treatment and bypassed the requirement to deconvolute individual or interactive toxicities generated by continuous combinatorial therapy of RT with IL2 and IL12. All dogs were monitored for 48 hours after intratumoral cytokine dosing for symptoms of toxicity and had periodic blood draws performed for cellular and chemistry analyses.

Primary tumor volumes at the time of the first intratumoral dose had a median volume of 7.5 cm³ (min: 0.5, max: 43.4), although the highest dose cohort (“5×”) included a dog with a primary tumor volume near the upper end of our eligibility criteria (Fig. 1B). Responses to therapy were evaluated through comparative and serial assessments of CT scans of primary tumor and associated regional metastatic lymph nodes identified at the baseline (pretreatment) with subsequent CT scans performed on day 28 and day 84. Rapid primary tumor volume reduction occurred in 13/15 (86.7%) malignant melanomas at day 28 scan after just two doses of cytokine therapy and a single RT dose (Fig. 1C). At day 84 CT scan, performed 2 weeks after the final (sixth) dose of intratumoral cytokine treatment, primary tumor responses were found to be stable or have further improved in 10/13 (76.9%) surviving dogs (Fig. 1D). Two patients were euthanized before the day 84 CT tumor measurement because of progression of their primary and/or metastatic tumor sites. These tissues were collected for additional analysis, which is detailed later in this study.

Treated pet dogs were followed after the 12-week treatment period to monitor the durability of their responses and assess overall survival. As of the time of writing (January 2024), the median survival regardless of tumor stage is 256 days, with three dogs still alive the past 2 years (Fig. 1E; Supplementary Fig. S2A and S2B). This is in contrast to the reported median survival of 65 days for dogs with untreated oral melanoma (68) and 147 days for OMM dogs treated with 9 Gy × 4 RT (63). Two dogs were euthanized because of unrelated issues (age/quality of life; development of sinonasal chondrosarcoma) nearly a year after completing treatment. Interestingly, there appeared to be no correlation between the cytokine dose level and overall survival (Supplementary Fig. S3A–D). Of the dogs alive nearly 1,000 days after treatment, the local response to therapy was rapid and robust, with less treatment morbidity than curative-intent surgical removal of OMM (Fig. 1F and G). Overall, the objective responses observed in these canine patients with advanced stage and heterogeneous primary tumors were favorable, and further corroborate and extend upon the documented anticancer activities demonstrated in mouse models treated with the same collagen-binding cytokine approach (35, 37).

The clinical promise of IL2 and IL12 cytokines has been limited by the toxicities observed at therapeutically effective doses (9, 10, 15, 16, 69, 70). As such, evaluating if the collagen-anchoring approach would ameliorate cytokine-driven toxicities at doses capable of promoting antitumor responses in pet dogs was paramount and translationally relevant. Analysis of whole blood at intervals following the first and second doses of intratumoral cytokine therapy indicated minimal elevation of systemic alanine transaminase (ALT) levels for most patients tested at the lowest three dose levels, with ALT levels normalizing prior to administration of each subsequent cytokine dose (Fig. 2A). The predominant adverse events observed were mostly grade 1 and 2 across dose-level cohorts, with the most commonly occurring events being associated with hemoglobinemia, thrombocytopenia, lethargy, anorexia, and elevation in ALT and alkaline phosphatase (ALP) levels (Table 2). The owner of one 2×-dose-level dog with elevated ALT chose not to pursue the sixth dose of cytokine treatment. Select dogs demonstrated elevated ALT in the 3.3× dose cohort and responded well to s-adenosylmethionine and silybin to mitigate hepatocyte toxicity and normalize liver function. More clinically significant ALT elevation and symptoms consistent with cytokine release syndrome (i.e., thrombocytopenia, hypoproteinemia, severe lethargy, and pyrexia) were observed in the 5× dose cohort (Table 2). These patients received supportive care including intravenous fluids and dexamethasone SP (0.5 mg/kg, IV) and fully recovered after treatment. A reduction in subsequent doses to this cohort to 3.3× was instituted to minimize discomfort and health risks in these patients.

