Female C57BL/6 (B6) mice aged 6–8 weeks were from the National Cancer Institute/Charles River Program (Wilmington, Massachusetts, USA). MDA5 KO, TLR3 KO and mice expressing diphtheria toxin receptor (DTR) under CD11c promoter (CD11cDTR) were from Jackson Laboratory. Interferon (IFN)αβ receptor-deficient (IFNαβR KO) mice were originally provided by P. Marrack (National Jewish Health, Denver, Colorado, USA) and bred in our facility. Mice deficient of IFNαβR in endothelial cells were produced by crossing Tie2-Cre (Tek-Cre, Jackson Stock No: 004128) mice with IFNaR1 flox mice (Jackson Stock No: 028256). To generate BM chimeric mice, CD45.1 congenic B6 mice received one dose of total body irradiation (1000 rads) followed by an intravenous injection of ~1×10⁷ BM cells from appropriate KO mice or CD11cDTR mice (all CD45.2) and 60 days later reconstitution efficiency was checked using flow cytometry. For CD11c+ cell depletion, mice reconstituted with CD11cDTR BM were injected intraperitoneally with 100 ng diphtheria toxin (DT)/mouse −2, 0 and +2 days of the PICLC injection. All animal care and experiments were conducted according to our institutional animal care and use committee (IACUC) guidelines. Lewis lung carcinoma (LLC) cells were the A9F1 subclone15 provided by L. Eisenbach (Weizmann Institute of Science, Rehovot, Israel). Mouse melanoma B16F10 cells were provided by Alan Houghton (Memorial Sloan Kettering Cancer Center, New York, New York, USA). Human umbilical vein endothelial cells (HUVECs) were obtained from Lonza. The mouse endothelial cell lines from brain (bEnd.3) and heart (H5V) were from American Type Culture Collection.

PICLC (Hiltonol) is a clinical grade formulation of poly-IC stabilized with poly-L-lysine and carboxymethylcellulose manufactured by Oncovir. High molecular weight (hmw) poly-IC and poly-AU were purchased from InvivoGen. Poly-IC/PEI was prepared by mixing hmw poly-IC with in vivo-JetPEI (Polyplus) at a nucleotide:PEI ratio of 6 in a 5% glucose solution. Antimouse PD-L1 monoclonal antibody (mAb) (clone, 10F.9G2) was from BioXcell. Mouse recombinant Interferon-beta and human recombinant Interferon-α2 were purchased from BioLegend. CXCL10 ELISA kits were purchased from R&D Systems. EliSpot plates were from eBioSciences. Dispase I (neutral protease, grade I) was purchased from Sigma-Aldrich. Fluorescence-conjugated Abs for flow cytometry were from Biolegend or eBioscience.

Mice were injected subcutaneously with 5×10⁵ LLC, or 3×10⁵ B16F10 cells in a shaved rear flank. Seven days (B16F10 tumor model), or 9 days (LLC tumor model) later, poly-IC, PICLC or JETPEI/poly-IC were administered intravenously, intramuscularly or intratumorally at 50 µg/dose. The administration of PICLC was repeated 3 times, 5 days apart. Anti-PD-L1 mAb (200 µg/dose) was administered intraperitoneally on days 1 and 3 after each PICLC administration. Tumor growth was monitored every 2–3 days in individually tagged mice by measuring two opposing diameters with a set of calipers. Mice were euthanized when the tumor area reached 400 mm². Results are presented as the mean tumor size (area in mm²±SD for every treatment group at various time points until the termination of the experiment.

Established tumors and normal tissues (pancreas, lungs and hearts) were removed from mice, dissociated and treated with 1 mg/mL collagenase and DNase in phosphate-buffered saline (PBS) for 1 hour at 37°C. Cells were washed twice, and red blood cells were lysed using red blood cell lysis buffer, followed by filtration through a 40-μm strainer. For CD8 T cell detections, tumor/pancreas single suspensions were stained with anti-CD45 (Alexaflour700), anti-MHC-II (FITC) and anti-CD8 (Pacific Blue or PercpCy5.5). For myeloid-derived suppressor cells (MDSCs) detection, tumor single suspensions were stained with anti-CD45 (Alexaflour700), anti-MHC-II (FITC), anti-CD11b (PercpCy5.5) and anti-Gr1 (PE). For in vivo CD45 staining,16 tumor-bearing mice were injected with 5 µg APC-conjugated anti-CD45 (clone EM-05) in 200 µL PBS. Five minutes later, mice were euthanized, and blood and tumors were collected for in vitro CD45 staining using Alexaflour700-conjugated anti-CD45 (clone 30-F11). For measuring CXCL9 and vascular cell adhesion molecule 1 (VCAM-I) expression in tumor, heart/lung VECs B6 mice were inoculated subcutaneously with 5×10⁵ B16F10 cells 10 days before intravenous administrations of PBS (No Tx), 50 μg poly-IC, or 50 μg PICLC on days 10 and 14. Tissues were harvested on day 15 and resuspended in Dispase-I, minced and dissociated in a gentleMACS (Miltenyi Biotec), at 37°C for 20 min. The cell digest was filtered through a 70 μm cell strainer, washed and stained with anit-CD45, anti-CD31 to specifically gate on VECs (CD45-negative, CD31-positive). Anti-VCAM-I was included to measure cell surface expression of this adhesion molecule. For chemokine detection, cells were incubated with 1 μL/mL GolgiPlug (BD Bioscience) at 37°C for 6 hours and followed by cell surface staining with anti-CD45 and anti-CD31 mAbs (for negative and positive gates, respectively) and with intracellular staining with anti-CXCL9 mAb. Cell florescence analyzes were performed in LSRII cytometers (BD Biosciences). Data analysis was performed using FlowJo software (V.8.5, TreeStar).

