Gram-Negative Microflora Dysbiosis Facilitates Tumor Progression and Immune Evasion by Activating the CCL3/CCL5–CCR1–MAPK–PD-L1 Pathway in Esophageal Squamous Cell Carcinoma

Esophageal squamous cell carcinoma (ESCC), which accounts for more than 90% of esophageal cancer cases in China, is a common upper gastrointestinal tract malignant tumor and is responsible for an estimated 512,500 new cases and 49,570 deaths worldwide in 2020 (1). Almost half of the cases happened in China, and the incidence of ESCC is still increasing in China (2). ESCC has become a worldwide healthcare concern and represents an aggressive malignancy. Therefore, more profound research on the molecular mechanism of the tumorigenesis and development of ESCC might contribute to more effective management and treatment of ESCC, which is of great clinical value.

ESCC had a significant relationship with nutritional status and eating habits. Also, the above factors might affect esophagus colonizing microflora composition (3). Gram-negative (G−) bacilli are found to be the dominant microflora in the ESCC microenvironment. The continuous stimulation of G− bacilli may cause tumorigenesis. Researchers have found that intratumoral bacteria could predict the treatment efficacy of neoadjuvant chemotherapy combined with immunotherapy in the ESCC (4). Another research introduced that Fusobacterium nucleatum-Dps could be bound to the PD-L1 gene promoter activating transcription factor-3 to transcriptionally upregulate PD-L1 expression in ESCC (5). All these results indicated that the tumor colonizing G− flora might affect effects of immunotherapy in ESCC. Lipopolysaccharide (LPS) is one of the main components of the cell wall of G− bacilli (6). LPS has been reported to promote the proliferation, migration, and stemness of various tumors including ESCC (7–9). However, no study has investigated whether LPS could induce the immune evasion of ESCC.

In recent years, esophageal squamous carcinoma immunotherapy has attracted tons of attention. Many research studies focused on PD-L1 expression. PD-L1 expression was believed to be the most promising biomarker for immune checkpoints inhibitors (10). PD-L1 can cause T-cell dysfunction and failure, which prevents cytotoxic T cells from effectively targeting tumor cells through combining to the programmed death receptor 1 (PD-1) on T cells, and thus promotes the tumorigenesis activity (11, 12). Research involving 5,257 patients from 10 randomized controlled trials of immunotherapy for advanced ESCC demonstrated that immunotherapy showed improvements in survival outcomes (HR, 0.71), progression-free survival, and overall response rate (HR, 0.78; OR, 1.5). The analysis showed that the benefit in ESCC was higher in the case of PD-L1 CPS ≥ 10 (HR, 0.60 vs. 0.83 in PD-L1 < 10; P = 0.009), suggesting a promising role for PD-L1 as molecular biomarkers in ESCC (10). It was reported that neoadjuvant immunotherapy with PD-L1 antibody combined with surgery is expected to improve the survival rate (13). However, many patients who received immune checkpoint inhibitors still suffered from abnormally upregulated PD-L1 expression. One of the possible reasons for this was acquired immune resistance. Tumor-infiltrating lymphocytes induce the expression of PD-L1 and form a tumor-suppressing microenvironment, which plays a key role in PD-1/PD-L1 inhibitor resistance. The presence of primary and acquired resistance to this blocking therapy limits its efficacy (14). Therefore, when PD-L1 expression increases as a result of therapy such as gefitinib treatment, re-selection of drug targets should be considered. It is very necessary to study the new possible factors regulating PD-L1 expression.

Chemokines are a specialized family of chemotactic cytokines with molecular weights ranging from 8 to 14 kDa (70–120 amino acids), and a wide variety of cells like macrophages and T and B lymphocytes could release them (15). They potently mediate inflammation by recruiting and activating specific leukocyte subpopulations. CC chemokine receptor 1 (CCR1) is a widely studied G protein–coupled receptor target expressed on multiple cell types. CCR1 overexpression has been described in several types of cancer and is associated with increased immunosuppressive cell infiltration and metastasis (16).

