Proinflammatory S100A8 Induces PD-L1 Expression in Macrophages, Mediating Tumor Immune Escape

Inflammation is a protective response of the organism to harmful stimuli by exogenous pathogens or endogenous signals (such as damaged cells or “dangerous” signals), thus resulting in the elimination of the initial cause of injury, the clearance of necrotic cells, and tissue repair. However, nonresolving inflammation is viewed as the driving factor in many diseases, including atherosclerosis, obesity, pulmonary fibrosis, rheumatoid arthritis, and cancer (1). Inflammation plays significant roles at the whole process of cancer development, including initiation, promotion, malignant conversion, invasion, and metastasis. Inflammation also mediates tumor immune escape and affects responses to therapy (2).

Ag-specific T cell responses are controlled by the costimulatory and coinhibitory signals (3). PD-1 is the most important coinhibitory signal and has been under intensive studies for decades. Tumor cells or immune cells expressing PD-1 ligands (PD-L1) on their surface use the PD-1 pathway to evade an effective antitumor immune response. Blockade of PD-1 or PD-L1 is a very successful immunotherapy strategy in the field of oncology (4). To predict the efficacy of anti–PD-1 or PD-L1 therapy, we need to understand the mechanisms controlling PD-L1 expression (5). The PD-1/PD-L1 axis is a critical determinant of physiological immune homeostasis. PD-L1 gene expression has been shown to be controlled by inflammatory signalings (6), consistent with the physiological role of the PD-1/PD-L1 axis in suppressing T cell activation. A few soluble inflammatory factors produced by immune cells had been identified as inducers of PD-L1, such as IFN-γ (7, 8), TNF-α (9), TGF-β (10), IL-1β (11), TLR4 (12), IL-6 (13), etc. These discoveries suggested a significant role of inflammatory stimuli in regulation of PD-1/PD-L1 (5).

S100A8 and S100A9 are Ca²⁺ binding proteins belonging to the S100 family and play a decisive role in the development of inflammation. S100A8 and S100A9 are damage-associated molecular pattern proteins released by neutrophils and monocytes. They can form a heterodimer or homodimer both in vitro and in vivo, but these two proteins may have different functions regulated by different mechanisms (14). S100A8 released in the microenvironment binds to TLR4 receptor (15), RAGE receptor (16), or other receptors and activates downstream pathways to exert proinflammatory or immunosuppressive effects. High S100A8 and S100A9 expression levels are characteristic of inflammatory conditions (17), but they also exert important roles in tumorigenesis. Upregulation of S100A8 and S100A9 has been found in multiple types of cancer (18). We previously provided evidence that multiple key pathways (such as NF-κB and STAT3 signalings) were specifically involved in different phases of the inflammation–cancer link (19) and revealed a novel mechanism in which inflammation-induced S100A8 promoted colon tumorigenesis by activating an Akt1–Smad5–Id3 axis (20), and reported anti-S100A9 Ab can effectively alleviate the inflammation-induced tumorigenesis in a mouse model (21).

In this study, we asked whether S100A8 or S100A9, as an important inflammatory mediator (16), can regulate PD-L1 expression in the tumor microenvironment and thus contribute to their decisive role in the inflammation–cancer link.

The murine macrophage cell line RAW264.7, the murine colon carcinoma cell line CT26, and the human monocytic cell line THP-1 were purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). They were cultured in RPMI 1640 medium supplemented with 10% FBS. THP-1 cells were treated with 25 ng/ml PMA for 48 h to obtain macrophage-like differentiated THP-1 cells (dTHP-1). Stable S100A8 and PD-L1 short hairpin RNA (shRNA) knockdown lines were generated by infecting RAW264.7, CT26 or bone marrow–derived macrophage (BMDM) cells with lentiviruses pGLV3/H1/GFP+Puro vector (Genepharma, Shanghai, China), which harbors mouse S100A8 shRNA targeting sequence 5′-TCAGAGAATTGGACATCAATACTCGAGTATTGATGTCCAATTCTCTGA-3′, mouse PD-L1 shRNA targeting sequence 5′-GGTGCGGACTACAAGCGAATC-3′. S100A8 knockdown was confirmed by quantitative real-time PCR (RT-qPCR) (Fig. 1G, Supplemental Fig. 1A), and PD-L1 knockdown was confirmed by Western blot (Supplemental Fig. 1B). Plasmids transfection was performed by using Lipofectamine 3000 (Invitrogen, Carlsbad, CA). The cell lines were authenticated by the short-tandem repeat method. They were not contaminated by Mycoplasma before or after experiments.

