Blockade of PD-1/PD-L1 Pathway Enhances the Antigen-Presenting Capacity of Fibrocytes

Fibrocytes, a distinct population of collagen-producing, monocyte-derived cells, exhibit diverse morphological and molecular characteristics of both macrophages and fibroblasts (1–3). In the peripheral circulation, fibrocytes account for ∼0.5% of the leukocyte pool. Morphologically, fibrocytes appear as adherent, large, spindle-shaped cells on in vitro culture. At the time of writing, there is no single specific marker to identify fibrocytes, the combination of intracellular collagen staining along with the expression of hematopoietic marker CD45 and chemokine receptor CXCR4 is thought to be appropriate to isolate fibrocytes (2, 4–8).

In health, fibrocytes home to the sites of injury and promote wound healing by producing proinflammatory cytokines (IL-6 and IL-8), chemokines (CXCL3 and CXCL4), and collagen (2, 9, 10). However, fibrocytes have also been reported to involve in the pathogenesis of various fibrotic diseases, such as interstitial lung disease, bronchial asthma, cardiovascular disease, and rheumatoid arthritis (6, 11–14). In addition, fibrocytes have been demonstrated to function as a potent APC (15). The surface expression of immune molecules (CD80 and CD86) and MHC class II are thought to potentiate the immune response by fibrocytes (16–18).

However, there is an increasing body of evidence suggesting that fibrocytes from cancer patients suppress the T cell function and, as a result, support tumor progression. Fibrocytes instigate lung metastasis by inducing the influx of Ly-6C⁺ monocytes via CCL2 (19). The production of indolamine oxidase by fibrocytes is another of their immunosuppressive properties (20). Based on these data, it is unclear whether fibrocytes stimulate or suppress the T cell function.

We and others have reported the proangiogenic property of fibrocytes via the production of angiogenic factors, including fibroblast growth factor 2 (FGF-2) (21, 22). In the tumor microenvironment, under treatment with the anti–VEGF-A Ab bevacizumab, we found a remarkable increase in tumor-infiltrating fibrocytes that mediated acquired resistance in mice and humans (21).

Recent advances in cancer immunotherapy have resulted in the development of immune checkpoint inhibitors (ICIs) of several Abs against CTLA4 and programmed cell death 1 (PD-1) and its ligand PD-L1, which have been approved for various malignancies (23, 24). In addition, clinical trials involving combination therapy with anti–PD-1/PD-L1 Abs and other modalities have demonstrated promising efficacy against some cancers (25, 26). Among these, the combination of ICIs with antiangiogenic agents, which eventually increase the infiltration of fibrocytes into tumors, has been expected, whereas an immunological background showing additive or synergistic effects remains unclear.

It is of importance to understand the role of immune checkpoint molecules expressed in various types of immune cells, including fibrocytes in the tumor microenvironment. In the current study, we analyzed the expression and role of PD-L1 in the Ag-presenting capacity of both human and murine fibrocytes in vitro. We also compared the Ag-presenting and T cell–stimulatory ability of fibrocytes with those of dendritic cells (DCs) and several types of macrophages. Finally, we examined the immune regulatory function of PD-L1–expressing cancer-associated fibrocytes derived from PBMCs of patients with human lung adenocarcinoma and MC38 tumor in mice.

C57BL/6 female and BALB/C male mice were purchased from The Jackson Laboratory and Charles River Laboratory, respectively. OT-I Rag1^−/− transgenic mice expressing a TCR specific for OVA_257-264/Kb were kindly provided by Prof. K. Yasutomo (Tokushima University, Tokushima, Japan). All mice were maintained under specific pathogen-free conditions in the animal research center of Tokushima University, and all animal experiments were approved by the animal research committee of Tokushima University and performed in accordance with our institution’s guidelines for animal care and use.

Human fibrocytes were isolated according to previously published methods (7). Informed consent was obtained from all volunteers, and the protocol was approved by the institutional review board of Tokushima University Hospital (no. 2838). Briefly, PBMCs were isolated from the peripheral blood of healthy volunteers using Ficoll density centrifugation. The isolated cells were cultured in DMEM supplemented with 20% FBS, penicillin, and streptomycin on fibronectin (no. 1030-FN; R&D Systems, Minneapolis, MN)–coated, 150-mm cell culture dishes (BD Pharmingen, Franklin Lakes, NJ). After 7–8 d of incubation at 37°C with 5% CO₂, only strongly adherent cells (washed three times with sterile PBS) were determined to be circulating fibrocytes by a flow-cytometric analysis. A flow-cytometric analysis revealed that harvested fibrocytes expressed CD45, collagen I, and CXCR4 and showed >90% purity.

