Autocrine Complement Inhibits IL10-Dependent T-cell–Mediated Antitumor Immunity to Promote Tumor Progression

The complement system is a major component of innate immunity but also regulates many aspects of the adaptive immune response (1). Complement receptors are expressed on T lymphocytes (2–5), and locally produced C3a and C5a bind to their receptors on CD4⁺ T cells to promote differentiation and cell survival, and regulate IFNγ and IL17 production (2, 6). Intracellular generation of C3a is required for human CD4⁺ T-cell survival (7), although the role of complement on regulatory CD4⁺ T cells (Treg) is complex. Complement has been shown to decrease the immunosuppressive capacity mediated by Tregs, whereas others suggest that complement promotes Treg generation (3–5, 8).

Given its important role in innate and adaptive immunity, complement has long been assumed to play an active role in tumor immune surveillance. However, mounting evidence suggests that complement signaling may inhibit antitumor immunity in humans (9). Clinical studies have found that complement levels in patients' plasma or tumors positively correlate with tumor size and poor outcome in lung cancer, colorectal cancer, neuroblastoma in children, ovarian cancer, carcinomas of the digestive tract, brain tumors, and chronic lymphocytic leukemia (9). Furthermore, recent experimental results show that mice lacking complement components (C3, C4, or C5aR) or treated with complement inhibitors exhibit tumor resistance or suppressed metastasis (8, 10–12). Complement may also inhibit antitumor immunity by recruiting myeloid-derived suppressor cells (MDSC; refs. 8, 10, 12) or by inhibiting natural killer (NK)–cell activation (13).

Despite the well-known inhibitory function on many types of cells (14), IL10 enhances the effector function of activated human and mouse CD8⁺ T cells by increasing cytolytic activity (15–17), and has been used to boost antitumor immunity (18–22). These studies demonstrate that IL10 activates and expands CD8⁺ tumor-infiltrating lymphocytes (TIL) and enhances their antitumor activity in situ. However, conflicting reports suggest that IL10 may promote tumor growth and progression (23, 24). Given its pleiotropic function, the source of IL10 may be critical in promoting CD8⁺ TIL antitumor activity. Generating IL10-producing CD8⁺ TILs within the tumors and autocrine IL10/IL10R interaction are a preferable approach for IL10-based tumor immunotherapy, as this avoids the inhibitory effect of IL10 on other cell types. However, CD8⁺ TILs do not produce IL10 under normal conditions in tumor models (24, 25), and it is not clear whether IL10 can be induced or how induction would be achieved. Furthermore, the antitumor effect of IL10-producing CD8⁺ T cells is also not known. In this study, we aimed to identify the molecular regulator of IL10 production in CD8⁺ TILs and characterize the roles of complement and IL10 in the regulation of CD8⁺ TIL antitumor T-cell responses.

A previous report by Tandem and colleagues found that a fraction of the viral-specific CD8⁺ T cells in the central nerve system expresses IL10 at the peak of coronavirus infection (26). These IL10⁺CD8⁺ T cells express higher levels of IFNγ and TNFα compared with IL10⁻CD8⁺ T cells and protect mice from chronic encephalomyelitis. The authors analyzed the gene expression profiles in sorted IL10 eGFP⁺ or eGFP⁻ effector CD8⁺ T cells. To investigate the molecular regulation of IL10 production in effector CD8⁺ T cells, we analyzed the gene expression profiles of IL10⁺CD8⁺ T cells, based on the data by Tandem and colleagues (26). IL10⁺CD8⁺ T cells had a very distinct gene expression pattern compared with that in IL10⁻CD8⁺ T cells (Fig. 1A; ref. 26). The higher Il10 mRNA level in GFP⁺CD8⁺ T cells than that in GFP⁻CD8⁺ T cells indicates that GFP expression faithfully reflected IL10 expression (Supplementary Fig. S1A). These differentially expressed genes are likely comprised of regulators and/or effectors of IL10. Pathway analysis revealed that genes involved in the complement pathway were highly enriched among the genes differentially expressed between IL10⁺ and IL10⁻CD8⁺ T cells (Fig. 1B). The mRNA expression levels of several complement components and their receptors were upregulated in the IL10⁺CD8⁺ T cells (Fig. 1C and D and Supplementary Fig. S1B). These data suggest that complement signaling pathways may be involved in the regulation of IL10 expression in CD8⁺ T cells.