To correlate the observed clinical activity and potential toxicity with pharmacodynamic biomarkers, profiling of the systemic chemokine/cytokine responses to combination RT with intratumoral cytokine treatment was performed. Similar response dynamics to those previously reported were observed, wherein IL12 drives elevation of systemic levels of IFNγ, with a delay in the elevation of IL10 (Fig. 2B; refs. 51, 57, 71–73). Peak levels of IFNγ were mostly consistent among the lowest three dose cohorts, but spiked significantly higher at the more toxic 5× dose level. To confirm that circulating elevations of IFNγ and IL10 were biomarkers of intratumoral cytokine activities and not an epiphenomenon of ionizing radiation or injection site trauma, an additional cohort of four dogs receiving only a single dose of RT (9 Gy) and sham intratumoral saline injection was analyzed and no measurable concentrations of circulating IFNγ or IL10 was identified (Supplementary Fig. S4A–D). Moreover, a cohort of three dogs receiving intratumoral cytokine only without RT demonstrated similar dynamics of IFNγ and IL10 changes following treatment, providing further evidence that the dynamic responses observed via multiplex serum profiling are IL2/IL12 mediated rather than being due to the combination of RT with intratumoral cytokine treatment (Supplementary Fig. S4A–D). Biological antitumor responses were also observed in this cohort of dogs treated without radiation, demonstrating the critical role of IL2 and IL12 in potentiating regression of advanced melanoma in pet dogs (Supplementary Tables S3 and S4; Supplementary Fig. S5A and S5B).

Given the importance of IFNγ both directly on tumor cells and in facilitating productive antitumor immune responses (74–77), an estimation of the systemic exposure of patients to IFNγ via AUC was calculated. The analysis provided some evidence of immune tachyphylaxis, in which the response to intratumoral cytokine therapy seems to have diminished by the sixth dose, relative to the responses to the initial doses of therapy (Fig. 2C). This is most pronounced in the 2× dose cohort, although some pet owners elected to not continue treatment with six doses of intratumoral cytokine therapy because of complete regression of the local tumor site concurrent with some adverse toxicities [2/4 (50%) dogs], confounding the statistical comparisons at the 3.3× dose level. The phenomenon of immunologic defervescence has been difficult to study in murine models but has been noted in human patients, highlighting the potential utility to examine various treatment regimens in dogs to minimize tachyphylaxis. Characterization of antidrug antibody responses that could attenuate immunostimulatory activities to collagen-anchored cytokines found the existence of antibodies but not at levels high enough to explain the magnitude of reduced IFNγ response at the final dose timepoint (Supplementary Fig. S6A–C).

Finally, patient body temperatures were measured during the posttreatment monitoring phase, and it was observed that most dogs became mildly febrile regardless of the dose level (Fig. 2D). These mild symptoms did not require medical intervention and were often accompanied by transient inappetence and lethargy among the patients during the monitoring phase. Overall, the responses potentiated by therapy were well tolerated at the 1× and 2× dose levels, with dose-limiting toxicities first observed at the 3.3× dose level in a subset of patients but among a majority of patients at 5×.

Many pet dogs enrolled in this trial presented with metastatic lesions, providing an opportunity to examine whether local treatment of the primary tumor with IL2, IL12, and RT could promote locoregional responses at untreated metastatic sites, an important outcome for intratumoral therapies. CT measurements were obtained for metastatic lymph nodes and measured for radiologic response in comparison with their pretreatment volumes (Fig. 3A). Following treatment of primary tumors with RT and intratumoral cytokines, 3/10 (30%) dogs displayed a partial response at metastatic lymph nodes. Two additional dogs achieved SD during the treatment period, for an overall biologic response rate to combination therapy in 5/10 (50%) dogs. Two dogs were euthanized prior to the day 84 measurement; one because of suspected progression of brain/CNS metastases and another for significant progression of lung metastases. For a subset of the responding patients, appreciable regional edema was present at metastatic lymph node sites at the interim (day 28) measurement.

For one patient, a pretreatment FNA of the tumor-draining lymph node and a subsequent FNA of the same regional lymph node in the day-28 CT scan were obtained (Fig. 3B). Prior to treatment, this lymph node was completely effaced with disease, as detected via the absence of immune cells and the majority presence of cancerous melanocytes and extracellular melanin (Fig. 3C). After two doses of intratumoral cytokine and single RT treatment, the lymph node CT scan indicated a robust decrease in metastatic regional lymph node volume (−35.3%; partial response) and concurrent immunologic clearing of melanoma cells and pigmentation (Fig. 3D). We observed the presence of polymorphonuclear cells, likely neutrophils, in the FNA, many of which had phagocytosed tumor cell debris and melanin. One additional patient had detectable lung metastasis at the time of presentation and trial enrollment (Fig. 3E). Although this dog ultimately succumbed to progressive metastatic disease, there was evidence of at least one regressing lung metastasis lesion during treatment (Fig. 3F). This mixed abscopal response may be due to underlying genetic differences between primary and disseminated disease, as well as among differing clonally derived lung metastases (78–80). However, the locoregional response of metastatic disease to combined intratumoral IL2/IL12 and single-dose RT treatment is consistent with an immune-mediated mechanism of action and similar to prior reports of combined radiation with immunotherapy (81–85).