BAY11-7082 (no. S1523), LY294002 (no. S1737), U0126 (no. S1901), and LPS (no. S1732) were purchased from Beyotime (Jiangsu, China). S3I-201 (no. S1155) was from Selleck (Shanghai, China), and TAK-242 (no. HY-11109), FPS-ZM1 (no. HY-19370), actinomycin D (no. HY-17559), and cycloheximide (no. HY-12320) were from MedChemExpress (Monmouth Junction, NJ). Recombinant mouse TNF-α protein was purchased from R&D Systems (Minneapolis, MN). Recombinant mouse S100A8 protein (no. P4345), mouse S100A9 protein (no. P4348), and human S100A8 protein (no. H00006279-P01) were purchased from Abnova (Walnut, CA). The purity of recombinant S100A8 and S100A9 was >95% and >90%, respectively, by SDS-PAGE. The endotoxin level in the recombinant protein was lower than 1 endotoxin unit/μg, a rule-of-thumb criterion for a safe protein product, as measured by quantitative Endpoint Chromogenic Limulus Amebocyte Lysate test (Lonza, Basel, Switzerland). Anti-TLR4/MD2 complex (no. MTS510) was purchased from Abcam (Cambridge, MA). The full-length coding sequence of murine TLR4 was subcloned into the mammalian expression vector pcDNA3.1+.

cDNA was synthesized from 2 μg total RNA with a Revert Aid RT Reverse Transcription Kit (Thermo Scientific, Waltham, MA). The RT-PCR analysis was performed using a Bio-Rad CFX96 Real-Time System (Bio-Rad, Hercules, CA). Mouse or human GAPDH was amplified in parallel as an internal control. Expression of each gene was quantified by measuring cycle threshold values, and the 2^–ΔΔCt method was used to calculate relative changes in gene expression. The primers for quantitative PCR (qPCR) were listed in Supplemental Table I.

Cells were cultured in the presence or absence of 5 μg/ml rS100A8 for 48 h, then fixed in 1% formaldehyde for 15 min. Cells were lysed in SDS lysis buffer and then sonicated to shear chromatin into 200–300 bp DNA fragments. The chromatin fragments were subjected to immunoprecipitation with anti-H3K27me3, anti-H3K27ac, or anti-H3K4me3 (Cell Signaling Technology [CST]) or a control unrelated IgG. The coimmunoprecipitated DNA was purified with the GeneJET PCR Purification Kit (Thermo Fisher Scientific). Occupancy of the promoter and other regulatory regions was measured by qPCR with the following primer sets by using 1/100 of the chromatin immunoprecipitation (ChIP) DNA. The base pair location is in reference to transcription start site (TSS) of mouse PD-L1 (Fig. 4A): R1, Forward 5′-GTTTCATTATGTCGAGGAAC-3′, Reverse 5′-ATCTCTTTCAAGCCCTTTCT-3′; R2, Forward 5′-GATGGGGAATCGGATGGTAA-3′, Reverse 5′-TCGCAGCTGGCCAAACTAGC-3′, R3, Forward 5′-TTCTGAGACCCTAGCCCTGG-3′, Reverse 5′-CCCTCAAAGAAACAGTCCAG-3′, R4, Forward 5′-TCTGGACTGTTTCTTTGAGG-3′, Reverse 5′-CCTTAATTCCAGTACTCAGG-3′, R5, Forward 5′-AATTAAGGCGTGTGTCACCG-3′, Reverse 5′-TCCTCAAAGTTCCTCGACAT-3′, R6, Forward 5′-ATGAGGATATTTGCTGGCAT-3′, Reverse 5′-CAAACTGAATCACTTGCTCA-3′, R7, Forward 5′-TGTGGCAGGAGAGGAGGACCTT-3′, Reverse 5′-GCGTGATTCGCTTGTAGTCCG-3′.

