C3aR Signaling Inhibits NK-cell Infiltration into the Tumor Microenvironment in Mouse Models

Natural killer (NK) cells can mediate direct killing of transformed cells without prior antigenic exposure or priming (1). Efficient homing of NK cells into the tumor microenvironment (TME) is essential for effective tumor cell killing (2, 3). NK-cell trafficking is controlled by various chemokines, complement fragments, and adhesion molecules (4–6). Therefore, impaired NK-cell trafficking would have a direct impact on an effective innate immune response against tumors (7).

Although the immune system is effective in limiting tumor cell growth and survival, tumor cells frequently develop immune evasion strategies that allow them to escape from immune-mediated killing, particularly from NK cells. Studies have shown that NK-cell numbers are reduced in malignant tissues compared with nonmalignant tissues, and that tumors with infiltrating NK cells correlate with a better prognosis (2, 8, 9). Analysis of the NK cells found in the TME of many solid tumors is dominated by noncytotoxic, cytokine-producing CD56^bright NK cells rather than cytotoxic CD56^dim NK cells, providing an explanation for ineffective NK-cell control of tumor growth (9). Therefore, approaches to enhance cytotoxic NK-cell infiltration into the TME represent a reasonable strategy to control tumor growth and progression.

Tumor-derived C3a (complement component 3a) is implicated in the immunosuppression, growth, and inflammation of various tumor models (10). C3a plays a critical role in the recruitment of leukocytes into the TME (11). In contrast, the role of C3a on NK-cell infiltration into tumors and NK-cell antitumor activity has not been previously explored. Cytotoxic CD56^dim NK cells show increased C3aR expression compared with cytokine-producing CD56^bright NK cells (12). However, little is known mechanistically about how C3a and its receptor, C3aR, regulate cytotoxic NK-cell recruitment and activity in solid tumors.

C3aR is a G protein–coupled receptor (GPCR), and GPCR stimulation can induce activation of integrin molecules (13). Integrins play important roles in immune cell extravasation, infiltration, and homeostasis (14, 15). For example, the conformational state of lymphocyte function-associated antigen-1 (LFA-1 or α_Lβ₂; CD11a/CD18) regulates immune cell migration, extravasation, and activation (15–17). Earlier reports demonstrate an increased expression of LFA-1 in NK cells and decreased expression of other β₂ integrin receptors, such as CD11b/CD18 (CR3) and CD11c/CD18 (CR4; refs. 18, 19). Higher expression of LFA-1 is found in the cytotoxic CD56^dim NK subset compared with CD56^bright NK subset and T cells (20). The increased expression of LFA-1 is important because adhesion and migration of IL2-activated NK cells are modulated by LFA-1 (19, 21, 22). Interestingly, CD56^dim NK cells also show increased expression of C3aR compared with CD56^bright NK cells (12). The interaction between C3aR and LFA-1 in regulating cytotoxic NK-cell infiltration into tumors has not been rigorously explored.

This study investigated the role of the C3aR on NK-cell infiltration into the TME. The interaction between C3aR and LFA-1 in modulating LFA-1 conformation and its contribution to tumor evasion from NK-cell immunosurveillance was also investigated in preclinical models. These studies provide evidence that inhibition of C3aR activation or disruption of LFA-1 conformational changes can lead to increased infiltration of NK cells and effective immunotherapy.

Details about the reagents used in this study can be found in Supplementary Table S1.

All procedures involving animals and their care were approved by the Institutional Animal Care and Use Committee of Stanford University in accordance with institutional and NIH guidelines. Six- to eight-week-old female C57BL6, Balb/c, and C3aR-knockout (KO) mice were used in all experiments. C57BL6 (Strain code 027) and Balb/c (Strain code 028) mice were purchased from Charles River. C3aR-KO animals were purchased from Jackson laboratory (stock no. 005712). Animals were housed for 1 week to acclimate prior to the commencement of experimental procedures. All animal experiments were performed according to institutional guidelines and approved animal protocol. Power calculation and experimental models from previous publications were used as reference to determine the sample size.

Blood samples from anonymous healthy donors were obtained from Stanford Blood center and used to isolare NK cells. Human total peripheral blood mononuclear cells (PBMC) were collected by density-gradient centrifugation using Ficoll-Paque PLUS (catalog no. 17-1440-02, GE Healthcare). The manufacturer's protocol was followed to separate NK cells from total PBMCs using the EasySep human NK Negative Selection Kit (catalog no. 19055, StemCell Technologies). Mouse NK cells were isolated from splenocytes using Ficoll-Paque premium (catalog no. 17-5446-02, GE Healthcare) and EasySep mouse NK Negative Selection Kit (catalog no. 19755 and 19855, StemCell Technologies). Purified NK cells were incubated in supplemented RPMI1640 (Gibco, catalog no. 3180089) supplemented with 10% FBS (Omega Scientific, Inc., catalog no. FB-02), 1% penicillin/streptomycin (Gibco, 15140122) and l-Glutamine (Gibco, catalog no. 25030081) and containing 1,000 units/mL recombinant human (rh)IL2 (catalog no. 200-02, PeproTech) at 37°C and 5% CO₂ for four days before use.