Similar to the pivotal phase III clinical trials with T-VEC (86), out of concern that longitudinal sampling of the primary treated tumors could confound results by introducing additional paths for intratumoral dose egress, we did not profile the immune response to therapy during treatment. However, building upon our prior characterization of the immune-mediated response to collagen-anchored cytokines in canine soft-tissue sarcoma and murine tumors (35, 51), we highlight an anecdotal case of long-term antitumor response after the completion of treatment in oral melanoma which presumably involved immune activity.

One patient had a strong primary tumor response while on therapy but displayed slow growth of that tumor in the year following treatment completion (Fig. 3G). However, at the 12-month follow-up appointment after treatment, the primary tumor was no longer visible and was later confirmed to be absent via CT (Fig. 3H and I), as well as by histopathology (Fig. 3J). Additional immunohistochemistry for Melan-A further confirmed the absence of disease in the gingival tissue of this patient at day 529 (Supplementary Fig. S7). Although examples of spontaneous human tumor regressions have been reported (87, 88), they are quite rare (∼10⁻⁵; ref. 87). The slow posttreatment tumor growth may correspond to a state of immune equilibrium, leading eventually to tumor elimination, similar to other immunotherapy approaches (89, 90).

Identifying and understanding which factors, if any, had contributed to poor response to the combined RT plus intratumoral cytokine treatment regimen was further studied. Toward this goal, FFPE-processed primary and metastatic tumor tissues from eight dogs that were euthanized for PD were advanced for detailed histologic and genomic evaluations. No clear trends were observed between the overall survival of these progressor patients and immune infiltration status profiled through immunohistochemistry for CD3 and Iba1 (Supplementary Fig. S8A–G). Extracted RNA from these tissue sections were profiled using the NanoString nCounter platform (Fig. 4A). A hierarchical cluster of pathway-specific gene expression emerged that encompassed the coordination of innate and adaptive immunity, including T-cell, B-cell, and macrophage functions, as well as antigen presentation (Fig. 4B). Within this cluster, varied expression among the progressor dogs was observed and additional unsupervised clustering of the antigen presentation gene set yielded two clusters of four dogs each (Fig. 4C). Given that tumor dysregulation of antigen presentation and response to IFNγ is a common immune evasion mechanism (75, 91, 92), the expression of MHC class I–related genes was examined and a significant difference in B2m and Dla79 transcripts between the clusters of progressor dogs was identified (Fig. 4D). This result suggested that the first cluster of dogs may have had impaired MHC-I expression, at least among a partial population within the heterogeneous tumor. Broader comparisons in gene expression between these two cohorts indicated greater expression of effector lymphocyte-associated genes such as Slamf6, Ctsw, and Trgc3, as well as interferon-inducible genes such as Ido1, Gbp5, and Cxcl10 among the MHC-I higher expression cohort, cluster 2 (Fig. 4E).

Intriguingly, the most differentially expressed gene was for the Fas ligand (Faslg) and may represent a consistent mechanism of immune escape within the cohort of progressor dogs (cluster 2) with greater B2m expression. It has been established that peripheral expression of Fas ligand on multiple cell types in response to inflammatory stimulus promotes deletion of auto-reactive T lymphocytes (e.g., peripheral tolerance; ref. 93), so we examined whether there were compositional differences in the immune compartments from the tumors of the progressor dog cohorts. Using CIBERSORTx (60), the relative abundance of immune cell populations from the bulk NanoString profiling data was estimated. Tumors with reduced B2m expression were accompanied by greater populations of canonical tumor-suppressive immune cells (i.e., “M2” polarized macrophages and neutrophils), whereas dogs with higher MHC-I antigen presentation had more activated macrophages and CD4 T lymphocytes (Fig. 4F). Together, these differences likely contributed to the poorer prognosis of patients with reduced MHC class I antigen presentation, regardless of the tumor stage at presentation (Fig. 4G; log-rank hazard ratio: 4.472).