We then tested whether complement signaling could regulate IL10 production in effector CD8⁺ T cells during tumor development. We crossed Il10 reporter mice (termed Tiger mice), in which an IRES-GFP cassette was inserted between the stop codon and polyadenylation signal of the Il10 gene (27), with C3-deficient mice (28) and inoculated wild-type and C3^−/−Il10 reporter mice with B16 melanoma. We examined IL10 production in CD8⁺ TILs. Interestingly, approximately 10% of CD8⁺ TILs expressed high levels of IL10 in C3^−/− mice, whereas, as expected, no IL10-producing CD8⁺ TILs were detected in the wild-type Tiger mice (Fig. 1E). Importantly, CD8⁺ T cells in the draining lymph nodes (dLN) of either wild-type Tiger or C3^−/− Tiger mice did not produce IL10 (Supplementary Fig. S1C). These data demonstrate that complement signaling inhibits IL10 production in CD8⁺ TILs, and its removal is sufficient to promote IL10 production in the tumor microenvironment but not in the periphery.

Exogenous IL10 activates CD8⁺ TILs and promotes their antitumor activity (21, 22). Given that CD8⁺ TILs produced IL10 in the absence of complement, complement-deficient mice may be resistant to tumor development due to an enhanced antitumor activity of CD8⁺ TILs. Indeed, B16 melanoma growth was dramatically slower in C3-deficient mice than that in wild-type mice (Fig. 2A–C). C3^−/− mice were also resistant to E0771 breast cancer development (Fig. 2D–F). These results are consistent with previous reports showing resistance to tumor development or metastasis when complement signaling is disrupted in vivo (8, 10–12).

Because CD8⁺ TILs from C3^−/− mice produce IL10, the tumor resistance by C3^−/− mice may be due to an IL10-induced effector function. Indeed, CD8⁺ T cells were the dominant cell population in TILs from C3^−/− mice, representing an approximate 4-fold increase in their numbers compared with those from wild-type mice (Fig. 2G). These CD8⁺ TILs exhibited multipotency as demonstrated by a simultaneous increase in their IFNγ and TNFα expression (Fig. 2H). To further determine whether the enhanced antitumor response in C3^−/− mice is T-cell–mediated, we crossed C3^−/− mice with TCRα^−/− mice (29) and investigated tumor growth in the C3^−/−TCRα^−/− mice. As expected, TCRα^−/− mice had impaired antitumor immunity (Fig. 2I). Removal of mature T cells from C3^−/− mice resulted in a complete loss of their tumor resistance (Fig. 2I). These results suggest that the enhanced antitumor immunity in C3^−/− mice was likely mediated through the enhanced CD8⁺ CTL-mediated killing.

Previous studies suggest that complement inhibits antitumor immunity by recruiting MDSCs or preventing NK activation in different tumor types or different animal models (8, 10, 12, 13). We quantified MDSCs in B16 melanoma from tumor-bearing mice. Comparable numbers of myeloid cells in the dLNs of both types of mice were observed (Supplementary Fig. S2A). The percentages of CD11b⁺GR1⁺ cells, which contain both CD11b^hiGR1^hi neutrophils and CD11b⁺GR1^dim MDSCs, were comparable in tumor-infiltrating leukocytes from the tumors growing in either wild-type or C3^−/− mice (Fig. 3A). Most of these cells were CD11b^hiGR1^hi neutrophils (Fig. 3A). These data suggest that MDSC-mediated immunosuppression might not be a major cellular mechanism in B16 melanoma in C3^−/− mice.