These pieces of evidence indicated that CCR1 is closely related to tumor immunotherapy. However, it remains largely unknown whether CCR1 is overexpressed in ESCC tissues and whether CCR1 promotes ESCC immune evasion.

The compound 4-nitroquinoline N-oxide (Cat. N8141; Sigma) was dissolved in drinking water with 100 μg/mL concentration as the carcinogen. Six-week-old male C57BL/6 mice were given this water for 16 weeks to induce ESCC. After that, regular drinking water was given for 8 weeks. Then, we euthanized the mice and collected esophageal samples.

Sequencing of microbial 16S amplicons was performed by TIANGEN Biotech Co, Ltd. Briefly, total genomic DNA from samples was extracted using CTAB or SDS method. DNA concentration was determined using the NanoDrop spectrophotometer. According to the concentration, DNA was diluted to 1 ng/μL using sterile water. rRNA genes (16S/18S) were amplified using the specific primer with the barcode. Samples with bright main strip band between 400 and 450 bp were chosen for further experiments. Amplicons were pooled in equal proportions and purified using TIANgel Purification Kit (TIANGEN Biotech). Sequencing libraries were generated using TIANSeq Fast DNA Library Prep Kit (Illumina; TIANGEN Biotech). The library quality was assessed on the Qubit@ 2.0 Fluorometer (Thermo Fisher Scientific) and Agilent Bioanalyzer 2100 system. At last, the library was sequenced on the Illumina platform using the 2 × 250 bp paired-end protocol.

Human ESCC KYSE150 (RRID: CVCL_1348) and TE1 (RRID: CVCL_1759) and Jurkat cells (RRID: CVCL_0065) were obtained from the Cell Bank of Chinese Academy of Sciences. To prevent Mycoplasma contamination, cells were tested every 2 months using a PCR-based method suggested by Uphoff and Drexler (17). The supernatant medium without antibiotics after 7 days of cell culturing whose DNA was extracted and purified using silica-gel columns (TIANGEN) was collected. After that, PCR assays were performed with hot-start Taq DNA polymerase according to the protocol provided by the manufacturer. PCR products were separated on a 1.3% agarose-TAE gel (Sangon Biotech) containing 0.3 mg/mL ethidium bromide, and the results were visualized with a UV transilluminator. If the tested medium contained Mycoplasma, an additional band at 515 to 525 bp could be observed. Only the Mycoplasma-negative cells are kept for subsequent experiments. The above cells were cultured in RPMI-1640 (Gibco) with 10% FBS and 1% antibiotics. ESCC cells and Jurkat cells were cocultured in transwell rooms (14111, LABSELECT). A total of 1 × 10⁶ KYSE150 or TE1 cells were seeded in the under room, and 5 × 10⁵ Jurkat cells were seeded in the upper room with 1 mL RPMI-1640 in both upper and under rooms. A total of 2.5 μg/mL antihuman CD3 (300331, BioLegend, RRID: AB_11147368) and antihuman CD28 (302933, BioLegend, RRID: AB_11150591) were used to activate Jurkat cells in all groups. LPS (L4391, Merck) was dissolved in double distilled water to overexpress PD-L1 cells, and BX471(HY-12080, MCE) was dissolved in DMSO to block CCR1 expression in KYSE150 cells and TE1 cells. PBS or 5 μg/mL LPS or 10 μg/mL LPS was added in the PD-L1 overexpressed groups and DMSO or 250 nmol/mL BX471 or 500 nmol/mL was added in the CCR1 blockage groups. All cell lines used in our research were short tandem repeat authenticated by Biowing Biotech.