To explore why the cohort of dogs with higher class I antigen presentation and reduced abundance of immunosuppressive immune populations (cluster 2) still progressed after therapy, gene expression was examined within tissue collected from metastatic tumor sites. Using a gene set describing common genetic mutations that enable immune escape in primary or metastatic tumor tissues (80), the differences in expression between cohorts is diminished in metastatic tumors (Fig. 4H). This suggests that metastatic tumors from dogs with higher MHC-I expression at their primary tumors may have been preferentially seeded by tumor subpopulations with greater genetic immune escape, such as MHC-I loss of heterozygosity. We further examined gene expression between cohorts at their primary and metastatic tumors across an annotated set predictive of human response to checkpoint inhibitors (94). We found that only primary tumors of higher expression MHC-I dogs are expected to have a positive response to immunotherapy, consistent with the observed local response, but metastatic progression of these patients follows our combined cytokine treatment (Fig. 4I). Overall, these results motivate exploration of treatment combinations to overcome dysfunctional MHC class I antigen presentation in tumors to extend therapeutic benefit to a greater population of pet dogs, with the intention that lessons gleaned from comparative oncology studies can be quickly pivoted to accelerate novel immunotherapeutic strategies to benefit human patients with cancer.

Mechanisms of primary and adaptive resistance to immunotherapy contribute to the lack of clinical benefit for a majority of patients with cancer who are treated with antagonistic, checkpoint-inhibiting antibodies (5). As a result, there have been attempts to combine these therapies with agonistic, or immune-stimulating, agents to overcome tumor resistance mechanisms and drive more durable responses (95, 96). Cytokines, such as IL2 and IL12, are one class of agonistic therapies that have shown great promise against human cancers, but suffer from unacceptable toxicities due to their activation of immune cells throughout the body (15, 16). Approaches to restrict the activity of potent cytokines to tumor have gained momentum, one of which includes the retention of engineered cytokines to tumor extracellular matrix following intratumoral injection (25, 35–37). We and others have previously reported on the safety improvements provided by this strategy of anchoring cytokines to tumor collagen in both mice and pet dogs (25, 35, 51), but the efficacy in advanced canine tumors had been previously unexplored.

In this work, we evaluated the efficacy of tumor-localized IL2 and IL12 cytokines in pet dogs with advanced OMM to potentially predict success of clinically translating this approach. As dogs share key physical features and tumor biology with humans, they have gained traction as models for human comparative oncology (43, 45, 48). Here, we have observed encouraging results for both the antitumor efficacy and tolerability of single-dose RT with repeat intratumoral IL2 and IL12 cytokines. Primary tumor responses were often rapid and durable, with 256-day median survival across all treated cohorts; significantly longer than the historical 65-day median for untreated canine oral melanoma (68). Moreover, many of these responses were observed among dogs in the non-toxic 1× and 2× cohorts, suggesting that the tumor-localization strategy via retention to tumor collagen is clinically promising for safely and effectively treating human malignancies. Locoregional responses at metastatic sites driven by intratumoral therapy achieved an overall biologic response against combined tumor and metastases in 10/13 (76.9%) dogs receiving full therapy, with partial responses in 8/13 (61.5%) dogs (Supplementary Fig. S9). This result provides early evidence that intratumoral treatment with collagen-bound cytokines may potentiate systemic antitumor immunity in pet dogs with naturally occurring cancers. Importantly, these canine tumors develop under evolving tumor immune evasion and suppression mechanisms analogous to those in humans, suggesting that this engineered cytokine approach may achieve similar responses in human clinical trials.

Profiling of dogs that progressed while, or soon after, receiving the RT plus intratumoral cytokine treatment revealed that dysfunctional antigen presentation may contribute to the rapid progression of canine malignant melanoma. This complements a growing list of canine tumor features that overlap with the human hallmarks of cancer, including sustained proliferative signaling (97) and mutations to oncogenic driver or tumor suppressor genes (49, 98). Although less definitive, the dogs with higher MHC class-I expression may have progressed because of tumor microenvironment–induced dysfunction of immune cells. With the observation that Faslg and Ido1 are more highly expressed by these tumors, we suspect that the combination cytokine therapy was actively promoting an antitumor response met by immune counter-regulation, as we have observed previously in canine soft-tissue sarcomas (51). The combination of IL12/IL2 has been described to upregulate the expression of Fas ligand on draining lymph node lymphocytes (99), which, while aiding their ability to kill malignant tumor cells, could contribute to eventual lymphocyte fratricide or suicide (100). This mechanism might contribute to our observation of tachyphylaxis in some of the dogs (Fig. 2C). Moreover, the mixed response between primary tumors and metastatic sites may manifest from the varied gene expression landscape and erected barriers to immune function observed between these metastatic tumors and their primary tumor counterparts, suggesting that systemic therapies (such as anti-PD1 antibodies) may be necessary to leverage cytotoxic effector cells primed by local intratumoral therapy (35, 51).