Human cancer cell lines MDA-MB-231, MCF7, BT474, HCC1143, T47D, AU565, and HCT116 were obtained from ATCC and used within 1 year of purchase. Cells were cultured in supplemented DMEM (Gibco, catalog no. 12800082) containing 10% FBS, 1% penicillin/streptomycin, and l-Glutamine. Mouse cancer cell lines MC38, EO771 (courtesy of Peter MacCallum Cancer Institute and Olivia Newton-John Cancer Research Institute, Heidelberg, Australia), CT26 (courtesy of Matt Bogyo, Stanford University, Stanford, CA), and JC (courtesy of Chris Contag, Michigan State University, East Lansing, MI) were cultured in supplemented RPMI (10% FBS, 1% penicillin/streptomycin/l-glutamine). Cells were cultured at 37°C and 5% CO₂. Cell lines were authenticated (short tandem repeat), and cultures were checked regularly to ensure no Mycoplasma contamination (Lonza, catalog no. LT 07-318).

A total of 50 μg fibronectin (Corning, catalog no. 354008) from stock solution (1 mg/mL) was mixed with 950 μL migration medium [RPMI with 1% BSA (Sigma-Aldrich, catalog no. A7030)]. A total of 5 μm pore size, 6.5 mm diameter transwell inserts (Corning, catalog no. 3421) were coated with 50 μL of fibronectin and incubated for 1 hour at room temperature. Excessive fibronectin was pipetted off carefully without damaging the inserts. Blocking was achieved with 50 μL of migration medium for 30 minutes incubation at room temperature. Excessive blocking solution was pipetted off carefully. 600 μL control migration medium or chemoattractant [MCP-1 or SDF (stromal-derived factor-1)] supplemented media or cancer cell conditioned media (conditioned media were obtained from 90% confluency of corresponding cancer cell culture) was added to the bottom well. The coated inserts were placed carefully into the transwells, and isolated human or mouse NK cells in 100 μL of migration medium were placed in the top well. For control experiments, 50 ng/mL human MCP-1, 100 ng/mL mouse SDF-1, and 2 μg/mL C3a were used. Cultures were incubated at 37°C and 5% CO₂ for 2 hours to allow for NK-cell migration. After incubation, inserts were removed carefully, and the bottom well content was suspended using micropipette for uniform distribution of migrated cells. Cells were counted using a hemocytometer, and the migrated cells were normalized to the input cell number and percentage of migration was calculated.

For C3aR inhibitor study, 1 million cells of the cancer cell line EO771 were injected orthotopically in the mammary fat pad of C57BL/6 or 1 million cells of the cancer cell line MC38 were injected subcutaneously into the flank of C57BL/6 mice. Animals were checked every 24 hours. Animal weight and tumor size were measured using weighing balance and caliper, respectively. Inhibitor treatment was started once tumors reached approximately 100 mm³ in size. Randomization of animals was done to ensure approximately equal size of tumors in the different treatment groups. The working concentration of the C3aR-specific inhibitor SB-290157 (TOCRIS, catalog no. 6860) was 10 mg/kg in 90% PBS, 5% DMSO, and 5% ethanol, which was prepared from a stock solution of 20 mg/mL in DMSO. When tumor volume reached 100 mm³, SB-290157 was intraperitonially administered once every 12 hours as described previously (23). Control animals were treated with vehicle (90% PBS, 5% DMSO, and 5% ethanol). After 5 days of treatment, animals were sacrificed, and tumor tissues were collected and processed (described below) for analysis.

For the NK depletion study, the subcutaneous MC38 model was used. A total of 150 μg anti-NK1.1 (BioXcell, catalog no. BEOO36) or IgG control (BioXcell, catalog no. BEOO85) was administered to deplete NK cells 2 days prior to tumor cell inoculation, the day of tumor cell inoculation, and every fourth day after tumor inoculation. SB-290157 was administered as described above to the IgG control and NK1.1-depleted groups once the tumors reached 100 mm³. We also used 30 μL rabbit polyclonal anti–asialo-GM1(FUJIFILM Wako Pure Chemicals Corporations, catalog no. 986-10001) to deplete NK cells in C3aR-KO mice. Anti–asialo-GM1 or saline was injected intraperitoneally 1 day prior to, and on the day of, tumor cell inoculation as well as every fifth day after tumor inoculation.

To test a syngeneic model in BALB/c mice, CT26 cells (10⁶ in 50 μL of PBS) were injected subcutaneously into the flanks of BALB/c wild-type (WT) or C3aR-KO animals. JC cells (10⁶ in 50 μL of PBS) were administered orthotopically into to the mammary fat pad of BALB/c WT or C3aR-KO mice.