Expanded Tregs in dLNs and TILs are associated with tumor immunosuppression. Complement signaling regulates Treg differentiation through C3a and C5a receptors in CD4⁺ T cells (3). However, no difference was found in the percentage of Tregs in dLNs or TILs from wild-type and C3^−/− mice (Fig. 3B and Supplementary Fig. S2B). Although the percentage of CD4⁺ TILs was comparable between wild-type and C3^−/− mice (data not shown), more CD4⁺ TILs from C3^−/− mice expressed effector cytokines (Supplementary Fig. S2C). To test the direct impact of IL10 on CD4⁺ T-cell function, we purified human and mouse CD4⁺ T cells and activated them by anti-CD3/CD28 antibodies in vitro with or without IL10. IL10 did not obviously alter the effector status of either human or mouse CD4⁺ T cells in vitro (Supplementary Fig. S2D and S2E). These results suggest that the enhanced effector phenotype in CD4⁺ TILs is likely due to an indirect effect in the tumors from C3^−/− mice. Furthermore, the percentage of total or activated NK cells in the TILs from C3^−/− mice did not change compared with wild-type mice (Fig. 3C). In addition, we observed no difference in the cell populations of NK cells, Th17 cells, macrophages, or dendritic cells in the tumor-infiltrating leukocytes between wild-type and C3^−/− mice (data not shown).

To test whether IL10 is essential for the enhanced antitumor immunity in C3^−/− mice, we crossed C3^−/− mice with Il10^−/− mice (30) to generate double-mutant mice and examined tumor development in these mice. Deletion of Il10 in C3^−/− mice completely abolished their enhanced tumor resistance to B16 melanoma (Fig. 4A–C) as well as E0771 breast cancer (Fig. 4D–F). However, Il10 deletion in the wild-type background did not result in altered antitumor immunity compared with wild-type mice (Fig. 4A–C), suggesting that IL10 may not be involved in antitumor immunity in these tumor models when complement signaling is intact, as the complement signaling prevents IL10 production in CD8⁺ T cells (Fig. 1E).

We tested whether recombinant human IL10 could enhance the function of TILs from patients with cancer. TILs isolated from human lung tumors started to expand and enter logarithmic growth phase after 2 weeks of culture in the presence of IL2 (Fig. 5A). IL10 robustly enhanced the proliferative capacity of TILs when added with IL2, whereas IL10 alone did not drive TILs into the cell cycle (Fig. 5A). To test the tumor killing of the expanded TILs directly, we cocultured the in vitro–expanded TILs with autologous primary tumors. Compared with IL2-expanded TILs, TILs expanded by IL2 plus IL10 induced a rapid and more effective killing of primary tumor cells (Fig. 5B). Furthermore, CD8⁺ TILs expanded in vitro with IL2 plus IL10 had a much enhanced expression of IFNγ and TNFα (Fig. 5C and D). However, the expression of T-cell exhaustion markers, such as programmed death 1 (PD-1), LAG3, and TIM3, was not altered by IL10 (Supplementary Fig. S3). Together, these results suggest that IL10 may be applied as a T-cell growth factor for in vitro expansion of human TILs. To further understand how IL10 promotes human CD8⁺ TIL function, we performed a mRNA expression profiling assay in IL2-expanded CD8⁺ TILs from human lung tumors with or without IL10. IL10/IL2-cultured CD8⁺ TILs displayed a different gene expression pattern compared with that in IL2-cultured CD8⁺ TILs (Fig. 5E). IL10 upregulated genes related to several signaling pathways, including T-cell receptor (TCR) signaling (Fig. 5F), Notch signaling (Fig. 5G), cell cycle (Fig. 5H), and killing activity (Fig. 5I). The effect of IL10 on the expression of chemokine genes was variable depending on the specific chemokine (Fig. 5J). These results are consistent with published mouse data (21), which also suggests that IL10 expands CD8⁺ TILs and promotes their killing activity by upregulation of perforin and granzymes in vivo and in vitro. Together, these data indicate that IL10 plays a similar role in human and mouse CD8⁺ T cells that will facilitate the translation of mouse studies to human clinical trials.

