Collagen-anchored interleukin-2 and interleukin-12 safely reprogram the tumor microenvironment in canine soft tissue sarcomas

Abstract

Purpose:

Cytokine therapies such as interleukin-2 and −12 suffer from impractically small therapeutic windows driven by their on-target, off-tumor activity, limiting their clinical potential despite potent antitumor effects. We previously engineered cytokines that bind and anchor to tumor collagen following intratumoral injection, and sought to test their safety and biomarker activity in spontaneous canine soft tissue sarcomas (STS).

Methods:

Collagen-binding cytokines were canine-ized to minimize immunogenicity and were used in a rapid dose-escalation study in healthy beagles to identify a maximum tolerated dose. Ten client-owned pet dogs with STS were then enrolled into trial, receiving cytokines at different intervals prior to surgical tumor excision. Tumor tissue was analyzed through immunohistochemistry (IHC) and Nanostring RNA profiling for dynamic changes within treated tumors. Archived, untreated STS samples were analyzed in parallel as controls.

Results:

Intratumorally administered collagen-binding IL-2 and IL-12 were well-tolerated by STS-bearing dogs, with only Grade 1/2 adverse events observed (mild fever, thrombocytopenia, neutropenia). IHC revealed enhanced T cell infiltrates, corroborated by an enhancement in gene expression associated with cytotoxic immune function. We found concordant increases in expression of counterregulatory genes that we hypothesize would contribute to a transient antitumor effect, and confirmed in mouse models that combination therapy to inhibit this counterregulation can improve responses to cytokine therapy.

Conclusions:

These results support the safety and activity of intratumorally-delivered, collagen-anchoring cytokines for inflammatory polarization of the canine soft tissue sarcoma tumor microenvironment. We are further evaluating the efficacy of this approach in additional canine cancers, including oral malignant melanoma.

INTRODUCTION

Despite a myriad of promising studies on novel anti-cancer agents in mouse tumor models, their likelihood of clinical translation has remained low(1-3). Challenges with study design and/or execution logistics partially contribute to their lack of translational success; however, most clinical trial failures result from unexpected toxicities or lack of anticipated benefit(3,4). Preclinical mouse models of human cancers, including syngeneic, patient-derived xenograft, and genetically-engineered mouse models offer statistically-powered insight into the potential effect and mechanism of action of anti-cancer therapies (or treatment combinations). However, as these models incompletely capture key aspects of human tumor biology–particularly the complex interactions between the tumor, immune, and stromal cellular compartments over time–potential toxicities and modes of action can be indiscernible or overlooked. As a landmark example, the anti-CD28 agonist antibody TGN1412 caused severe cytokine release syndrome in every treated patient after demonstrating negligible toxicity in murine tumor models, a difference now attributed to the lack of an effector memory CD4+ T cell population in mice(5,6). Moreover, we have come to recognize that a surprising number of anti-cancer agents achieve efficacy through off-target toxicity and not their intended mechanism of action(7). As such, use of additional preclinical tumor models that offer improved prediction of human safety, magnitude of benefit, and mechanism of action of anti-cancer agents would complement murine tumor models and improve the likelihood of successful clinical translation.

Spontaneous cancers that arise in outbred companion animals such as pet dogs have gained considerable attention as a bridge between murine and human cancer trials, due to their similarities with certain human tumors(8-10). Canine malignancies mirror several features of human cancer by developing over time in the context of an intact, experienced immune system and with recurrence and metastatic behaviors(11,12). Moreover, owing to the body size and metabolic rates of dogs, comparative physiologic properties (e.g. diffusional distances) in canine tumors can better predict the pharmacokinetic and pharmacodynamic profiles of anti-cancer agents than mouse cancer models(13). Particularly for the evaluation of novel immunotherapies, in which the cancer-immune set point dictates response, canine tumors’ heterogeneous microenvironment and immune resistance mechanisms offer significant biological relevance(9,13,14). With estimates of over 4 million new cancer diagnoses in pet dogs each year in the United States alone, canine tumors are an abundant, human-relevant model for specific tumor histologies to evaluate the activity and safety of anti-cancer agents, immunotherapies, and combination dosing regimens(15).

We sought to investigate in canine cancers our previously-reported strategy of anchoring pro-inflammatory cytokines to tumor-associated extracellular matrix collagen following intratumoral injection(16,17). Interleukins −2 (IL-2) and −12 (IL-12) have promising antitumor potential driven by their ability to expand and stimulate cytotoxic T cells and natural killer (NK) cells, but their clinical development has been limited by narrow therapeutic windows caused by on-target, off-tumor activation of circulating immune cells(18-20). To minimize this toxicity, several strategies to limit the off-tumor activity of IL-2 and IL-12 are under development, including conditionally-active cytokines(21-23), biased engineered IL-2 receptor agonists(24-26), and tumor-targeting antibody-cytokine fusions(27-31). However, the accumulation and exposure of cytokines in the tumor microenvironment remains limited with these strategies(17,32,33). As a result, intratumoral injection of immunotherapies has gained traction as an approach to increase tumor exposure while reducing systemic toxicity(34-37). We and others have found that intratumorally-injected cytokines potentiate stronger anti-tumor immunity in mice when engineered to persist in the tumor via anchoring to the tumor extracellular matrix component collagen(16,17,38,39).