Fibrocytes were also isolated from patients with lung cancer (adenocarcinoma) without any chemotherapy/radiotherapy/intervention.

The isolation of murine fibrocytes follows a previously published protocol (4). In brief, lungs removed from female C57BL/6 mice were minced with scissors and incubated with DMEM containing 1 mg/ml BSA (no. A9418; Sigma-Aldrich, St. Louis, MO), 1 mg/ml collagenase type 1 (no. LS004196; Worthington, Lakewood, NJ) and 500 μg/ml DNase 1 (no. LS002139; Worthington) for 1 h at 37°C. The single-cell suspensions from whole lungs were incubated in DMEM supplemented with 20% FBS in 150-mm, fibronectin-coated dishes. After 5 d, the trypsinized cells were incubated with anti-CD45 Ab coupled to magnetic beads (130-052-301; Miltenyi Biotec, Auburn, CA) for 15 min at 4–8°C. CD45⁺ populations were purified using an autoMACS instrument (Miltenyi Biotec) according to the manufacturer’s instructions. In flow cytometry, intracellular staining revealed that more than 95% of these cells were positive for collagen 1; these were therefore used as fibrocytes.

Detection of human fibrocytes by flow cytometry has been described previously (7, 27). Briefly, FITC-conjugated anti-CD45 (1:25 dilution, HI30; BioLegend, San Diego, CA) and PE-conjugated anti-CXCR4 Abs (1:25 dilution, 12G5; eBioscience, San Diego, CA) were used to stain the surface molecules before proceeding to intracellular staining. Cells were then fixed and permeabilized (BD Pharmingen) for staining with biotin-conjugated anti–collagen I Ab (1:100 dilution, no. 600-406-103; Rockland, Limerick, PA) and PE-Cy7 streptavidin (1:500 dilution, no. 405206; BioLegend).

The expression of immune checkpoint molecules on fibrocytes has been documented with regard to anti-human PE CD274 (PD-L1, MIH-1), CD273 (PD-L2, MIH18), CD152 (CTLA-4, L3D10), CD366 (TIM-3, F38-2E2), CD40 (5C3), CD80 (2D10.4), CD86 (IT2.2), HLA-ABC (W6/32), and HLA-DR (LN3) Abs and mouse IgG (mIgG) 1 k (P3.6.2.8.1) (eBioscience). Anti-human PE Galectin-9 (9M1-3), CD54 (HA58) Abs (BioLegend), and anti-human PE-Cy7 CD279 (PD-1, EH12.1) Ab and mIgG1 (MOPC-21) (BD Pharmingen), anti-human PE CD3 (UCHT1), CD19 (HIB19), and CD14 (61D3) Abs (eBioscience) were also used.

To identify murine fibrocytes, cells were fixed and permeabilized for intracellular staining. Before permeabilization, PE-conjugated Ab to CD45 (1:25, 30-F11; eBioscience) and FITC-conjugated Ab to CXCR4 (1:25 dilution, no. 557967; BD Biosciences) were applied. Biotin-conjugated anti–collagen I Ab (1:100 dilution, 600-406-103; Rockland) and PE-Cy7 Streptavidin (1:500 dilution, 405206; BioLegend) were applied to detect intracellular collagen simultaneously with CD45 and CXCR4.

For murine immune checkpoint expression, anti-murine PE CD274 (MIH5), CD273 (122), CD40 (1C10), CD86 (GL1), and CD54 (YN1/1.7.4) Abs and rat IgG2a (eBR2a) (eBioscience), and anti-murine PE CD80 (1G10; life Technology, Carlsbad, CA), PE Galectin-9 (RG9-35; BD Pharmingen), PE MHC-H-2K^b (28-8-6), PE-Cy7 anti-mouse I-A/I-E (M5/114.15.2), PE CD152 (UC10-4B9), PE CD366 (TIM-3) and PE CD279 (RMPI-14) Abs, and rat IgG2b (RTK 4530) (BioLegend) were used.