Reference transcriptomes and statistical analysis were performed by TIANGEN Biotech Co, Ltd. The NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific) was used to determine the concentration and evaluate the purity of RNA samples. Agilent 2100 Bioanalyzer and 2100 RNA Nano 6000 Assay Kit (Agilent Technologies) were used to evaluate the integrity of RNA samples. Briefly, the transcriptome sequencing library was constructed through RNA randomly fragmentation, cDNA strand 1/strand 2 synthesis, end repair, A-tailing, ligation of sequencing adapters, size selection, and library PCR enrichment. The clustering of the index-coded samples was performed on a cBot Cluster Generation System using TruSeq PE Cluster Kit v3-cBot-HS (Illumina) according to the manufacturer’s instructions.

Cells at a concentration of 800,000/mL in the logarithmic growth phase were seeded in a 6-well plate and incubated overnight. The cells were then scratched with a 200 μL sterile pipette tip and incubated with 2 mL RPMI-1640 without serum. Mobility was calculated using the following equation: (0-hour scratch width–24-hour scratch width)/0-hour scratch width × 100% and (0-hour scratch width–48-hour scratch width)/0-hour scratch width × 100%.

For the Transwell invasion assays, a Transwell system (12 μm pore size, BD Biosciences) was used. The upper chamber side of the polycarbonate membrane was lined with a layer of matrix glue. A total of 6 × 10⁵ cells were seeded in the top chamber of each insert and cultured in serum-free medium. RPMI-1640 with 5% FBS was added to the lower chamber. After 72 hours, migrating cells were fixed, stained, and quantified for a total of five fields per membrane.

The required cells were plated in a 96-well plate with 5,000 cells per well, the day before the experiment. After the cells adhered, CCK8 (TargetMOI, KA251771) was added to the wells, and the absorbance was measured after incubating for 2 hours at 37°C. The absorbance was measured after 24, 48, and 72 hours.

Total RNA was extracted using TRIzol reagent (R0016, Beyotime Biotechnology). The PrimeScript RT reagent kit (TIANGEN Biotech Co. Ltd.) was used to convert RNA into cDNA. qRT-PCR analysis was conducted using 2 × Hot-Start Taq PCR MasterMix (TIANNGEN Biotech Co., Ltd.) on a fluorescent PCR device (7500, Thermo Fisher Scientific). The primers were as follows: PD-L1 (forward: 5′- GCT​GCA​CTA​ATT​GTC​TAT​TGG​GA -3′ and reverse: 5′- AAT​TCG​CTT​GTA​GTC​GGC​ACC- 3′). CCR1 (forward: 5′- GAC​TAT​GAC​ACG​ACC​ACA​GAG​T-3′ and reverse: 5′-CCA​ACC​AGG​CCA​ATG​ACA​AAT​A -3′), CCL3 (forward: 5′- TGT​ACC​ATG​ACA​CTC​TGC​AAC -3′: and reverse: 5′- CAA​CGA​TGA​ATT​GGC​GTG​GAA -3′) and CCL5 (forward: 5′-CCT​CGC​TGT​CAT​CCT​CAT​TGC​T-3′ and reverse: 5′- CTT​CTC​TGG​GTT​GGC​ACA​CAC​T-3′). The results were expressed using the 2−△△Ct method.

The cells were lysed in RIPA lysis buffer (Beyotime, P0013B) supplemented with protease inhibitor cocktail (Topscience, C0001) and phosphatase inhibitor cocktail I + II (Topscience, C0001 and C002). The bicinchoninic acid method was used to determine protein concentrations. The protein samples were separated by SDS-PAGE and transferred to a nitrocellulose membrane (FFP24, Beyotime Biotechnology). The membranes were blocked and incubated with the following primary antibodies: PD-L1(13684, Cell Signaling Technology, RRID: AB_2687655), CCR1(LS-C483302-50, LSBio), CCL3(23-032, ProSci), CCL5(515502, BioLegend, RRID: AB_2071299), MAPK (4690, Cell Signaling Technology, RRID: AB_390779), p-MAPK (4370, Cell Signaling Technology, RRID: AB_2315112), and actin (AA128, Beyotime Biotechnology). Then, the membranes were incubated with the appropriate horseradish peroxidase–conjugated secondary antibody (goat antirabbit IgG, 1:3,000; goat antimouse IgG, 1:4,000; CWBIO). Bands were detected using the ECL Western Blotting Detection System.