In vivo inhibition of LFA-1 was performed through intraperitoneal administration of 30 μL of BIRT377 (25 mg/kg, TOCRIS, Cat# 4776) dissolved in DMSO on alternate days beginning once the tumors reached a volume of 100 mm³. Control mice were administered an equal volume of vehicle (DMSO). After four doses (post tumor injection days: 12, 14, 16, and 18), animals were sacrificed, and tumor tissues were collected and processed as described below for further analysis.

Tumor samples were collected in 5 mL of ice-cold serum-free RPMI. Samples were chopped using a razor blade and collected in 50 mL conical tubes, and the solution volume was adjusted to 5 mL with ice-cold RPMI. A total of 20 μL of liberase (catalog no. 5401020001, Sigma-Aldrich) and 20 μL of DNase (catalog no. D4527-10KU, Sigma-Aldrich; 20 μL for every 5 mL sample) were added to each sample and kept in a shaker for 40 minutes (225 rpm at 37°C). The enzymatic reaction was terminated after 40 minutes by adding 10 mL of ice-cold RPMI supplemented with 10% FBS. Dissociated tissue samples were mashed through 40 μm sterile nylon strainer (catalog no. 431750, Corning) using a 5 mL syringe. The collected cells were centrifuged with excess RPMI medium at 1,500 rpm for 10 minutes. The cell pellets were then processed for flow cytometry analysis.

Freshly isolated or IL2-activated human or mouse NK cells, or extracted cells obtained from tumor tissues, were washed twice with FACS buffer [PBS supplemented with 1% BSA (Sigma-Aldrich, catalog no. A7030)]. Fc blocking was done for 10 minutes at 4°C. Antibodies or isotype controls were added, and the samples were kept at 4°C in the dark for 45 minutes. The samples were washed with FACS buffer at 1,500 rpm for 5 minutes, and the supernatant was discarded. The final volume was adjusted to 150 μL with FACS buffer before analysis. To identify the high-affinity status of LFA-1, an untagged primary antibody specific to epitope 24 (mAB24; catalog no. 13219, Abcam) and secondary antibody to mAB24 were used. Flow cytometry study was blinded during acquisition and analysis. Information about Fc blocking and primary/secondary antibodies is available in Supplementary Table S1.

MDA-MB-231, MCF7, BT474, HCC1143, T47D, AU565, and HCT116 were grown in phenol red–free DMEM (Gibco, catalog no. 31053028) medium supplemented with 10% FBS and 1% antibiotic solution in T75 flasks. The supernatants were collected when the confluency reached 90%. The human C3a ELISA kit from Hycultbiotech (catalog no. HK354-01) was used for C3a quantification according to the manufacturer's protocol.

“μ-slide chemotaxis” microfluidic devices (ibidi, catalog no. 80326) were coated with fibronectin (50 μg/50 μL mixed with 950 μL of migration medium) and incubated at room temperature for 1 hour. The excessive fibronectin was removed, and the device was filled with 0.4% BSA in RPMI for blocking. The BSA-blocked device was kept at room temperature for 1 hour for further use. Meanwhile, microscopic stage (DMi8, Leica) temperature was maintained at 37°C using Okolab microscope incubator for live cell imaging. NK cells (0.4 million cells for each experiment) were loaded into the middle inlet. Recombinant human (rh) C3a (R & D, catalog no. 3677-C3-025; 2 μg/mL mixed with migration medium, 0.4% BSA in RPMI), conditioned medium from MDA-MB-231, HCT116, and T47D, C3a-neutralized conditioned medium from MDA-MB-231 and HCT116 (supernatant collected when the confluency is about 90%), or migration medium alone was infused into the device through two side inlets using 200 μL pipette. NK-cell migration was recorded by time-lapse microscopy at three frames/minute for 20 minutes under a Leica DMi8 using 10X object. The acquired images were analyzed using ImageJ's (NIH open source) “analyze plugin-manual tracker” to obtain position of migration. The Chemotaxis and Migration tool was used to calculate the average velocity and cell tracks.

IL2-activated human NK cells were attached in the microfluidic channel as described above and exposed to C3a-containing medium (2 μg/mL mixed with migration medium, 0.4% BSA in RPMI) or IgG control for 20 minutes. The cells were fixed [PFA 4% in PBS (Santa Cruz Biotechnology, catalog no. 30525-89-4)], washed, and blocked with 10% goat serum (Thermo Fisher Scientific, catalog no. 31872) for 45 minutes at room temperature. Primary rabbit anti-human CD11a (catalog no. ab52895, Abcam), primary rabbit anti-human CD18 (phospo-T758; catalog no. ab63388, Abcam), primary mouse anti-human CD11a+CD18 antibody (ref. 24; catalog no. ab13219, Abcam), or respective control antibody (Rabbit IgG and Mouse IgG1) was diluted as suggested by the provider, infused into the microfluidic devices, and incubated for overnight at 4°C. Samples were then washed three times and incubated with diluted secondary antibody (1:200 dilution; anti-mouse goat IgG, catalog no. A32723 or anti-rabbit goat IgG, catalog no. A32721, Thermo Fisher Scientific) in the dark at room temperature for 1 hour. The samples were washed three times and exposed to DAPI in the dark at room temperature for 10 minutes. After washing, samples were preserved at 4°C for confocal microscopy analysis. Images were taken using Leica SP8 confocal system at 10X object (zoom 5).