To further explore the clinical potential of this intratumoral anchoring approach, we generated canine versions of IL-2 and IL-12 collagen binding cytokines and evaluated their safety and activity in pet dogs bearing soft tissue sarcomas (STS). This cancer type is surgically excisable, enabling assessment of the cytokine-driven changes in the TME through immunohistochemistry and Nanostring RNA profiling(40). We observed significant, but transient, increases in tumor immune infiltrates driven by the activity and regulatory response to intratumoral cytokines. Through comparative analysis in the mouse B16F10 model, we investigated treatment combinations to attenuate immune counterregulation and the resultant impact on survival efficacy. Taken together, our studies suggest that the development of novel immunotherapeutic agents like our collagen-binding IL-2 and IL-12 cytokines can be bolstered through comparative oncologic evaluation in both mouse and canine tumor models to understand safety, mechanism of action, and synergistic dosing combinations. These collective insights accelerate and derisk the human clinical translation of this novel immunotherapy and other similar intratumoral agents.

MATERIALS AND METHODS

Cell lines and mouse studies

The cell lines HEK293-F (Gibco; RRID:CVCL_D603), B16F10 (ATCC Cat# CRL-6475, RRID:CVCL_0159), CTLL-2 (ATCC Cat# TIB-214, RRID:CVCL_0227), and HEK Blue IL-12 (Invivogen; RRID:CVCL_UF31) were cultured according to vendor instructions. All cells were maintained in 5% CO₂ at 37°C and tested negative for mycoplasma. B16F10 cells were confirmed to be negative for rodent pathogens (IMPACT I test, IDEXX) prior to murine tumor studies. Six- to eight-week old female C57BL/6 mice (Taconic) were purchased and used for murine Nanostring and combination therapy efficacy experiments. All mouse studies were conducted under approval of the MIT Committee on Animal Care in accordance with federal, state, and local guidelines. Additional details on study designs and dosing can be found in the Supplementary Methods S1.

Collagen-binding IL-2 and IL-12 cloning, expression, and characterization

We cloned and expressed canine-ized cytokines IL-2 and IL-12 fused to the collagen-binding domain LAIR1 as detailed in the Supplementary Methods S1 (sequences in Supplementary Table S1). To assess IL-2 bioactivity, 10,000 CTLL-2 cells (ATCC Cat# TIB-214, RRID:CVCL_0227) were plated per well in incomplete T-cell media (RPMI 1640, 10% fetal bovine serum, 2mM L-glutamine, 1mM sodium pyruvate) and incubated for 48h at 37°C with dilutions of the IL-2 fusion protein. After incubation, cell viability was assessed using the CellTiter-Glo 2.0 assay (Promega). The HEK Blue IL-12 (Invivogen; RRID:CVCL_UF31) assay was performed according to vendor instructions using dilutions of the collagen-binding IL-12 cytokine. Collagen-binding of both cytokines was evaluated through ELISA with rat collagen I-coated plates and an anti-hexahistidine detection antibody (Abcam Cat# ab1269, RRID:AB_299333) as previously reported(16). Low endotoxin levels (<5 EU/kg/dose) were confirmed for each cytokine batch using the Endosafe Nexgen-PTS system (Charles River Labs) with typical values <1 EU/mg protein.

Dose escalation in healthy beagles

We sought to characterize the safety of collagen-binding IL-2 and IL-12 to establish a well-tolerated treatment dose for tumor-bearing dogs. Four beagles were enrolled into the study protocol approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Illinois at Urbana-Champaign (UIUC). Using reported values for single-agent maximum tolerated doses of IL-2 in humans(41-43), we allometrically scaled dosing to dogs (SI-T2)(44). Similarly, a maximum dose of IL-12 was selected based upon prior assessment of an antibody-IL-12 fusion protein tolerated by melanoma-bearing dogs(45). The first beagle was treated intradermally with the predicted maximum doses of 173.6 μg/kg cLAIR-CSA-cIL2 and 0.417 mg/m² cIL12-CSA-cLAIR and was clinically monitored for response. Periodic blood draws by jugular venipuncture were performed for whole blood profiling and cytokine/chemokine analysis (Supplementary Methods S1) at timepoints before dosing, as well as 2-, 4-, 8-, 24-, and 48-hours post-treatment. Additional beagles were dosed at 1/10th and 1/100th of the predicted maximum doses to assess tolerability of the therapy at lower doses. Any symptoms consistent with acute toxicity were treated appropriately.

Trial eligibility and enrollment of pet dogs

Ten client-owned pet dogs with histologically confirmed soft tissue sarcoma (STS) including peripheral nerve sheath tumors, myxosarcoma, and perivascular wall diagnoses were enrolled in this study (Supplementary Figure S1, Supplementary Table S3, Supplementary Table S4). Eligibility criteria required dogs to have no evidence of metastatic disease (nodal or distant), primary tumors to be amenable to surgical removal, and all parameters of kidney and liver function (such as creatinine, ALT) to be less than 1.5x upper normal limits prior to study entry. All patient owners provided written consent before enrollment and all procedures were performed in accordance with the study protocol approved by the UIUC IACUC. FFPE STS tissue samples surgically removed from treatment naïve pet dogs served as controls, and were randomly selected from existing, archived tissues maintained in the Veterinary Diagnostic Laboratory, College of Veterinary Medicine at UIUC. These untreated STS cases were matched for comparable histologic grade (I or II) and overlapping histotypes.