For extracellular staining, cells were incubated with the indicated combinations of Abs for 30 min at 4°C. Cells were then washed and analyzed on a BD FACSCalibur. The following Abs were used: PE rat anti-mouse CD274 (i.e.PD-L1) (no. 558091; BD Pharmingen), κ Isotype control (no. 100621; BioLegend), allophycocyanin hamster anti-mouse CD3e (no. 561826; BD Pharmingen), V450 rat anti-mouse CD8a (no. 560471; BD Pharmingen), allophycocyanin rat anti-mouse CD107a (no. 560646; BD Pharmingen), PE anti-human/mouse Granzyme B (no. 372207; BioLegend), allophycocyanin /Cy7 anti-mouse CD86 Ab (no. 105029; BioLegend), allophycocyanin anti-human CD206 (no. 321109; BioLegend), CD11b FITC M1/70 (no. 557396; BD Pharmingen), CD3e mAb (no. 16-0031-85; eBioscience), CD28 mAb (no. 16-0281-85; eBioscience), LEAF Purified anti-mouse CD274 Ab (no. 124304; BioLegend), and LEAF Purified rat IgG2b. The data were analyzed using FlowJo software.

Extracts of cells lysed with RIPA buffer were cleared by centrifugation. Lysates (50 μg of protein) were subjected to SDS-PAGE, and the separated bands were transferred to a polyvinylidene difluoride membrane (Millipore, Billerica, MA) that was then probed with various Abs. The following Abs were used: anti-AKT (no. 9272; CST, Danvers, MA), p-AKT (T308) (no. 9274; CST), p-AKT (Ser473) (no. 9271; CST), p-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) (no. 4377; CST), p-TAT3 (Tyr705) (no. 9145; CST), STAT3 (no. 9139; CST), NF-κB p65 (no. 4764; CST), p-NF-κB p65 (Ser536) (no. 3036; CST), Histone H3 (no. 4499; CST), α-Tubulin (no. 66031-1-Ig; Proteintech, Wuhan, China), PD-L1 (no. 179711; R&D systems), and GAPDH (no. MAB3740; Millipore). Cytoplasmic and nuclear fractions were isolated with nuclear and cytoplasmic protein extraction kits (Beyotime).

RAW264.7 cells were treated with 10 μg/ml actinomycin D, and stability of mouse PD-L1 mRNA was measured by RT-qPCR. Data are normalized to time 0 h when actinomycin D was added. RAW264.7 cells were treated with 20 μg/ml cycloheximide (CHX) at indicated intervals and analyzed in Western blotting to monitor the PD-L1 protein stability.

All animal care and euthanasia protocols were approved by the Institutional Animal Care and Use Committee of Central South University (Changsha, China). In step 1, monocytes were isolated from C57B/L6 mouse bone marrow through a monocyte isolation kit (no. 130-100-629; Miltenyi Biotec, Bergisch Gladbach, Germany). Then, M-CSF (10 ng/ml, 3 d) was used to induce the monocytes to become the BMDMs. The BMDMs were identified with anti-CD11b and F4/80 Abs by flow cytometry. BMDMs were treated with 5 μg/ml S100A8 (or PBS as negative control) for 36 h before cocultured with T cells. In some cases (Fig. 5B–D), BMDMs were treated with anti–PD-L1 Ab for 1 h. In step 2, splenic T cells isolated from C57B/L6 mouse were measured using the intracellular dye CFSE (BD Pharmingen). T cells were stained with 1 μM CFSE at 37°C 5 min, followed by quenching in RPMI 1640 (with 10% FBS). Fluorescence was analyzed by flow cytometer. A total of 5 × 10⁵ cells T cells were plated in 96-well plates bound with anti-CD3 and anti-CD28 Abs (0.5 μg/ml, each). For step 3 autologous T cells (from step 2) were cocultured with BMDMs (from step 1) (BMDM–to–T cell ratio = 1:5) subpopulations for 72 h. After 72 h, cell cultures were stained for surface markers with CD3 or with CD8 and CD107a in PBS (with 2.5% FBS). The proportion of T cell proliferation was evaluated by CFSE dilution through flow cytometry. Unstimulated T cells were used as the negative control.