Cells were treated according to instructions and analyzed by BD FACSCelesta. Then, the following antibodies were used: PD-L1-PECy7 (329717, BioLegend, RRID: AB_2561686), CD3-BV421 (300433, BioLegend, RRID: AB_10897105), CD107a-PE (328608, BioLegend, RRID: AB_1186040), Ki67-APC (350513, BioLegend, RRID: AB_10959326), LAG3-APC-Cy7 (369329, BioLegend, RRID: AB_2734419), and PD-1-PE-Cy7(329918, BioLegend, RRID: AB_2159324).

Four-week-old male BALB/c-nude mice (Strain NO. D000521) were purchased from GemPharmatech. ESCC cell lines KYSE150 and TE1 were cocultured with Jurkat cells with PBS or LPS (10 μg/mL) or LPS (10 μg/mL) with BX471 (500 nmol/mL) in vitro. A total of 2 × 10⁶ ESCC cells and 4 × 10⁴ Jurkat cells were injected subcutaneously into one side of mice groin. Six weeks later, the mice were executed with cervical dislocation, and the size of tumors was measured by caliper fo00 statistical analysis. For LPS or LPS with BX471 pretreated groups, cells were treated with LPS or LPS with BX471 for 1 week before the assay, and the resuspension was also supplementary with LPS or LPS with BX471 of corresponding concentration. During the next 30 days, the irritants were continuously injected into the tumor. The experiments complied with the ARRIVE guidelines, and the care for the animals was in accordance with NIH guide for the care and use of laboratory animals.

Five human ESCC tissues were generated by the Zhongshan Hospital. Samples of ESCC (five cases) were analyzed for PD-L1, CCR1, CCL3, and CCL5 expression. Clinicopathologic parameters, including age at diagnosis, histopathologic type, tumor size, and nodal status, were evaluated. All patients had not received chemotherapy or radiotherapy. All specimens were obtained after obtaining informed consent under an institutional review board–approved protocol of the Zhongshan Hospital. The primary antibodies against PD-L1, CCR1, CCL3, and CCL5 were used (LS Bio, LS-C107430-50; LS Bio, LS-C813821-50; LS Bio, LS-C384561; and LS Bio, LS-C416972-50).

We selected 10 patients with pretreatment biopsy. Then, they were given preoperative anti–PD-1 therapy and 2 to 4 cycles of paclitaxel-based chemotherapy. R0 resection surgery was performed, and the biological samples were collected. Pathologic complete ypT0N0 or carcinoma in situ on the basis of histologic evaluation of the surgical samples was defined as pathologic complete response (pace). We leveraged single-cell RNA sequencing in ESCC tissues from the 10 patients. We obtained written informed consent from all the patients. The studies were conducted in accordance with Declaration of Helsinki, and the studies were approved by Institutional Review Board and Ethics Committee of Zhongshan Hospital, Fudan University. The ethics approval number is B2021-129.

R software version 3.63, R package Gene Expression Signature (RRID: SCR_000455), R package DSG-seq (RRID: SCR_001104), SPSS software version 25.0, FlowJo (RRID: SCR_000410), and GraphPad Software version 9.0 (RRID: SCR_002798) were used to assess the data. Data are presented as the mean ± SEM. Student t test was used to compare the differences between two groups. Correlations between two variables were analyzed by Pearson's correlation. Kaplan−–Meier survival curves and log-rank test were employed to depict overall survival (OS). The differences in tumor volume among groups were analyzed by ANOVA, followed by a two-tailed Student t test. The statistical significances of groups are represented as *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001.