For the detection of DCs, anti-human PE CD83 (HB15e; eBioscience) and anti-murine PE CD11c (N418; BioLegend) Abs were used, respectively, for humans and mice. Monocyte-derived macrophage polarization to M1 and M2 was documented with anti-human PE CD14 (61D3), CD80 (2D10.4), CD86 (IT2.2), CD163 (GH1/61), and CD206 (19.2) Abs (eBioscience) in humans and anti-mouse FITC F4/80 (BM8), PE CD80 (1G10), CD86 (GL1), CD206 (MR6F3; eBioscience), CD200R (OX-110; BioLegend), and CD11b Abs (M1/70; BD Pharmingen) in mice. The stained cells were analyzed by flow cytometry using a BD LSRFortessa (BD Biosciences) for acquisition and were analyzed with the FlowJo software program (Tree Star, Ashland, OR).

In human MLR, CD8⁺ T cells were isolated from the PBMCs of healthy donors with a CD8⁺ T cell isolation kit (no. 130-045-201; Miltenyi Biotec). Human CD8⁺ T cells were plated in a 96-well plate (flat bottom; Falcon) coated with anti-CD3 Ab (OKT3, 5 μg/ml, 37°C for 2–3 h; eBioscience) and/or soluble anti-CD28 Ab (CD28.2, 2 μg/ml; eBioscience) at a density of 1 × 10⁵ cells per well. Fibrocytes or DCs were added to CD8⁺ T cells at 1 × 10⁴ cells per well after irradiation (25 Gy) with a Hitachi Medico MBR-1520A-TW (Hitachi, Chiba, Japan) or treatment with mitomycin C (25 μg/ml) at 37°C for 30 min (no. M4287; Sigma-Aldrich). Control cultures contained CD8⁺ T cells, fibrocytes, or DCs alone.

In murine MLR, splenocytes were used after depleting RBCs with RBC lysis buffer (no. R7757; Sigma-Aldrich). CD8⁺ splenic T cells were harvested with a CD8a⁺ T cell isolation kit, as described (no. 130-104-075; Miltenyi Biotec). CD8⁺ T cells (2 × 10⁵ per well) were stimulated with fibrocytes or DCs (2 × 10⁴ per well) in a 96-well, flat-bottom plate (Falcon) coated with anti-CD3e Ab (145-2C11, 10 μg/ml) and/or soluble anti-CD28 Ab (37.51, 2 μg/ml) (eBioscience).

After 72 and 48 h of coculture, cells were pulsed with 1 μCi/well of ³H-thymidine (³H-TdR) (PerkinElmer, Boston, MA) for 15 and 12 h before termination in human and murine MLRs, respectively, and DNA synthesis was measured by a ³H-TdR incorporation with a β-scintillation counter (Aloka instrument).

In blocking experiments for MLRs, the antagonistic Abs, for CD80 (2D10.4), CD86 (IT2.2), or CD54 (HA58) (10 μg/ml) in humans and CD80 (16-10A1), CD86 (GL1), or CD54 (eBioKAT-1) in mice were used with the respective control Abs, mIgG1 (P3.6.2.8.1; eBioscience), rat IgG2a (eBR2a), or Armenian Hamster IgG (eBio299Arm). In all experiments, fibrocytes were preincubated with the corresponding Abs at 37°C for 1 h before being added to CD8⁺ T cells.

To determine the effect of PD-L1/PD-1 interaction on CD8⁺ T cell proliferation, recombinant human PD-L1 (hPD-L1)–Fc chimera protein (no. 156-B7; R&D Systems) was added to CD8⁺ T cells (1 × 10⁵ cells per well) stimulated with precoated anti-CD3 Ab and anti-CD28 (1 μg/ml). Likewise, BALB/C splenic CD8⁺ T cells were plated at 2 × 10⁵ cells per well with anti-CD28 (2 μg/ml), which had previously been coated with anti-CD3e Ab (10 μg/ml), and the recombinant murine (rm) PD-L1–Fc chimera protein (no.1019-B7; R&D Systems) was added at the designated concentrations.

