Abstract
Immune checkpoint inhibitors have revolutionized the treatment of unresectable hepatocellular carcinoma (HCC), but their impressive efficacy is seen in just a fraction of patients. One key mechanism of immunotherapy resistance is the paucity of dendritic cells (DC) in liver malignancies. In this study, we tested combination blockade of PD-1 and CXCR4, a receptor for CXCL12, a pleiotropic factor that mediates immunosuppression in tumors. Using orthotopic grafted and autochthonous HCC models with underlying liver damage, we evaluated treatment feasibility and efficacy. In addition, we examined the effects of treatment using immunofluorescence, flow cytometric analysis of DCs in vivo and in vitro, and RNA sequencing. The combination anti-CXCR4 and anti–PD-1 therapy was safe and significantly inhibited tumor growth and prolonged survival in all murine preclinical models of HCC tested. The combination treatment successfully reprogrammed antigen-presenting cells, revealing the potential role of conventional type 1 DCs (cDC1) in the HCC microenvironment. Moreover, DC reprogramming enhanced anticancer immunity by facilitating CD8+ T-cell accumulation and activation in the HCC tissue. The effectiveness of anti-CXCR4/PD-1 therapy was compromised entirely in Batf3 knockout mice deficient in cDC1s. Thus, combined CXCR4/PD-1 blockade can reprogram intratumoral cDC1s and holds the potential to potentiate antitumor immune response against HCC.
Introduction
Liver cancer is the sixth most commonly diagnosed cancer worldwide and the third most common cause of cancer-related death, with increasing incidence and mortality rates in the United States and Europe (1). Hepatocellular carcinoma (HCC) represents most liver cancer cases. Surgery is a potentially curative treatment, but many patients are not eligible due to underlying liver disease or high tumor burden. As such, systemic drug therapy is vital in managing unresectable HCC. Sorafenib, a multiple kinase inhibitor (MKI), has been the first-line therapy for over a decade, followed more recently by other MKIs. However, resistance to MKI monotherapy develops rapidly (2). Combining immune checkpoint inhibitors (ICI) and VEGF inhibitors has emerged as a breakthrough therapy. However, further advances are required, with responses of less than 30% and high relapse rates among initial responders (3). Therefore, new combination therapies to enhance the durability of immune responses after ICIs are urgently needed.
Dendritic cells (DC) can be categorized into two main subsets: conventional DCs (cDC) for antigen presentation and plasmacytoid DCs for IFN production. Among cDCs, conventional type 1 DCs (cDC1) are potent antigen-presenting cells and accelerate subsequent anticancer immune responses, whereas cDC2s stimulate CD4+ T-cell responses via MHC class II (4). Recent findings have highlighted that in addition to these, there is an entirely distinct subset of CCR7+ DCs that partially share features with both cDC1s and cDC2s (5, 6). Of these, cDC1s are essential in triggering CD8+ CTL activation and tumor cell identification and destruction through antigen cross-presentation (7). Defective antigen cross-presentation is a substantial cause of cancer immunosurveillance failure (8). Moreover, accumulating evidence in recent years highlights the importance of cross-presenting cDC1s in harnessing the potential of ICI therapy (9). The liver microenvironment has been found to induce a paucity of DCs in tumors compared with other organs (10–13). The mechanisms that suppress cDC1s in HCC remain unknown, but inefficient distribution for encountering tumor antigens and insufficient crosstalk between DCs and CTLs may contribute to immune evasion in HCC.
CXCL12 and its receptor CXCR4 are crucial for the interaction between cancer cells and their tumor microenvironment (TME). CXCR4 is present in various cell types, including immune, endothelial, and stem cells, with increased levels generally present in cancer cells (14). CXCL12 is also expressed by multiple immune cells, endothelial cells, stromal fibroblasts, and stem cells, and cancer cells often produce CXCL12 (15). CXCR4 signaling activation has been linked with cancer treatment resistance; tumor growth, invasion and metastasis; myeloid-derived suppressor cell recruitment; and angiogenesis (16). CXCL12/CXCR4 axis inhibition is being tested in clinical trials for various cancers, including HCC, in which CXCR4 expression levels are associated with tumorigenesis, progression, and poor outcomes (17, 18).
Previous reports have shown that the CXCL12/CXCR4 axis promotes fibrotic and immunosuppressive TME formation in HCC (16). However, CXCR4 inhibition alone has been largely ineffective across multiple cancer models, suggesting that CXCR4 targeting should be tested in combination with other modalities. In this study, we tested the effect of specifically blocking CXCR4 on standard anti–PD-1 immunotherapy in HCC models. Most previous studies and clinical trials tested AMD3100, a small-molecule antagonist of CXCR4, which has a short half-life and a complex mechanism of action. We reasoned that using antibodies with a long half-life and specificity would help elucidate the relationship between CXCL12/CXCR4 signaling and the HCC TME, reduce off-target effects, and improve effectiveness. In this study, we discovered the potential of anti-CXCR4 to reprogram the immunologically “cold” TME of HCC to an immunologically “hot” one as a result of effects on DCs and the benefits when combining them with anti–PD-1, a current standard therapy for patients with HCC.