This research was approved by Zhongshan Hospital Ethics Committee at Fudan University in Shanghai, (Approval No. B2017–153). The Institutional Review Committee at Zhongshan Hospital, Fudan University (Shanghai, China), granted consent for all animal procedures and ensured that all animals received standard care (approval number: IBCB-SPF1843). We obtained written informed consent from all the patients. We strictly followed the declaration of Helsinki in conducting this work, and it also conformed with all applicable ethical guidelines for using animals in research.

The data used to support the findings of this study are available in this published article and either online (10.6084/m9.figshare.2527214810.6084/m9.figshare.25272148) or from the corresponding author upon request.

A total of 10 male C57BL/6 mice were used to induce ESCC. The ESCC tissue and normal esophageal epithelial tissue were used for 16S rRNA sequencing (RNA-seq), and three groups of them were effectively measured. Firmicutes, Proteobacteria, and Bacteroidota were the dominant microflora in both ESCC tissues and normal esophageal epithelial tissues (Fig. 1A). Although the main types of microflora in ESCC tissues and normal esophageal epithelial tissues were almost the same, the distribution numbers of each type of bacterium showed significant differences. The proportion of Gram-positive (G+) bacterium Firmicutes and G− bacterium Proteobacteria was 92.81% and 4.5% in normal esophageal epithelial tissue, and 39.09% and 57.03% in ESCC tissue, respectively (Fig. 1B and C; P < 0.05), which presented significant discrepancies. Furthermore, we detected the detailed genus of the microflora in mice normal esophageal epithelial tissues and ESCC tissues. We found that in normal esophageal epithelial tissues (NEEC), Lactobacillus was the dominant microflora in mice NEEC tissues. The proportion of Lactobacillus in mice normal esophageal epithelial tissues and ESCC tissues was 80% and 10%, respectively. Meanwhile, Rodentibacter had a significant presence in mice ESCC tissues, accounting for 55%. Other dominant microflora in the ESCC tissues included Gemella and Streptococcus (Fig. 1D and E). We analyzed G+ and G− bacterium abundance in mice NEEC tissues and ESCC tissues. G− bacterium abundance in ESCC tissues was 69%, whereas in NEEC tissues, it was 8.5%. The results demonstrated that G− abundance significantly upregulated in ESCC tissues (Fig. 1F).

To compare the bacterial abundance between mice and humans, we analyzed ESCC microflora data from TCMA. Both mice ESCC and human ESCC showed higher abundance of G−bacteria. Mice ESCC contained 55% G− bacteria and 25% G+ bacteria, whereas human ESCC tissue had 48% G− bacteria and 41% G+ bacteria (Fig. 1G). These results demonstrate significant microflora discrepancies in normal esophageal epithelial tissues and ESCC tissues in both mice and humans, with G− bacteria exhibiting much higher abundance in ESCC tissues.

It has been reported that G− bacteria usually act on esophageal tissues or tumors through LPS (5, 8). In order to investigate the potential role of LPS of G− in ESCC and T-cell cocultured system, we performed CCK8 assays to detect the proliferation of both ESCC cell lines. The proliferation of ESCC cells had a positive relationship with LPS concentration (Fig. 2A). We also stimulated ESCC cell lines in different LPS concentrations in transwell room to perform invasion assays. The result showed significantly stronger invasion ability of KYSE150 and TE1 under LPS stimulation (Fig. 2B; P < 0.05). Scratch assays were performed, and the results showed that KYSE150 and TE1 cells showed faster heal of scratch under LPS stimulation compared with PBS, which was positively correlated to LPS concentration (Fig. 2C). All above results proved that LPS stimulation could enhance the proliferation, invasion, and migration of ESCC cell lines.