We next determined whether endogenous C3 expressed by CD8⁺ T cells inhibits their IL10 production. We transplanted CD4⁺ and CD8⁺ T cells from wild-type or C3^−/− naïve mice into C3^−/−TCRα^−/− receipients and inoculated the mice with B16 melanomas (Fig. 6A). Tumors in mice receiving C3^−/−CD8⁺ T cells developed more slowly than those in mice receiving wild-type CD8⁺ T cells (Fig. 6B), indicating that complement C3 expressed by CD8⁺ T cells inhibits their antitumor activity through an autocrine mechanism. In addition, C3aR and C5aR transcripts were upregulated in IL10-producing effector CD8⁺ T cells (Fig. 1D). We then examined the surface expression of C3aR and C5aR on CD8⁺ TILs. Peritoneal macrophages and splenic neutrophils, which express high levels of C3aR and C5aR (Supplementary Fig. S4A and S4B), were used as controls. CD8⁺ T cells from dLNs of naïve mice or tumor-bearing mice did not express C3aR or C5aR (Fig. 6C and D, left two plots). However, approximately 20% of the CD8⁺ TILs from melanomas expressed C3aR and C5aR (Fig. 6D, right two plots). A similar percentage of CD8⁺ TILs from an E0771 breast cancer model also expressed C3aR and C5aR (Fig. 6E and F). Moreover, CD8⁺ TILs isolated from human liver tumors expressed high levels of C3aR and C5aR (Fig. 6G).

It has been reported that complement signaling is activated and anaphylatoxins are generated in the tumor microenviroment (10). We found that anaphylatoxins C3a and C5a were detected in freshly isolated B16 tumors from wild-type mice (Supplementary Fig. S4C). To test the role of locally produced C3a and C5a in regulating IL10 expression in CD8⁺ TILs, we overexpressed both subunits of carboxypeptidase N (CPN) 1 and 2 in B16 melanoma cells (Supplementary Fig. S4D and S4E). Local overexpression of CPN inactivated anaphylatoxins (C3a and C5a) in the tumors from wild-type mice (Supplementary Fig. S4C) and restored IL10 production in CD8⁺ TILs to an extent similar to that in CD8⁺ TILs from tumors of C3^−/− hosts (Figs. 1E and 6H). In addition, blockade of C3aR and C5aR using antagonists also restored IL10 production in vivo (Fig. 6I), indicating that locally produced C3a and C5a inhibit IL10 production in CD8⁺ TILs. To determine whether anaphylatoxins inhibited IL10 expression through C3aR and C5aR on CD8⁺ T cells, we purified and activated CD8⁺ T cells in vitro and determined their C3aR and C5aR expression. Interestingly, C3aR and C5aR were detected on activated CD8⁺ T cells only after long-term in vitro culture (Supplementary Fig. S4F and S4G). C3aR or C5aR in vitro blockade alone using antagonist slightly increased IL10 expression in activated CD8⁺ T cells, whereas combined blockade of both receptors further enhanced IL10 expression in these cells (Fig. 6J). Taken together, the in vivo and in vitro results demonstrate that C3aR and C5aR have redundant suppressive functions in IL10 production by CD8⁺ T cells.

We further investigated whether blockade of the engagement of anaphylatoxins to their receptors could inhibit the development of established tumors using complement receptor antagonists. SB 290157 is a nonpeptide small compound that was developed as a selective antagonist of C3aR (31). PMX205, the cyclic hexapeptide hydrocinnamate-(L-ornithine-proline-D-cyclohexylalanine-tryptophan-arginine), is a well-defined C5aR antagonist (32). Pharmacologic blockade of C3aR and C5aR by SB 290157 and PMX205 suppressed tumor growth in wild-type mice, and the efficacy was IL10-dependent (Fig. 6K; Supplementary Fig. S4H). These data demonstrate that C3aR and C5aR play important roles in complement-mediated suppression of IL10 production and antitumor immunity. The results suggest that C3aR and C5aR expressed on CD8⁺ TILs may serve as novel immune checkpoint receptors that may be targeted for further immunotherapy of cancers.