Trial design and intratumoral dosing of cytokines in soft tissue sarcomas

Ten pet dogs with STS were recruited to three cohorts that received the tolerated dose of collagen-binding cytokines, followed by surgical excision of tumors. These cohorts varied in the interval between cytokine treatment and surgical excision; cohorts 1 and 2 (n=3/group) had tumors removed either 2 or 8 days after treatment. Cohort 3 (n=4) received a second dose of intratumoral cytokines 6 days after the first dose, with tumor resection performed 2 days later. All dogs were clinically monitored at the UIUC Veterinary Teaching Hospital for 48 hours following cytokine dosing with scheduled blood draws performed. Doses of cLAIR-CSA-cIL-2 (17.4 μg/kg) and cIL-12-CSA-cLAIR (41.7 μg/m²) were prepared from frozen protein aliquots and combined in a total volume not exceeding 0.5 mL in sterile saline. An insulin syringe with 29G, ½ inch needle 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. Adverse events were monitored for and graded according to published veterinary standards and were treated appropriately(46).

Canine tumor resection, processing, and immunohistochemistry

Canine STSs were surgically excised at the specified intervals following intratumoral dosing of collagen-anchored cytokines, as described above. Resected tumor tissues were fixed in 10% neutral buffered formalin prior to an extended paraffin processing and embedding protocol. Inflammatory cells corresponding to positive CD3 (T lymphocyte; Biocare CP215C), PAX5 (B lymphocyte; Abcam Cat# ab109443, RRID:AB_10862070), and Iba-1 (macrophage; Biocare Cat# CP 290 B, RRID:AB_10583150) immunohistochemical staining were quantified on an Olympus BX46 microscope using a high-power 40x objective with a field number (FN) of 26.5 mm. This FN number indicates each high-power field is 0.34 mm², and to reach a desired total examined area of 2.38 mm², seven fields were examined for every stained slide within each cohort. Cells with both positive staining and an observable nucleus or nuclear silhouette were counted, cell counts were recorded and average results for each stain within each cohort were calculated. All STS samples were histologically evaluated and classified by a single board-certified veterinary pathologist.

Nanostring RNA profiling

RNA was isolated from 10μm formalin-fixed, paraffin-embedded (FFPE) samples from resected canine STSs or mouse B16F10 tumors using an RNEasy FFPE Kit and deparaffinization solution (Qiagen). Isolated RNA was examined by Bioanalyzer (Agilent) for assessment of fragment size prior to hybridization with Nanostring probe sets. Canine RNA samples were hybridized with the Canine IO nCounter Panel code set for 22h at 65°C per manufacturer’s instructions. Murine RNA samples were hybridized with the Mouse PanCancer Immune Profiling Panel code set under the same conditions. Following hybridization, samples were loaded into the analysis cartridge and scanned at maximum resolution using Nanostring PrepStation and Digital Analyzer. Raw RCC count files are available at GSE218914.

Canine RCC count files were normalized using Nanostring nSolver software after background thresholding using the mean of 8 negative control probes and batch correction against a panel standard control. No batch correction was required for murine analysis as only a single scanning cartridge was used. 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 false discovery rates (FDR) of differentially-expressed genes (DEGs).

Statistical analysis

Statistics were performed using Prism v9 (GraphPad). Power calculations were not performed to determine treatment group sizes, but sample numbers and experiment replicates were chosen based on prior published results to be sufficient for significant results from appropriate statistical tests. The details of statistical analysis have been provided in the descriptions for figures.

Data availability statement

The data generated in this study are available within the article and its supplementary files. Nanostring expression data for mouse and canine tumor responses to IL-2 and IL-12 therapy has been made publicly available in Gene Expression Omnibus (GEO) at GSE218914.

RESULTS

Development of canine-specific collagen-binding interleukin-2 and −12 cytokines.

Analogous to our reported murine constructs, we designed canine interleukin-2 (cIL-2) and interleukin-12 (cIL-12) cytokine fusion proteins to bind tumor collagen following intralesional injection(16,17). Canine IL-2 or IL-12 were linked to the collagen-binding ectodomain of canine leukocyte-associated immunoglobulin-like receptor 1 (cLAIR1), and canine serum albumin (CSA) to aid protein expression and increase molecular weight for improved tumor retention(17) (Figure 1A; Supplementary Table S1). We recombinantly expressed and purified monomeric cIL12-CSA-cLAIR and cLAIR-CSA-cIL2, then confirmed their molecular weights by SDS-PAGE and the minimal presence of protein aggregates by size-exclusion chromatography (Figure 1B-C). The correct folding of canine-ized IL-12 cytokine was confirmed using the HEK Blue IL-12 reporter cell line, indicating the canine protein possessed similar potency to its murine analogue (Figure 1D). We validated the activity of our IL-2 cytokine through the proliferation of the IL-2 dependent CTLL-2 murine cell line, noting that the canine cytokine does not cross-react with the high-affinity murine IL-2Rα (Figure 1E). Finally, we confirmed both canine cytokine fusion proteins bound collagen via ELISA, albeit with a weaker affinity than that reported of murine LAIR (Figure 1F, Supplementary Figure S2).

Figure 1. Characterization and activity of canine collagen-binding cytokines.

(A) Schematic of IL-2 and IL-12 (single-chain) fusion proteins. (B) Purified canine cytokines on a non-reducing SDS-PAGE gel with Coomassie blue (Simple Blue) stain. (C-D) Size exclusion chromatograms of 100 μg purified canine cytokines using a Superdex 200 Increase 10/300 GL column. Black bars above the trace signify the fractions isolated, filtered, frozen and used for therapeutic injections. (E) IL-12 bioactivity as measured by the absorbance readout for the HEK-blue IL12 reporter cells. (F) IL-2 bioactivity measured via CTLL-2 proliferation. CTLL-2 cells were incubated with indicated protein for 2 days in incomplete media, then cell viability was measured using CellTiter-Glo. (G) LAIR fusion protein binding to rat collagen type I was measured by enzyme-linked immunosorbent assay. MSA (E, G) or mLAIR-MSA (F) serve as a nonspecific protein controls.

