Dual Programmed Death Receptor-1 and Vascular Endothelial Growth Factor Receptor-2 Blockade Promotes Vascular Normalization and Enhances Antitumor Immune Responses in Hepatocellular Carcinoma

Cells and Culture Conditions

We used two murine cell lines: HCA-1, established in our laboratory,27, 28 and RIL-175 (a p53/Hras mutant HCC cell from C57Bl/6 mice, a kind gift from Dr. Tim Greten, National Institutes of Health).29 HCA-1 was maintained in Dulbecco’s modified essential medium (DMEM) with 10% fetal bovine serum (FBS) and pyruvic acid. RIL-175 was maintained in DMEM with 20% FBS with pyruvic acid.

Hypoxic condition cell culture was performed in a Brunswick Galaxy incubator at 1% for 48 hours. Organoid culture was performed using primary cells from tumor tissue. Tumor tissues were resected and minced with a scalpel. Minced tissue fragments were incubated in Hank’s balanced salt solution (HBSS) with 15 µg/mL of collagenase and 1.5 mg/mL of hyaluronidase for 30 minutes at 37°C. Digested tissues were passed through a 70-µm cell strainer and washed twice with HBSS. After counting the cell concentration, cells were diluted to 1 million cells/mL with DMEM with 10% FBS. The cell suspension was poured into a Nano culture plate (MH type; SCIVAX, Organogenix, Japan) for 72 hours to form organoids at 37°C with 5% CO₂. After the confirmation of organoid formation by light microscopy, 50 µg/mL of DC101 and/or 20 µg/mL of anti-IFN-γ blocking antibody was added to the medium, and organoids were placed in the specific experimental culture condition. Primary endothelial cells were obtained from a single-cell suspension of RIL-175 tumors. CD31⁺ cells were sorted using biotin-conjugated CD31 antibody (clone MEC13.3; Biolegend) and antibiotin microbeads (Miltenyi, Germany) by magnetic sorting. Endothelial cells were cultured on the type I collagen–coated cell culture dishes, using Endothelial Cell Growth Medium MV2 (Promo Cell).

HCC Models

The orthotopic HCC mouse models were described elsewhere.30 The HCC cells (1 million 1:1 in Matrigel; Mediatech/Corning, Manassas, VA) were grafted into mice matching their genetic background—HCA-1 in C3H mice and RIL-175 in C57Bl/6 mice. Tumor growth was monitored by high-frequency ultrasonography twice a week. In the macrophage-stimulating 1 (Mst1^–/–)Mst2^f/– model31 adeno–cyclization recombination at a concentration of 10⁷ pfu was injected through the tail vein in 3-week-old mice. To induce liver damage, 100 µL of 20% CCl₄ (Sigma-Aldrich, St. Louis, MO) was administered orally from 8 to 12 weeks.30 When tumors reached ~5 mm in diameter, mice were randomly assigned to a treatment group (n = 6-8). Treatments were administered intraperitoneally thrice per week for 14 days at doses of 10 mg/kg (anti-PD-1 antibody, anti-CD4 antibody, and immunoglobulin G [IgG] control) and 10 mg/kg (low) or 40 mg/kg (standard) (mouse antibody DC101). All animal experiments were performed after approval by the Institutional Animal Care and Use Committee of the Massachusetts General Hospital.

Reagents

DC101 antibody was kindly provided by Eli Lilly and Company (Indianapolis, IN). Mouse anti-PD-1 antibodies (clone RMP-014) and anti-CD4 antibodies were purchased from BioXcel (West Lebanon, NH). Control rat IgG was purchased from Jackson ImmunoResearch (West Grove, PA). Anti-IFN-γ antibody (clone XMG1.2) was purchased from Thermo Fisher Scientific (Waltham, MA).

Flow Cytometry

Tumor tissues were resected and minced, and fragments were incubated in HBSS with 15 µg/mL of collagenase and 1.5 mg/mL of hyaluronidase for 30 minutes at 37°C. Digested tissues were passed through a 70-µm cell strainer and washed twice with phosphate-buffered saline (PBS)/0.5% bovine serum albumin. Single-cell suspensions were incubated with antimouse CD16/32 antibody (clone 93; Biolegend, San Diego, CA) for 5 minutes prior to staining for immune cell markers for 15 minutes at room temperature. For the intracellular markers, cells were fixed and permeabilized with either the FoxP3/Transcription Factor Staining Buffer Set (eBioscience/Thermo Fisher Scientific, Waltham, MA) or BD Cytofix/Cytoperm Kit (BD Biosciences, San Jose, CA), according to the manufacturers’ protocols. For cytokine staining, harvested cells were incubated in Roswell Park Memorial Institute medium with cell activation cocktail with brefeldin A (Biolegend) for 6 hours at 37°C, and stimulated cells were stained as described above. Monoclonal antibodies used for flow-cytometric analysis were CD45 (30-F11), CD3e (145-2C11), CD4 (RM4-5), CD8 (53-6,7), PD-1 (29F.1A12), IFN-γ (XMG1.2), and forkhead box P3 (FoxP3; FJK-165).