To assess the effect of ICIs on fibrocyte-induced CD8⁺ T cell proliferation, we introduced autologous and allogeneic MLR. Autologous studies designated fibrocytes (1 × 10⁴ per well) and CD8⁺ T cells (1 × 10⁵ per well) from the same donor (human) and the same mouse species (C57BL/6). Anti-CD3 Ab was used to mimic antigenic stimulation (OKT3, 5 μg/ml), and (145-2C11, 10 μg/ml) was added to human and murine autologous MLRs. Atezolizumab (a humanized anti–hPD-L1 Ab; Chugai Pharmaceutical, Tokyo, Japan) and anti-DNP human IgG1 (N297A) (Acro Biosystem, Newark, DE) or Ultra-LEAF Purified Human IgG1 isotype control recombinant (QA16A12; BioLegend) were applied at a concentration of 20 μg/ml for preincubation with fibrocytes at 37°C for 1 h in both autologous and allogeneic human MLRs. In allogeneic studies, fibrocytes were exposed to the allogeneic CD8⁺ T cells of different donors (human) and different mouse species. Purified CD8⁺ T cells (2 × 10⁵ per well) isolated from the spleen of male BALB/c mice were cocultured at a 10:1 ratio with female C57BL/6 fibrocytes (2 × 10⁴ per well). Anti–PD-L1 Ab (10F.9G2; Bio X Cell, Lebanon, NH) and LEAF Purified Rat IgG2b isotype control (RTK4530; BioLegend) were used at a concentration of 20 μg/ml for preincubation with fibrocytes in murine MLRs. To explain the PD-1 blocking effect, anti–human PD-1 Ab (EH12.2H7; BioLegend) and mIgG1 k isotype control (P3.6.2.8.1; eBioscience), at concentrations of 10 μg/ml, were preincubated with CD8⁺ T cells for 1 h at 37°C before mixing with fibrocytes. Similarly, anti-mouse PD-1 Ab (BE0146; Bio X Cell) with isotype control rat IgG2a at a concentration of 10 μg/ml (BE0089; Bio X Cell) were used. All experiments described the coculture stands for 72 and 48 h for human and murine MLR, respectively, before the cell proliferation assessment.

Fibrocytes derived from the lungs of OT-I Rag^−/− mice were pulsed with 2 μg/ml Ova peptide (SIINFEKL; Sigma-Aldrich) at 37°C for 4 h in 10% RPMI 1640 and washed extensively five times with sterile PBS before further application. Pulsed fibrocytes were further preincubated with anti–PD-L1 Ab and control rat IgG2b at 37°C for 1 h after recounting. In this study, H-2K^b TRP-2 peptide (SVYFFVWL; MBL Life Science) was used as control (28).

Human peripheral blood-derived monocytes were isolated from healthy donors by magnetic sorting using a human monocyte isolation kit (no. 130-050-201; Miltenyi Biotec) according to the manufacturer’s protocol). CD14-positive cells were stimulated with recombinant human (rh) GM-CSF (no. 074-05603; Wako Pure Chemicals, Osaka, Japan) and IL-4 (no. 098-03964; Wako Pure Chemicals) for 7 d; in the last 48 h, they were additionally stimulated with TNF-α (no. 203-15263; Wako Pure Chemicals) at a concentration of 50 ng/ml to harvest mature DCs.

M1 and M2 macrophages were obtained after 6 d of culturing in medium (RPMI 1640) with 10% FBS supplemented with 50 ng/ml rhGM-CSF or rhM-CSF (no.138-16101; Wako Pure Chemicals), respectively. For coculturing with CD8⁺ T cells, DCs or each type of macrophage were pretreated with mitomycin C at 15 or 25 μg/ml, respectively, for 30 min at 37°C.

To assess the fibronectin effect, these cells were plated in fibronectin-coated (no. 1030-FN; R&D Systems) six-well plates following the above protocol.

Murine monocyte precursor cells were obtained by flushing bone marrow out from the femur and tibia of 6–10-wk age of C57BL/6 mice. To generate murine DCs, bone marrow cells were cultured with murine rGM-CSF (no. 415-ML/CF; R&D Systems) and IL-4 (no. 214-14; PeproTech, Rocky Hill, NJ) for 5 d at a 20 ng/ml, then 2 additional d with GM-CSF, IL-4, and TNF-α (no. 315-01A, as 20 ng/ml; PeproTech). Bone marrow cells were also polarized into M1 and M2 macrophages by culturing with rmGM-CSF and rmM-CSF (no. 416-ML; R&D Systems), respectively, at 20 ng/ml for 6 d (29, 30).

OT-I Rag^−/− murine splenic CD8⁺ T cells were purified using a CD8a⁺ T cell isolation kit. Fibrocytes and DCs were also collected from the same species and pulsed with Ova protein (albumin of chicken egg white, no. 5503; Sigma-Aldrich) in different concentration (1 and 4 mg/ml) for 4 h at 37°C. After washing extensively, fibrocytes or DCs (1 × 10⁴ per well) were added to CD8⁺ T cells (1 × 10⁵ per well) at a 1:5 ratio. In some experiments, pulsed fibrocytes were preincubated with anti–MHC class I (H-2K^b, clone: AF6.88.5.5.3, no. BE0121; Bio X Cell) (10 μg/ml) or anti–PD-L1 Ab (20 μg/ml) with control mIgG2a (clone: C1.18.4, no. BE0085; Bio X Cell) at a 1:10 ratio for 1 h before starting MLR. Fibrocytes were pulsed with Ova protein for 2 h to determine the PD-L1 blocking effect on CD8⁺ T cells by cross-presenting fibrocytes. Cellular proliferation was evaluated by a ³H-TdR incorporation assay.