Identification of canine maximum tolerated dose (MTD) of anchored cytokines.

We allometrically scaled published IL-2 and IL-12 single agent doses from prior studies in humans and dogs to assess the safety and tolerability of our engineered cytokines in a cohort of healthy dogs at a potentially human-relevant dose(41-45,47). As our intended dosing was a combination of IL-2 and IL-12 injected intratumorally, we predicted that adverse events might be observed at lower doses than either monotherapy alone due to the interacting effects of these proinflammatory agents. As such, we set our ceiling dose for investigation at the allometrically scaled doses (173.6 μg/kg cLAIR-CSA-cIL2; 0.417 mg/m² cIL12-CSA-cLAIR) and also tested doses at 1/10th and 1/100th of this dose (Supplementary Table S2). In the absence of injectable tumors in healthy animals, beagles were injected intradermally at each of the dose levels and received a second dose 14 days later, if well-tolerated by clinical observation (Figure 2A; n=4 beagles with n=9 doses administered).

Figure 2. Dose-escalation study in healthy beagles determines maximum tolerated dose (MTD).

(A) Dogs were injected intradermally (i.d.) with escalating doses of cytokines on days 0 and 14. A single dog was treated at each increasing dose level until toxicity was observed (MTD), then 2 additional dogs (one de-escalated, one naïve) were treated at the previous dose (n=4 total dogs). In total, 9 doses of cytokine were administered. (B) Heatmaps of white blood cell count (WBC), neutrophil count, platelet count, and alanine transaminase (ALT) levels measured via bloodwork at indicated timepoints (NR: normal range). (C) Serum from the indicated time points after dosing was collected and analyzed for cytokines and chemokines. Heatmap columns show average data for sera from each dose level (n per group as indicated), reported as log10 fold change in concentration compared to pre-treatment (t=0hr). Raw data are in Supplementary Data File.

Treatment at the ceiling dose level (henceforth “10X MTD”) in one dog led to symptoms consistent with cytokine release syndrome (CRS) including Grade 2/3 adverse events(46) such as leukopenia, elevated serum ALT, fever, and vomiting (Figure 2B). This patient was treated with intravenous fluids and dexamethasone to manage CRS and restore normal body function. The lower doses of cytokines (0.1X MTD and 1X MTD), despite causing transient leukopenia and thrombocytopenia, were clinically well-tolerated with no evidence of serum ALT elevation and only Grade 1 fever and inappetence (SI-T3). We further examined serum analytes to profile potential biomarkers of response to the combination cytokine treatment (Figure 2C, Supplementary Figure S3). Interferon-gamma (IFN-γ) is a biomarker for IL-12 activity(48), and was observed to rise in a dose-dependent manner with our cytokine dose. Furthermore, we observe a delayed increase in serum IL-10 levels following treatment, irrespective of dose level. Both of these analytes have been reported in response to a tumor-targeted IL-12 immunocytokine tested in pet dogs(45). Chemokines KC-like and MCP1 were also elevated at the 10X MTD, but unexpectedly IP10 was not found to be elevated at any dose level, despite its strong correlation with IL-12 therapy(48). From these results, we selected the intermediate dose tested (1X MTD equivalent to 1/10th allometric ceiling dose) as the optimal biologic dose for investigation in tumor-bearing pet dogs.

Collagen-binding IL-2 and IL-12 cytokines are safely tolerated by pet dogs with sarcoma.

Having established a tolerated treatment dose, we enrolled 10 client-owned pet dogs with soft tissue sarcomas (STS) to evaluate the safety and activity of collagen-binding cytokines. These pet dogs presented with various STS subtypes that were amenable to surgical excision (Supplementary Table S3, Supplementary Table S4). Dogs were randomly assigned to the three treatment cohorts, which varied in the interval between intratumoral cytokine dosing and tumor resection (Figure 3A). Cohort 1 (treated two days prior to surgery) was designed to determine the short-term effects of treatment, and Cohort 2 (treated eight days prior to surgery) was intended to probe the durability of that effect. Cohort 3 (treated eight and two days prior to surgery) allowed characterization of the booster effect of cytokine stimulation on any immune infiltrates that result from the first dose. The cohort of untreated dogs represented a set of archived STS tumor tissue samples of varying histologic grade (I, II, and III) available at UIUC.

Figure 3. Collagen-anchored cytokines show safe toxicity profile in soft tissue sarcoma dogs.

(A) Dogs were injected intratumorally (i.t.) with cytokines (yellow arrows), then surgically resected (black arrow) on the schedules shown (patient information in SI-T3, T4). All dogs dosed at 17.4 ug/kg cLAIR-CSA-cIL-2 and 0.041 mg/m² cIL-12-CSA-cLAIR. Dogs were monitored for 48 hours post-treatment. (B) Heatmaps of white blood cell (WBC), neutrophil, and platelet count, and alanine transaminase (ALT) levels measured via bloodwork at indicated timepoints (NR: normal range). (C) Body temperature was measured at indicated time points. (B-C) Heatmap rows represent individual patients. (D) Serum from the indicated time points after dosing was collected and analyzed for cytokines and chemokines. Heatmap columns show average data for sera from each dose level (n per group as indicated), reported as log10 fold change in concentration compared to pre-treatment (t=0hr). Bloodwork values and serum analyte concentrations are available in the Supplementary Data File.