Cell Sorting

For the magnetic bead sorting of endothelial cells or myeloid cells, cells were first stained with either biotin-conjugated anti-CD31 (clone MEC13.3) or anti-CD11b (clone M1/70) and anti-CD11c antibodies (clone N418) (all Biolegend). Biotin-labeled cells were separated by magnetic sorting using antibiotin microbeads (Miltenyi). For macrophage collection, the CD45⁺CD11b⁺Ly6C-F4/80⁺ population was sorted using a FACS Aria cell sorter.

Immunohistochemistry and Immunofluorescence

Resected tumor tissues were fixed in 4% paraformaldehyde and embedded either in paraffin or in optimal cutting temperature compound and frozen. Vascular endothelial staining was performed using anti-CD31 antibody (clone DIA-310; Dianova, Germany, for immunohistochemistry [IHC], and Millipore, St. Louis, MO, for immunofluorescence [IF]), pericytes were identified with antibodies to α-smooth muscle actin (SMA) (Sigma-Aldrich), PD-L1 expression was detected using anti-PD-L1 antibodies (eBioscience), and hypoxia was detected with anti–carbonic anhydrase IX (CA-IX) antibody (Abcam, Cambridge, UK). Specimens were incubated with fluorescence (cyanine 3 and Alexa 647)–conjugated antihamster or antirabbit secondary antibodies, as appropriate (Jackson ImmunoResearch). Cell nuclei were identified with 4′,6-diamidino-2-phenylindole (DAPI; ProLong Gold Antifade Mountant with DAPI; Thermo Fisher Scientific). The total number of vessels and pericyte-covered vessels and the fraction of hypoxic tumor area were counted in five random fields using a ×400 magnification lens. These data were analyzed using ImageJ (US National Institutes of Health) and PhotoShop (Adobe Systems Inc., San Jose, CA) software. Organoids were fixed in 0.4% formaldehyde for 15 minutes. Then, organoids were washed 3 times with PBS and immunostained. IF images were analyzed using a confocal microscope (Fluoview FV100; Olympus, Center Valley, PA).

RNA Sequencing Analysis

When tumors reached 200 mm³, three to five samples per group were collected from mice treated with control IgG, anti-PD-1 antibody alone, anti-PD-1 antibody in combination with DC101, and anti-PD-1 antibody in combination with DC101 and anti-CD4 antibodies. Mice were sacrificed, and total RNA was isolated from the freshly isolated tumor tissues using Qiagen kits. Total RNA was sequenced at Molecular Biology Core Facilities, Dana Farber Cancer Institute (Boston, MA), and used for Gene Set Enrichment Analysis.

Cytokine Analysis

Culture supernatants were collected after the specific experimental treatment. Samples were assayed in duplicate using the MSD proinflammatory Panel I, a highly sensitive multiplex enzyme-linked immunosorbent assay (ELISA) for quantitatively measuring 10 cytokines—IFN-γ, interleukin (IL)-1β, IL-2, IL-4, IL-6, IL-8, IL-10, IL-12p70, IL-13, and tumor necrosis factor alpha—from a single small sample volume (25 μL) using electrochemiluminescence-based detection (MesoScale Discovery, Gaithersburg, MD).

Statistical Analysis

All statistical analyses were performed using Stata software, version 14.1 (Stata Corp, College Station, TX). Error bars indicate standard error of the mean. Differences were considered significant at P < 0.05. Quantitative variables were compared using the Student t test. Analysis of experiments with more than two groups was performed using one-way analysis of variance with Scheffe’s correction for multiple comparisons. Median overall survival (OS) was estimated using the Kaplan-Meier method. Statistical analyses in the survival experiments were performed by the Cox proportional hazards model, and hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated.