Images were taken using Leica SP8 confocal system at 10X object (zoom 5) using the LAS X life science microscope software. TIFF files were exported for further analysis. For fluorescence intensity analysis of pCD18 and mAB21, images were converted to binary images in the ImageJ software and intensity was measured using the “Analyze particles” algorithm available in ImageJ.

Transcriptomic data for breast invasive adenocarcinoma and colorectal adenocarcinoma were obtained from The Cancer Genome Atlas (TCGA; Broad GDAC Firehose-Broad Institute, https://gdac.broadinstitute.org/) and converted to a format compatible with the CIBERSORT software (24). The algorithm was run using a signature gene file (LM22) containing expression data for 22 genes that effectively identify specific immune cell types. Calculated percentages of NK cells computed by this method for tumors and normal tissues were compared using GraphPad Prism.

NK cells were isolated and activated with IL2 for 96 hours as described above. The cells were collected and washed with ice-cold PBS at 4°C (1,500 rpm for 10 minutes). A standard immunoprecipitation protocol was followed as described in a previous publication (25). A total of 1 μg of anti-human C3aR (catalog no. SC-133172, Santa Cruz Biotechnology) and 1 μg of anti-human CD11a (catalog no. 350602, BioLegend) were used in 1 mg of total protein for pulldown. Pulled down eluates were processed for Western blotting as described below.

For mouse NK-cell immunoprecipitation, mouse NK cells were treated with IL2 for 96 hours as described. The activated mouse NK cells were processed as described above. Pulldown assays were performed for C3aR using anti-C3aR (catalog no. SC-133172, Santa Cruz Biotechnology). Polyclonal anti-CD11a polyclonal antibody (catalog no. BS 20370R, Bioss) and anti-C3aR polyclonal antibody (catalog no. BS 2955R, Bioss) were used for immunoblotting.

Samples were run in 4% to 20% Criterion TGX Precast Midi Protein Gel (Bio-Rad, catalog no. 5671093). Transfer onto a polyvinylidene difluoride membrane was performed using Bio-Rad wet transfer with Tris-glycine buffer containing 10% methanol and 0.01% SDS. The membrane was blocked with 5% milk in PBST (0.1% v/v Tween-20) at room temperature for 1 hour. Primary antibodies were diluted in 5% BSA in PBST and incubation was performed at 4°C overnight. Anti-CD11a (catalog no. ab52895, Abcam) and polyclonal anti-C3aR (catalog no. BS 2955R, Bioss) were used for immunoblotting of human samples. Polyclonal anti-CD11a (catalog no. BS 20370R, Bioss) and polyclonal anti-C3aR (catalog no. BS 2955R, Bioss) were used for immunoblotting of mouse samples. The membranes were washed with PBST (3× for 10 minutes) and secondary antibody (1:5,000 dilution) incubation was performed in 5% milk at room temperature for 1 hour. The membranes were washed and probed with Clarity ECL Western Blotting Substrate (Bio-Rad) or Super Signal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific) in a ChemiDoc XRS System (Bio-Rad) using Image Lab software (Bio-Rad).

A total of 50,000 tumor cells (MDA-MB-231, HCT116, or CT26) in 100μL were seeded in 96-well plates and cultured for 48 hours. After 48 hours, human or mouse NK cells were added to each well (1:1 ratio of NK and tumor cells; cells were trypsinized from additional wells for cell counting). The mixture was incubated for 4 hours at 37°C in 5% CO₂. The Promega CytoTox 96 Non-Radioactive Cytotoxicity Assay was used according to the provider's procedure (catalog no. G1780) The percentage killing was estimated against total lysis.

One-way ANOVA was performed if the variable was one independent variable. Two-way ANOVA was performed when there were two nominal predictor variables. The Sidak model was used to compare set of means, and the Tukey model was used to compare every mean with every other mean (three or more sets).

The data that support the findings of this study are available within the article and its Supplementary Data files.

Activated NK-cell infiltration into the TME is a marker for better prognosis in a number of solid tumors (9). However, the infiltration of cytotoxic NK cells in solid tumors is generally low. Because previous clinical studies report elevated serum C3a concentration and decreased NK-cell infiltration in invasive breast cancer and colorectal adenocarcinomas, we selected these two tumor types for analysis (26–29). We used CIBERSORT to quantify the differences in mRNAs associated with NK cells as a surrogate means of assessing infiltrating activated NK cells in tumor and normal tissues from TCGA data (24). We first compared NK-cell gene expression changes in 98 normal breast tissues and 250 breast tumor tissues. Activated NK-cell infiltration within total immune cells was 2.7% in normal breast tissues and 0.7% in tumor tissues (P < 0.0001; Fig. 1A). We also compared normal colon tissues and colorectal tumor tissues (n = 51 each). Activated NK-cell infiltration was 0.8% in normal colon tissues and 0.28% in tumor tissues (P = 0.0025; Fig. 1B). Although there were differences in the frequency of activated NK cells between normal tissues, activated NK-cell infiltration was significantly decreased in malignant tissues compared with normal tissues. Kaplan–Meier survival analysis of TCGA data for invasive breast cancer showed that increased NK-cell infiltration, based on expression of activating NKG2D (KLRK1), associated with increased overall survival (Fig. 1C).