Materials and Methods
Cells and culture condition
We used two murine cell lines: HCA-1, established in our laboratory from a C3H mouse (19), and RIL-175 (RRID: CVCL_B7TK; a p53/Hras-mutant HCC cell from C57Bl/6 mouse, a kind gift from Dr. Tim Greten, NIH). HCA-1 cells were maintained in DMEM (Sigma-Aldrich, #D8900) with 10% FBS (HyClone, #SH30071.03) and pyruvic acid (Sigma-Aldrich, #107360). RIL-175 cells were maintained in DMEM with 20% FBS and pyruvic acid. Cell line authentication and Mycoplasma contamination testing (using MycoAlert Mycoplasma Detection Kit, Lonza, #LT07-318), were performed before all experiments. Cell lines were typically used at a passage number of approximately 3 to 7.
Mouse models’ orthotopic HCC and liver damage
All animal experiments were performed at least in duplicate after approval by the Institutional Animal Care and Use Committee of the Massachusetts General Hospital.
Treatments
Treatments were administered intraperitoneally every 3 days for 21 days at 10 mg/kg [anti–PD-1, Dan G. Duda – Massachusetts General Hospital, Cat. # mPD1-4H2-mg1-D265A clone 6A1_RAS_Ab, RRID: AB_3476760; anti-CXCR4, Dan G. Duda – Massachusetts General Hospital, Cat. # mCXCR4.8-mG1 clone 6A4_RAS_Ab, RRID: AB_3477163; and IgG control, Bio X Cell, Cat. # BP0089, RRID: AB_1107769] in survival studies. For the time-match study, all the mice with orthotopic grafted models were sacrificed on day 8, and the mice with the autochthonous model were sacrificed on day 13 after starting the treatment when objective responses to immunotherapy were apparent.
Reagents
Bristol Myers Squibb provided the anti-mouse CXCR4 and anti-mouse PD-1, as well as isotype-matched rat IgG per the sponsored research agreement.
Flow cytometry analysis
Tumor tissues were resected and minced, and fragments were incubated in RPMI medium (Cat. #10-040-CV, Corning) with collagenase (Cat. #C4-BIOC, Sigma-Aldrich), dispase (Cat. #17105-041, Gibco), and DNase I (Cat. #10104159001, Roche) for 30 minutes at 37°C. Digested tissues were passed through a 70-μm cell strainer and washed with Hank’s Balanced Salt Solution/5 mmol/L EDTA. Single-cell suspensions were incubated with anti-mouse CD16/32 (clone 93; Invitrogen, Cat. #14-0161-82, RRID: AB_467133) before staining for immune cell markers for 15 minutes at room temperature. Cells were fixed and permeabilized before staining for the intracellular markers with FoxP3/Transcription Factor Staining Buffer Set (eBioscience/Thermo Fisher Scientific) according to the manufacturer’s protocols. For cytokine staining, harvested cells were incubated in RPMI medium with a cell activation cocktail with brefeldin A (BioLegend) for 4 hours at 37°C, and stimulated cells were stained as described above. mAbs used for flow cytometric analysis were CD45 (BioLegend, Cat. # 103140, RRID: AB_2562342), TCRb (BioLegend, Cat. # 109224, RRID: AB_1027648), CD8 (BioLegend, Cat. # 100752, RRID: AB_2563057), Ki67 (BioLegend, Cat. # 151218, RRID: AB_2910305), IFNγ (BioLegend, Cat. # 505808, RRID: AB_315402), Granzyme B (GzmB; BioLegend, Cat. # 515403, RRID: AB_2114575), CD103 (BioLegend, Cat. # 121422, RRID: AB_2562901), XCR1 (BioLegend, Cat. # 148218, RRID: AB_2565231), MHC class II (BioLegend, Cat. # 107630, RRID: AB_2069376), CD11b (BioLegend, Cat. # 101243, RRID: AB_2561373), CD11c (BioLegend, Cat. # 117308, RRID: AB_313777), CD19 (BioLegend, Cat. # 115528, RRID: AB_493735), Ly6G (BioLegend, Cat. # 127622, RRID: AB_10643269), NK1.1 (BioLegend, Cat. # 108730, RRID: AB_2291262), and Ly6C (BioLegend, Cat. # 128024, RRID: AB_10643270). All data were collected on a BD LSRFortessa or FACSAria IIIu (BD Biosciences) instrument and analyzed using FlowJo software (TreeStar).