Dual VEGFR-2/PD-1 Blockade is Effective in Orthotopic HCC Models in Mice

We tested combination therapy in orthotopic models of HCC in mice with (chemically induced) underlying liver damage and examined the impact of the dose of the AA agent. We treated mice with established HCC (5-6 mm in diameter) with the anti-VEGFR-2 antibody DC101 at two different doses (AA-low, 10 mg/kg, and standard AA, 40 mg/kg, both thrice a week), anti-PD-1 antibody (ICB), or their combination. Using HCA-1 in C3H mice, we found that in the single-agent treatment groups there was a significant growth delay only in the AA group compared to control and nonsignificant delays in the others; however, combinations of AA and ICB therapy—irrespective of the dose—induced comparable significant growth delays (Supporting Fig. S1A). In addition, there was a significant increase in median OS in the combination therapy groups, which was not seen in any of the monotherapy arms (Fig. 1A). The lack of survival advantage in the AA group was due to lack of control of metastatic disease, despite delay of the primary tumor growth; the metastatic burden was reduced in all the other groups at the experimental endpoint (Supporting Fig. S1B). Tumor growth inhibition was also seen in the genetically engineered RIL-175 (p53/Hras) and Mst-mutant HCC models (Supporting Fig. S1C,D). Next, we tested the impact of dual AA/ICB in C57Bl/6 mice bearing RIL-175 HCCs—sensitive to anti-PD-1 therapy. We found that the combination therapy produced the longest median OS (Fig. 1B).

Dual VEGFR-2/PD-1 Blockade Reprograms the Immune Microenvironment and Promotes Antitumor Immunity in HCC

Having reproduced the clinical benefit seen with combination immunotherapy/AA in mouse models, we first examined how therapy impacted lymphocyte infiltration. We first analyzed the number of tumor-infiltrating leukocytes as a fraction of total CD45⁺ immune cells by flow cytometry. The number of tumor-infiltrating CTLs was not affected by treatment with ICB alone or AA-low/ICB (Supporting Fig. S2). AA decreased the proportion of infiltrating CTLs, even when combined with ICB in RIL-175 HCCs (Fig. 2A; Supporting Fig. S2A,B). However, the overall fraction of infiltrating CD4⁺ T cells and the CD8⁺/CD4⁺ T-cell ratio were not affected in any of the combination groups (Supporting Fig. S2C,D). Moreover, the fraction of FoxP3⁺ Tregs was significantly decreased in all treatment groups, which translated into increased CTL/Treg ratios in the ICB-containing groups but not in the AA alone groups (Fig. 2B,C; Supporting Fig. S2E,F). These findings were confirmed in the Mst-mutant model of HCC, where we found a decrease in Tregs and an increase in CD8⁺/CD4⁺ and CTL/Treg ratios (Supporting Fig. S2G-I). Furthermore, when evaluating CTL function by evaluating IFN-γ expression, we found it to be significantly increased in all ICB-containing groups (Supporting Fig. S2J). Finally, using IHC, we found that the number of tumor-infiltrating CTLs was significantly increased in the combination treatment groups compared to AA treatments alone (Fig. S2L).

We also evaluated the changes in myeloid cell populations posttreatment using flow cytometry. AA-low treatment increased the fraction of F4/80⁺CD80⁺ and/or CD86⁺ antitumor “M1-type activated” tumor-infiltrating macrophages (TAMs) (Supporting Fig. S3). ICB treatment alone did not change the fraction in these populations; however, AA/ICB combinations increased the fraction of M1-TAMs (Supporting Fig. S3A). Moreover, AA/ICB combinations also decreased the fraction of both protumor F4/80⁺CD206⁺ “M2-type activated” TAMs (Supporting Fig. S3B) and CD11b⁺–chemokine (C-C motif) receptor 2–positive (CCR2⁺) monocytic cells (Supporting Fig. S4C); these effects appeared to be primarily mediated by AA (Supporting Fig. S3D-F).

To examine more broadly the immune effects of AA/IBC therapy, we performed RNA sequencing and found that ICB (alone or with AA) significantly up-regulated immune pathways (including proliferation, function/activity, and migration) compared to control-treated HCC tissues. Moreover, AA/IBC therapy selectively up-regulated pathways related to myeloid cells and B cells that were not enhanced by ICB alone (Fig. 2E; Supporting Table S1).

PD-L1 and PD-1 Expression is Increased After VEGFR-2 Blockade in HCC In Vivo

We have reported that VEGFR2 inhibition with the multitargeted tyrosine kinase inhibitor (TKI) sorafenib increases PD-L1 expression in these HCC models.9 Thus, we examined whether specific VEGFR-2 blockade has any impact on PD-L1 expression in HCC tissue in vivo. Using flow cytometry, we found that the number of nonimmune (CD45^–) cells expressing PD-L1 was increased in tumors after VEGFR-2 blockade alone or in combination with ICB (Fig. 3A; Supporting Fig. S4A). The high expression of PD-L1 in both cancer and stromal cells was confirmed by IHC (Fig. 3B; Supporting Fig. S4B). Of note, double staining using anti-PD-L1 and anti-F4/80 antibodies showed that PD-L1 was predominantly expressed in TAMs in control and AA-low treated tumors, while AA-treated tumors exhibited PD-L1 expression in both TAMs and nonimmune cells (Fig. 3C). Interestingly, AA significantly increased the fraction of CD4⁺ cells expressing PD-1 and the PD-1 expression level (Fig. 3D-F).