In step 1, 2 × 10⁵ CT26 mouse tumor cells were injected s.c. into the flank of 5-wk-old male BALB/C mice. After 18 d, mice were sacrificed to harvest CT26 tumor-primed T cells from the spleen by using positive isolation kit (Miltenyi Biotec). In step 2, BMDMs were harvested from male BALB/C mice. Half of the BMDMs were treated with mouse recombinant S100A8 protein (5 μg/ml) for 36 h, and half of them were treated with PBS as negative control. In step 3, 1.2 × 10⁵ CT26 cells were injected s.c. into the flank of 5-wk-old male NOD/SCID mice. After 8 d, the NOD/SCID mice were injected i.p. with 1) PBS (as a negative control), 2) tumor-primed CTLs (5 × 10⁶ each mice) and BMDMs (5 × 10⁵ each mice), 3) CTLs and S100A8-pretreated BMDMs, 4) CTLs and S100A8-pretreated BMDMs (lentivirus was used to knock down the expression of PD-L1 gene in BMDMs), or 5) CTLs and S100A8-pretreated BMDMs plus anti–PD-L1 Ab. Each group has seven mice. Eighteen days after CT26 cells injection, the NOD/SCID mice were sacrificed to observe tumor mass. Tumor volume was evaluated with the following formula: A × B² / 2, where A is the largest diameter and B is the perpendicular diameter.

For human cancer samples, the association between S100A8, PD-L1, CD86, and CD163 expression levels and patient survival data (from The Cancer Genome Atlas [TCGA] dataset) were analyzed by the human protein atlas Web site (https://www.proteinatlas.org) using Kaplan–Meier scanner. The association between S100A8 expression levels and PD-L1, CD86, and CD163 levels were analyzed by the Cbioportal Web site (http://www.cbioportal.org/) using Spearman correlations.

A total of 104 human colorectal cancer and 58 adjacent control colorectal tissue paraffin sections were obtained from The Second Xiangya Hospital and The Third Xiangya Hospital of Central South University from 2004 to 2012. Collections and use of tissue samples were approved by the ethical review committees of the Second and Third Xiangya Hospital of Central South University and were in accordance with the Declaration of Helsinki. The paraffin-embedded sections were cut 4 μm thick and then deparaffinized and rehydrated. Immunohistochemical staining was performed to detect the expression of S100A8 and PD-L1 as described previously (20). The following Abs were used: S100a8 (1:1000 dilution, no. ab92331; Abcam), and PD-L1 (1:500 dilution, no. ab205921; Abcam). Immunostained slides were observed under a microscope and were scored using the percentage of positive cells as the basis for the (0–4) score. Scores were as follows: 0: no staining; 1: <10% positive cells; 2: 11–50% positive cells; 3: 51–75% positive cells; 4: >75% positive cells. Individual samples were evaluated by at least two pathologists in a blinded manner; those expression scores of ≥2 were defined as high expression, and those <2 were defined as absence or low expression.

Statistical analyses were performed using SPSS 18 and Prism 8.0. For normally distributed data, significance of mean differences was determined using two-tailed paired or unpaired Student t tests; Two-way ANOVA was used to compare mouse tumor volume data among different groups. Correlations were determined by Pearson r coefficient. Log-rank and Wilcoxon tests were used to analyze the mouse survival data. A p value < 0.05 was considered statistically significant.

To investigate the potential relationship between S100A8 or S100A9 and PD-L1, we treated different cells with either S100A8 or S100A9 recombinant proteins. S100A8 or S100A9 can increase the PD-L1 mRNA expression levels in mouse macrophage cells RAW264.7 (Fig. 1A, 1B). However, this effect had not been observed in the mouse cancer cells (Fig. 1C, 1D). Because the effect of S100A8 on PD-L1 induction is much stronger comparing to that of S100A9 (Fig. 1A, 1B), we mainly focused on S100A8–PD-L1 regulation in this study.

S100A8 induced PD-L1 expression in a time-dependent manner (Fig. 1E), and it also can induce PD-L1 expression in human macrophage-like differentiated THP-1 cells (Fig. 1F). In contrast, inhibition of endogenous S100A8 expression by shRNA can decrease the PD-L1 expression significantly in macrophages (Fig. 1G, Supplemental Fig. 1A). By means of flow cytometry, we revealed that S100A8 can induce cell surface PD-L1 protein expression in both RAW264.7 cells and mouse BMDMs (Fig. 1H, 1I).

