Targeted delivery of IL2 to the tumor stroma potentiates the action of immune checkpoint inhibitors by preferential activation of NK and CD8⁺ T cells

Abstract

Recombinant human interleukin-2 (IL2) is being considered as a combination partner for immune checkpoint inhibitors in cancer therapy, but the product only has a narrow therapeutic window. Therefore, we used F8-IL2, an antibody-IL2 fusion protein capable of selective localization to the tumor site, in combination with antibodies against murine CTLA-4, PD-1, and PD-L1. In immunocompetent mice bearing CT26 tumors, the combination of F8-IL2 with CTLA-4 blockade was efficacious, leading to increased progression-free survival and protective immunity against subsequent tumor re-challenges. The combination with anti–PD-1 induced substantial tumor growth retardation, but tumor clearance was rare, whereas the combination with anti–PD-L1 exhibited the lowest activity. A detailed high-parametric single-cell analysis of the tumor leukocyte composition revealed that F8-IL2 had a strong impact on NK cell activity without collateral immune activation in the systemic immune compartment, whereas CTLA-4 blockade led to significant changes in the T-cell compartment. Leukocyte depletion studies revealed that CD8⁺ T and NK cells were the main drivers of the therapeutic activity. We extended the experimental observations to a second model, treating MC38 tumor-bearing mice with F8-IL2 and/or CTLA-4 blockade. Only the combination treatment displayed potent anti-cancer activity, characterized by an increase in cytolytic CD8⁺ T and NK cells in tumors and draining lymph nodes. A decrease in the Treg frequency within the tumors was also observed. The results provide a rationale for the combined use of engineered IL2 therapeutics with immune checkpoint inhibitors for cancer therapy.

Introduction

Interleukin 2 (IL2) is important for the activation, differentiation, and expansion of T and NK cells (1). Recombinant human IL2 was the first cancer immunotherapeutic product approved by the FDA, on the basis of durable objective responses observed in a subset of patients with metastatic melanoma or renal cell carcinoma (2,3). However, the systemic administration of high-dose IL2 may cause substantial toxicity with potentially life-threatening side effects, thus, limiting this treatment option to a small group of patients, who are sufficiently fit. The short half-life of IL2 (4) and the absence of tumor targeting specificity (5–7) also limits the therapeutic potential of IL2 in the clinic.

Various strategies have been considered in order to increase the therapeutic index of IL2. For example, Nektar Therapeutics has developed NKTR-214, an IL2 derivative featuring an average of six releasable polyethylene glycol (PEG) chains (4). NKTR-214 has shown promising activity when used in combination with antibodies specific to CTLA-4 or PD-1 (4). An alternative strategy for the enhancement of IL2 activity and specificity relies on the antibody-based delivery of this immunostimulatory payload to the tumor environment (5,6,8,9). Various antibody formats have been considered for fusion protein development and several products have been moved to clinical trials (10–14).

Our laboratory has previously described two fusion proteins (F8-IL2 and L19-IL2), featuring antibodies in diabody format fused to IL2 and expressed in mammalian cells (6,9,15), which recognize the alternatively-spliced EDA and EDB domains of fibronectin, respectively. These extra-domains of 91 amino acids are conserved from mouse to human (16) and are expressed in the majority of aggressive solid tumors and lymphomas while being undetectable in normal tissue, with the exception of the female reproductive system (placenta, endometrium, and some ovarian vessels) (17). Both F8-IL2 and L19-IL2 have shown single-agent activity in various immunocompetent mouse models of cancer (6,9,18,19), and these products can potently synergize with certain cytotoxic agents (15,18), external beam radiation (20), intact immunoglobulins working through ADCC mechanisms (7), as well as other antibody-cytokine fusions (19,21,22). The disease-homing properties of the parental L19 and F8 antibodies have been extensively characterized in mouse models and in patients using nuclear medicine procedures (9,18,23–26).

In this work, we studied the anti-cancer activity of F8-IL2 in combination with antibodies directed against murine PD-1, PD-L1, and CTLA-4 in immunocompetent mice bearing CT26 tumors. The immune checkpoint inhibitors against the PD-1/ PD-L1 (nivolumab, pembrolizumab, atezolizumab, durvalumab, avelumab) or CTLA-4 (ipilimumab, tremelimumab) pathway served as murine surrogates for successful biopharmaceuticals, which are increasingly being used for the treatment of patients with various types of malignancies (27). The activity of F8-IL2 was strongest when used in combination with a CTLA-4 blocker, leading to complete tumor eradication and induction of anti-cancer protective immunity in all treated animals. A detailed multiplex analysis of the leukocyte composition in tumors, spleen, and lymph nodes of mice, which had received F8-IL2 and CTLA-4 blockade (alone or in combination), revealed the critical role played by activated NK cells and CD8⁺ T cells. Therapeutic activity was lost when either CD8⁺ T cells or NK cells were depleted. The results of our study support the use of targeted IL2 in combination with checkpoint inhibitors for cancer therapy.