IHC and immunofluorescence
Immunofluorescence (IF) was performed on frozen sections of optimal cutting temperature–embedded tissues. Each section was prepared at 8 μm thickness for immunostaining. The sections were washed with PBS and blocked with 10% normal donkey serum (Cat. #017-000-121, Jackson ImmunoResearch Laboratories, RRID: AB_2337258) for an hour at room temperature. Anti-CD11c (Cell Signaling Technology, #97585, 1:200, RRID: AB_2800282), anti-XCR1 (BioLegend, 109402, 1:200, RRID: AB_2910282), anti-CD8 (Biorbyt orb348907, 1:200, RRID: AB_3371712), or anti-CXCR4 (Abcam, ab124824, 1:200, RRID: AB_10975635) were applied as primary antibodies overnight at 4°C, followed by the reaction with appropriate secondary antibodies [Cy3 AffiniPure donkey anti-rat IgG (Cat. #712-165-153, Jackson ImmunoResearch Laboratories, RRID: AB_2340667), Cy3 AffiniPure Donkey Anti-Rabbit IgG (Cat. #711-165-152, Jackson ImmunoResearch Laboratories, RRID: AB_2307443), Alexa Fluor 647 AffiniPure donkey anti-rat IgG (Cat. #712-605-153, Jackson ImmunoResearch Laboratories, RRID: AB_2340694), and Alexa Fluor 647 AffiniPure donkey anti-rabbit IgG (Cat. #711-605-152, Jackson ImmunoResearch Laboratories, RRID: AB_2492288)] for 1 hour at room temperature. All slides were mounted with ProLong Gold Antifade with 4′,6-diamidino-2-phenylindole (Thermo Fisher Scientific). Analysis was performed using a laser scanning confocal microscope (Olympus, FV-1000, RRID: SCR_020337) in five random fields at ×400 magnification. For samples from orthotopic HCC, the area of the tumor edge was defined as within 600 μm of the tumor-normal liver border, and each area of the tumor core or edge was evaluated separately. Samples from autochthonous HCC were randomly assessed for each multicentric lesion. These data were analyzed with Fiji is just ImageJ (FIJI, RRID: SCR_002285) and Photoshop (Adobe Systems Inc., RRID: SCR_014199).
IHC was performed on paraffin sections. Each section was prepared at 5 μm thickness for immunostaining. Slides were deparaffinized in xylene for 10 minutes, rehydrated through a graded alcohol series, placed in an endogenous peroxide blocker for 5 minutes, and washed with PBS. The sections were blocked with blocking solution (5% skim milk, 1% Triton X-100) for an hour at room temperature. Anti-Ki-67 primary antibody (Millipore, AB9260, RRID: AB_2142366) was applied overnight at 4°C, followed by peroxidase-conjugated anti-rabbit IgG (Abcam, ab214880, RRID: AB_3106917) as the secondary antibody for 30 minutes at room temperature. The sections were then developed with 3,3′-diaminobenzidine color solution (Cat. #ab64238, Abcam) for 3 minutes at room temperature, and the slides were counterstained with hematoxylin (Cat. #72511, Epredia) sequentially for 30 seconds. Each slide was analyzed in five random fields under ×200 magnification using a brightfield microscope (Olympus, BX40). The Ki-67 index of HCC growth (in the autochthonous model) was defined as the percentage of Ki-67+ nuclei in each field, and it was determined using ImmunoRatio plugins on FIJI by dividing the total intensity of positive nuclei by that of all nuclei in the field.
RNA sequencing
Tumors were collected from three mice treated with control IgG, two with anti–PD-1 alone, two with anti-CXCR4 alone, and three with anti–PD-1 in combination with anti-CXCR4. Total RNA was extracted from freshly isolated tumor tissues using RNeasy Mini kits (Cat. #74104, Qiagen). The RNA quality examination and sequencing library construction were conducted at the Molecular Biology Core Facilities of the Dana-Farber Cancer Institute. In brief, libraries were prepared using KAPA mRNA HyperPrep strand-specific sample preparation kits (Cat. #08098123702, Roche) from 200 ng of purified total RNA according to the manufacturer’s protocol on Biomek i7 (B87585, Beckman Coulter). The finished double-stranded DNA libraries were quantified using a Qubit 4 fluorometer (Cat. #Q33238, Qubit) and 4200 TapeStation System (Cat. #G2991BA, Agilent). The final pool of libraries was sequenced on Illumina NovaSeq X Plus (Cat. #20084804, Illumina) with single-end mode at the Dana-Farber Cancer Institute Molecular Biology Core Facilities. Raw sequencing data were quality controlled using FastQC (RRID: SCR_014583). After quality control, Cutadapt (RRID: SCR_011841) was used to remove low-quality bases and adaptor contaminations. The quality of the resulting clean data was examined again using FastQC software. Subsequently, Hisat2 (RRID: SCR_015530) was utilized to align the clean data to the mouse reference genome mm10, obtained from the Illumina iGenomes database. Following data mapping, SAM and BAM files were manipulated using SAMTOOLS (RRID: SCR_002105). The number of reads aligned to gene features was counted using HTSeq-count (RRID: SCR_011867) from the HTSeq package (RRID: SCR_005514). Differentially expressed genes (DEG) were identified by edgeR (RRID: SCR_012802) using a cutoff of |log(fold change)| > 1 and a P value <0.01. Gene set enrichment analysis (GSEA) was performed using GSEA software (RRID: SCR_003199) with standard gene sets from the Molecular Signatures database (MSigDB). The deconvolution analysis of bulk RNA sequencing (RNA-seq) data was performed by TIMER2 (http://timer.cistrome.org) and xCell (https://github.com/dviraran/xCell).