S100A8 and S100A9 participate in the inflammatory response by binding to TLR4 (22) or RAGE (23) and thus triggering the TLR4- or RAGE-mediated inflammatory pathways. To determine through which receptor S100A8 inducing PD-L1 transcription, we used TLR4 inhibitor TAK-242 or RAGE inhibitor FPS-ZM1 to treat the cells. The inhibitor TAK-242 reversed S100A8-induced PD-L1 upregulation but not the inhibitor FPS-ZM1 (Fig. 2). In both RAW264.7 cells (Fig. 2A, 2B) and BMDMs (Fig. 2C), we observed the same phenomenon. In contrast, overexpression of TLR4 in RAW264.7 cells enhanced the effect of S100A8 on PD-L1 induction (Fig. 2D, 2E). This result suggested that TLR4 is needed in the S100A8-induced PD-L1 expression.

Through blocking mRNA synthesis with actinomycin D, and blocking protein synthesis with cycloheximide, we found that S100A8 treatment did not affect the PD-L1 mRNA and protein half-life (Supplemental Fig. 2). Multiple signaling pathways are reported to be activated by the S100A8 stimulation (24–26). We observed that AKT, ERK, STAT3, and NF-κB signalings are activated upon S100A8 stimulation in RAW264.7 cells (Fig. 3A). Then, specific signaling inhibitors are employed to explore which signaling pathways are downstream of S100A8 to induce PD-L1 transcription. To our surprise, blocking each of these signaling can suppress the induction of PD-L1 by S100A8 stimulation in RAW264.7 cells (Fig. 3B–D). In addition, a variety of transcription factors have been reported to be associated with increased PD-L1 transcription (27–33). We also observed that S100A8 treatment promoted nuclear translocation of STAT3 and p65 (NF-κB) (Fig. 3E), the two most important transcription factors in the inflammation–cancer link. These results imply that S100A8 increase PD-L1 transcription through multiple signalings such as AKT/PI3K, ERK/MAPK, STAT3, and NF-κB.

To investigate the chromatin characteristics of the PD-L1 gene promoter, we analyzed the cell type–specific DNAse I hypersensitivity of this DNA region through the ENCODE Project (34). Monocytes showed DNAse I hypersensitivity peaks in the PD-L1 gene region (Supplemental Fig. 3A), whereas many cancer cells did not show these peaks in this region. These results suggested a transcriptionally active open chromatin (euchromatin) configuration in the PD-L1 gene of monocytes but a heterochromatic (repressive) chromatin in the PD-L1 gene regions of many types of cancer cells. ENCODE RNA sequencing analysis revealed very limited PD-L1 expression in monocytes, suggesting that without stimulation, monocytes normally do not express the PDL1 gene (Supplemental Fig. 3B). Specific histone modifications closely associated with transcriptionally active or inactive chromatin. Two important histone modifications linked with active chromatin are H3K4me3 and H3K27ac (35). The ENCODE ChIP sequencing data (36) showed high levels of H3K4me3 and H3K27ac enrichment in the PD-L1 promoter region in monocytes but not in many types of cancer cells (Supplemental Fig. 3C). However, the repressive mark H3K27me3 enrichment has no significant difference between monocytes and cancer cells (Supplemental Fig. 3C). So, results from the ENCODE analysis of chromatin state at the PD-L1 promoter locus in different cell types indicate that PD-L1 promoter harbors a transcriptionally active chromatin configuration in monocytes but not in cancer cells.