Complete blood counts assessed before and two-days after cytokine treatment indicated that roughly half of the dogs had Grade 1/2 leukopenia and thrombocytopenia (Figure 3B). Only one dog had elevated serum ALT levels, and received IV fluids. All dogs receiving a second dose of cytokine did not display more severe adverse reactions measured through complete blood counts. Clinically, all dogs receiving intratumoral cytokines were found to tolerate the therapy well, with only minor lethargy, inappetence, and fever reported in a few patients (Figure 3C, Supplementary Table S3). Additional screening of serum analytes through multiplex assay revealed a common response pattern of elevated IFN-γ, followed by elevated IL-10, in the sarcoma-bearing dogs, similar to that seen in healthy beagle dogs (Figure 3D, Supplementary Figure S4).

Tumor-retained IL-2 and IL-12 drive immune infiltration to canine soft tissue sarcomas.

Canine STSs typically display a paucity of adaptive immune cell infiltrates and are thus classified as immunologically “cold” or “desert” phenotypes(49). Given the increased serum chemokine levels following treatment with collagen-binding cytokines (Figure 3D), we sought to assess whether treatment changed the immune composition of canine STSs. Resected tumors from the treated cohorts, as well as archived untreated STS samples, were stained with antibodies against canine CD3, Iba-1, and PAX5 (Figure 4). PAX5 for B cell staining revealed no difference between cohorts, despite regions of B cell aggregates in one treated sample (Supplementary Figure S5). Quantification of CD3+ cells confirmed the absence of a significant population of T cells in untreated STS tumors, which increased following treatment with intratumoral IL-2 and IL-12 (Figure 4A). Tumor-associated macrophages often contribute to an immunosuppressive TME, but can be functionally polarized towards an antitumor phenotype through exposure to IL-12 and/or IFN-γ exposure(50-52). Untreated STS samples display a notable macrophage population, and following treatment with two doses of cytokines there was a significant increase in Iba-1+ cells (Figure 4B). As the polarization of these macrophages and other immune infiltrates cannot be inferred from their morphological features alone, we used bulk gene expression profiling to understand the activities of the aggregated cell populations.

Figure 4. Cytokine treatment increases abundance of tumor-infiltrating lymphocytes and monocytes.

Immunohistochemical staining for CD3 and Iba-1 was performed on a single section from each patient within each cohort. Representative images for a single patient in each cohort are shown. Inflammatory cells corresponding to (A) CD3+ (T lymphocytes), (B) Iba-1+ immunohistochemical staining were quantified. The average count for each stained slide (1 per patient) with each cohort was calculated by taking the mean of the positive cell count from seven fields per sample. Each field of quantification is 0.34 mm². Scale bar = 100 μm.

Anti-tumor effector phenotypes are displayed by sarcoma immune infiltrates.

Using RNA isolated from FFPE tissue samples from treated and untreated patients, we found that intratumoral administration of collagen-binding cytokines produced significant, but transient, upregulation of many genes included in the Nanostring panel relative to untreated controls (Figure 5A). Notably, gene expression in cohort 2, 8 days following cytokine treatment, is not significantly different from untreated controls. Cohorts 1 and 3, which received their final cytokine treatment two days prior to tumor resection, displayed remarkable overlap in their patterns of differentially expressed genes (DEGs), including the immune recruiting factors Cxcl10 and Ccl8, as well as T and NK cell-related genes Gzmb, Ncr1, and Nkg7 (Figure 5B).

Figure 5. Nanostring RNA profiling of canine soft tissue sarcomas.

(add more description) (A) Volcano plots for cohorts 1 (C1), 2 (C2), and 3 (C3). Fold-change is determined relative to untreated control data. Genes associated with significant p-adj values (< 0.05) are highlighted in red and quantified for each cohort. (B) Venn diagram of all significant differentially expressed (DE) genes highlighted in (A), grouped by cohort in which they are found. C2 displays no significant DE genes and is therefore not represented. The top 10 to 20 DE genes per category are listed by absolute magnitude of fold-change in expression. (C) Pathway scoring for Nanostring annotated pathways involved in canine immune response. Pathway scores are calculated as the first principal component of the pathway genes’ normalized expression. Heatmap columns represent individual patient soft tissue sarcomas. (D) Z-scored expression data for top DE genes (15 max) in select pathways associated with IL-2/IL-12 activity. Heatmap rows represent individual patients. (E) Normalized expression (log 2) of counter-regulatory genes (one-way ANOVA with Dunnett’s multiple comparisons test; *: p<0.05, **: p<0.01, ***: p<0.001, ****: p<0.0001).

Pathway enrichment analysis confirmed that intratumoral cytokine therapy led to expression of genes associated with several immune functions, specifically in cohorts 1 and 3 that received therapy two days before resection (Figure 5C, Supplementary Figure S6). Probing individual genes in pathways classically associated with IL-2 and IL-12 activity, including antigen processing, cytotoxicity, and immune cell function, we observed consistent upregulation of expression among treated dogs in cohorts 1 and 3 (Figure 5D). Immune pathways (Figure 5C) or individual genes (Figure 5D) upregulated in cohorts 1 and 3 were unamplified in cohort 2 or the untreated control samples.

The lack of persistent expression of these immune effector genes at the late timepoint of 8 days following single-dose intratumoral cytokine therapy observed in cohort 2 dogs was surprising, but consistent with the homeostatic tendency of the immune system to self-limit through regulatory mechanisms. We examined the gene expression data for known checkpoint targets of immune function, including the programmed cell death receptor 1/ligand 1 axis (Pdcd1, Cd274) and cytotoxic T-lymphocyte antigen-4 (Ctla4; Figure 5E). We also examined the immune counter-regulatory enzyme indoleamine 2,3-dioxygenase (IDO1, gene: Ido1), which is interferon-inducible and catabolizes the essential amino acid tryptophan, inhibiting T cell proliferation and promoting CD4+ differentiation to T_reg phenotypes(53-55). Expression of checkpoint molecules increased in response to collagen-anchored cytokine treatment, although none as markedly as IDO1, which was amplified by multiple orders of magnitude 2 days following cytokine treatment. As IDO1 can contribute to tumor escape and immunosuppression(56), this enzyme likely contributed to the return to baseline of immune effector gene expression 8 days after cytokine treatment.