Materials and Methods

Cell lines, animals, and tumor models

The CT26 colon carcinoma (CRL-2638), the F9 teratocarcinoma (CRL-170), and the CTLL-2 (TIB-214) cell lines were obtained from ATCC, expanded, and stored as cryopreserved aliquots in liquid nitrogen. The MC38 cell line was a kind gift of Prof. Onur Boyman. Cells were grown according to the supplier’s protocol and kept in culture for no longer than 2 weeks. The CT26 colon carcinoma cells were expanded and cultured in RPMI1640 (Gibco) supplemented with 10 % fetal bovine serum (Gibco, 10 % FBS) and the F9 teratocarinoma cell line in DMEM (Gibco) containing 10 % FBS. The MC38 cell line was grown in advanced DMEM (Gibco) with 10 % FBS and 1 % Ultraglutamine (Lonza). CTLL2 cells were cultivated in RPMI 1640 (Gibco) supplemented with 10% FBS, 1 % Ultraglutamine, 25 mM Hepes (Gibco), 0.05 mM β-mercaptoethanol (Sigma Aldrich) and human IL2 (60 Units/mL; Roche Diagnostics). Authentication of the cell lines included check of post-freeze viability, growth properties, morphology, test for mycoplasma contamination, isoenzyme assay, and sterility testing and were performed by the cell bank before shipment.

Seven- to eight-week-old female BALB/c mice were purchased from Charles River (Germany). Eight-week-old female 129/Sv and C57BL/6JRj mice were purchased from Janvier (France). On the day of tumor cell injection, exponentially growing F9 teratocarinoma, CT26 colon carcinoma, or MC38 cells were harvested, repeatedly washed, and resuspended in saline prior to injection. CT26 colon carcinoma cells (2x 10⁶ cells per mouse, using BALB/c mice), MC38 cells (1x10⁶ cells per mouse, using C57BL/6JRj) or F9 teratocarinoma cells (12x10⁷ cells per mouse, using 129/Sv mice) were implanted subcutaneously in the right flank. Mice were monitored daily and tumor volume was measured with a caliper. Tumor volume was calculated as follows: (length [mm] x width [mm] x width [mm])/2. Animals were euthanized when tumor volumes reached a maximum of 2000 mm³, weight loss exceeded 15 %, tumors were ulcerated, or 24 hours after the last injection for the biodistribution and the multiplex flow cytometry analysis (on day 13 after tumor cell injection). Tumors, draining lymph nodes, and spleen were harvested for flow cytometry analysis. To assess the biodistribution of the fusion protein, tumor, liver, lung, spleen, heart, kidney, intestine, and blood was collected.

All experiments were performed under a project license granted by the Veterinäramt des Kanton Zürichs, Switzerland (27/2015) in agreement with Swiss regulations. Animals were maintained in pathogen-free facilities at the Swiss Federal Institute of Technology (ETH Zurich) and procedures were approved by the ETH Zurich Institutional Animal Care and Use Committee.

Antibodies for therapy experiments

The F8-IL2 immunocytokine was produced as previously described (9). In brief, a stable CHO-S cell line (Invitrogen) was incubated for 6 days in a shaking incubator at 31°C. F8-IL2 was purified from the culture medium by a protein A affinity column (Protein A agarose beads, Sino Biologicals Inc.). The commercial anti–PD-1 (clone 29F.1A12), anti–PD-L1 (clone 10F.9G2), anti–CTLA-4 (clone 9D9) were purchased from BioXCell. The products had previously been extensively studied in immunocompetent models of cancer (28,29). Rat anti-CD4 (clone GK1.5, BioXCell), rat anti-CD8 (clone YTS169.4, BioXCell) and rabbit anti-Asialo GM1 (Wako Chemicals) antibodies were used for in vivo depletion, as described below.

In vitro characterization of F8-IL2

F8-IL2 was analyzed using SDS-PAGE in non-reducing and reducing conditions, size exclusion chromatography (Superdex 200 Increase, 10/300 GL, GE Healthcare), and surface plasmon resonance analysis (BIAcore S200, GE Healthcare) on an EDA antigen-coated CM5 BIAcore sensor chip (GE Healthcare). The biological activity of F8-IL2 and IL2 (Proleukin, Roche) was determined as described before (9). Briefly, 10x10⁶cultured CTLL-2 cells were starved in CTLL2 culture medium (RPMI 1640 (Gibco) supplemented with 10% FBS, 1 % Ultraglutamine, 25 mM Hepes (Gibco), 0.05 mM β-mercaptoethanol (Sigma Aldrich)) without IL2 for 24 hours. Starved CTLL-2 cells (2x 10⁴/well) were seeded in 96-well plates in CTLL2 culture medium containing varying concentrations (5x10^-12 – 10^-9 M IL2 equivalents) of IL2 equivalents (F8-IL2 or as positive control purchased human IL2 (Roche Diagnostics) in triplicates. After 48 hours at 37°C, cell proliferation was determined with the Cell Titer 96^® Aqueous One Solution (Promega) according to the manufacturer’s instructions by measuring the OD at 490 nm and 620 nm. Percent proliferation was calculated as follows: % proliferation = (OD_490-62^treated-OD_490-620^medium)/(OD_490-62^untreated-OD_490-620^medium) x 100 %.

Quantitative biodistribution study

Quantitative biodistribution was used to assess the in vivo targeting performance of F8-IL2 (n = 4) as described previously (26). Purified F8-IL2 was radiolabeled with iodine 125 (Perkin Elmer) using the Iodogen method. Immunocompetent 129/Sv mice (n = 3) were injected with 12x 10⁷ F9 teratocarcinoma cells subcutaneously into the right flank. Mice were monitored daily, and tumor volume was measured with a caliper. Tumor volume was calculated as follows: (length [mm] x width [mm] x width [mm])/2. When the tumors reach approx. 350 mm³, 0.86 μg of radio-iodinated F8-IL2 was injected intravenously into the lateral tail veins. After 24 hours, mice were sacrificed, tumor, liver, lung, spleen, heart, kidney, intestine and blood collected, weighed, and the radioactivity was counted using a Cobra γ counter (Packard, Meriden, CT, USA). The radioactivity of tumors and organs was expressed as percentage of injected dose per gram of tissue (% ID/ g ± SD).