Immune checkpoint blockade with monoclonal antibodies targeting PD-1 expressed on CD8⁺ TILs results in an objective antitumor response in select patients (33). We investigated the roles of the PD-1/programmed death ligand 1 (PD-L1) pathway in complement/IL10-mediated antitumor immunity. CD8⁺ TILs from C3^−/− mice expressed comparable levels of PD-1 to those from wild-type mice (Fig. 7A). IL10 did not modulate PD-1 expression on activated CD8⁺ T cells in vitro (Fig. 7B). Tumor cells from C3^−/− mice expressed similar levels of PD-L1 compared with that from wild-type mice (Fig. 7C and D). In addition, PD-L1 expression on tumor cells was not changed upon culture in IL10-containing medium (Fig. 7E). These results demonstrate that IL10 does not obviously regulate PD-1/PD-L1 expression in CD8⁺ T and tumor cells. It has been reported that PD-L1 stimulates activated T cells to produce IL10 (34). We generated a PD-L1–deficient B16 melanoma cell line using CRISPR/Cas9-mediated gene editing (Fig. 7F). IL10 expression in CD8⁺ TILs from PD-L1–deficient B16 tumors resembled that in CD8⁺ TILs from unmanipulated B16 tumors (Figs. 1E and 7G), suggesting that PD-1 signaling does not regulate IL10 production in CD8⁺ T cells in vivo. Based on these findings, we tested the effect of combined blockade of complement signaling and PD-L1/PD-1 signaling pathways. PD-L1–deficient tumors developed more slowly than the wild-type tumors in C3^−/− mice (Fig. 7H). To further investigate the theraputic effect of combined blockade of PD-1 and complement receptors on established tumors, we implanted B16 melanoma cells into a cohort of wild-type B6 mice and randomized the tumor-bearing mice into four groups after 6 days of tumor development. We then treated with control antibody, anti–PD-1 blocking antibody, C3aR and C5aR antagonists, and a combination of anti–PD-1 blocking antibody with C3aR and C5aR antagonists. In line with published data (35), the anti–PD-1 blocking antibody alone did not show antitumor effects in the B16 melanoma model (Fig. 7I). However, combined treatment with anti–PD-1 blocking antibody plus C3aR and C5aR antagonists further enhanced the antitumor effect mediated by C3aR and C5aR blockade (Fig. 7I).

This work has identified complement signaling as an important regulator of endogenous IL10 production in CD8⁺ TILs. Although IL10 is largely considered an inhibitory cytokine, extensive evidence also supports its role in enhancing antitumor immunity by activating and expanding CD8⁺ TILs in situ (21, 22). Based on the findings from animal studies, a phase I clinical trial is under way to test the therapeutic role of recombinant human IL10 in different types of cancers (NCT02009449). Given its pleiotropic functions, however, systemic use of exogenous IL10 may cause adverse effects. Indeed, injection of pegylated IL10 causes lymphocyte and monocyte infiltration and apoptosis of epithelial cells in the liver and pancreas in mice (21). Therefore, promoting endogenous IL10 production in CD8⁺ TILs may be an effective approach to boost host antitumor immunity. However, CD8⁺ TILs do not produce endogenous IL10 under normal conditions in different tumor models (24). We demonstrate that CD8⁺ TILs produce endogenous IL10 when complement signaling is removed.

Our data suggest that complement proteins are activated locally, producing C3a and C5a. C3a and C5a then interact with their receptors on CD8⁺ TILs, which inhibits the expression of IL10. Several lines of evidence support this pathway. First, IL10 is produced by CD8⁺ TILs in mice lacking C3 expression or in tumors expressing CPN for local inactivation of anaphylatoxins. Second, the complement receptors C3aR and C5aR are expressed on CD8⁺ TILs. Consistent with a previous genetic tracing study (36), we found that C3aR and C5aR are not expressed on peripheral CD8⁺ T cells. The observation that IL10⁺CD8⁺ T cells express genes related to the complement signaling pathway suggests a feedback regulatory loop between IL10 and complement gene expression. The fact that coexpression of C3aR and C5aR on CD8⁺ TILs and C3aR blockade alone is not sufficient to inhibit tumor development suggests that C3aR and C5aR play a redundant role in suppressing IL10 expression. Our in vitro data also indicate that blockade of both receptors results in further enhancement of the expression of IL10 than blockade of single receptor. Although it was reported that some cancer cells produce complement proteins (37), we did not detect C3a and C5a from ex vivo–cultured B16 melanomas. Thus, C3a and C5a are primarily provided by the host. C3 is the central mediator of both classic and alternative complement activation pathways. C3b is an essential component of C5 convertase in both classic (C4b2a3b) and alternative (C3bBbC3b) pathways. C3 deficiency impedes the cleavage of C5 to C5a in both classic and alternative complement activation pathways. Although it has been shown that C5 activation can occur in the absence of C3 under certain conditions (38), our results demonstrate that C5 is not activated in the tumor microenvironment in the absence of C3. Indeed, C5a was barely detected in the tumors from C3^−/− mice.