Intratumoral treatment of murine tumors results in similar polarization of tumor microenvironment observed in dogs.

We performed comparative assessment in the B16F10 mouse melanoma model, to test the ability of this syngeneic tumor cell line model to recapitulate the response to therapy observed in canine tumors—particularly, whether IDO1 and other immune checkpoints are also upregulated in response to intratumoral cytokine treatment. This model is considered immunologically “cold”, similar to canine STS tumors(49,57,58). Intratumoral cytokine treatments and tumor collections were performed among murine cohorts at intervals matching those for canine patients (Figure 6A). RNA profiling revealed that treatment of B16F10 with intratumoral collagen-binding cytokines led to significant changes in gene expression across all treated cohorts (Figure 6B). These DEGs were overrepresented across a number of immune function pathways, similar to the responses observed among canine STS patients (Figure 6C; Supplementary Figure S7). Although the immune pathways enriched in dogs and mice were predominantly overlapping, the response kinetics were intriguingly different between the two. While the dog STSs returned to a less-inflamed phenotype after 8 days, the mouse melanomas remained inflamed following the same delay after treatment. Moreover, immune pathway enrichment scores following the second cytokine dose in mice was greater than after the first dose, while the enrichment scores for dog cohorts 1 and 3 remained similar. These differences suggest that spontaneous dog tumors have developed more powerful immunosuppression during the protracted course of tumor development and growth, while rapidly-growing mouse tumors appear to more additively accumulate cytokine treatment effects across consecutive doses.

Figure 6. Nanostring RNA profiling of murine melanoma reveals conserved and distinct responses to cytokine therapy.

(A) Mice were inoculated with 10⁶ B16F10 cells, injected intratumorally (i.t.) with cytokines (yellow arrows), then euthanized for tumor analysis (black arrow) on the schedules shown. mLAIR-MSA-mIL-2 dose was 0.11 nmol (10.8 ug), mIL-12-MSA-mLAIR dose was 14 pmol (2 ug). (B) Volcano plots for murine cohorts 1 (C1), 2 (C2), and 3 (C3). Fold-change is determined relative to untreated control data. Genes associated with significant p-adj values (< 0.05) are highlighted in red and quantified for each cohort. (C) Pathway scoring for Nanostring annotated pathways shared by murine and canine panels (pathway gene sets are not identical, but represent functionally similar sets). (D) Z-scored gene expression for homologous genes to those contained in the human 18-gene Tumor Inflammation Signature(60), which is known to correlated with human clinical response to ICB(59). (E) Comparison of differential expression of shared genes in canine soft tissue sarcoma and murine melanoma tumors, cohort 1 (C1) and cohort 3 (C3). Genes that are significantly differentially expressed (p <0.05) in canine and/or murine cohorts are shown. (F) Venn diagram of all significant differentially expressed (DE) genes shown in (D), grouped by cohort and species. The top 20 DE genes shared by both species and both cohorts (left) and genes that are only significantly differentially expressed in either canine cohorts (right, top) or murine cohorts (right, bottom) are shown. (G) Normalized expression (log 2) of counter-regulatory genes (one-way ANOVA with Dunnett’s multiple comparisons test). (H) Tumor survival for mice inoculated with B16F10 tumors, treated with cytokines ± aPD-1, aCTLA-4, or IDOi. Survival P values were determined by log-rank (Mantel-Cox) test (*: p<0.05, **: p<0.01, ***: p<0.001, ****: p<0.0001).

To explore whether the common inflammatory gene expression responses to cytokine therapy in mice and dogs might correlate with human signatures, we examined an 18-gene Tumor Inflammation Signature (TIS)(59) that has been shown to correlate with PD-1 blockade responses for samples in the TCGA(60), and in the KEYNOTE-028 clinical trial(61). We identified 17 of 18 homologous genes in the dog and mouse Nanostring data. Cohorts 1 and 3 of the dogs and cohort 1, 2, and 3 of the mice exhibit upregulation across the majority of genes in this signature (Figure 6D). This suggests that the inflammatory response to anchored intratumoral cytokine therapy in both the mouse model and the clinical dog patients are consistent with a TME state that in human tumors favors response to checkpoint blockade therapy.

To probe whether any species-specific genes may be responsible for more persistent immune function in the mouse model, we compared DEGs for the two species across the single- and two-dose cohorts (Figure 6E). The majority of genes were similarly upregulated by both mouse and canine models, with a few notable exceptions involving genes associated with cells of myeloid origin. Marco and Arg1 were differentially expressed following therapy only in murine tumors, and have been described as markers of immunosuppressive, M2-like macrophages(62,63). Saa1 is known to recruit and maintain an immunosuppressive neutrophil population and in addition to Ido1, is differentially expressed only in the canine tumors(64). While the species-specific differences in differential gene expression appear to impact the myeloid compartment, a signature of 97 genes was common across dosing cohorts as well as mouse and canine tumor models (Figure 6F). This signature includes IFN-γ inducible genes to control antigen processing and presentation (Psmb8, Psmb10, Irf1, Irf8, Ciita, Nlrc5, Tap1), as well as genes associated with cytotoxic T and NK cells (Gzmb, Gzma, Il2ra, Il2rb, Ncr1), suggestive that the cytokine treatment polarizes the TME in both models towards an antitumor state. However, the differential myeloid responses to treatment may represent species- or cancer-specific processes of tumor resistance towards this polarized state to evade immune destruction.