To unravel the potential molecular mechanisms of esophageal residential microflora in ESCC, we conducted reference transcriptome assays on ESCC tissues from mice. As shown in the heatmap and volcano map, ESCC tissue and normal tissue exhibited significant discrepancies in gene expression (Fig. 3A). Upon comparing mRNA expression between ESCC tissue and normal esophageal epithelial tissues from mice, we identified 2,091 upregulated genes (log FC > 0.5, P < 0.05) and 1,790 downregulated genes (log FC < −0.5, P < 0.05; Fig. 3B). Multiple studies have demonstrated that TLR4 acts as the receptor of LPS and upregulates PD-L1 (18). To explore potential regulators of PD-L1 in ESCC and its relationship with LPS stimulation, we analyzed TLR4-correlated genes (GSE53625) and differentially expressed genes (DEG) of ESCC stimulated by LPS F(GSM2807474). Through cross-comparing the above data, we found 781 upregulated TLR4-correlated genes, 6,037 upregulated genes from ESCC stimulated by LPS, and 2,091 upregulated genes from reference transcriptomes (Fig. 3C and D). Among them, eight genes intersected, and CCR1 was closely related to PD-L1 (Fig. 3E). Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis revealed that ESCC tissue exhibited greater enrichment in cytokine–cytokine receptor interaction (Fig. 3F and G). Gene set enrichment analysis showed that the MAPK signaling pathway was significantly activated in ESCC tissues (NOM P-value < 0.05, FDR q-value < 0.25; Fig. 3H). Based on the results of our research, we further detected whether CCR1 may cause OS differences in patients with ESCC. We used http://gepia2.cancer-pku.cn to perform Kaplan−Meier analysis based on gene expression. CCR1 high expression patients had 85 months of survival, which is lower than patients with low CCR1 expression. Macrophage inflammatory protein-1α (CCL3) and C-C motif chemokine ligand 5 (CCL5) are the ligands of CCR1. We combined CCR1 with TLR4, CCR1 with PD-L1, and CCR1 with CCL3, CCL5, and PD-L1. The results showed that all high expression combinations had lower OS (Fig. 3I). The correlation of CCR1 and TLR4 (R = 0.48, P < 0.05), CCR1 and PD-L1 (R = 0.24, P < 0.05), CCL3 and PD-L1 (R = 0.29, P < 0.05), and CCL5 and PD-L1 (R = 0.38, P < 0.05) was close (Fig. 3J). These results suggest that CCR1 may be overexpressed in ESCC with LPS stimulation and may be connected to tumor immune evasion.

To evaluate the molecular mechanism underlying CCR1 gene regulation with LPS stimulation and its regulation on ESCC immune evasion, we tested CCR1 expression in KYSE150 and TE1 cells by RT-PCR. After LPS stimulation, CCR1, its ligands CCL3 and CCL5, and PD-L1 mRNA expression were significantly increased and positively correlated with LPS concentration (Fig. 4A). Elevated CCR1, CCL3, CCL5, and PD-L1 protein levels were confirmed by Western blot (Fig. 4B). Research has proved that MAPK phosphorylation can induce tumor growth by rebuilding the tumor metastatic microenvironment in a CD4⁺ T cell, IFNγ, and a macrophage-dependent manner (19). p-MAPK protein levels are upregulated in LPS stimulation assays (Fig. 4C). These results suggested that LPS stimulation facilitated CCR1 expression, which activated MAPK phosphorylation and upregulated PD-L1. BX-471 is a piperazine-based CCR1 antagonist. It was able to displace the CCR1 ligands MIP-1α, RANTES, and MCP-3 with high affinity (Ki = 1–5.5 nmol/L; ref. 20). KYSE150 and TE1 cells were treated under different concentrations of BX471 for 48 h. BX471 effectively inhibited CCR1 and downregulated PD-L1 in both mRNA and protein expression (Fig. 4D and E). Western blot showed that CCL3 and CCL5 expression remained invariant, and p-MAPK expression reduced significantly (Fig. 4E and F). Flow cytometry proved that LPS stimulation induced PD-L1 expression, whereas BX471 suppressed it (Fig. 4G and H). These data demonstrated that CCR1 had a positive correlation with PD-L1 and mediated PD-L1 mRNA and protein expression through MAPK phosphorylation. To figure out whether CCR1 might affect ESCC malignant biological properties, we performed CCK8 assays on KYSE150 and TE1 cell lines. After CCR1 inhibition, the proliferation of ESCC cells significantly downregulated and had a negative relationship with BX471 concentration. It suggested that CCR1 inhibition resulted in decreased ESCC cell proliferation (Fig. 4I). We also stimulated ESCC cell lines in different BX471 concentrations in transwell room to perform invasion assays. The result demonstrated that the invasion ability of KYSE150 and TE1 was weakened when CCR1 was inhibited (Fig. 4J; P < 0.05). Scratch assays were performed to detect the migration of ESCC cell lines. CCR1 inhibition in KYSE150 and TE1 cells showed less or no advantage in healing of scratch (Fig. 4K). All above results proved that CCR1 inhibition could suppress the migration, invasion, and proliferation of ESCC cell lines.