MC38 tumor cells (1 × 10⁶) were inoculated s.c. into syngeneic C57BL/6 mice. One group (n = 6) was treated with i.p. injection of semaxanib (SU5416, VEGFR-2 inhibitor; Abcam, Cambridge, U.K.), 10 mg/kg/d from day 7 to 21. The tumor tissues collected were minced and suspended to ensure single cells on day 21. The cells were counted by using a hemocytometer, and 6 × 10⁷ cells were subjected to magnetic sorting of CD45⁺ cells with an autoMACS instrument. Tumor-derived CD45⁺ cells were plated in fibronectin-coated, 100-mm dishes supplemented with 20% FBS containing DMEM. After 7 d, adherent cells were labeled with anti-mouse CD45⁺ microbeads again to purify tumor-infiltrating fibrocytes. About 50% of the adherent population expressed collagen I and CXCR4 in addition to CD45 by a flow-cytometric evaluation. The number of fibrocytes counted, treated with anti–PD-L1 Ab as 20 μg/ml (Bio X Cell), and mixed with CD8⁺ T cells originated from healthy BALB/c mice in a 96-well plate to proceed allogeneic MLRs.

Data are presented as the mean ± SEM. The statistical analyses were performed using a one-way ANOVA, followed by Tukey multiple-comparison post hoc test, using the Prism software program (version 8.0; GraphPad Software, La Jolla, CA). All p values <0.05 were considered to indicate statistical significance. All the experiments were triplicated unless otherwise specified.

First, we examined the cell surface expression of coinhibitory checkpoint molecules like PD-L1, PD-L2, PD-1, CTLA-4, Galectin-9, and TIM-3 at the same time as costimulatory molecules such as CD40, CD80, CD86, CD54, and both HLA molecules on murine and human fibrocytes. Flow cytometry revealed that human fibrocytes derived from PBMCs expressed high levels of PD-L1, CD86, and CD54 in addition to both HLA class I (HLA-ABC) and class II (HLA-DR) (Fig. 1A). Regarding other coinhibitory molecules, human fibrocytes expressed the lower level of galectin-9 and TIM-3, which depended on the donor, whereas no PD-L2, PD-1, or CTLA-4 expression was found. Notably, the CD80 expression on fibrocytes was very low. When we analyzed murine fibrocytes, the expression pattern of coinhibitory and stimulatory molecules was similar to that in human fibrocytes, although the expression of galectin-9, TIM-3, CD40, and MHC class II was not detected (Supplemental Fig. 1A). In addition, the level of CD86 on fibrocytes in mice was relatively low in comparison with that in humans. The high expression of PD-L1 is a characteristic finding of fibrocytes.

IFN-γ is a known inducer of PD-L1 via the activation of the JAK/STAT/IRF1 downstream signaling pathway (31). Thus, harvested fibrocytes were cultured with IFN-γ (10 ng/ml) for 48 h in vitro and analyzed by flow cytometry. As a result, treatment with IFN-γ was found to significantly upregulate the expression of PD-L1 (Fig. 1B, Supplemental Fig. 1B). The expression of CD86, CD54, HLA-ABC, and HLA-DR was also enhanced by IFN-γ. However, CD80 failed to upregulate the expression, even with IFN-γ stimulation, in both murine and human fibrocytes (Fig. 1B, Supplemental Fig. 1B).

To examine the stimulatory activity induced by fibrocytes in T cells, fibrocytes were cocultured with CD8⁺ T cells in several experimental conditions, and the activity was measured with ³H-TdR incorporation of T cells. First, we used an allogeneic human system of MLR with fibrocytes and CD8⁺ T cells harvested from PBMCs (Fig. 2A). As shown in Fig. 2A, fibrocytes showed stimulatory activity of the proliferation of CD8⁺ T cells. However, DCs had higher capacity to stimulate CD8⁺ T cells than fibrocytes. Similarly, murine fibrocytes augmented the growth of CD8⁺ T cells, although the activity was lower than that of DCs (Supplemental Fig. 2A).