To determine whether S100A8 affects chromatin changes in the promoter region of macrophage PD-L1, we designed primer sets to test recruitment of H3K4me3, H3K27me3, and H3K27ac to five different regions of the PD-L1 promoter and two regions of the PDL1 TSS by ChIP-qPCR (Fig. 4A). H3K4me3 and H3K27ac are modifications that promote gene transcription, whereas H3K27me3 is a marker for repressive transcription, and all of them have been shown to be involved in PD-L1 transcription. We assessed the recruitment of H3k4me3, H3K27ac, and H3K27me3 to the PD-L1 promoter by ChIP-qPCR analysis and found that H3K4me3 and H3K27ac were recruited to this region in an S100A8-dependent fashion, whereas H3K27me3 recruitment was decreased upon S100A8 treatment in RAW264.7 cells (Fig. 4B–D). We also compared the effect of S100A8 treatment on H3K4me3 and H3K27me3 recruitment to PD-L1 promoter region between macrophages (RAW264.7) and tumor cells (CT26). S100A8 increased the H3K4me3 recruitment and decreased the H3K27me3 recruitment in macrophages but not in tumor cells (Fig. 4E, 4F). The histone modifications (H3K4me3 and H3K27me3) in the PD-L1 promoter region are sensitive to S100A8 stimulation in the RAW264.7 cells but not in the CT26 cells.

The most important function of PD-L1 is to suppress T cell proliferation and function. We cocultured mouse-extracted T cells with mouse BMDMs for 72 h and then measured the T cell proliferation abilities under the influence of S100A8 treatment. Coculture with BMDMs promoted T cells proliferation (Fig. 5A), but this effect was compromised if the BMDMs were pretreated with S100A8 (comparing panel 3 versus 1 in Fig. 5B) (i.e., S100A8 treatment induced an immunosuppressive function of BMDMs on T cells). Addition of anti–PD-L1 Ab in this system had no significant effect on T cells proliferation (comparing panel 2 versus 1 in Fig. 5B); however, if the BMDMs were pretreated with S100A8, the anti–PD-L1 Ab can rescue the T cell proliferation (comparing panel 4 versus 3 in Fig. 5B). This result suggested S100A8-treated BMDMs exert T cells suppression through PD-L1 induction. Lysosomal-associated membrane protein 1 (CD107a) and granzyme B (GZMB) are important markers for judging the activation and function of CD8⁺ T cells (37, 38). T cells expanded with S100A8-pretreated BMDMs showed significantly reduced activation compared with the untreated BMDMs by measuring the granzyme B and CD107a-positive T cells (comparing panel 3 versus 1 in Fig. 5C, 5D). Anti–PD-L1 Ab can rescue the T cell activation under the influence of S100A8 stimulation (comparing panel 4 versus 3 in Fig. 5C, 5D).

We next questioned whether S100A8 treatment can modulate the macrophages polarization (i.e., M1 or M2 switch) (39). S100A8 dramatically increased the expression levels of the molecular markers of M1 macrophages as well as M2 (Fig. 5E, 5F), which is different from LPS, a classic M1 polarization stimulator (40). This result implies that S100A8-stimulated macrophages are not simply polarized to M1 or M2, and a more complex mechanisms may be existed in this process. We thus compared the effect of S100A8 and LPS on the macrophage polarization. Like LPS, S100A8 can increase the expression of proinflammatory molecules (TNF-α, IL-6, IL-12); however, it also can increase the expression of immunosuppressive molecule (IL-10), which is different from LPS, and S100A8 has significant stronger effect on the induction of PD-L1 expression than that of LPS (Fig. 5G). At the same time, S100A8 can positively induce the expression of itself, so that molecules can accumulate rapidly in the inflammatory process and exert stronger regulatory functions (Fig. 5G). Taken together, S100A8 treatment makes the macrophages becoming immunosuppressive and has a complex effect on their polarization.

To extend these results from in vitro to in vivo, we used a mouse xenograft tumorigenesis model to assay the effect of S100A8 on BMDMs. We performed adoptive transfer experiments with BMDMs and tumor-primed cytotoxic T cells. BMDMs were harvested from the health BALB/C mice, and the tumor-primed cytotoxic T cells were generated from the spleen of the CT26 tumor–bearing BALB/C mice. Before adoptive transfer experiment, CT26 tumor cells were s.c. injected into NOD/SCID mice on day 0. On day 8, adoptive transfer was performed with BMDMs and tumor-primed cytotoxic T cells mixture. On day 18, NOD/SCID mice were sacrificed to observe tumor mass. As Fig. 6 shows, adoptive transfer with BMDMs and tumor-primed cytotoxic T cells inhibited the tumor growth comparing to PBS injection, whereas the S100A8-pretreated BMDMs reversed this effect. Both shRNA targeting PD-L1 and anti–PD-L1 therapy can abate the effect of S100A8 treatment. This result suggested that S100A8-treated BMDMs suppressed the cytotoxic function of the tumor-primed T cells in vivo through PD-L1.