Therapy study and in vivo depletion of NK cells and CD4⁺ and CD8⁺ T cells

On the day of tumor cell injection, exponentially growing CT26 colon carcinoma or MC38 cells were harvested, repeatedly washed, and resuspended in saline prior to injection. CT26 colon carcinoma cells were implanted subcutaneously in the right flank of BALB/c mice using 2x 10⁶ cells per mouse. MC38 cells were implanted subcutaneously in the right flank of C57BL/6JRj mice using 1x 10⁶ cells per mouse. Mice were monitored daily and tumor volume was measured with a caliper. Tumor volume was calculated as described before. When tumors reached a suitable volume (approx. 80 mm³), mice were randomly divided into different treatment groups and injected into the lateral tail vein. Mice received three injections of 30 or 45 μg F8-IL2 or 200 μg of a checkpoint inhibitor (anti–PD-1, anti–PD-L1, or anti-CTLA-4), phosphate-buffered saline, or the combination of F8-IL2 and a checkpoint inhibitor at intervals of 72 hours. In the combination group, mice received F8-IL2 followed by a checkpoint inhibitor after 24 hours. For the in vivo depletion of NK, CD4⁺, and CD8⁺ cells, CT26 colon carcinoma-bearing mice (n= 5 per group) were repeatedly injected intraperitoneally with 30 μL anti-Asialo GM1 (day 2, 5, and 8 after tumor implantation), 250 μg anti-CD4 or 250 μg anti-CD8 antibodies (day 2, 5, 8 and 11 after tumor implantation). An additional group (n = 5) was injected with an anti-CD4 on day -1, 2, 4, and 8 after tumor injection. A saline group (n = 5) and a treatment group (n = 5) without depletion were included as controls. Animals were euthanized when tumors reached a maximum of 2000 mm³. Flow cytometry of the spleen was used to verify the successful depletion of the specific leukocyte population.

Tumor re-challenge

Mice that cleared tumors were injected subcutaneously with 2x 10⁶ CT26 colon carcinoma or 1x 10⁶ MC38 cells on day 62 and day 60, respectively, after the first injection of tumor cells. As controls, naïve BALB/c mice (n = 5) and naïve C57BL/6JRj (n= 6) were injected with the same tumor cells to monitor tumor growth and cell viability after injection.

Immunohistochemical analysis of EDA expression in CT26 and MC38 tumors

CT26 and MC38 tumors were excised and immediately embedded in frozen section medium (Thermo Scientific). Staining was performed on 10 mm cryosections fixed in ice-cold acetone. Primary antibodies in small immunoprotein format (F8 and KSF, kindly provided by Dr. Rémy Gebleux) were detected with a rabbit anti-human IgE (1:1000, Dako Agilent) and in a second step with Alexa Fluor 488–coupled anti-rabbit (1: 200, Invitrogen). An anti-CD31 (1:100, MEC 13.3, BD Biosciences) was detected with an Alexa Fluor 594–coupled anti-rat antibody (1:200, Invitrogen). Sections were counterstained with DAPI (SigmaAldrich) and mounted with fluorescent mounting medium (Dako Agilent). Slides were then analyzed with an Axioskop2 mot plus microscope (with a 20x/0.5 objective lense, Zeiss) and documented with an AxioCam color camera (Zeiss), using the AxioVision software (4.7.2. Release, Zeiss).

Flow cytometry

For a multiparameter flow cytometry analysis, mice were sacrificed and tumor, spleen, and draining inguinal lymph nodes were excised 24 hours after the last injection. Tumors were transferred into gentleMACS C tubes (Miltenyi) in RPMI and digested with DNase (30 μL/tumor, Sigma) and liberase (60 μL/tumor, Roche) for 40 minutes at 37°C using the gentleMACS Dissociator (Miltenyi) according the manufacturer’s instruction (program 37C_m_TDK_1). After centrifugation, the pellet was resuspended in 40% Percoll (GE Healthcare) and the solution was filtered using a 100 μm cell strainer (Greiner-bio-one). 75% Percoll (10 mL for each 10 mL cell suspension) was slowly poured onto the 40% Percoll 40 cell suspension. After Percoll gradient centrifugation, the buffy coat was collected. Cells were washed twice and stained (Supplementary Table S1).

Spleen were digested as described for the tumors (using program 37C_m_SDK_1). After the digestion, the erythrocytes were lysed using RBC lysis buffer (BioLegend TM). The cells were spun down and resuspended in PBS to yield a concentration of 4x10⁶cells/mL. Single-cell suspensions were directly used for flow cytometry staining (Supplementary Table S1).