Blockade of immune counterregulatory processes improves anti-tumor responses to IL-2 and IL-12.

Upregulation of immune counterregulatory genes in canine STS tumors was hypothesized to contribute to the transience of gene expression across antitumor immune pathways. We characterized the same checkpoint genes in the B16F10 model and observed similar trends in the upregulation of Pdcd1, Cd274, and Ctla4 following treatment (Figure 6G). Notably, collagen-binding cytokines did not cause the expression of Ido1 to increase to the same extent in mouse tumors as in canine tumors. This result may contribute to the persistence of antitumor immune cell signatures in mouse cohort 2 that was not seen in dogs.

We next asked whether treatment combinations of cytokines with agents to inhibit immune counterregulation would synergize for better tumor control. Since canine STSs are typically removed surgically, we could not ethically withhold this therapeutic intervention to study long-term responses of these tumors to combination therapy. As such, and for expediency, we used the mouse B16F10 melanoma model to examine survival efficacy. Not only does this model serve as a bridge to the immunologically “cold” canine STS, but also to canine oral melanoma for which we have begun investigation for the same intratumoral cytokine therapy protocol(65).

Mice were treated with monotherapies of collagen-binding IL-2 and IL-12 cytokines, anti-PD1, anti-CTLA4, the IDO1 inhibitor indoximod (1-methyl-D-tryptophan), or combinations of the cytokines with each respective agent (Figure 6H). B16F10 is known to be resistant to anti-PD1 treatment(66,67), which was confirmed in comparison to PBS-treated mice. IDO1 inhibitor alone also provided no survival benefit, consistent with prior reports as a single agent in other mouse tumors(68,69). Monotherapy anti-CTLA4 led to tumor rejection in nearly half of all treated mice, similar to collagen-binding IL-2/IL-12 cytokines alone. Interestingly, the combination of cytokines with anti-CTLA4 or anti-PD1 enhanced survival and led to 90% tumor rejection rates. However, the combination with indoximod provided no additional survival benefit over cytokines alone. Given that the treatment-induced upregulation in gene expression in mice of Ctla4, Pdcd1, and Cd274 was greater in magnitude than Ido1, the observed synergy may manifest from the tumor dependence on that particular counterregulatory mechanism. The significant difference in IDO1 upregulation in cytokine-treated murine vs canine tumors indicates that the mouse model is unlikely to accurately predict effects of IDO inhibition in dogs. By contrast, the murine treatment trials suggest that anti-CTLA4 or anti-PD-1 treatment may provide significant synergy with anchored cytokine immunotherapy in canine STS, and perhaps other tumor types as well.

DISCUSSION

Durable responses to checkpoint blockade(70,71) are observed in only a minority of patients, driving a consensus that rationally-designed combination strategies will be necessary for immunotherapy to overcome treatment resistance mechanisms erected by the tumor microenvironment and provide benefit to the patient majority(72,73). Thus far, potent agonistic immunotherapies such as cytokines have produced unacceptable toxicity due to on-target, off-tumor activation of immune cells(18-20,74). As such, we and others have described a strategy to endow immunostimulatory cytokines with the ability to anchor to tumor collagen, preventing their systemic escape and exposure following intratumoral administration(16,38). This strategy has proven to reduce toxicity in murine tumor models while enhancing efficacy, raising confidence in the clinical translation of this approach.

Advancing beyond murine models, we turn here to canine soft tissue sarcomas as a closer analogy to human tumors for assessment of collagen-binding cytokine safety and activity(12,13,40,49). Canine versions of IL-2 and IL-12 collagen-binding cytokines were recombinantly produced and tested in a cohort of healthy beagle dogs to identify both a maximum tolerated dose and biomarkers of response. Consistent with prior canine testing of a tumor-targeted IL-12 immunocytokine(45), our treatment led to a transient increase in serum IFN-γ levels followed by a delayed increase in serum IL-10. IL-10 is a known mediator of IFN-γ driven effects on the immune system, specifically to prevent autoimmunity coordinated by the helper T cell compartment(75). Overall, the patients treated at or below the established MTD tolerated the treatment with only minor symptoms. The highest tested dose led to symptoms consistent with CRS, likely due to the combined toxicity encountered from IL-2 with IL-12, but also the lack of a dense, tumor-associated collagen in the intradermal tissue space to which these cytokines could bind.

We observed that our cytokine treatment was similarly well-tolerated by all cancer-bearing dogs in the treatment cohorts, with only one patient displaying an elevated serum ALT level. At this dose level, the tumor microenvironment was significantly modified: increased infiltrates of T cells and macrophages were observed through immunohistochemistry in these tumors, and bulk gene expression profiles were consistent with antitumor effector functions of the immune system following treatment. Comparative gene expression in treated mouse tumors demonstrated near-identical pathway enrichment to the canine tumors following intratumoral cytokine therapy, which was previously found to drive enhanced CD8+ T cell infiltration to B16F10 melanoma and potentiated significant tumor control(16). Notably however, the persistence of gene expression associated with immune function in the canine tumors was significantly shorter-lived than in mice, which we hypothesize is due to a concordant increase in expression of immune counterregulatory factors such as PD1 (Pdcd1), PDL1 (Cd274), CTLA4 (Ctla4), and in particular IDO1 (Ido1), whose transcripts were upregulated by two to three orders of magnitude.