PD-L1 Expression in HCC Cells is Up-Regulated Upon VEGFR-2 Blockade in Endothelial Cells in Part Through IFN-γ Production

To gain insight into the mechanism of PD-L1 up-regulation after DC101 treatment, we used organoid three-dimensional cultures in vitro. Single suspension of cells was obtained after enzymatic digestion of tumor tissues and then incubated in Nano-cultured plates under normoxic conditions. The cells formed heterotypic tumor organoids consisting of multiple cell types, including endothelial cells, after 3 days; and then organoids were cultured in hypoxic conditions for 48 hours. While PD-L1 expression was detectable, it was only slightly increased in hypoxic conditions (Supporting Fig. S4C,D). However, when we treated the organoids with blocking doses of DC101 (50 μg/mL),32 PD-L1 expression was substantially up-regulated in tumor and stromal cells (Fig. 4A,B). In this HCC model, VEGFR-2 expression was restricted to endothelial cells and to a subset of myeloid cells (Supporting Fig. S5). To elucidate the nature of the paracrine interaction leading to PD-L1 up-regulation after VEGFR-2 blockade, we next depleted certain cell lineages from the original single-cell suspension prior to organoid formation. First, we depleted all VEGFR-2⁺ cells from the single-cell suspension using magnetic beads and treated the resulting organoids with DC101 under hypoxic conditions. DC101 treatment failed to up-regulate PD-L1 expression in these conditions (Supporting Fig. S4E). Next, we depleted endothelial cells using anti-CD31 antibodies or myeloid cells using anti-CD11b and anti-CD11c antibodies and generated organoids that were treated with DC101 in hypoxic conditions. In these conditions, PD-L1 up-regulation was not seen in organoids where endothelial cells were depleted but was readily detectable in organoids where myeloid cells were depleted (Supporting Fig. S4E). Collectively, these results indicate that up-regulation of PD-L1 in HCC is dependent on VEGFR-2 blockade in endothelial cells, particularly in hypoxic conditions, which mimic the in vivo effects of AA.

To examine the mechanism of PD-L1 expression regulation in a paracrine manner by endothelial cells, CD31⁺ cells were sorted from a single-cell suspension of RIL-175 tumors and cultured in vitro. We collected supernatant of endothelial cell culture treated with or without DC101 and evaluated inflammatory cytokine production by multiplexed ELISA. DC101 treatment enhanced most notably IFN-γ production by the endothelial cells (Supporting Fig. S4F). Because IFN-γ is a key inducer of PD-L1 expression,33 we treated the organoids with DC101 and blocked IFN-γ using a blocking antibody. IFN-γ blockade decreased PD-L1 expression by half in HCC endothelial cells (Fig. 4C,D), indicating that the HCC cell PD-L1 up-regulation by DC101 targeting of VEGFR-2 in endothelial cells is in part mediated by IFN-γ expression by endothelial cells.

ICB Therapy Promotes Vascular Normalization When Combined With AA Treatment

Because immunostimulatory reprogramming and vascular normalization may be reciprocally regulated in HCC,22 we next examined the effects of treatment on vascular structure and function in size-matched tumors. We found that ICB increased the total and pericyte-covered microvessel density (MVD) in established HCCs (Fig. 5A-C). Surprisingly, dual AA/ICB treatment further increased total and pericyte-covered MVD and prevented an increase in hypoxia in HCC, in contrast to the effect of AA alone (Fig. 5D,E). These data demonstrate that ICB promotes durable vascular formation and normalization even when used in combination with AA.

Because combination treatment did not increase CTL infiltration or activation over ICB alone but increased PD-1 expression in CD4⁺ cells, we next sought to decipher whether CD4⁺ cells play a role in the modulation of vascular structure and function in HCC. Of note, a recent study showed that CD4⁺ cells mutually regulate vascular and immune effects in developing tumors in an IFN-γ-dependent manner.22 We found that CD4⁺ cell depletion abrogated the normalized vessel formation seen after AA/ICB treatment and resulted in increased hypoxia (Fig. 5). Strikingly after CD4⁺-cell depletion during AA/ICB compared to AA/ICB with intact CD4⁺ cells, RNA sequencing revealed that 25/235 significantly down-regulated pathways were related to blood vessel formation, maturation, and normalization (Supporting Fig. S6 and Table S1). Thus, CD4⁺ cells promote vascular formation and remodeling after ICB despite potent AA in HCC (Fig. 6).