Formalin-fixed paraffin-embedded tissues were sectioned at a 5 μm thickness. Sections were subsequently deparaffinized in xylene, rehydrated with decreasing concentrations of ethanol, and boiled in citrate buffer for antigen retrieval. Sections were assessed with a rat antimouse CD8α (Alexa Fluor 488) mAb, mounted in aqueous mounting media with 4′,6-diamidino-2-phenylindole (DAPI, ThermoFisher Scientific) and visualized under a phase contrast and fluorescence imaging by EVOS fluorescence microscope.

Primary human (HUVECs) or murine (bEnd.3 and H5V) vascular endothelial cell lines cells were stimulated with poly-IC, poly-AU or PICLC as indicated in the figure legends at 37°C in a humidified 5% CO₂ incubator for 24 hours. Type I interferon (IFN-I) detection was performed using with HEK-Blue IFN-α/β cells (human) or B16-Blue IFN-α/β cells (murine) from InvivoGen according to manufacturer’s protocol using mouse IFN-β or human IFN-α2 for the standard curves.

Statistical significance to assess surface markers expression and cytokines production were performed using Student’s t-tests or one-way analysis of variance (ANOVA) as appropriate. Statistical significance to assess the antitumor efficacy was performed using two-way ANOVA. Results are presented as mean±SD. Statistical analyses were performed using GraphPad Prism (V.7).

While developing an intravenous peptide-based therapeutic vaccination strategy in mouse tumor models using poly-IC as the adjuvant and an ICI (anti-PD-L1 antibody), we observed a significant therapeutic effect in the control group receiving a non-TAg peptide, injected with poly-IC and ICI.5 Throughout these studies, we used a stabilized formulation of poly-IC (PICLC) containing poly-lysine and carboxymethylcellulose that protects this synthetic double-stranded RNA (dsRNA) from RNAse degradation.17 Subsequently, we reported that repeated intravenous administration of PICLC and anti-PD-L1 antibody alone (without any peptide) had a remarkable therapeutic effect mediated by CD8 T cells in several mouse tumor models.18 Moreover, in the case of LLC, which expresses low levels of PD-L1, repeated administration of PICLC alone was able to control tumor growth. Nevertheless, the mechanisms involved in the antitumor activity of the PICLC therapy remained unknown. We first hypothesized that PICLC could be functioning as an immune adjuvant potentiating CD8 T cell responses elicited by dendritic cells (DCs) presenting TAgs released by tumor cells. PICLC could also promote intratumoral CD8 T cell proliferation and survival. If these assumptions were correct, we predicted that intratumoral administration of PICLC should elicit superior antitumor effects as compared with systemic administrations (intravenous or intramuscular injections), which we routinely had used. To our surprise, intratumoral PICLC was significantly less effective in controlling growth of LLC tumors as compared with intravenous or intramuscular administrations (figure 1A). Also, DC depletions prior to intravenous PICLC therapy did not completely eliminate the antitumor efficacy (figure 1B). These results suggested that some of PICLC’s antitumor effects were taking place outside of the tumor parenchyma. Thus, we considered the possibility that PICLC could be facilitating the traffic and infiltration of effector CD8 T cells into the tumors. Mice bearing LLC tumors were treated twice with PICLC intravenous or intratumoral and tumors were resected the day after the second treatment and examined by flow cytometry for the presence of CD8 T cells. In agreement with our prediction, there was a substantial increase in the frequency of CD8 T cells in tumors from mice that received intravenous PICLC as compared with intratumoral PICLC or untreated control mice (figure 1C). In addition to these findings, the frequency of CD11b/Gr1-positive myeloid cells was significantly reduced in the tumors from mice intravenously treated with PICLC as compared with the other groups. The increase in tumor infiltrating CD8 T cells and decrease of CD11b/Gr1 cells induced by intravenous PICLC was also observed in DC-depleted mice (figure 1D).