We next investigated C3aR expression on human NK cells. Flow cytometry analysis showed C3aR expression in 75% of total IL2-activated NK cells (Fig. 1D). Building on our analysis of clinical breast and colorectal tumor samples described above, we measured C3a in supernatants from the breast and colorectal tumor cells grown in culture for 48 hours and found detectable concentrations of C3a in supernatants from MDA-MB-231, MCF7, BT474, and HCT116 cells (Fig. 1E).

To investigate the role of C3aR on NK-cell migration, we performed fibronectin-coated transwell assays and quantified NK-cell migration against conditioned media from tumor cells. The results showed decreased NK-cell migration in presence of conditioned media from MDA-MB-231 and HCT116 cells. Addition of anti-human C3a or blocking of C3aR by SB-290157, a small molecule antagonist of C3aR, resulted in a 3-fold increase of NK-cell migration (Fig. 1F). These results supported a negative role of C3aR on NK-cell migration. We also found that C3a neutralization did not affect NK-cell cytotoxicity against tumor cells (Supplementary Fig. S1A and S1B).

To further investigate the role of C3aR on NK-cell migration, we performed live cell migration assays in a two-dimensional (2D) microfluidic platform. The microfluidic channel was coated with fibronectin and then IL2-activated NK cells were attached (see Materials and Methods). These microfluidic-based assays also demonstrated an inhibitory role of C3a on NK-cell migration. Nonpolarized NK cells did not migrate in C3a-containing environments (Fig. 2B and D; Supplementary Fig. S1C). A control experiment with NK cells exposed to IgG alone showed polarization of NK cells based on pseudopod and uropod formation and increased migration (Fig. 2A and C; Supplementary Fig. S1D). The average migration velocity was approximately 3-fold higher in untreated cells compared with C3a-treated cells (Fig. 2E). We also found that NK cells migrated poorly when exposed to supernatants from MDA-MB-231 cells (Fig. 2F; Supplementary Fig. S1E). When a neutralizing antibody against C3a is added to MDA-MB-231 cells, we found significantly increased NK-cell migration and velocity (Fig. 2G and H; Supplementary Fig. S1F). Similar results were obtained for supernatants from HCT116 cells (Fig. 2I–K; Supplementary Fig. S1G and S1H).

Because most of the in vitro studies were performed in human cells, we needed to first determine whether C3aR had similar effects in mouse NK cells. The purity of isolated NK cells from human enriched leukocyte was 95% and 90% from mouse spleen (Supplementary Fig. S2A and S2B). We confirmed the expression of C3aR on activated NK cells obtained from C57BL/6 mice (Supplementary Fig S2C). The fibronectin-coated transwell assays confirmed the inhibitory role of C3a and C3aR on murine NK-cell migration in the presence of conditioned media from EO771 and MC38 cell lines compared with untreated controls (Supplementary Fig. S2D).

After implanting EO771 tumor cells orthotopically and MC38 tumor cells subcutaneously in C57BL/6 mice and allowing the tumors to reach approximately 100 mm³ in size (see Materials and Methods), we evaluated the role of C3aR on NK-cell exclusion in these tumors by treating with SB-290157 every 12 hours for 5 days (23). We found significantly decreased tumor growth in SB-290157–treated mice compared with vehicle-treated mice for both EO771 and MC38 models (Fig. 3A and B, and 3E and F). SB-290157–treated animals also exhibited significantly decreased tumor weights compared to vehicle-treated animals (Fig. 3C and G). Flow cytometry showed a 2- to 3-fold increase in NK-cell infiltration in the tumor samples obtained from SB-290157–treated mice (Fig. 3D and H; Supplementary Fig. S3). C3aR inhibition showed no significant difference in T-cell infiltration in E0771 tumors (Supplementary Fig. S4A and S4B) and increased T-cell infiltration in MC38 tumors (Supplementary Fig. S4C and S4D), there. Flow cytometry also showed increased infiltration of macrophages with C3aR inhibition in both tumor models (Supplementary Fig. S4E and S4F). We did not observe weight loss or any abnormalities in the control or treated animals. Taken together, these in vivo results support a role of C3aR on NK-cell exclusion in the TME. We further investigated the effect of inhibiting C3aR with SB-290157 on NK-cell migration using the microfluidic platform. We treated NK cells with or without SB-290157 in the presence of C3a and observed increased velocity in the SB-290157–treated NK cells (Fig. 4A–C).