Our learnings from each group of progressing dogs provide actionable insights for future combination treatments to test alongside the intratumoral cytokine approach. To this end, we are interested in evaluating the combination of checkpoint inhibitors with the RT-plus-intratumoral cytokine treatment in future studies, particularly along the PD1/PDL1 axis, because of recent encouraging responses with these immunotherapies observed in local and advanced canine melanoma (101–104). Our prior work with canine soft-tissue sarcomas indicated that checkpoint blockade might relieve counter-regulatory responses to intratumoral cytokine therapy, which we confirmed in the murine B16F10 tumor model (51). However, resistance to intratumoral IL2 and IL12 therapy via beta-2-microglobulin loss and, subsequently, dysfunctional antigen presentation, seem to overlap with known resistance mechanisms to checkpoint inhibitors (91, 105). As a result, future screening of canine B2m and MHC class-I–associated gene expression prior to trial enrollment could help accrue patients into separate, more rationally designed, combination treatments. For dogs with reduced or dysfunctional antigen presentation, there have been strategies reported for combining immunotherapies with epigenetic drugs to remove silencing of beta-2-microglobulin and restore MHC-I expression (106–108), in addition to strategies to engage innate immune cells for direct tumor-cell killing (109–111) or to coordinate their licensing of antigen-independent killing by CD8⁺ lymphocytes (112, 113). Finally, given our observation of tachyphylaxis in response to repeat cytokine dosing and reports of the importance for immune rest in engineered CAR-T therapies (114), we are interested in exploring longer intervals between cytokine doses or separately administering individual cytokines to minimize activation-induced cell death or induced dysfunction of primed CD8⁺ T cells.

Overall, this work highlights the benefit of preclinical evaluation of a novel immunotherapy alongside current standard of care in a more human-analogous cancer model than mouse tumors. While statistical power of such a trial in pet dogs is more limited, we argue that the value gained in predictive efficacy, safety, and resistance to therapy is obtained at dramatically lesser expense and greater speed than a corresponding human clinical trial. Exploitation of canine trials as a bridge from murine studies to the clinic should be expanded to reap these benefits more widely. Certain methodology to maximize value from these canine cancer trials stands to gain from broader investigation as well. We recognize that a primary limitation of this study is the lack of longitudinal sampling from canine tumors to characterize the evolution of antitumor responses and resistance to cytokine treatment. Through comparative oncologic testing, we anticipate a greater likelihood of future clinical success for our collagen-binding cytokine approach, as well as more broadly for other novel immunotherapies investigated in pet dogs with cancer.

N. Momin reports personal fees from Cullinan Therapeutics outside the submitted work, as well as a patent for US20200102370A1, licensed to Cullinan Therapeutics. E. Fink reports grants from NIH during the conduct of the study. K.D. Wittrup reports other support from Cullinan Therapeutics outside the submitted work. No disclosures were reported by the other authors.

J.A. Stinson: Conceptualization, formal analysis, investigation, writing–original draft, writing–review and editing. M.M.P. Barbosa: Conceptualization, formal analysis, investigation, writing–original draft, writing–review and editing. A. Sheen: Conceptualization, formal analysis, investigation, writing–review and editing. N. Momin: Conceptualization, formal analysis, investigation, writing–review and editing. E. Fink: Formal analysis, investigation, writing–review and editing. J. Hampel: Investigation, writing–review and editing. K.A. Selting: Formal analysis, investigation, writing–review and editing. R.L. Kamerer: Investigation, writing–review and editing. K.L. Bailey: Formal analysis, investigation, writing–review and editing. K.D. Wittrup: Conceptualization, formal analysis, funding acquisition, investigation, writing–original draft, writing–review and editing. T.M. Fan: Conceptualization, formal analysis, funding acquisition, investigation, writing–original draft, writing–review and editing.

We gratefully thank all of our pet dog owners for their consent and willingness to participate in this investigational trial. This study was supported by National Cancer Institute grant R01CA271243 (T.M. Fan and K.D. Wittrup), NIBIB grant R01EB031082 (K.D. Wittrup), and GRFP (A. Sheen). The authors would like to acknowledge Dr. Amy Schnelle and the Tumor Engineering and Phenotyping Shared Resource at the Cancer Center at Illinois for assistance with histology and immunohistochemistry analysis.