A second dose of intratumoral cytokines delayed by one week was sufficient to rescue the gene expression of most of the key immune effector function pathways. Episodic exposure of this kind may help overcome natural homeostatic return to an immunosuppressed state, naturally preventing overstimulation and immune exhaustion(76,77). These counterregulatory processes can be thwarted pharmacologically with rationally chosen combinations that target these known resistance pathways(72,78). While gene expression analysis suggests these regulatory processes promote a return of IL-2/IL-12 treated dogs to the baseline of an untreated, archived STS cohort, one limitation of this work is the lack of paired pre-treatment samples for longitudinal analysis by patient. Immunologically, we had concern that serial tru-cut biopsies could introduce additional inflammation that would confound our interpretation of cytokine-induced changes within the tumor, while introducing practical concerns about leakage of the bolus volume and increasing the potential for patient discomfort or self-mutilation. In future work, we intend to explore methods to enable us to collect multiple tissue timepoints during treatment to better understand the dynamics of response upon modified or combination dosing schemes.

As we have shown, the counterregulatory mechanisms in response to intratumoral IL-2 and IL-12 treatment are not fully concordant between canine soft tissue sarcoma and murine melanoma models. Primarily, Ido1 expression emerged as one of the most differentially-expressed genes in canine tumors following treatment while it was only minimally expressed in the syngeneic mouse B16F10 model. Various cell types, including cancer cells, can express IDO1, but it has been well-documented that myeloid-derived macrophages and dendritic cells upregulate IDO1 in response to IFN-γ, limiting T cell responses as a mechanism of tolerance(79-81). As such, the different Ido1 response in canine and mouse tumors may be due to underlying differences in the pre-treatment cellular composition of these cancers. We observed a significant tumor macrophage population in our canine STS samples while it is known that macrophages and monocyte-derived suppressor cells (MDSC) comprise only a minor fraction of murine B16F10 tumors(57,58). This absence of an IDO1+ tumor macrophage population likely contributes to the lack of therapeutic improvement for B16F10 tumors when combining the IDO1 inhibitor, indoximod, with our collagen-binding cytokines. However, synergy was observed in treatment of murine tumors between the checkpoint inhibitors anti-PD1 and anti-CTLA4 with our cytokine therapy, which do have marked upregulation of their respective genes following IL-2/IL-12 treatment in mice. We conclude that the murine B16F10 model system is simply not an appropriate model for canine STS with respect to IDO1 counterregulatory responses, and that IDO1 inhibition remains of potential interest to be tested in combination with collagen-binding intratumoral IL-2 and IL-12 therapy in canine STS.

These findings illustrate the value in performing comparative study of novel immunotherapies and combinations in spontaneous canine cancers alongside traditional mouse models. As extensively reviewed elsewhere, mouse models of cancer are an invaluable tool to statistically-power investigation of anticancer agents, but do not recapitulate key hallmarks of human cancers(47,82,83). Particularly for the investigation of immunotherapies, most syngeneic and genetically-engineered mouse models do not undergo the continuous immunoediting processes that influence the development and progression of human cancers(84,85). This, amongst other factors, can yield variable efficacy between treatment strategies tested in mice and humans. Notably, the IDO1 inhibitor epacadostat failed to synergize with anti-PD1 therapy in a phase III trial in metastatic melanoma(86,87), despite remarkable combinatorial benefit in some murine models(88). Screening of spontaneous canine melanomas, for example, could yield information about the heterogeneity in basal Ido1 expression in these tumors, enabling prediction and validation of combination efficacy before undertaking expensive human trials. By comparatively testing our intratumoral cytokines in mouse and canine tumor models, we have gained insights into the variable immune counterregulatory responses within tumors which provide context for the selection of treatment combinations. Through greater access to canine tumor models and expanded reagent availability, the broader cancer immunotherapy field could benefit from more predictive assessment of drug safety and activity at human-relevant doses and TME conditions. Arguments for the applicability of both canine and murine studies to human cancer are supported by the strong overlap of gene upregulation across a clinically relevant human Tumor Inflammation Signature(59) for both the dog and mouse responses to immunotherapy (Figure 6D).

Here, we have demonstrated the ability of collagen-binding IL-2 and IL-12 cytokines to safely reprogram the TME of canine soft tissue sarcomas towards an antitumor state. Combining learnings from murine models and canine tumors enables an efficient treatment build-test-learn design cycle which should accelerate the de-risking of clinical immunotherapy development. Building upon prior efficacy demonstrated in murine melanoma(16,17) and the strong intratumoral biomarker responses observed in the present work, we have commenced a trial of intratumoral collagen-binding cytokine therapy in canine oral melanomas(65).

STATEMENT OF TRANSLATIONAL RELEVANCE.

Successful translation of novel cancer therapies could be accelerated through the inclusion of tumor models that accurately recapitulate natural evolution and malignant transformation processes operative in human tumor development. Spontaneous cancer in pet dogs provides an underutilized opportunity to assess the safety and activity of investigational cancer therapies in tumors that arise following years of immunoediting. Particularly for the evaluation of immunotherapies, canine tumors enable the assessment of clinical potential in the context of an experienced, and often senescent, immune background. Beyond efficacy, such evaluation provides meaningful insight into tumor resistance mechanisms that could influence eventual human clinical success. Herein, we characterize immune activities generated by intratumoral injections of engineered collagen-binding cytokines IL-2 and IL-12 into naturally-occurring canine soft tissue sarcomas, and demonstrate through comparative assessment in mouse tumors the differential learnings from each model and their combined role in guiding rational design of treatment combinations with greater expected efficacy.