Lymph nodes were smashed using the plunger of a 1 mL syringe and filtered into a polystyrene tube with a cell strainer cap (StemCell Technologies). After centrifugation, cell pellets were resuspended in RPMI complemented with 10% FBS. Non-specific binding was blocked using anti-CD32/CD16 (BioLegend TM) followed by the flow cytometry antibody panel described in Supplementary Table S1. For intracellular staining, cells were fixed and permeabilized with fixation/ permeabilization solution (Thermo Scientific), according to the manufacturer’s instructions. Cells were acquired on a Symphony flow cytometer (BD Biosciences). Data was analyzed using FlowJo (version 10.0.8, TreeStar Inc). For an unbiased analysis, we reduced the high-dimensional dataset into two dimensions through t-distributed stochastic neighbor embedding (t-SNE) in combination with FlowSOM metaclustering as described (30).

Statistical analysis of murine tumor, lymph node, and spleen datasets

Data were analyzed using Prism 6.0 (GraphPad Software, Inc.). Statistical significance of in vivo experiments was determined with a regular two-way ANOVA test with the Bonferroni post-test. We utilized unsupervised validated clustering approaches (FlowSOM, CellCNN, Citrus) to discern between different cell populations. For FlowSOM metaclustering, flow cytometer data were compensated, exported with FlowJo software (version 10.0.8, TreeStar Inc.) and normalized using Cyt MATLAB (version 2017b). An unbiased analysis was performed as described (30).

Results

Product characterization

The fusion protein composed of the F8 antibody specific to the extra-domain A of fibronectin and human IL2 in diabody format (F8-IL2) was produced as previously described by our group (9) and purified to homogeneity. Fig. 1A shows a schematic illustration of the non-covalent homodimer, formed by the antibody-IL2 fusion in diabody format. The product was homogeneous and had the estimated molecular size of approximately 42 kDa in SDS-PAGE and gel-filtration analysis [Fig. 1B,C], bound avidly to the cognate antigen EDA of fibronectin [Fig. 1D], and retained the same in vitro activity as commercial human recombinant IL2 in a proliferation assay of murine T lymphocytes [Fig. 1E]. A radio-iodinated preparation of F8-IL2 selectively localized in solid tumor lesions of F9 teratocarcinomas compared with liver, lung, spleen, kidney, heart, intestine, and blood 24 hours after intravenous administration. At that time point, a mean of 9.3 ± 2.9 percent injected dose per gram of tumor (% ID/g) was observed, with a tumor:blood ratio of 42.5 [Fig. 1F]. The F8 antibody stained the stroma of CT26 and MC38 tumors, whereas the KSF antibody (directed against hen egg lysozyme and serving as negative control) did not exhibit a specific staining. CD31 was used as the target for staining blood vessels to have a better understanding of the tumor structure [Supplementary Fig. S1A,B]. Collectively, our findings showed that the F8-IL2 construct retained the activity of recombinant IL2, bound with high affinity to the cognate EDA antigen, and selectively localized to tumors in vivo.

Figure 1. In vitro and in vivo characterization of F8-IL2.

(A) Schematic domain assembly of the F8-IL2 in diabody format. (B) SDS-PAGE analysis of purified F8-IL2 under nonreducing (NR) and reducing (R) conditions. M = Marker. (C) Size-exclusion chromatography profile. (D) BIAcore analysis of F8-IL2 on EDA-coated chip. (E) CTLL2 cell proliferation assay. The biological activity of the IL2 moiety was assessed by its capability to stimulate proliferation of CTLL2 cells. Starved CTLL2 cells were seeded in triplicates in 96-well plates in culture medium supplemented with different concentrations of F8-IL2 or recombinant IL2. Shown is the mean ± SEM of triplicates. (F) Biodistribution study of F8-IL2 in F9 teratocarcioma bearing mice. Mice were injected with radio-iodinated F8-IL2 (0.86 μg), sacrificed after 24 hours, organs excised, and the radioactivity counted. The radioactivity of each organ is expressed as injected dose per gram of tissue (% ID/ g ± SE).

Therapy experiments

For therapy studies, we used immunocompetent mice bearing subcutaneous CT26 carcinomas. These tumors had previously been reported to respond to immune checkpoint inhibitors (directed against murine PD-1, PD-L1, and CTLA-4) when used as monotherapy (28,31). Mice bearing subcutaneously grafted CT26 tumors were treated when lesions had reached approximately 80 mm³. F8-IL2 was administered on days 5, 8, and 11 at a dose of 45 μg, whereas immune checkpoint inhibitors were given on days 6, 9, and 12 at a dose of 200 μg [Fig. 2]. F8-IL2 displayed a potent tumor growth inhibition compared to the saline control, but complete remission was observed only in a small proportion of treated mice. Similarly, antibodies directed against CTLA-4, PD-1, and PD-L1 had single-agent activity, but also only rarely led to complete tumor control [Fig. 2]. The combination of F8-IL2 with CTLA-4 blockade led to complete and durable responses in all treated animals. A potent activity was also observed for the combination with anti–PD-1 antibodies, even though most tumors eventually regrew. The lowest therapeutic activity was observed for the PD-L1 combination. All treatments were very well tolerated, as evidenced by the absence of body weight loss [Fig. 2].

Figure 2. Therapeutic performance of F8-IL2 and checkpoint inhibitors in immunocompetent BALB/c mice bearing subcutaneous CT26 colon carcinoma.