Cell proximity analysis
The distance between cDC1s and CD8+ T cells was analyzed using QuPath software (RRID: SCR_018257) as follows: XCR1+ and CD8+ cells in IF images captured using a 400× magnification field of view were classified using the random forest algorithm in this software. In brief, annotations to the fluorescent stains of representative multiple images were assigned for cell classification based on multiple measurements. The distance from each XCR1+ cell to the nearest CD8+ cell was then calculated using detection centroid distance. The calculated distances and their averages were plotted using a GraphPad Prism (RRID: SCR_002798).
Bone marrow–derived cell culture and analysis
Bone marrow (BM) cells were collected from the femurs of 6 to 8 weeks old mice and cultured in BM-derived cell culture medium, which was composed of RPMI 1640, 1% penicillin–streptomycin (Cat. #15070063, Gibco), 55 μmol/L 2-mercaptoethanol (M6250, Sigma-Aldrich), 10% FBS (Cat. #100-106, Gemini Bio-Products), GM-CSF (20 ng/mL, Peprotech, Cat. #315-03), and IL-4 (10 ng/mL, Peprotech, Cat. #214-14). The adherent cells were collected on days 3 and 6. Both nonadherent cells in the supernatant and loosely adherent cells were collected on day 9 and used for experiments. Collected DCs were stained following the same method as in the vivo study (see “Flow cytometry analysis”). Collected cells were incubated with anti–PD-1 (10 μg/mL, Dan G. Duda – Massachusetts General Hospital, Cat. # mPD1-4H2-mg1-D265A clone 6A1_RAS_Ab, RRID: AB_3476760) and anti-CXCR4 (1 μmol/L, Dan G. Duda – Massachusetts General Hospital, Cat. # mCXCR4.8-mG1 clone 6A4_RAS_Ab, RRID: AB_3477163) for 24 hours and collected for flow cytometric analyses. mAbs used for flow cytometric analysis were, CD45 (BioLegend, Cat. # 103155, RRID: AB_2650656), CD3 (BioLegend, Cat. # 100349, RRID: AB_2565841), MHC class II (BioLegend, Cat. #107647, RRID: AB_2565978), CD11b (BioLegend, Cat. # 101226, RRID: AB_830642), CD11c (BioLegend, Cat. # 117307, RRID: AB_313776), F4/80 (BioLegend, Cat. #, RRID: AB_893502), B220 (BioLegend, Cat. #103222, RRID: AB_313005), Ly6G (BioLegend, Cat. #127622, RRID: AB_10643269), CD80 (BD Biosciences, Cat. #740888, RRID: AB_2740537), CD86 (BD Biosciences, Cat. #564198, RRID: AB_2738663), and CXCR3 (BioLegend, Cat. # 126511, RRID: AB_1088994).
Statistical analysis
Statistical analyses were performed using GraphPad Prism software (GraphPad Software, Inc. Version 10.2.0), and data were presented as the mean ± SD. Specifically, the two-tailed unpaired Student t test (parametric) was used for all comparisons between two groups, except for blood test comparisons, which were analyzed using the Mann–Whitney test (nonparametric). A one-way ANOVA, followed by the Tukey post hoc test (parametric), was used to compare all three or more groups. The Kaplan–Meier method was used to generate survival curves underlying the log-rank test and Cox proportional hazard model, and HR and 95% confidence interval were calculated for statistical survival analyses for murine models. To analyze changes in body weight and tumor volume, we used two-way ANOVA with the Bonferroni test (comparisons between two groups, parametric) or Tukey test (comparisons between three or more groups, parametric).
Data availability
Data are available upon reasonable request from the corresponding author (Dan G. Duda). The RNA-seq data are available in the NCBI Gene Expression Omnibus database under the accession number GSE280861.
Results
Combined anti-CXCR4/anti–PD-1 treatment induces significant anticancer effects
We first investigated the feasibility and effectiveness of combining anti-CXCR4 and anti–PD-1 treatments (CXCR4/PD-1) in established preclinical murine HCC models, including the orthotopically implanted RIL-175 (p53/Hras mutant) and HCA-1 HCC models with induced liver damage (19, 21). Tumor growth was monitored by ultrasound imaging. After HCCs were established (i.e., reached 5–6 mm diameter), tumor-bearing mice were randomized to treatment with either an anti–PD-1 (10 mg/kg i.p. every 3 days), an anti-CXCR4 (10 mg/kg i.p. every 3 days), combination of anti–PD-1 and anti-CXCR4 (10 mg/kg i.p. every 3 days), or an isotype control IgG (10 mg/kg, every 3 days; Supplementary Fig. S1A). We found that the combination therapy was feasible, with no apparent toxicity (based on mouse appearance and absence of body weight loss). Moreover, the combination therapy group showed a significant reduction of tumor volume and an increase in median overall survival compared with all other treatment groups both in the anti–PD-1 therapy–sensitive model (RIL-175 in C57Bl/6 mice) and the anti–PD-1 therapy–resistant model (HCA-1 in C3H mice; Fig. 1A and B; Supplementary Fig. S1B and S1C). Finally, we tested the feasibility and efficacy of dual CXCR4/PD-1 blockade in autochthonous murine HCCs induced using Cre-adenovirus injection in Mst1−/−Mst2f/− mice with induced liver damage (Supplementary Fig. S2A; ref. 19). The combination therapy resulted in a significant tumor growth suppression effect, as shown by ultrasound imaging and pathologic evaluation of liver nodules (Fig. 1C and D; Supplementary Fig. S2B). These results demonstrate that anti-CXCR4/anti–PD-1 combination therapy is feasible in murine models of HCC with liver damage, which mimics the clinical presentation of some human HCCs, and provides superior outcomes compared with anti–PD-1 therapy alone.