Our results have also established a mechanistic link between two major types of experimental and clinical observations. The first concerns the role of complement in host antitumor immunity. As discussed in the introduction, extensive experimental studies and clinical findings suggest that complement inhibits host antitumor immunity. The second relates to the role of IL10 in host antitumor immunity. It was observed that IL10- or IL10R-deficient mice are susceptible to tumor development (39, 40), and exogenous IL10 can boost host antitumor immunity (18–22, 41). Our data suggest that complement signaling inhibits IL10 production in CD8⁺ TILs and prevents the CD8⁺ TILs from becoming potent effectors. Autocrine production of complement by CD8⁺ T cells is sufficient to exert this effect. Previous studies also demonstrate that complement inhibits antitumor immunity by recruiting MDSCs and preventing NK-cell activation in different tumor models (8, 10, 12, 13). Thus, complement signaling–mediated inhibition of host antitumor immunity is multifactorial. We did not observe MDSC infiltration in a B16 melanoma model in either wild-type or C3^−/− mice. Furthermore, there was no increase in the total or activated NK cells from C3^−/− mice. These data suggest that MDSC-mediated immunosuppression may be tumor-type specific. The inhibition of NK function by complement through MDSCs may thus also depend on the tumor type. Previous publications have demonstrated conflicting roles of complement on Tregs (3–5, 8). However, the percentages of Tregs in dLNs or TILs from wild-type and C3^−/− mice were comparable. Therefore, the effect of complement on Tregs appears to be context-dependent (12).

Our study suggests several novel strategies for cancer immunotherapy. First, a combined blockade of complement signaling by antagonists to C3aR, C5aR, and anti–PD-1 may enhance the clinical efficacy of anti–PD-1 mAb therapy. It is worth noting that the C5aR antagonist PMX205 has been successfully used in a phase I clinical trial for the treatment of inflammatory disorder (42). Second, a targeted strategy to deliver IL10 to CD8⁺ TILs by generating an anti–PD-1–IL10 or anti-CTLA4–IL10 fusion protein may improve the efficacy of anti–PD-1 antibody by blocking PD-1 expressed on TILs and PD-L1 expressed on antigen-presenting cells and tumor cells (43) and promoting the differentiated CD8⁺ TILs into more potent effectors. Finally, the addition of IL10 in the expansion protocol of TILs for adaptive cellular therapy may enhance the number and potency of the effector T cells needed for a long-term durable response (44).

C57BL/6J (wild-type) mice (stock number: 000664), C3^−/− mice (N7, Stock No: 003641), TCRα^−/− mice (N13, Stock No: 002116), Il10^−/− mice (N13, Stock No: 002251), and Il10 GFP reporter (Tiger) mice (N10, stock number: 008379) were purchased from The Jackson Laboratory. C3^−/− mice were further backcrossed to C57BL/6J for 5 generations (N12). N indicates the number of backcrossed generations. Six-to-8-week-old mice were used for all experiments. All mouse strains are of the C57BL/6 genetic background. Mice were housed in a specific pathogen-free facility in the Duke University Medical Center and used according to protocols approved by the Duke University Institutional Animal Care and Use Committee. The B16-F10 murine melanoma cell line was purchased from the ATCC in 2011. E0771, a murine mammary adenocarcinoma, was a gift from Dr. Scott A. Gerber (University of Rochester) in 2013. The mouse tumor cell lines used in this article have not been authenticated by us.

A total of 5 × 10⁴ to 4 × 10⁵ B16F10 or 1 × 10⁶ E0771 tumor cells in 100 μL PBS were s.c. inoculated into 6-to-8-week-old mice. Tumor development was monitored daily or every other day. The mice were sacrificed when their tumor volumes reached 2,000 mm³.

Tumors were cut into small (<3 mm) pieces and incubated in 5 mL dissociation solution [RPMI medium supplemented with Collagenase type I (200 U/mL) and DNase I (100 μg/mL)] for 30 minutes at 37°C. Samples were mixed by pipetting and vortexing every 10 minutes during the incubation. Collagenase (C0130) and DNase I (ND25) were purchased from Sigma-Aldrich. CD45 was used to distinguish tumor-infiltrating leukocytes from other cells, and different antibodies targeting surface markers were used for flow cytometry analysis. For intracellular cytokine staining, cells were cultured in complete RPMI for 4 to 6 hours with phorbol 12-x02011;myristate 13-x02011;acetate, ionomycin, and Brefeldin A (Cell Activation Cocktail; BioLegend).