T cells are the key tumor-suppressing cells by forming physical contact with malignant tumor cells and inducing the death of tumor cells through activating their intracellular signals (21, 22). Immune evasion occurs when PD-L1 on tumor cells and PD-1 on T cells are activated and combined (23). To figure out the role of LPS stimulation and CCR1 inhibition on T-cell immune evasion, we cocultured T cells with ESCC cell lines with single PBS, single LPS 10 µg/mL, and LPS 10 µg/mL with BX471 500 nmol/mL and performed flow cytometry to examine PD-1 on T cells. CCK8 assays were performed to measure the proliferation of T cells. The results demonstrated that LPS suppressed proliferation of T cells, whereas CCR1 inhibition could upregulate it. The proliferation of T cells had a negative correlation with LPS stimulation, whereas it was positively correlated to BX471 concentration (Fig. 5A and B). As was shown in our results, LPS stimulation led to the highest PD-1 expression in T cells. The proportions of PD-1 positive cells were 65.5% and 66.3%. In the CCR1 inhibition subset, the number of PD-1 positive cells significantly reduced (52.4% and 51.2%). The PBS control group had almost the same PD-1 expression of the CCR1 inhibition group (52.3% and 51.8%; Fig. 5C, P < 0.05). Next, we examined lymphocyte-activation gene 3 (LAG 3), the marker reflecting T-cell exhaustion, and observed the same trend. The proportions of LAG 3 positive cells in LPS stimulation were 53.8% and 51.5%. BX471 subset and PBS control group had nearly the same proportion of LAG3 positive cells, and LAG3 expression was significantly downregulated after CCR1 inhibition (Fig. 5D). On the other hand, limited cytotoxicity and proliferation of T cells can also reflect the immune evasion of the tumors (24). CD107a and INFγ were tested to measure the cytotoxicity of T cells. LPS stimulation suppressed CD107a and INFγ expression on T cells, whereas CC1 inhibition induced their expression. The proportion of CD107a and INFγ positive cells significantly increased in CCR1 inhibition groups (68.5% and 67.7%; 64.1% and 61%) compared with LPS stimulation groups (58.7% and 58.5%; 56.9% and 53.2%; P < 0.05; Fig. 5E–G). Ki67 was tested, and the results verified CCK8 assays. Higher Ki67 expression was tested in CCR1 inhibition subsets compared with LPS stimulation subsets or PBS control groups (P < 0.05; Fig. 5H), which suggested that CC1 inhibition promoted T-cell proliferation, whereas LPS stimulation suppressed it. These data proved that LPS stimulation induced T-cell immune evasion and exhaustion, whereas CCR1 inhibition could enhance T-cell cytotoxicity and proliferation.