Combined anti-CXCR4/anti–PD-1 antibody treatment showed significant anticancer effects and survival benefits in orthotopic models of HCC with liver damage. A and B, Tumor growth delay (A) and overall survival distributions (B) after a 3-week treatment with the anti-CXCR4 antibody, anti–PD-1 antibody, their combination, or IgG (isotype control) administered intraperitoneally at doses of 10 mg/kg diluted in PBS thrice weekly in RIL-175 murine HCC orthotopic grafts in C57Bl/6 mice with CCl4-induced liver fibrosis. C and D, Confirmation of potent anticancer effects for the combination therapy in the autochthonous HCC model (induced in Mst1−/−Mst2F/− mice with underlying liver fibrosis), as measured by tumor volume using ultrasound (C) and the tumor proliferation index (D) at day 14. N = 8 or 9 mice in all groups, with experiments performed at least in duplicate. Data are the mean and SD (A, C, and D); two-way ANOVA (A) or log-rank test (B) or one-way ANOVA with Tukey’s test (C and D). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Combo, combination; Ctrl, control.
Combined anti-CXCR4/anti–PD-1 antibody treatment showed significant anticancer effects and survival benefits in orthotopic models of HCC with liver damage. A and B, Tumor growth delay (A) and overall survival distributions (B) after a 3-week treatment with the anti-CXCR4 antibody, anti–PD-1 antibody, their combination, or IgG (isotype control) administered intraperitoneally at doses of 10 mg/kg diluted in PBS thrice weekly in RIL-175 murine HCC orthotopic grafts in C57Bl/6 mice with CCl4-induced liver fibrosis. C and D, Confirmation of potent anticancer effects for the combination therapy in the autochthonous HCC model (induced in Mst1−/−Mst2F/− mice with underlying liver fibrosis), as measured by tumor volume using ultrasound (C) and the tumor proliferation index (D) at day 14. N = 8 or 9 mice in all groups, with experiments performed at least in duplicate. Data are the mean and SD (A, C, and D); two-way ANOVA (A) or log-rank test (B) or one-way ANOVA with Tukey’s test (C and D). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Combo, combination; Ctrl, control.
Combining CXCR4 and PD-1 blockade increases antigen presentation and intratumoral DCs
We next repeated the experiment and collected tumor tissues from the treated mice bearing time-matched RIL-175 murine HCC. To understand the underlying mechanisms of how the combination therapy remodels the immune TME, we collected tumors from the RIL-175 models that received the isotype control, single treatment, or combination treatment and performed RNA-seq analysis. The gene expression of the tumors in different groups was separated from each other by using principal component analysis; the principal component (PC) analysis suggested differential remodeling of the TME after treatment (Fig. 2A). The PC1 contributed 90.3% of the variable, and most genes in the PC1 were associated with antigen processing and presentation. Next, we compared the DEGs between each treatment group and the isotype control group. We identified 100 DEGs that were only observed after the combination therapy (Fig. 2B). GSEA of these 100 DEGs showed a significant increase in pathways related to the T-cell receptor signaling pathway (Fig. 2C). Furthermore, T-cell activation–related signaling pathways were elevated in the combination-therapy group (Fig. 2D). Next, the expression patterns of genes related to DCs, which play a critical role for tumor antigen presentation, were compared among the groups. The combination-therapy group exhibited higher expression levels of cDC1-related genes (Fig. 2E). In addition, genes related to DC maturation and migration were upregulated in the combination-therapy group (Supplementary Fig. S3). These data suggest that anti-CXCR4 and anti–PD-1 combination therapy enhances antigen presentation and DC frequency in the TME of HCC.
Transcriptomic analysis data from RIL-175 murine HCC after anti-CXCR4 therapy combined with PD-1 blockade. A, Principal component analysis of gene expression profiles of murine HCC tissues from the control and treatment groups. B, Venn diagram representing the significantly DEGs between the treatment and control groups. C, GSEA analysis showing T-cell receptor signaling pathway upregulated in the combination therapy. D, Bar plot representing the top 10 most upregulated or downregulated GO terms using GSEA. The GO terms are ranked by the NES. The color of the bar plot indicates the transformed P values. E, The heatmap represents the gene expression of DC-related genes. Combo, combination; Ctrl, control; GO, Gene Ontology; GOBP, Gene Ontology biological process; KEGG, Kyoto Encyclopedia of Genes and Genomes; NES, normalized enrichment score; pDC, plasmacytoid DC.