Next, we determined which costimulatory molecules play an important role in CD8⁺ T cell–stimulatory activity by fibrocytes with the same model used in Fig. 2A. As shown in Fig. 2B, the enhancement of the growth of CD8⁺ T cells by human fibrocytes was blocked with anti-CD86 or -CD54, but not -CD80. These results were confirmed in the murine system, whereas anti-CD80 Ab inhibited the growth of CD8⁺ T cells stimulated by murine fibrocytes (Supplemental Fig. 2B).

The ability of fibrocytes to activate CD8⁺ T cells was also examined using autologous CD8⁺ T cells activated with anti-CD3 and -CD28 Abs. Both human and murine fibrocytes enhanced the growth of CD8⁺ T cells, and the stimulatory activity was blocked by anti-CD86 and -CD54 Abs (Fig. 2C, Supplemental Fig. 2C). However, CD80 was not involved in the activity in human or mice (Fig. 2C, Supplemental Fig. 2C), suggesting that the stimulatory activity of CD8⁺ T cells by fibrocytes was mainly mediated through CD86 and not CD80.

To investigate the involvement of the PD-1/PD-L1 pathway on fibrocyte-mediated activation of CD8⁺ T cells, we first investigated the effect of addition of an exogenous soluble form of hPD-L1 (hPD-L1–Fc). As shown in Fig. 3A, the addition of recombinant hPD-L1–Fc inhibited the activation of human CD8⁺ T cells induced by anti-CD3 Ab in a dose-dependent manner, indicating that the PD-1/PD-L1 pathway was involved in the activation of CD8⁺ T cells. Subsequently, to clarify the effect of PD-L1 blockade in the human system, we examined the effect of a humanized anti–PD-L1 Ab, atezolizumab, which has been approved for cancer immunotherapy in the clinical setting. In autologous MLR, the addition of atezolizumab significantly enhanced the fibrocyte-induced CD8⁺ T cell proliferation (Fig. 3B). In allogeneic MLR, atezolizumab also enhanced the growth of CD8⁺ T cells stimulated by human fibrocytes (Fig. 3C). When we examined the effect of anti–PD-1 Ab, the addition of anti–human PD-1 Ab, like anti–PD-L1 Ab, upregulated the proliferation of CD8⁺ T cells induced by human allogeneic fibrocytes (Fig. 3D). In brief, the interruption of PD-1/PD-L1 interaction augments the growth of CD8⁺ T cells stimulated by fibrocytes.

To further define the PD-L1 function, we cross-checked PD-L1 blocking in murine MLRs. The addition of rmPD-L1–Fc prevented the proliferation of CD8⁺ T cells mediated by anti-CD3 and -CD28 Abs in a dose-dependent manner (Supplemental Fig. 3A). The addition of anti–PD-L1 Ab further enhanced the proliferation of CD8⁺ T cells stimulated by murine fibrocytes and anti-CD3 Ab (Supplemental Fig. 3B). Next, in allogeneic MLR, CD8⁺ T cells derived from BALB/c mice were stimulated by fibrocytes derived from C57BL/6 mice. As shown in Supplemental Fig. 3C, anti–PD-L1 Ab significantly enhanced the fibrocyte-induced proliferation of CD8⁺ T cells. Similarly, anti–PD-1 Ab enhanced the growth of CD8⁺ T cells induced by allogenic fibrocytes (Supplemental Fig. 3D).

We next compared the expression levels and functions of coinhibitory molecules (particularly PD-L1) of fibrocytes with DCs or macrophages. We therefore generated DCs and M1 and M2 macrophages by culturing with GM-CSF and IL-4, GM-CSF alone, or M-CSF alone, respectively, from CD14⁺ monocytes in humans and from bone marrow cells in mice. Like fibrocytes, M2 macrophages showed a spindle shape in culture (data not shown). Flow cytometry also showed that the expression pattern of coinhibitory molecules, including PD-L1, in fibrocytes resembles that in DCs and M1 macrophages (Supplemental Fig. 4). In comparison with DCs, a distinct difference existed in the expression of CD80. DCs expressed high levels of CD80 without any stimulation, whereas fibrocytes did not. These results indicated that the phenotype of fibrocytes was quite unique among macrophage-lineage cells. Next, we examined the functional differences of the PD-L1 expression on fibrocytes, DCs, and M1 and M2 macrophages with allogeneic MLR. Among these cells, DCs showed the highest stimulatory capacity of CD8⁺ T cells, whereas anti–PD-L1 Ab, unlike fibrocytes, did not mediate further enhancement of the growth of CD8⁺ T cells in human or mice (Fig. 4A, 4B, Supplemental Fig. 5A, 5B). Interestingly, anti–PD-L1 Ab tended to augment the growth of CD8⁺ T cells stimulated by M1 macrophages in humans (Fig. 4C), but not in mice (Supplemental Fig. 5C), although the stimulatory activity by both macrophages was apparently lower than that by fibrocytes in humans and mice (Fig. 4C, 4D, Supplemental Fig. 5C, 5D).