Having determined that PD-L1 is an S100A8-responsive gene in macrophages, we investigated their expression correlation in the tumor tissues. We performed bioinformatics analysis on the gene expression profiling dataset from TCGA. The results demonstrated that S100A8 expression positively correlates with PD-L1 (Fig. 7A) in the tissues of colorectal cancers. We also noticed a positive correlation between S100A8 and CD86 or CD163 (Fig. 7B, 7C), suggesting that the amount of S100A8 is significantly correlated the infiltration of macrophages in tumor tissues. In addition, higher expression levels of S100A8, PD-L1, CD86, and CD163 are associated with poorer disease-free survival of colorectal cancer patients (Fig. 7D–G). We further validated S100A8 and PD-L1 expression levels in human CRC specimens by means of immunohistochemistry. S100A8 and PD-L1 were highly expressed in 104 cases of CRC tissues compared with normal colorectal tissues; meanwhile, S100A8 expression was positively correlated with PD-L1 level (Supplemental Fig. 4, Supplemental Table II).

We asked one more question, whether the expression levels of S100A8 could affect the cancer patients’ response to the anti–PD-1 therapy. Indeed, regarding patients with melanoma, renal cell carcinoma, and bladder cancer who get anti–PD-1 therapy, unresponsive patients have higher S100A8 expression comparing to that of the responsive patient, although the significant p value is bigger than 0.05 due to the limited patient numbers (Supplemental Fig. 5).

Anti–PD-1 or PD-L1 has been a successful antitumor therapy strategy in recent years. Understanding the mechanism that modulating the PD-L1 or PD-1 expression in the tumor microenvironment is required for better therapy strategy design and side effects prevention. Multiple mechanisms responsible for the upregulation of PD-L1 expression in the tumor microenvironment has been revealed, including inflammation mediators IFN-γ, TNF-α, TGF-β, IL-10, etc. Those discoveries raise a concept that inflammation not only stimulates cell proliferation, which can lead to tumorigenesis as a consequence, but also represses the host’s antitumor immunity ability, which makes the inflammation become more vicious than we previously described.

S100A8 and S100A9 are inflammatory mediators released by neutrophils and monocytes, as well as other cell types. We previously reported the S100A8 and S100A9 were upregulated dramatically throughout the colitis-associated cancer process and therefore gained our focus (19, 20). S100A8 is generally considered to promote the development of inflammation, but there are some different aspects appeared. Under aseptic inflammatory conditions, S100A8 is able to reduce macrophage response to LPS stimulation via a methyltransferase-dependent chromatin methylation pathway in a TLR4-dependent manner (41). Bone marrow precursor cells of tumor patients overexpress S100A8/S100A9, thereby inhibiting the differentiation and maturation of DCs and macrophages, increasing the number and function of myeloid-derived suppressor cells (42). In a mouse lung carcinogenesis model, blocking S100A8 inhibited myeloid-derived suppressor cell activity and restored T cell–mediated antitumor immune responses (43). These discoveries revealed another aspect of S100A8 in modulation of inflammation and antitumor immunity. In this study, we reported that S100A8 can significantly upregulated PD-L1 expression in the macrophage cells and thus endowed the cells an immunosuppressive function.

We found that S100A8 can dramatically induce PD-L1 expression; however, the effect of S100A9 is not that significant. Recently, Cheng et al. (44) reported S100A9 induced PD-1/PD-L1 expression in myelodysplastic syndromes through the transcription factor c-Myc. Although S100A8 and S100A9 usually formed a heterodimer, their functions are varied in different circumstances (14). It is interesting that S100A8 can induce PD-L1 expression in macrophages but not in cancer cells, which may relate to different epigenetic modification patterns of the PD-L1 gene promoter region among the different cell types. Our ChIP experiment result revealed that S100A8 treatment can modulate the H3K4me3 and H3K27me3 modifications in the macrophages (RAW264.7) but not in the tumor cells (CT26). This discovery suggesting the discrepancy of histone modification patterns between macrophages and cancer cells are important for PD-L1 regulation. Similar finding has been reported for CTLA-4 regulation (45).