Transcriptomic analysis data from RIL-175 murine HCC after anti-CXCR4 therapy combined with PD-1 blockade. A, Principal component analysis of gene expression profiles of murine HCC tissues from the control and treatment groups. B, Venn diagram representing the significantly DEGs between the treatment and control groups. C, GSEA analysis showing T-cell receptor signaling pathway upregulated in the combination therapy. D, Bar plot representing the top 10 most upregulated or downregulated GO terms using GSEA. The GO terms are ranked by the NES. The color of the bar plot indicates the transformed P values. E, The heatmap represents the gene expression of DC-related genes. Combo, combination; Ctrl, control; GO, Gene Ontology; GOBP, Gene Ontology biological process; KEGG, Kyoto Encyclopedia of Genes and Genomes; NES, normalized enrichment score; pDC, plasmacytoid DC.
Intratumoral cDC1s mediate the benefit of adding CXCR4 blockade to anti–PD-1
To verify the results of transcriptomic data and examine the causal role of DC subsets, we first performed a flow cytometric analysis using time-matched tumor tissues from the four treatment groups. We found no significant difference in the overall frequency of cDC1s or cDC2s between the groups (Supplementary Fig. S4A and S4B). However, when examining the distribution and expansion of DCs in the TME by histologic analysis, we found a significant increase in cDC1s in the center of tumors but not in the edge of the tumors after the combination therapy, suggesting that the combination therapy promotes the infiltration of cDC1s. This effect was detected in orthotopic and autochthonous murine HCC models (Fig. 3; Supplementary Fig. S5A and S5B). Moreover, we found that the expression of CXCR4 on cDC1s was predominantly detected in tumors compared with the surrounding liver tissue and that anti–PD-1 treatment enhanced further CXCR4 expression in intratumoral cDC1s (Supplementary Fig. S5C). This suggests that the added benefit of anti-CXCR4/anti–PD-1 treatment is primarily cDC1-dependent.
Classical cDC1 subset infiltration inside HCC increases after combined treatment with CXCR4 and anti–PD-1 blockade. A–D, Combining the anti-CXCR4 antibody with anti–PD-1 induces a significant increase in the accumulation of cDC1 (defined as XCR1, CD11c double-positive cells) orthotopic grafted (A and B) and autochthonous (C and D) HCC models. All the mice with orthotopic grafted models were sacrificed on day 8, and the mice with autochthonous tumors were sacrificed on day 13 after starting the treatment, at the time of objective immune responses. Representative IF in A and C (arrowheads point to cDC1) and quantitative analysis of IF data in B and D. N = 7–8 mice in all groups. Scale bars, 50 μm (A) and 100 μm (C). Data are the mean and SD (B and D); one-way ANOVA with Tukey’s test (B and D).*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Combo, combination; Ctrl, control; DAPI, 4′,6-diamidino-2-phenylindole.
Classical cDC1 subset infiltration inside HCC increases after combined treatment with CXCR4 and anti–PD-1 blockade. A–D, Combining the anti-CXCR4 antibody with anti–PD-1 induces a significant increase in the accumulation of cDC1 (defined as XCR1, CD11c double-positive cells) orthotopic grafted (A and B) and autochthonous (C and D) HCC models. All the mice with orthotopic grafted models were sacrificed on day 8, and the mice with autochthonous tumors were sacrificed on day 13 after starting the treatment, at the time of objective immune responses. Representative IF in A and C (arrowheads point to cDC1) and quantitative analysis of IF data in B and D. N = 7–8 mice in all groups. Scale bars, 50 μm (A) and 100 μm (C). Data are the mean and SD (B and D); one-way ANOVA with Tukey’s test (B and D).*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Combo, combination; Ctrl, control; DAPI, 4′,6-diamidino-2-phenylindole.
To determine whether cDC1s mediate the anticancer effect of the combined anti–CXCR4/PD-1 therapy, we next performed survival studies using Batf3−/−/C57Bl/6 mice, which specifically lack cDC1, and compared the results we obtained using C57Bl/6 WT mouse (Supplementary Fig. S6A). RIL-175 murine HCC-bearing Batf3−/−/C57Bl/6 mice showed a deficiency of cDC1, as confirmed by flow cytometry (Supplementary Fig. S6B). We found that the therapeutic efficacy of the combination treatment was abrogated in mice deficient in cDC1s, leading to a mortality rate comparable to that observed in the WT control group, in part due to the increased local and distant HCC dissemination in Batf3−/−/C57Bl/6 mice (Fig. 4A; Supplementary Fig. S6C–S6F).
Combined treatment with anti-CXCR4 and anti–PD-1 antibodies promoted the activation and proliferation of cytotoxic T cells via cDC1. A, The survival benefit of the combination treatment in murine HCC was compromised when tumors were grown in Batf3 knockout (cDC1-deficient) mice. B–G, Combination therapy using anti-CXCR4 and anti–PD-1 antibodies induces proliferation and accumulation of activated cytotoxic T cells in the TME of HCC: frequency (B) and absolute numbers (C) of tumor-infiltrated CD8 T cells; frequency (D), representative plots (E), and absolute counts (F) of GzmB+ and Ki67+ proliferating (G) CD8+ T cells in HCC tissues from mice in the control IgG, monotherapy, or combination-treatment groups. Data are the mean ± SD. (B–G). H and I, Representative fluorescent photomicrographs of orthotopic grafted HCC models of mice treated with each demonstrated condition stained with DAPI (blue), anti-CD8 antibody (green), and anti-XCR1 antibody (red; H) and the average distance between CD8+ T cells and DC1s in HCC tissues (I). J, The frequency of CXCR3+ cells in DCs in vitro. N = 7–10 mice in all groups. Scale bars, 50 μm (H). Most experiments were performed at least in duplicate. One-way ANOVA with the Tukey test (A and B). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Combo, combination; Ctrl, control; DAPI, 4′,6-diamidino-2-phenylindole; NS, not significant; GzmB, Granzyme B.