To address the question as to whether the fibronectin used in the culture of fibrocytes affects the PD-L1 blocking effects on DCs and M1 and M2 macrophages, we investigated the expression of immune checkpoint molecules on DCs and M1 and M2 macrophages with fibronectin stimulation (Supplemental Fig. 6A). These results indicated that no change was found in their expressions, including PD-L1. Moreover, in an allogeneic MLR, DCs and M1 and M2 macrophages failed to enhance the proliferation of CD8⁺ T cells with atezolizumab (Supplemental Fig. 6B), like the data in Fig. 4.

Moreover, to confirm the role of the PD-1/PD-L1 pathway in Ag-specific CD8⁺ T cell stimulation by fibrocytes, we used OVA-specific OT-1Rag^−/− transgenic mice. In this system, fibrocytes pulsed with OVA peptide (SIINFEKL), but not the relevant TRP-2 peptide, and stimulated the proliferation of CD8⁺ T cells from OT-1Rag^−/− transgenic mice (Fig. 5A). Anti–PD-L1 Ab significantly enhanced the proliferation of CD8⁺ T cells from OT-1Rag^−/− transgenic mice (Fig. 5A). These data suggest that the PD-1/PD-L1 pathway was involved in an Ag-specific activation of CD8⁺ T cells by fibrocytes.

Cross-priming of naive CD8⁺ T cells, in which APCs can phagocytose an exogenous protein and present it via MHC class I, termed cross-presentation, is crucial for adaptive cell-mediated immunity. The cross-presentation of fibrocytes was examined using OT-1Rag^−/− mice. Fibrocytes and bone marrow–derived DCs were pulsed with OVA protein in different concentrations. Importantly, OVA protein–pulsed fibrocytes have shown the ability of CD8⁺ T cell proliferation, like DCs (Fig. 5B), and this activity was strongly inhibited by anti–MHC class I Ab (Fig. 5C). These results clearly indicate that fibrocytes have the ability of cross-presentation, similarly to DCs. Furthermore, anti–PD-L1 Ab significantly enhanced the growth of CD8⁺ T cells mediated by OVA protein–pulsed fibrocytes (Fig. 5D). Again, these data suggest that PD-L1 plays an important role in cross-presentation by fibrocytes.

Finally, we confirmed the APC function of fibrocytes derived from PBMCs of patients with lung adenocarcinoma or MC38 murine tumors. Therefore, we ascertained the blocking effect of atezolizumab on fibrocyte-induced proliferation of CD8⁺ T cells. As shown in Fig. 6A, treatment with atezolizumab reproducibly enhanced the proliferation of CD8⁺ T cells stimulated with fibrocytes derived from three patients with lung adenocarcinoma. To provide further evidence, we purified tumor-infiltrating fibrocytes from MC38 tumor syngeneic with C56BL/6 mice. The flow-cytometric analysis showed that ∼50% of attached cells were CD45 positive, and the phenotype of these cells were compatible to those of fibrocytes (Fig. 6B). Addition of anti–PD-L1 Ab clearly enhanced the proliferation of CD8⁺ T cells stimulated with fibrocytes derived from MC38 tumors with or without treatment of VEGFR-2 inhibitor (semaxanib) (Fig. 6C).

In the current study, we demonstrated that fibrocytes express immune checkpoint molecules, including PD-L1, and the blockade of PD-L1 showed the significant enhancement of the APC function to stimulate CD8⁺ T cells in several models in mice and humans. In addition, we found that fibrocytes isolated from peripheral blood of patients with lung cancer and murine tumor also had APC function that was significantly enhanced by anti–PD-L1 Ab.