Mice received intravenous injections of either 45 μg F8-IL2 (blue squares), 200 μg of a checkpoint inhibitor (anti–CTLA-4, anti–PD-1 or anti–PD-L1, red triangles), 45 μg F8-IL2 followed by 200 μg of a checkpoint inhibitor (pink x) or PBS (black circles). The left column depicts tumor volume versus time. The right column shows the corresponding body weight changes over time. Data represent tumor volume ± SEM or body weight change (%) ± SEM. n = 5 mice per group. CR = complete response. Arrows indicate the day of injection of F8-IL2 (blue arrows) and checkpoint inhibitors (red arrows). Colored stars indicate days on which a mouse of the corresponding group with the same color code had to be euthanized according to the termination criteria. A regular two-way ANOVA test with the Bonferroni post-test was performed to determine the statistical significance between the groups (*, p<0.05, **, p<0.01; ****, p<0.0001). For the CTLA-4 study on day 14: PBS vs each treatment group p = ***, F8-IL2 vs CTLA-4 non-significant, F8-IL2 vs Combo p= ***, Combo vs CTLA-4 p = ***. For the PD-1 study on day 19: PBS vs each treatment group p= ***, F8-IL2 vs PD-1 non-significant, F8-IL2 vs Combo p=*, PD-1 vs Combo non-significant. For the PD-L1 study on day 19: PBS vs each treatment group p=***, the difference between the treatment groups is non-significant (p> 0.05).

To determine whether the combination of F8-IL2 in combination with CTLA-4 blockade would also be effective in a second tumor model, we treated immunocompetent mice bearing MC38 colon carcinomas. This model had previously been reported to be associated with an immunosuppressive phenotype and does not respond well to immune checkpoint inhibitors used as a monotherapy (28). Mice bearing subcutaneously grafted MC38 colon carcinomas were treated when lesions had reached approximately 75 mm³. F8-IL2 was administered on days 5, 8, and 11 at a dose of 30 μg, and anti–CTLA-4 was given on days 6, 9, and 12 at a dose of 200 μg. The combination of CTLA-4 blockade with F8-IL2 substantially reduced the tumor growth, whereas the individual agents were significantly less potent (p<0.0001) [Fig. 3]. Mice without measurable tumor masses at day 62 or day 60 after the tumor implantation were re-challenged with CT26 or MC38 cells. In all cases, the tumors did not grow back [Supplementary Fig. S2], indicative of protective anti-cancer immunity and immune memory, presumably directed against the AH1 antigen (31–35).

Figure 3. Therapeutic activity of the F8-IL2/CTLA-4 blockade combination against subcutaneous MC38 colon carcinoma.

Mice were challenged with 1x10⁶ MC38 colon carcinoma cells, and treatment was started when tumors reached a size of approximately 75 mm³. Mice received intravenous injection of either F8-IL2 (45 μg; blue squares, on day 5, 8, and 11 after tumor cell implantation), an anti–CTLA-4 (200 μg; red triangles, on day 6, 9, and 12), or F8-IL2 (30 μg) followed 24 hours later by the checkpoint inhibitor (200 μg; pink x, on day 5, 8, and 11) or PBS (black circles, on day 5, 6, 8, 9, 11, and 12). The left graph depicts tumor volume versus time, and the right graph displays the body weight change versus time. Data represent tumor volume or body weight change ± SEM. n = 4 mice per group. **, p<0.01; ****, p<0.0001 (regular two-way ANOVA with the Bonferroni post-test).

Mechanistic studies

To gain mechanistic insights on the anti-cancer activity played by different types of leukocytes, we performed a multiparameter flow-cytometric characterization of leukocytes using a panel of 23 fluorochromes [Supplementary Table S1]. We analyzed the leukocyte populations present in tumors, spleens, and draining lymph nodes in CT26 tumor–bearing mice at day 13 after tumor implantation (i.e., 24 hours after the last injection) [Figs. 4-5, Supplementary Figs. S3-S4]. Fig. 4 shows the phenotype of the tumor-infiltrating leukocytes from all treated groups, defined as saline (PBS), F8-IL2, anti–CTLA-4 (aCTLA-4), and combination (Combo) treatment. For an unbiased analysis, we reduced the high-dimensional dataset into two dimensions through t-distributed stochastic neighbor embedding (t-SNE) that stochastically displayed CD45⁺ cells from the 4 groups [Fig. 4, Supplementary Fig. S3A] in combination with FlowSOM meta-clustering. Heatmaps show the normalized marker expression of each population (black and white) or each condition (red) in the tumor [Fig. 4]. We identified 7 main clusters with this approach [Fig. 4A]. An increased density (cell number/ mg tumor) of myeloid, NK cells, and CD4⁺ T cells in tumors from mice that received F8-IL2 or anti–CTLA-4 (alone or in combination) was observed [Fig. 4A]. All these treatments also doubled the density of CD8⁺ T cells and increased Tregs approximately three-fold [Fig. 4A]. Differences in CTLA-4, KLRG1, PD-1, and Ki67 positivity in Tregs were observed in the Combo group compared with PBS-treated mice [Fig. 4B, Supplementary Fig. S3B]. A finer analysis of T-cell marker expression in different treatment groups revealed that CTLA-4 blockade (alone or in combination) led to increased PD-1 positivity in CD4⁺ T cells [Fig. 4B]. In those cells, CD25 was upregulated in all treatment groups compared to saline [Supplementary Fig. S3B]. Analysis of CD8⁺ T cells showed that CTLA-4 treatment (alone or in combination) was the most efficient in triggering the expression for TIM-3 [Fig. 4B]. Unbiased analysis of the CD8⁺ T-cell compartment (using FlowSOM and Citrus) displayed a lower frequency of PD1⁺TIM3^–Ki67^–CD8⁺ T cells in mice treated with CTLA-4 blockade (alone or in combination with F8-IL2) [Fig. 4C, Supplementary Fig. S3C]. However, in those mice, the frequency and the expression for PD-1, TIM-3, and Ki67 were increased, suggesting local T-cell activation and expansion [Fig. 4C, Supplementary Fig. S3C]. Further analysis (using FlowSOM and Citrus) of the NK cell compartment showed an increased frequency and expression for KLRG1, CD11b, and EOMES in the PBS-treated group, corresponding to a more mature phenotype and a slower in vivo turnover rate (36) [Fig. 4D,E, Supplementary Fig. S3D].