Combined treatment with anti-CXCR4 and anti–PD-1 antibodies promoted the activation and proliferation of cytotoxic T cells via cDC1. A, The survival benefit of the combination treatment in murine HCC was compromised when tumors were grown in Batf3 knockout (cDC1-deficient) mice. B–G, Combination therapy using anti-CXCR4 and anti–PD-1 antibodies induces proliferation and accumulation of activated cytotoxic T cells in the TME of HCC: frequency (B) and absolute numbers (C) of tumor-infiltrated CD8 T cells; frequency (D), representative plots (E), and absolute counts (F) of GzmB+ and Ki67+ proliferating (G) CD8+ T cells in HCC tissues from mice in the control IgG, monotherapy, or combination-treatment groups. Data are the mean ± SD. (B–G). H and I, Representative fluorescent photomicrographs of orthotopic grafted HCC models of mice treated with each demonstrated condition stained with DAPI (blue), anti-CD8 antibody (green), and anti-XCR1 antibody (red; H) and the average distance between CD8+ T cells and DC1s in HCC tissues (I). J, The frequency of CXCR3+ cells in DCs in vitro. N = 7–10 mice in all groups. Scale bars, 50 μm (H). Most experiments were performed at least in duplicate. One-way ANOVA with the Tukey test (A and B). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Combo, combination; Ctrl, control; DAPI, 4′,6-diamidino-2-phenylindole; NS, not significant; GzmB, Granzyme B.
Furthermore, we performed in vitro experiments to examine whether the combination treatment had a direct effect on DCs. We expanded DCs from BM and treated them with the isotype control, anti-CXCR4, anti–PD-1, or the combination of anti-CXCR4 and anti–PD-1. We found that combination treatment increased the expression levels of the DC maturation markers CD80 and CD86 compared with the other treatments (Supplementary Fig. S5D and S5E). These data demonstrate that the added anticancer effect and survival benefit induced by adding CXCR4 blockade to PD-1 blockade is mediated by cDC1s.
Combined blockade of CXCR4 and PD-1 increases the activation of cytotoxic T cells
Tumor antigen presentation is mainly mediated by cross-presentation between cDC1s and CD8+ T cells, and it leads to more functional CD8+ T cells and anticancer effects. Thus, we examined whether the combination therapy activated cross-presentation by reprogrammed DCs and elicited CD8+ T-cell activation in the TME of HCC by flow cytometric analysis using size-matched tumor tissues. The results demonstrated that the combination therapy significantly increased the frequency and number of CD8+ tumor-infiltrating lymphocytes (Fig. 4B and C; Supplementary Fig. S7). Tumors from the combination therapy group also showed a higher frequency of GrzmB+CD8+ and IFNγ+CD8+ T cells than the monotherapy groups (Fig. 4D–F; Supplementary Fig. S8A–S8C). In addition, the combined blockade of CXCR4 and PD-1 increased the frequency of intratumoral Ki67+CD8+ T cells (Fig. 4G; Supplementary Fig. S8D). Next, we examined CD8+ T-cell distribution with respect to cDC1s. Histologic analyses showed an increased frequency of CD8+ T cells in the center of the tumor that were in closer proximity to cDC1s after combination treatment versus the other groups (Fig. 4H and I). In line with this finding, in vitro analysis of DCs showed increased expression of CXCR3, a receptor for chemotactic factors, after anti-CXCR4/anti–PD-1 therapy compared with anti–PD-1 monotherapy and control IgG groups (Fig. 4J). These findings show that adding anti-CXCR4 therapy to standard PD-1 blockade can increase the activation, expansion, and migration of intratumoral CD8+ T cells and cDC1s (Supplementary Fig. S9).
Discussion
Although recent advancements in combination therapy using angiogenesis inhibitors and immunotherapy have surpassed sorafenib as the standard of care for HCC (3), the therapeutic benefit remains limited, with almost all patients experiencing recurrences.