Similar results of the effects of systemic administration (intravenous or intramuscular) of PICLC on T cell infiltration were obtained with mice bearing B16 melanoma tumors (figure 2A). However, in this experiment we assessed whether the CD8 T cells detected in the tumors actually reside in the tumor parenchyma or in the blood irrigating the tumors. This was done by injecting APC-labeled anti-CD45 antibodies intravenously 5 min prior to tumor resections. This procedure has been shown to label blood circulating lymphocytes but not TILs.16 An example of a PICLC intravenously treated mouse is presented in figure 2B showing that ~93% of the CD8 in tumors appear to reside in the parenchyma. The increased presence of CD8 T cells in tumor parenchyma of mice intravenously treated with PICLC as compared with intratumoral treatment was corroborated by immunohistology (figure 2C). CD8 T cells purified from B16 tumors of intravenous PICLC-treated mice recognized tumor cells in IFN-γ release EliSpot assays (online supplemental figure S1). Next, we compared the ability of PICLC to enhance T cell infiltration into tumors versus a normal organ. As shown in figure 2D, the T cell infiltration enhancement of intravenously administered PICLC was only significant in LLC tumors and not in a normal organ (pancreas). Similarly, the decrease of CD11b/Gr1 cells by PICLC was observed only at the tumor site.

Poly-IC and PICLC function as agonists for the endosomal TLR3, but both can also stimulate the RIG-I-like dsRNA sensor MDA5 in circumstances when these polynucleotides are able to leak from endosomes into the cytoplasm. Activation of TLR3 and MDA5 results in the production of immunomodulatory cytokines such as IFN-I.19 To assess the roles of TLR3, MDA5 and IFN-I in the antitumor effects of intravenously administered PICLC, an experiment was performed in genetically deficient mice bearing LLC tumors. As shown in figure 3A, PICLC did not elicit antitumor responses in MDA5 and IFN-I receptor (IFNαβR)-deficient mice. In contrast, PICLC reduced tumor growth in TLR3-deficient mice to the same extent as with wild-type (WT) mice. These findings correlated directly with the levels of CD8 T cell tumor infiltrates, and inversely correlated with the frequency of CD11b/Gr1 myeloid cells (figure 3B). These results indicate that PICLC promotes CD8 T cell tumor infiltration and growth control via the production of IFN-I by stimulating MDA5. PICLC contains poly-lysine, a poly-cation that binds to negative charges in the polynucleotide’s backbone, which has been used as a transfection reagent to deliver DNA (and RNA) into the cytoplasm. Polycations such as poly-lysine and polyethyleneimine (PEI) facilitate cytoplasmic delivery of DNA and RNA via the proton pump effect, which ruptures endosomes.20 21 Thus, PICLC should exhibit higher antitumor effects as compared with poly-IC, and PEI should increase the antitumor effects of poly-IC. Indeed, as shown in figure 3C, both PICLC and poly-IC/PEI elicited significantly higher antitumor effects as compared with poly-IC. These results correlated with the increases of infiltrating CD8 T cells and decreases of CD11b/Gr1 cells (figure 3D).

Next, we assessed which cells responding to PICLC promote CD8 T cell tumor infiltration and antitumor effects. One possibility is that BM-derived cells capable of capturing PICLC such as DCs, monocytes and macrophages would be the main participants in this event. The results in figure 1B,D indicate that DCs may play some role but are not solely responsible in this process. There is evidence in the literature that stromal cells are also capable of responding to poly-IC and produce IFN-I via MDA5 stimulation.19 Thus, we generated several BM chimeric mice to evaluate the roles of MDA5 and IFNαβR expression on either BM derived or stromal cells in the antitumor effects and immune cell tumor infiltration of PICLC therapy. As shown in figure 4A, the most effective and significant antitumor effect of PICLC was observed when BM of MDA5-deficient mice was implanted into WT mice. BM of WT into either MDA5-deficient or IFNαβR-deficient hosts resulted in reduced, but still significant therapeutic antitumor responses by PICLC. These results correlated with tumor infiltrations of CD8 T cells (figure 4B). These findings indicate that both, immune cells derived from the BM and non-BM-derived stromal cells respond to PICLC and participate in the antitumor effects.