TILs from human lung tumors were in vitro–expanded as described in the Supplementary Methods for 20 days with 6,000 U/mL rIL2 and/or 100 U/mL rIL10. The expanded TILs were activated by incubation in 3 μg/mL anti-CD3/CD28 antibodies for 6 hours. The CD8⁺ TILs were enriched by negative selection using a kit (Stemcell Technologies). RNA was extracted using TRIzol reagent (Life Technologies) according to the instructions. Microarray analyses were performed using human U133A 2.0 arrays by the Duke University microarray facility. Raw intensities from the CEL files were analyzed using Partek Genomics Suite (Partek Incorporated) with standard background correction to generate an RMA (robust multiarray average intensity) on a log₂ scale for each probe set. Probe sets were filtered for “present” detection in two or more arrays, and the interquartile intensity range was >0.5. The filtered RMA intensities were then analyzed for differential expression using the advanced ANOVA, and differentially expressed genes were generated with/without FDR-adjusted P values using the Benjamini–Hochberg method.

RNA was isolated with Direct-Zol (Zymo Research) according to the manufacturer's protocol. Complementary DNA was synthesized with SuperScript III Reverse Transcriptase (Life Technologies). Quantitative real-time PCR was performed using a SYBR green–based assay (Applied Biosystems). For mRNA expression in tumor cells, β-actin mRNA was used for normalization across samples.

C3a receptor antagonist SB 290157 was purchased from EMD Millipore, and C5a receptor antagonist PMX205 was purchased from Selleck Chemicals LLC. For E0771 breast tumors, mice were treated intraperitoneally with SB 290157 at 10 mg/kg and PMX205 at 1 mg/kg (both in 5% DMSO/5% ethanol/90% PBS) twice a day from day 9 after tumor implantation. Control mice were treated with 5% DMSO/5% ethanol/90% PBS. For B16 melanoma, mice were i.p. injected with anti–PD-1 antibody (Clone: RMP1-14; Bio X Cell) at 200 μg/mouse twice a week, or SB 290157 and PMX205 as described in the E0771 model, or the combination of anti–PD-1 antibody with SB 290157 and PMX205.

The human TIL in vitro killing assay was performed using a real-time impedance assay using a Real Time Cell Analysis (RTCA) S16 (ACEA Biosciences). Primary autologous cancer cells (1.5 × 10⁴ cells/well) were seeded in an E-plate 16 and cultured for 2 days. The TILs were in vitro–expanded for 21 days and stimulated with anti-CD3/CD28 antibodies for 24 hours. Effector cells were added into each well cultured with tumor cells at a ratio of 20:1 (effector cell:cancer cell). Cancer cells and effector cells were cocultured in a 37°C CO₂ incubator, and real-time monitored by RTCA S16.

CD4⁺ and CD8⁺ T cells were negatively enriched using EasySep mouse CD4⁺ T-cell and EasySep mouse CD8⁺ T Enrichment Kits (Stemcell Technologies). One million mixed CD4⁺ and CD8⁺ T cells (2:1) were transferred by i.v. injection into 6-to-8-week-old gender-matched recipient mice. After 14 days of T-cell transfer, recipient mice were implanted s.c. with B16F10 melanoma, and tumor growth was monitored daily.