Several studies have investigated using CXCR4 signaling antagonists in treating solid tumors, including HCC. Despite this, these studies had limitations. One of the major limitations of CXCR4 antagonists for treating solid tumors is their lack of specificity and the short half-life of available drugs such as AMD3100 (22). To address these limitations, we evaluated the impact of a combination therapy of anti-CXCR4 and anti–PD-1, a standard of care for patients with HCC, in orthotopic and autochthonous preclinical HCC models. Our data suggest that anti-CXCR4 enhances the anticancer effects of anti–PD-1 therapy by increasing infiltration and accumulation of cDC1s and activated CD8+ T cells in the TME of HCC. A recent study showed that CXCR4 inhibition in combination with ICI enhances T-cell retention in the TME, which is consistent with our results (23).
cDC1s are well-known to play an essential role in anticancer immunity, but their role in the TME is not fully characterized and may depend on the organ microenvironment. Recent studies have demonstrated that cDC1s produce chemokines such as CXCL9/10, as seen in our RNA-seq analysis, which enhances the intratumoral infiltration of effector T cells. This, in turn, can improve the efficacy of PD-1/PD-L1 blockade (24). In addition, we recently showed that recurrent (metastatic) HCCs lack PD-L1 expression and have poor antigen presentation function in patients (25). This is consistent with the findings that tumors growing in the liver tissue contain fewer DCs, partly responsible for the resistance to anti–PD-1 therapy in colorectal metastases or melanoma (10, 25). In addition to chemokine production, effective antigen presentation by DCs is crucial for activating CD8+ T cells, which play a key role as effector cells in anticancer immunity (26). Thus, mediating DC accumulation and localization within the TME is of great interest therapeutically, particularly for anti–PD-1 therapy. The spatial distribution of immune cells within tumors is a critical aspect of the TME, and antigen-presenting cells exhibit intratumoral spatial heterogeneity in various cancers (27). In this study, we demonstrated that combination therapy with anti-CXCR4 and anti–PD-1 increased the frequency and proximity of cDC1s and CD8+ T cells, especially inside the HCC lesions. Immune cell infiltration into the tumor center is critical for interacting with tumor cells. Indeed, we found that the combination therapy led to increased infiltration and activation of CD8+ T cells and effective anticancer results, which were compromised entirely in mice lacking cDC1s.
In summary, our study demonstrates that combining anti-CXCR4 and anti–PD-1 is more effective than anti–PD-1 alone in orthotopic (grafted and autochthonous) murine HCC models and that the increased DC infiltration in the TME of HCC mediates the benefit of this combination therapy. This effect is associated with increased infiltration and activation of CD8+ T cells and prolonged survival. These insights into the role of CXCR4 in DCs may help improve the therapeutic efficiency of ICI-based therapy, especially against tumors growing in the liver, in which DCs are scarce.
Authors’ Disclosures
S. Morita reports grants from Harvard Medical School, Massachusetts General Hospital, during the conduct of the study. K. Shigeta reports grants from The Uehara Memorial Foundation during the conduct of the study and grants from Japan Society of Laparoscopic Colorectal Surgery and Japan Society for the Promotion of Science outside the submitted work. T. Ando reports grants from Takeda Science Foundation during the conduct of the study. D.G. Duda reports grants from Bristol Myers Squibb during the conduct of the study and grants from Bayer, Surface Oncology, and Exelixis outside the submitted work. No disclosures were reported by the other authors.
Authors’ Contributions
S. Morita: Conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, writing–original draft. P.-J. Lei: Data curation, software, formal analysis, visualization, methodology, writing–review and editing. K. Shigeta: Conceptualization, formal analysis, investigation, methodology, writing–review and editing. T. Ando: Investigation, methodology, writing–review and editing. T. Kobayashi: Formal analysis, investigation, visualization, methodology, writing–review and editing. H. Kikuchi: Investigation, methodology, writing–review and editing. A. Matsui: Validation, investigation, visualization, methodology, writing–review and editing. P. Huang: Resources, investigation, writing–review and editing. M.J. Pittet: Conceptualization, data curation, writing–review and editing. D.G. Duda: Conceptualization, resources, data curation, supervision, funding acquisition, visualization, methodology, project administration, writing–review and editing.
Acknowledgments
The authors would like to express their sincere gratitude to T. Mempel for useful discussions and M. Duquette, A. Khachatryan, H. Taniguchi, M. Galvan, and S. Roberge (all MGH) for outstanding technical support. This study was supported by Bristol Myers Squibb through a sponsored research agreement (D.G. Duda); Bristol Myers Squibb provided the murine antibodies for the treatment and financial research support for the preclinical studies. This work was supported by NIH grant R01CA260857 (D.G. Duda), NIH grant R01CA254351 (D.G. Duda), NIH grant R01CA247441 (D.G. Duda), NIH grant R03CA256764 (D.G. Duda), NIH P01CA261669 (D.G. Duda), Department of Defense Peer Reviewed Cancer Research Program (PRCRP) grant W81XWH-19-1-0284 (D.G. Duda), Department of Defense PRCRP grant W81XWH-21-1-0738 (D.G. Duda), Japan Society for the Promotion of Science (JSPS) Fellowship (S. Morita), Takeda Fellowship (T. Ando), MGH Fund for Medical Discovery Fundamental Research Fellowship Award (S. Morita), and Uehara Memorial Foundation Fellowship (K. Shigeta).
Note: Supplementary data for this article are available at Cancer Immunology Research Online (http://cancerimmunolres.aacrjournals.org/).
References
Supplementary data
Supplementary Methods
Supplementary Figure S1
Supplementary FIgure S2
Supplementary Figure S3
Supplementary Figure S4
Supplementary Figure S5
Supplementary Figure S6
Supplementary Figure S7
Supplementary Figure S8
Supplementary Figure S9