As to the nature of the stromal cells capable of responding to PICLC, we proposed that vascular endothelial cells (VECs) would be the most likely candidates that could promote the extravasation of T cells into the tumor parenchyma, since this is their function during inflammatory responses such as with infections. In addition, VECs are known to express various types of scavenger receptors capable of binding poly-anionic and poly-cationic compounds such as polynucleotides and poly-lysine, which would enable their capture and endocytosis. In view of this, we tested the ability of PICLC and poly-IC to stimulate the production of IFN-I in two mouse VEC lines (H5V and bEnd.3). Both VEC lines produced significant amounts of IFN-I when stimulated with PICLC but not with poly-IC (figure 4C). Stimulation of one of these cell lines (bEnd.3) with PICLC enhanced the expression of the adhesion molecule VCAM-I (figure 4D). Notably, incubation of these VECs with poly-IC or IFN-I (IFN-β) did not enhance the expression of VCAM-I (data not shown). Because these cells express TLR3 (data not shown) and did not respond to poly-IC, we believe that the production of IFN-I and enhancement expression of VCAM-I is likely mediated via MDA5.

In addition to the role of adhesion molecules, chemokines such as CXCL9 and CXCL10 participate in the recruitment of CD8 T cells into tissues and solid tumors.22–26 PICLC was effective in stimulating the production of CXCL10 in the bEnd.3 VEC line (figure 5A). Because INF-β also elicited the production of CXCL10 to roughly the same levels, we assume that the production of CXCL10 by PICLC in VECs is mediated through IFN-I. Since PICLC systemic administration enhanced T cell infiltration into tumors, but not into normal tissues, we isolated VECs from tumors and normal organs (lungs and hearts) after intravenous PICLC or poly-IC injections and measured the expression of VCAM-I and CXCL9 by flow cytometry. Both PICLC and poly-IC enhanced cell surface expression of VCAM-I in tumor VECs to a much higher extent as compared with VECs from normal tissues (figure 5B). Intracellular expression of CXCL9 was enhanced by PICLC and not by poly-IC in tumor VECs and no changes were observed in normal organ VECs. Both PICLC and poly-IC enhanced the expression of ICAM-I in tumor and normal organ VECs (data not shown).

We also assessed whether human VECs (HUVECs) could respond to PICLC stimulation. Primary HUVECs were stimulated with various concentrations of PICLC, poly-IC or poly-AU (a TLR3 agonist that does not stimulate MDA5). The results showed that PICLC was substantially more effective in inducing IFN-I secretion as compared with poly-IC, and that poly-AU did not stimulate the HUVECs (figure 5C). Both PICLC and poly-IC and to some extent poly-AU (but at much higher concentrations) induced the production of CXCL9 on HUVECs while all three compounds were able to increase the expression of VCAM-I (figure 5D). These results provide support to the conclusion that systemically administered PICLC enhances CD8 T cell-mediated antitumor responses by augmenting tumor T cell infiltration via MDA5 stimulation in VECs resulting in the production of IFN-I and T cell recruiting chemokines.

Since the role of IFN-I on VECs appears to be essential for the capacity of PICLC to promote tumor T cell infiltration, we examined the antitumor effects of PICLC and the presence of tumor T cell infiltrates in conditional IFNαβR-deficient mice. Using Tie2-Cre mice and Ifnar1^fl mice we generated mice deficient in the expression of IFNαβR in endothelial cells. As shown in figure 6A, intravenous administration of PICLC in Tie2-Cre/Ifnar1^fl/fl mice failed to reduce the tumor growth as compared with the mice injected with PBS. On the other hand, PICLC injected into Tie2-Cre/Ifnar1^fl/− control mice significantly decreased the tumor growth as compared with mice injected with PBS. Simultaneously, PICLC therapy resulted in increased CD8 T cell tumor infiltrates in the Tie2-Cre/Ifnar1^fl/− control mouse group but not in Tie2-Cre/Ifnar1^fl/fl mice (figure 6B).

Because IFN-I signals could play a role in the function and survival of CD8 T cells we examined the role of the IFNαβR on the CD8 T cells by generating T cell receptor OT-I (ovalbumin specific) transgenic mice on an IFNαβR background. Purified OT-I IFNαβR-deficient cells (CD45.1+/CD45.2+) were mixed at a 1:1 ratio with WT OT-I cells (CD45.1+) and were transferred into mice bearing subcutaneous B16 tumors expressing ovalbumin. After treating the mice with intravenous PICLC the tumors were examined for the presence of OT-I T cells. PICLC enhanced OT-I infiltrates into the tumor but no significant differences were observed between the percentages of OT-I IFNαβR-deficient cells and OT-I WT cells (online supplemental figure S2). Also, no significant differences in the ratio of these T cells were observed in spleens. These results indicate that the enhancement of CD8 T cell tumor infiltration by PICLC treatment is not influenced by IFN-I signals on the T cells.