To investigate the role of NK cells in tumor regression, we performed NK-cell depletion using anti-NK1.1 in the MC38 subcutaneous model (IgG-treated control group included). SB-290157 was administered every 12 hours to the NK1.1- and control-treated groups once the tumors reached approximately 100 mm³ in size (see Materials and Methods). Significantly increased tumor growth in the NK cell–depleted/SB-290157–treated group was seen compared with the SB-290157 alone group, demonstrating the importance of NK-cell infiltration in tumor control (Fig. 4D). SB-290157–treated animals exhibited significantly decreased tumor weights compared with NK cell–depleted/SB-290157–treated animals and vehicle-treated animals (Fig. 4E and F). Flow cytometry showed decreased NK-cell infiltration in NK cell–depleted/SB-290157–treated animals compared with IgG/SB-290157–treated group (Fig. 4G; Supplementary Fig. S3). We confirmed the NK-cell depletion in spleens obtained from NK1.1-depleted and control groups (Fig. 4H).

We also examined the role of C3aR on NK-cell migration using C3aR-KO NK cells. Knockout of C3aR in C3aR-KO animals was confirmed by flow cytometry and PCR (performed PCR as described by the vendor; Supplementary Fig. S5A and S5B). In transwell assays, increased migration of WT and C3aR-KO NK cells was observed in the presence of the chemokine CXCL12 (SDF). A 4-fold increase in NK-cell migration from C3aR-KO cells was seen compared with WT and controls when C3a was added (Supplementary Fig. S5C). We also found that IL2-activated C3aR-KO NK cells exhibited a 3- to 4-fold increase in migration compared with WT NK cells in tumor cell conditioned media (Supplementary Fig. S5D). Treatment of WT NK cells with SB-290157 increased NK-cell migration and was comparable with that found with C3aR-KO NK cells (Supplementary Fig. S5D). We also examined the role of C3aR in syngeneic JC and CT26 tumor models in BALB/c mice and found that growth of both tumor types was inhibited in C3aR-KO mice compared with WT BALB/c mice (Fig. 5A and C). Flow cytometry showed a 2- to 3-fold increase in NK-cell infiltration in tumors from C3aR-KO mice compared with WT controls (Fig. 5B and D; Supplementary Fig. S3). Immune analysis of the tumors also showed no significant difference in the CD8⁺ T cells and macrophages (Supplementary Fig. S6A and S6C) and an increase in CD4⁺ T cells in tumors obtained from C3aR-KO animals (Supplementary Fig. S6B). These studies indicate that C3aR mediates inhibition of NK-cell migration and tumor growth in syngeneic mouse models. We also used rabbit polyclonal anti–asialo-GM1 to deplete NK cells and then assessed the effects on tumor regression in C3aR-KO mice. Increased tumor growth in the NK cell–depleted C3aR-KO mice was observed compared with NK cell–infiltrated C3aR-KO tumors (Fig. 5E). Tumor growth in NK cell–depleted BALB/c and NK cell–depleted C3aR-KO mice was similar to saline-treated control BALB/c WT mice (Fig. 5E). Flow cytometry of harvested tumors showed elevated NK-cell infiltration only in the saline-treated C3aR-KO mice compared with anti–asialo-GM1–treated and saline-treated groups (Fig. 5F and G). In summary, these results demonstrate a critical role of the C3a/C3aR axis in regulating NK cells and tumor regression.

Integrins, specifically LFA-1, plays a critical role in activated NK-cell migration (15, 16, 21). Using the “STRING” portal, a computational-based method to predict protein–protein interactions (30), we found a strong (90%) interaction between C3aR and the integrin β₂ chain subunit, which typically binds to the integrin α chain subunit CD11a to form LFA-1 (Supplementary Fig. S7A and S7B). Because LFA-1 is predominantly expressed in cytotoxic NK cells and other β₂ integrins are expressed at reduced levels in NK cells and do not typically participate in NK-cell migration, we decided to investigate the role of LFA-1 in cytotoxic NK-cell migration into the TME (18–20). Because β₂ integrin is not exclusively found in LFA-1, we focused on the CD11a subunit, which is specific to LFA-1. We performed reciprocal coimmunoprecipitation (Co-IP) experiments with anti-human C3aR or Co-IP with anti-human CD11a and found that C3aR and LFA-1 coimmunoprecipitated (Fig. 6A).