Fibrocytes are known as a bone marrow–derived, collagen-producing cells (1–3). Thus, functional analyses have been performed to evaluate their involvement in wound healing as well as fibrogenesis in various organs (4, 5, 7, 10–14). However, fibrocytes have also been demonstrated to show Ag-presenting capacity to activate CD8⁺ T cells (15–18) because they expressed costimulatory molecules such as CD86. In the current study, we further confirmed the expression levels of costimulatory molecules and then examined the expression of coinhibitory molecules, including PD-L1 in fibrocytes. As a result, fibrocytes were found to express a higher level of PD-L1. Notably, the PD-L1 expression in fibrocytes was functional because blocking with anti–PD-1/PD-L1 Abs clearly induced the upregulation of the proliferation of CD8⁺ T cells. However, the blockade of the PD-L1 expression in DCs did not enhance the growth of CD8⁺ T cells, although the PD-L1 expression level in DCs was higher than that in fibrocytes. Recently, Sugiura et al. (32) reported that cis–PD-L1/CD80 interactions on DCs block the binding of PD-L1 to PD-1. Therefore, PD-L1 did not work to inhibit T cell activation on DCs, and anti–PD-L1 Abs do not enhance the growth of CD8⁺ T cells, although the stimulatory activity of CD8⁺ T cells induced by DCs was highest among the APCs tested. We also found that M1, but not M2, macrophages stimulated the proliferation of CD8⁺ T cells in humans. These results suggest that the immune cells that are directly involved in the enhancement of antitumor immunity mediated by anti–PD-1/PD-L1 Ab in tumor sites may be fibrocytes or M1 macrophages, rather than DCs.

Recent advances in cancer immunotherapy have resulted in the development of several Abs against CTLA4 and PD-1 and its ligand PD-L1, which have been approved for use as ICIs in the treatment of various malignancies (23, 24). In addition, clinical trials involving combination immunotherapy of anti–PD-1/PD-L1 Abs with other modalities have demonstrated the promising efficacy of this regimen against some cancers (25, 26). Among these the combination of ICIs with antiangiogenic agents, including an anti-VEGF Ab, bevacizumab, has been expected and has shown durable efficacy in several cancers, including advanced hepatocellular carcinoma (33–37). We previously reported that therapy with anti-VEGF agents, including bevacizumab or inhibitors, increased the accumulation of fibrocytes in tumors in mice and humans (21). Therefore, we asked whether fibrocytes infiltrating into tumors play a role in additive/synergistic antitumor effects in combination with anti–PD-1/PD-L1 Abs and anti-VEGF agents.

In the current study, we recharacterized fibrocytes in comparison with DCs and macrophages. Fibrocytes have shown similarity to M2 macrophages because of their profibrotic properties (38, 39). However, from the immunological viewpoint, the expression pattern of PD-L1 and the effect of anti–PD-L1 Ab (in human MLR) on fibrocytes resembled those of M1 macrophages. These data suggest that fibrocytes are a unique population among macrophage-lineage cells. We also investigated the cross-presentation ability of fibrocytes pulsed with chicken OVA albumin (exogenous protein) in the context of OT-1Rag^−/− CD8⁺ T cells. Pulsed fibrocytes take up and process the exogenous OVA protein and present it to cytotoxic T cells via MHC class I, as evidenced by the blocking effect of anti–MHC class I Ab. The data clearly indicated the function of fibrocytes as APCs.

However, little is known about the APC function of fibrocytes isolated from tumor tissues as well as PBMC of cancer patients. Tumor-derived fibrocytes stimulated the proliferation of CD8⁺ T cells, and the activity was significantly enhanced by anti–PD-L1 Ab. Furthermore, we showed the stimulatory effects on CD8⁺ T cells by fibrocytes isolated from PBMC of patients with lung cancer. Treatment with anti–PD-L1 Ab reproducibly enhanced the T cell–stimulatory activity of fibrocytes from cancer patients. These data strongly suggest that fibrocytes function in tumor microenvironment as APCs to enhance antitumor immunity, especially when cancer patients were treated with anti–PD-1/PD-L1 Ab. Further study is required to investigate the other functions of fibrocytes as APCs, including the ability to migrate into lymph nodes, as well as the production of immune-related cytokines.

In the current study, we showed that fibrocytes had the ability to enhance tumor immunity when anti–PD-1/PD-L1 Ab was used, suggesting that fibrocytes play a role as APCs at least within tumor sites. Although novel mechanisms involved in therapy with PD-1/PD-L1 blockade have just emerged (40, 41), to identify and analyze fibrocytes in tumor tissues may provide a much deeper understanding of tumor immunity and be useful for the further development of cancer immunotherapy.

We thank Tomoko Oka and members in the Division of Transfusion Medicine and Cell Therapy, Tokushima University Hospital, for technical assistance. We also thank the members of the Nishioka laboratory for technical advice and fruitful discussions.

The authors have no financial conflicts of interest.