Figure 4. Multiparameter flow cytometric analysis of tumor immune cell infiltrates.

Single-cell suspensions from CT26 tumors (PBS n=3, F8-IL2 n=3, CTLA-4 n=4, Combo n=4 tumors per group) were stained with fluorochrome-conjugated antibodies. (A) t-SNE map showing the FlowSOM-guided meta-clustering (left) and quantification (cell number/ mg tumor) of live intratumoral CD45⁺ cell clusters in the different treatment groups (right). Each color represents a meta-cluster and is associated with a different immune population (left). (B) t-SNE map showing the FlowSOM-guided meta-clustering gated on TCRβ⁺ cells from live CD45⁺ cells (left), heatmap showing the median marker expression (value range: 0-1) for each defined population (middle; black and white), and among each T-cell subpopulation in the different conditions (right; red). (C) t-SNE map showing the FlowSOM-guided meta-clustering gated on CD8⁺ T cells from live CD45⁺TCRβ⁺ cells (left), heatmap showing the median marker expression for each defined population (middle; black and white), and frequencies of the four CD8⁺ T-cell subclusters among total CD8⁺ T cells within the different conditions (right). (D) t-SNE map showing the FlowSOM-guided meta-clustering gated on NK cells from live CD45⁺TCRβ^–NKp46⁺ cells (left), heatmap showing the median marker expression (value range: 0-1) for each defined population (middle), and frequencies of the three NK cell subclusters among total NK cells within the different conditions. (E) Median expression of selected cell markers shown for all intratumoral NK cells for each treatment. The experiment was performed once. Shown is the mean ± SEM. The Mann Whitney test was used to assess the statistical significance. *, p<0.05; **, p<0.01; ***, p<0.001

Figure 5. Multiparameter flow cytometric analysis of lymph node immune cell infiltrates.

Single-cell suspensions from CT26 tumor draining lymph nodes (PBS n=3, F8-IL2 n=4, CTLA-4 n=4, Combo n=4 draining lymph nodes per group) were stained with fluorochrome-conjugated antibodies. (A) t-SNE map showing the FlowSOM-guided meta-clustering and heatmap showing the median marker expression (value range: 0-1, black and white) for each defined population (left). Quantification (cell number) of TCRβ⁺ cell and heatmap showing the median marker expression in the different treated groups (right). Each color (left) represents a meta-cluster and is associated with a different immune population. Gating was performed on live CD45⁺TCRβ⁺ cells. (B) t-SNE map showing the FlowSOM-guided meta-clustering gated on CD8⁺ T cells from live CD45⁺TCRβ⁺ cells (left), heatmap showing the median marker expression (value range: 0-1) for each defined population (middle) and frequencies of the four CD8⁺ T-cell subclusters among total CD8 T cells within the different conditions. (C) t-SNE map showing the FlowSOM-guided meta-clustering gated on NK cells (live CD45⁺TCRβ^–NKp46⁺ cells, left), heatmap showing the median marker expression for each defined population (middle, black and white), and frequencies of the four NK cell subclusters among total NK cells within the different conditions (bottom). (D) Median expression of selected cell markers shown for all NK cells for each treatment. (E) Splenocytes (PBS n=3, F8-IL2 n=4, CTLA-4 n=4, Combo n=4 spleen per group) from mice bearing CT26 tumors were stained with fluorochrome-conjugated antibodies. Median expression of selected cell markers is shown for all NK cells for each treatment.

The experiment was performed once. Shown is the mean ± SEM. The Mann Whitney test was used to determine the statistics. *, p<0.05; **, p<0.01; ***, p<0.001.

Similarly, a detailed analysis of T cells and NK cells in tumor-draining lymph nodes was also performed based on t-SNE (displaying randomly cells from the 4 groups) in combination with FLowSOM meta-clustering [Fig. 5, Supplementary Fig. S4A-E]. The differences in the NK cell population were the most significant. We observed an increase in the frequency of NK cells in all treated groups compared with saline-treated mice [Fig. 5A]. As expected, the majority of CD8⁺ and CD4⁺ T cells had a naïve phenotype [Fig. 5B, Supplementary Fig. S4C,D]. However, the percentage of CD8⁺ effector-memory T cells was higher in mice treated with F8-IL2 (alone or in combination) although not significant [Fig. 5B, Supplementary Fig. S4C,D]. NK cells had an activated phenotype in mice treated with F8-IL2 (alone or in combination), as revealed by their CD11b⁺CD27⁺ status [Fig. 5C, Supplementary Fig. S4E]. Treatment with F8-IL2 (alone or in combination) led to an increase in proliferation (Ki67 positivity) [Fig. 5D, Supplementary Fig. S4E]. By contrast, NK cells in the anti–CTLA-4 monotherapy and saline groups exhibited a similar phenotype [Fig. 5C,D]. For the splenocytes, we observed similar results to what was detected in the lymph nodes [Fig. 5E, Supplementary Fig. S4F,G]. Mice treated with F8-IL2 (alone or in combination) showed increased Ki67, CD11b, and KLRG1 staining of NK cells, whereas the NK cells in the PBS- and CTLA-4–treated mice behaved similarly [Fig. 5E].