LFA-1 is activated by outside-in (ligand binding) and inside-out (activation by chemokine receptors and other GPCRs; refs. 31, 32) processes. The inside-out activation leads to a high-affinity LFA-1 conformation, exposing epitope 24, in the β₂ subunit, which is otherwise hidden in the inactive (bending) or intermediate active forms (16, 33, 34), and thus, can be detected with the antibody mAB24 (16). Activation of the β₂ subunit associates with cytoskeletal rearrangement and cell spreading involving α actinin (14, 33, 35). Inside-out signaling activates the β2 subunit and induces a conformational change in LFA-1, leading to cytoskeletal rearrangement in immune cells (14, 33, 35–37). The localization of low-affinity LFA-1 at the uropod is critical for efficient immune cell migration, whereas high-affinity LFA-1 inhibits migration (17, 38). Immunofluorescence showed that NK cells treated with IgG migrated and had focal localization of LFA-1 in the uropod (Fig. 6B), whereas NK cells exposed to C3a did not migrate and exhibited LFA-1 at the plasma membrane (Fig. 6C). C3a-exposed cells exhibited robust expression of epitope 24, indicating a high-affinity LFA-1 conformation (Fig. 6E; Supplementary Fig S7C) whereas control cells did not demonstrate epitope 24 staining (Fig. 6D). The time to induce a high-affinity LFA-1 conformation varies between attached NK cells and suspended NK cells. Attached NK cells in the microfluidic-based experiments showed a high-affinity conformation immediately after C3a treatment (Fig. 6E). In contrast, suspended NK cells required extended time (3-hour incubation) to demonstrate a high-affinity conformation (Supplementary Fig. S7D). During inside-out signaling, the β2 subunit is activated by phosphorylation at T758 (14, 33, 35–37). Immunofluorescence showed that T758 was phosphorylated (p-CD18 staining) only in C3a-exposed NK cells, supporting a role for C3a in inside-out signaling through the β2 subunit (Fig. 6F and G; Supplementary Fig. S7E). We confirmed the role of C3aR in the high-affinity LFA-1 formation by using SB-290157. Immunofluorescence did not show mAB24 reactivity in the C3a-treated NK cells in the presence of SB-290157 (Supplementary Fig. S7F). Taken together, these results demonstrate a direct effect of C3a/C3aR on LFA-1 localization in the uropod, β2 integrin phosphorylation, and formation of the LFA-1 high-affinity conformation.

To investigate the conformational change in LFA-1 in human NK cells, we treated IL2-activated human NK cells to receptor-saturating concentrations of human recombinant C3a. This increased epitope 24 staining, representing activation of domain I in the β2 subunit of LFA-1 and indicating high-affinity LFA-1 conformation (Supplementary Fig. S7D). To confirm the interaction between mouse C3aR and LFA-1, we immunoprecipitated C3aR from lysates of activated NK cells from C57BL/6 mice and probed for CD11a. Results indicated an interaction between C3aR and CD11a (Supplementary Fig. S8A). We also found that C3a increased the high-affinity conformation of LFA-1 in activated NK cells from WT mice, whereas C3a incubation did not alter LFA-1 conformation of activated NK cells from C3aR-KO mice, demonstrating a link between C3a/C3aR and the high-affinity formation of LFA-1 (Supplementary Fig. S8B).

Previous studies have shown that BIRT 377 acts as an allosteric inhibitor of the high-affinity LFA-1 conformation by blocking domain I in the β2 subunit (38, 39). In our microfluidic experiments, inhibition of the high-affinity conformation of LFA-1 with BIRT377 significantly increased the migration and velocity of C3a-treated NK cells compared with vehicle-treated cells (Fig. 7A–C; Supplementary Fig. S8C and S8D). Immunofluorescence also indicated the absence of high-affinity LFA-1 in the BIRT377-treated cells (Fig. 7D). Addition of BIRT377 in MDA-MB-231 supernatants increased the migration and velocity of NK cells exposed to the treated supernatant compared with control supernatant (Supplementary Fig. S8E–S8I). Taken together, these results revealed the inhibitory role of C3a-induced high-affinity LFA-1 on NK-cell migration. As a follow-up to the in vitro studies, we investigated the effect of BIRT377 on NK-cell infiltration and CT26 tumor growth. BIRT377 treatment resulted in significant CT26 tumor regression and increased NK-cell infiltration compared with untreated controls (Fig. 7E–G). Cytotoxicity assays showed that the addition of BIRT377 did not affect the cytotoxicity of activated mouse NK cells (Supplementary Fig. S8J). In summary, these studies demonstrate the requirement for the low-affinity LFA-1 conformation to increase NK-cell infiltration into tumors, which leads to tumor regression.

Increased infiltration of cytotoxic NK cells into tumors correlates with a better prognosis and more effective immune response (40, 41). In contrast to cytotoxic T cells, NK cells do not require prior priming and have a rapid cytotoxic effect on tumor cells (42). Unfortunately, the majority of solid tumors have low infiltration of NK cells (9). Previous studies have shown that cancer cells can secrete factors that create a suppressive microenvironment for infiltrating NK cells (9).

Only a minority of patients respond positively to immune checkpoint inhibitor therapy because in many cases, tumors lack a sufficient cytotoxic immune population and, therefore, derive little benefit from immune checkpoint inhibitor therapy (43). Hence, it will be important to study the role of checkpoint blockade therapy in modulating NK-cell activity when NK-cell numbers can be therapeutically increased in solid tumors.

Previous studies show increased C3a concentration in various solid tumors and serum samples obtained from human cancer patients (11). Paradoxically, although C3a inhibits NK-cell migration into tumors, C3a is also a chemoattractant for other immune cells (44). IL2 stimulation is sufficient to induce C3aR expression on NK cells. In contrast, cytotoxic T cells require inflammation and type-1 IFNs for C3aR expression (45). Consequently, C3aR expression on other immune cells may vary with the degree of inflammation and local cytokine environment. Therefore, it is critical to study the role of C3aR expression on other immune cells to evaluate this mechanism.