B16 cells were cultured in complete DMEM medium. Cells were passaged every 2 to 3 days with a ratio of 1:6 to 1:8. pLentiCRISPR V1 plasmid was a gift from Feng Zhang (Addgene plasmid # 49535. Current version is pLentiCRISPR V2, plasmid# 52961). Cas9 guide sequence for mouse PD-L1 (NCBI Accession number: GQ904196) was designed as 5′ AGCCTGCTGTCACTTGCTAC 3′ by the online program (http://crispr.mit.edu/). The two oligos were synthesized from IDT as 5′ CACCGAGCCTGCTGTCACTTGCTAC 3′ and 5′ AAACGTAGCAAGTGACAGCAGGCTC 3′ (bold indicates the BsmBI restriction site). The pLentiCRISPR V1 was digested by the BsmBI, and the annealed oligos were cloned into pLentiCRISPR V1 according to the protocol from Zhang laboratory. To make the lentivirus, pLentiCRISPR (with cloned sgRNA) were cotransfected into HEK293(F)T cells with the packaging plasmids pVSVg (AddGene #8454) and psPAX2 (AddGene #12260). For PD-L1 silencing, CRISPR control or CRISPR sgRNA targeting PD-L1 viruses were transduced into B16F10 cells according to the protocol. The transduced B16F10 cells were selected in complete medium containing 1 μg/mL puromycin (Invivogen) 48 hours after transduction. The culture medium was replaced every 48 hours. The expression of PD-L1 was determined by flow cytometry. More detailed information regarding guide RNA design and construct cloning can be found at http://crispr.genome-engineering.org.

Mouse CPN1 (NM_030703)

• Forward: 5′ GGTGGACCTGAACCGCAACTTC 3′

• Reverse: 5′ CGTTGGTGATGCCGTCTGGAA 3′

Mouse β-Actin (NM_007393)

• Forward: 5′ ACCTTCTACAATGAGCTGCG 3′

• Reverse: 5′ CTGGATGGCTACGTACATGG 3′

Complement C3a and C5a mouse ELISA kits (cat# ABIN415413 and ABIN415613) were purchased from Antibodies-Online. Tumors were carefully dissected from each mouse to keep them intact. The tumors were rinsed in cold PBS to remove blood thoroughly and weighed before homogenization. The tumors were then minced to small pieces and homogenized in PBS. The homogenates were centrifuged to obtain a cell-free supernatant. The quantities of C3a and C5a present in the supernatants, and in the medium from B16 cell culture, were determined by ELISA.

Mouse CD8⁺ T cells from lymph nodes were enriched using a negative selection kit (Stemcell Technologies; Cat# 19853) and activated by plate-bound anti-CD3 (2 μg/mL) and anti-CD28 (1 μg/mL) antibodies for 72 hours. The activated CD8⁺ T cells were then cultured in complete medium alone (control), 10 μmol/L SB290157 (C3aR antagonist, C3aRA), 10 μmol/L PMX205 (C5aR antagonist, C5aRA), or a combination of 10 μmol/L SB290157 and 10 μmol/L PMX205 for 6 days. The IL10 expression was determined by intracellular staining.

Freshly isolated human hepatocarcinoma and lung cancer biopsies as well as peripheral blood were provided by Hepatopancreatobiliary and Thoracic Surgery Departments, Beijing Cancer Hospital and Institute. The study was conducted in accordance with the Declaration of Helsinki and approved by the ethical committee of the Beijing Cancer Hospital and Institute. Patients signed an informed consent.

Data are presented as mean ± SEM. Results were analyzed by two-tailed Student t test or one-way ANOVA when multiple comparisons were made. Statistical significance was defined as P ≤ 0.05.

The referenced microarray data in the NCBI GEO public database are under accession number GEO: GSE25846 and GSE79127.

E.F. Patz Jr has given expert testimony for Cue Biologics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y. Wang, Y.-W. He

Development of methodology: Y. Wang, S.-N. Sun

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Wang, S.-N. Sun, Q. Liu, Y.-Y. Yu, K. Wang, B.-C. Xing, Q.-F. Zheng, M.J. Campa, E.F. Patz Jr, S.-Y. Li

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Wang, S.-N. Sun, E.F. Patz Jr, Y.-W. He

Writing, review, and/or revision of the manuscript: Y. Wang, M.J. Campa, E.F. Patz Jr, Y.-W. He

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Guo, S.-Y. Li

Study supervision: Y.-W. He

We thank Dr. Scott A. Gerber for the gift of the E0771 breast cancer cell line. We would like to thank the Duke Microarray Core facility (a Duke National Cancer Institute and a Duke Genomic and Computational Biology shared resource facility) for their technical support, microarray data management, and feedback on the generation of the microarray data reported in this article.

This project was supported by the US NIH AI074944 to Y.-W. He.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.