We also performed a multiparameter flow-cytometric characterization of leukocytes present in tumors and draining lymph nodes of MC38 bearing mice [Supplementary Figs. S5A-D, S6A-D]. Because the combination was very effective and almost led to a total regression in MC38 tumors, we could obtain only a few tumors for the flow-cytometric analysis. An increase of granzyme B (GZMB+) and Ki67 staining within the CD8⁺ T and NK cell population was observed after F8-IL2 treatment in the tumors and draining lymph nodes [Supplementary Figs. S5B-D, S6C,D]. We observed a similar high frequency of GZMB⁺CD8⁺ T and GZMB⁺ NK cells in the tumors of the combination group compared to F8-IL2 alone, but a decrease in the Treg frequency [Supplementary Fig. S5A-D]. In the draining lymph nodes, the frequency of Tregs was the highest when the mice were treated with F8-IL2 in combination with CTLA-4 blockade [Supplementary Fig. S6B]. Leukocytes present in the neoplastic masses and lymph nodes after CTLA-4 blockade displayed a phenotype similar to the one of leukocytes from saline-treated mice [Supplementary Figs. S5A-D, S6A-D].

Single-cell analysis of the leukocyte composition within the tumors and secondary lymphoid organs among the treatment groups suggested that alterations in the CD8⁺ and CD4⁺ T- and NK cell frequencies and phenotypes may influence the therapeutic outcome. To determine the relevant therapeutic cellular target, we depleted either CD4⁺, CD8⁺, or NK cells during therapy in CT26 colon carcinoma bearing mice. The in vivo depletion of either NK or CD8⁺ T cells completely abolished the therapeutic activity of F8-IL2 in combination with CTLA-4 blockade (CR in both groups 0/5) [Fig. 6]. CD4⁺ T-cell depletion before (CR 3/5) or after (CR 2/5) tumor implantation only had a minor influence on the therapeutic result.

Figure 6. In vivo depletion study.

CT26 colon carcinoma bearing mice received a total of three injections of F8-IL2 (45 μg) followed by anti–CTLA-4 (200 μg) 24 hours later. Depletion antibodies were injected intraperitoneally on days -1, 2, 5, 8, or 11 (black arrowheads). CD8⁺ T cell–depleted mice (250 μg) are depicted in brown triangles, NK cell–depleted mice in blue squares, and CD4⁺ T cell–depleted in purple circles. A group in which CD4⁺ T cells were depleted before tumor implantation is shown in red circles. A saline-treated (no depletion) negative control was included (black circles), as well as an undepleted positive control group that received F8-IL2 and anti CTLA-4 inhibitor (pink x). Black arrows indicate the day of injection of the therapeutic agents. Data represent mean tumor volume ± SEM. n = 5 mice per group (for NK cell depleted group, n = 4 from day 17 as indicated by a blue star due to the termination criteria). A regular two-way ANOVA test with the Bonferroni post-test was performed to determine the statistical significance between the groups (ns, non-significant p>0.05, *, p<0.05, **, p<0.01; ****, p<0.0001).

Discussion

In this study, we explored the therapeutic potential of immune checkpoint inhibitors in combination with F8-IL2, a tumor-targeting antibody-IL2 fusion protein. The combination of F8-IL2 with CTLA-4 and PD-1 blockade was extremely potent in immunocompetent mice bearing subcutaneous CT26 tumors, whereas the combination with PD-L1 blockade was less efficacious. The results may depend, in part, on the choice of antibodies that were used for the study. We based our investigations (i.e., antibody clone and dose) on a previous report by MedImmune, which extensively explored the performance and mechanism of action of 9D9 anti–CTLA-4 and 10F.9G2 anti–PD-L1 in various immunocompetent mouse models of cancer (28).

Recombinant IL2 has been studied in clinical trials in combination with ipilimumab for the treatment of metastatic melanoma with encouraging results (clinicaltrial.gov identifier: NCT00058279, NCT01856023) (37,38). NKTR-214 (a pro-drug version of IL2, that regains activity over time upon loss of polyethyleneglycol chains) (4) has shown promising preclinical and clinical results in combination with immune checkpoint inhibitors, thus, stimulating additional investigations with engineered forms of IL2. In this context, antibody-IL2 fusion proteins may be particularly attractive, as a selective localization of the cytokine payload at the site of disease leads to a considerable increase in therapeutic index in immunocompetent animals (6,18,39). IL2-based products may be ideally suited for combination with immune checkpoint inhibitors because IL2 can increase the activity not only of T cells, but also of NK cells.