Activated cytotoxic NK cells showed increased LFA-1 expression compared with other β₂ integrins, and this increased expression of both C3aR and LFA-1 inhibited cytotoxic NK-cell infiltration into the TME. Our mouse model experiments demonstrated the importance of the C3aR–LFA-1 axis in the regulation of cytotoxic NK-cell infiltration into TME. LFA-1 expression also varies among immune cells. Unlike CD56^dim NK cells, CD56^bright NK cells, and T cells express decreased LFA-1 (20). Decreased LFA-1 expression may be a reason for the lack of C3a/C3aR-mediated LFA-1 activation in CD56^bright NK cells, promoting better CD56^bright infiltration into the TME compared with CD56^dim cells. The mechanism of LFA-1 activation differs among various immune cells. For example, in T cells, LFA-1 activation requires T-cell receptor (TCR) signaling to induce the inside-out mechanism (19). Unlike NK cells, other immune cells vary in their responses to C3aR stimulation in different tumor models. Therefore, unlike NK cells, where C3aR alone activates LFA-1, other immune cells may require additional signaling to engage C3a/C3aR-mediated LFA-1 activation.

During immune cell migration, LFA-1 translocation to the retracting tail region (uropod) is critical (46). Low-affinity LFA-1 at the uropod promotes decreased adhesion of attached immune cells, facilitating their migration, whereas the high-affinity, active LFA-1 conformation at the uropod promotes adhesion of the uropod, leading to migration arrest (17, 47). Our microfluidic-based immunofluorescence experiments demonstrated the inhibitory role of C3a/C3aR in LFA-1 translocation to the uropod. Our results established that C3aR-mediated, inside-out induction of LFA-1 promotes a high-affinity LFA-1 formation. C3aR-KO NK cells had low-affinity LFA-1 formation, resulting in increased NK-cell infiltration into tumors. In T cells, additional signaling through the TCR or chemokine receptors is required to activate LFA-1 for synapse formation. Unlike T cells, NK cells express the intermediate LFA-1 conformation without any external stimuli. The presence of the intermediate LFA-1 conformation in NK cells is sufficient for cytolytic synapse formation through ICAM-1 expressed on target cells (48).

Previous publications have used BIRT377 to promote migration of elongated, non-migrating T cells (46). BIRT337, an allosteric LFA-1 inhibitor, prevents the formation of the high-affinity LFA-1 conformation (39). We confirmed that BIRT377 prevented the formation of the high-affinity active LFA-1 formation, and found it promoted NK-cell migration and infiltration. BIRT377 treatment, although useful in experimental studies, has some disadvantages, such as decreasing IL2 production, which could preclude it from human use (49). Because high-affinity LFA-1 is indispensable to mediate the synapsis formation between T cells and antigen presenting cells, the use of BIRT377 might result in a T-cell priming defect (50). Because of these limitations and others, it is imperative to develop a tumor-specific approach to target tumor-derived C3a or C3aR on NK cells to increase NK-cell infiltration into TME.

In summary, our experimental results define a role of the C3a/C3aR axis in NK-cell exclusion in solid tumors. C3a neutralization or genetic inactivation of C3aR overcomes a seemingly common cytotoxic NK-cell exclusion mechanism, promoting decreased tumor growth. Mechanistically, C3aR interaction with LFA-1 results in a high-affinity LFA-1 conformation, leading to decreased cytotoxic NK-cell infiltration into tumors. Therefore, targeting either the C3a/C3aR axis or inhibiting C3aR interaction with LFA-1 could provide a clinically feasible strategy for NK cell–mediated immunotherapy as a monotherapy or combination with other therapeutic methods.

A.J. Giaccia reports other support from Aravive and AKSO Bio outside the submitted work. No disclosures were reported by the other authors.

S. Nandagopal: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. C.G. Li: Validation, investigation, visualization, methodology. Y. Xu: Validation, visualization, methodology. Q.H. Sodji: Validation, visualization, methodology. E.E. Graves: Resources, supervision, validation, methodology, project administration, writing–review and editing. A.J. Giaccia: Resources, supervision, funding acquisition, validation, investigation, methodology, project administration, writing–review and editing.

The authors thank Kaushik Thakkar and Anne Hertz for their support with Western blots and PCR experiments. They thank Kitty Lee and the Stanford Cell Sciences Imaging Facility (CSIF) for their support with confocal microscopy. They also thank the Stanford Blood Center for providing human blood samples.

This work was supported by an MRC grant, NIH grants CA-67166, CA-198291, and CA-197713, the Sydney Frank Foundation, and the Kimmelman Fund (A.J. Giaccia), and NIH grants CA67166 and CA197136 (E.E. Graves). The project described was supported, in part, by award number 1S10OD010580-01A1 from the National Center for Research Resources (NCRR). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NCRR or the NIH.

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.