We dissected the signalings between S100A and PD-L1. S100A8 exhibits cytokine-like functions via binds and activates the receptors such as RAGE (46) and TLR4 (22). We noticed TLR4 is highly expressed in the monocytes but not in the cancer cells, whereas the RAGE expression is very limited in the monocytes (data from the ProteomicsDB), implying a limited role of RAGE in the S100A8/PD-L1 axis of the monocytes. Studies have found that IL-10 can upregulate the expression of PD-L1 through the Jak2–Stat3 axis (47, 48). IFN-γ upregulates PD-L1 through different pathways in different cells. For instance, in multiple myeloma, IFN-γ induces PD-L1 expression through TLR ligands (28), and in lung cancer cells, IFN-γ induces PD-L1 expression through the Jak/Stat pathway (49). Multiple signalings are reported to be activated by S100A8 stimulation, and in this study, we confirmed that AKT, ERK, STAT3, and NF-κB signalings are required for the S100A8-induced PD-L1 upregulation. We also showed that S100A8 can significantly increase the expression levels of TNF-α and IL-10, both of which are strong PD-L1 inducers. This raised a possibility that S100A8 is a main stimulator for PD-L1 induction in macrophages, and multiple signalings can be used by it to fulfill this mission. Investigating through which pathways S100A8 upregulates PD-L1 expression can provide theoretical basis for the future development of drugs to targeting S100A8-PD-L1 axis.

The infiltration of macrophages in tumor tissues to inhibit tumor immunity has been reported (50). We previously have reported that S100A8 can act as chemoattractant proteins by recruiting macrophages (20). The most important function of PD-L1 is to suppress T cell proliferation and function; in this study, we found that S100A8-pretreated macrophages effectively suppressed the T cell proliferation and cytotoxic function and that effect was attenuated by blocking PD-L1 on the macrophages. Unlike the LPS, a classic M1 polarization stimulator, S100A8 treatment did not simply induce the macrophages to polarizing to M1 or M2 phenotype. These findings imply the significance of the S100A8-induced PD-L1 in the macrophages under the circumstance of tumor microenvironment and indicate the difference between LPS and S100A8.

S100A8 often plays a proinflammatory role in the inflammatory response, yet the biological function of S100A8 is diverse. Our findings even further clarify its diversity: a double role in promoting inflammation and suppressing antitumor immunity. A few studies have reported the anti-inflammatory function of S100A8 under specific conditions to avoid tissue damage caused by overwhelming inflammation (51). Macrophages are involved in many biological processes, such as tissue development and tissue repair, defense against pathogens, chronic inflammation, and tumorigenesis, which implies that macrophages would be exceptionally plastic and adapt extremely well to their microenvironment to perform a vast array of cellular functions [2]. Under the influence of S100A8, the macrophages expressed high PD-L1 proteins on their surface, and thus could suppressed the cytotoxicity function of the T cells in the microenvironment. Our finding revealed a vicious side of the macrophages, “trained” by the inflammatory mediator such as S100A8, can repress the antitumor immunity of the host. Because S100A8 is highly expressed in many inflammation processes, the significance of this discovery is worthy of attention. Release of S100A8 has been associated with senescence and age (52–54); therefore, targeting the S100A8-PD-L1 axis may also modulate the senescence phenotype. In the “real world” of human cancer specimens, the expression levels of S100A8 and PD-L1 were extremely positively correlated, whereas the high expression of S100A8 and PD-L1 was associated with poor prognosis of cancer patients. The macrophage markers (CD86 and CD163) are also positively correlated to S100A8 expression.

The novelty of this study is that we identified, to our knowledge, that S100A8 can dramatically induce PD-L1 expression levels in macrophages but not in cancer cells. S100A8-trained macrophages lost their antitumor function and turned on their promoting-tumor function through expressing PD-L1 on their surface and thus repressing T cell cytotoxic function. This discovery provides new evidence, to our knowledge, that inflammation mediators can use the extreme plastic ability of the macrophages and induce them from antitumor function to promoting-tumor function. How to modulate this axis is a new challenge for the field of tumor immunotherapy.

We thank Professors Jie Zhou and Penghui Zhou for providing reagents and help.

The authors have no financial conflicts of interest.