All mice treated with a combination of F8-IL2 and CTLA-4 blockade exhibited a complete tumor remission and rejected subsequent challenges with CT26 tumors. A multiparameter FACS analysis of leukocytes in tumors, draining lymph nodes, and spleen provided insights on the relative contribution of individual therapeutic agents. The most significant effect of F8-IL2 (alone or in combination) was on the intratumoral density of NK cells. These leukocytes displayed an activated and more immature phenotype, as revealed by their low KLRG1, Eomes, and CD11b expression. CTLA-4 blockade (alone or in combination) was associated with a characteristic increase of CD8⁺ T cells in the tumor mass. These lymphocytes were PD1⁺ and TIM-3⁺ (revealing an exhausted phenotype), but also stained for Ki67 (indicating a proliferative potential). In the CD4⁺ T-cell population, CTLA-4 blockade (alone or in combination) led to PD-1 upregulation. Most probably, a fine balance between activation and inhibition signals determines the antitumor activity of these lymphocytes. The features of lymphocytes in secondary lymphoid organs were substantially different compared to tumor-infiltrating lymphocytes (TILs), in keeping with recent reports published with other therapeutic modalities (34,40). In line with the targeting result, the most significant effects of F8-IL2 on the leukocytes were seen in the TILs rather than the draining lymph nodes and spleen.

Single-cell analysis of the leukocyte composition within the tumor and secondary lymphoid organs between the treatment groups suggested that changes in the CD4⁺ and CD8⁺ T- and NK cell frequencies and phenotypes may influence the therapeutic outcome. The initial tumor regression was caused by both CD8⁺ T cells and NK cells, as evidenced by leukocyte depletion experiments. CD4⁺ T cells did not play a crucial role. Similar findings, showing an absolute requirement for CD8⁺ T-cell and NK cell activity have previously been reported for other antibody-cytokine fusions (including products based on IL12, IL4, TNF, IL15, GM-CSF) in various immunocompetent mouse models of cancer (15,18,20,21,35).

CT26 carcinomas have often been considered to be immunologically “hot tumors”, whereas other models (such as 4T1, B16F10, LL/2, and MC38) are characterized by a sparse T-cell infiltration and, thus, associated with a “cold” phenotype (28). The combination treatment with F8-IL2 and a checkpoint inhibitor against CTLA-4 was also efficacious in the MC38 model, whereas the individual agents displayed little activity when used in monotherapy. The combination treatment was associated with a rich leukocyte infiltrate in the tumor, with enhanced functional properties (e.g., cytolytic activity, as revealed by GZMB staining).

IL2-based therapeutics are complicated in cancer therapy because the antineoplastic activity could, in principle, be potentially mitigated by the potentiation of Treg activity (41). Indeed, recombinant human IL2 at ultra-low dose has shown encouraging effects for the treatment of host-versus-graft disease (42). The therapeutic activity of IL2 is likely to be context- and dose-dependent. In our setting, thanks to the antibody-based delivery to the tumor, the density of Tregs did not substantially vary between the treatment group in the CT26 colon carcinoma model. Similar observations have been reported in the bone marrow of patients with acute myeloid leukemia, treated with the F16-IL2 fusion protein directed against the alternatively-spliced A1 domain of tenascin-C (43).

We found that the treatment with the anti-CTLA-4 did not lead to the depletion of intratumoral Treg cells in the CT26 colon carcinoma model. Most probably, this is caused by the tumor model itself and the antibody that was chosen. Sergio Quezada and colleagues had previously shown that other anti-CTLA-4 antibodies (e.g., 9H10, IgG1SDALIE a mutated antibody with high binding affinity to CD16a) may be more effective than 4F10 and 9D9 in eradicating Tregs via an antibody-dependent cell-cytotoxicity mechanism (44,45). In general, IgG2a isotypes in the mouse are thought to be better depleting antibodies than IgG2b isotypes (such as 9D9)(46). Low levels of intratumoral antigen-presenting cells displaying cytotoxic Fc receptors may contribute to the poor depletion activity observed. We chose to use the 9D9 clone for our study because the impact of this antibody in many immunocompetent mouse models of cancer has been described in detail (28).

Immune checkpoint inhibitors have become foundational drugs for cancer therapy (27). In the clinic, PD-1 blockade appears to be more active and better tolerated than CTLA-4 blockade, possibly because approved products (nivolumab, pembrolizumab) may interfere with the negative regulation of T cells at the site of disease, rather than in secondary lymphoid organs (47). In the mouse, however, CTLA-4 blockade has often shown more potent anti-cancer activity compared to PD-1 and PD-L1 blockers (19,28,29).

Various antibody-IL2 fusions are currently being investigated in clinical trials (14). In addition to F8-IL2, we contributed to the development of L19-IL2 (6,7,22). Although the F8 antibody recognizes the alternatively-spliced EDA domain of fibronectin, L19 is specific to the EDB domain of fibronectin (48). L19-IL2 has been used for the treatment of patients with various types of malignancies both as monotherapy (49,50) and in combination with other modalities (13). The product is currently being studied in Phase III clinical trials in combination with L19-TNF for the treatment of patients with fully resectable stage IIIB,C melanoma (clinicaltrial.gov identifier NCT02938299) (13). Other companies have preferred to use IL2 fusions based on antibodies in full IgG format (51–53), sometimes with mutations on the IL2 moiety that reduce the affinity to the CD25 receptor (8).

The results of our study provide a rationale for the combined use of engineered IL2 therapeutics with immune checkpoint inhibitors for cancer therapy. The emerging notion that tumors, which do not respond to immune checkpoint inhibitors, still contain a large number of tumor-rejection antigens (e.g., neoepitopes) and of cognate T cells (54,55) provides a motivation to expand and stimulate these T-cell populations with experimental therapeutics.