To figure out how CCR1 may affect ESCC development and immune microenvironment in vivo, we cocultured ESCC cell line TE1 and Jurkat cell and stimulated them by PBS or LPS 10 µg/mL or LPS 10 µg/mL with BX471 (500 nmol/mL). The xenografts showed that LPS pretreatment could significantly induce tumor growth, whereas CCR1 inhibition suppressed tumor formation progress (Fig. 6A). We performed flow cytometry to detect immune evasion markers on both tumor and immune cells in the xenografts. The results double validated our hypothesis. As was shown in our results, LPS stimulation upregulated PD-L1 expression in tumor cells (Fig. 6B and C) and increased the expression of T-cell immune evasion and exhaustion markers PD-1 and LAG3 while downregulating cytotoxicity markers like CD107a and INFγ. CCR1 inhibition suppressed T-cell immune evasion and induced immune response (Fig. 6D–G).

We collected ESCC tissues from five patients. IHC staining was performed on ESCC tissues of the five patients. The IHC results were ranked from 0 to 3 points. In stage III, PD-L1 and CCL5 marked three points, and CCR1 and CCL3 marked four points. In stage II, PD-L1, CCL5, and CCL3 marked two points, and CCR1 marked three points. In stage I, PD-L1, CCR1, and CCL3 marked one point, CCL3 marked zero points (Fig. 7A). Correlation analysis was performed between IHC points and pathologic stages. The results demonstrated that PD-L1, CCR1, CCL3, and CCL5 expression had a positive correlation with tumor stages (Fig. 7B). In mice ESCC 16s RNA-seq assays, we found that the effects of microflora on potential functional pathways had significant differences and had centralization. The results showed that ESCC microflora may activate the phosphotransferase system and lots of other pathways related to tumorigenesis (Fig. 7C). In order to further figure out the role of microflora in the tumor microenvironment (TME) in human ESCC, we performed 16s RNA-seq and single-cell RNA-seq from 10 patients who received neoadjuvant immunotherapy from our hospital. The results showed significant G− abundance discrepancy. G− abundance was higher in SD patients (Fig. 7D; P < 0.05). We measured Tim3, LAG3, CD107a, and PD-1 in CD8⁺ T cells, and the results showed that Tim3, LAG3 and PD-1 were significantly upregulated in SD patients, whereas CD107a had no differences (Fig. 7E). To figure out the relationship between G− microflora and T cells, correlation analysis was performed between G− abundance and gene expression. The results demonstrated that Tim3, LAG3, and PD-1 had positive correlation with G− abundance in our patients (Fig. 7F). We also explored bacterium metabolites in ESCC tissues. We selected several metabolites closely related to G− bacterium and separated patients into two clusters. The results showed that high G− bacterium metabolite deposition led to higher accumulation of B cells and fewer CD4⁺ T cells in tumor cells (Fig. 7G). With all findings in our research, we built the potential mechanism of how G− bacterial might affect ESCC immunotherapy through CCL3/CCL5–CCR1–MAPK–PD-L1 pathway (Fig. 7H).

The TME disorder may cause reginal chronic inflammation, which is one of the most important factors in promoting tumorigenesis and development (24). Gram−bacteria produce plenty of LPS and the accumulation of LPS participates in multiple kinds of tumors. LPS can be used as the entry point of local inflammatory response to reflect the changes of local inflammatory microenvironment of tumor, and the regulatory effect of the microenvironment on tumor progression and therapeutic drug sensitivity. Enrichment of F. nucleate has been viewed as a common feature of human colon cancers and adenomas (25). In this current study, according to the 16s RNA-seq, we found that there existed microflora dysbacteriosis in ESCC tissues. The normal esophagus epithelial tissue had more G+ bacterium abundance, whereas G−bacterium is the dominant microflora in ESCC tissues. Microflora dysbacteriosis led to different metabolite accumulation such as LPS in cells, which may cause chronic inflammation and finally cause tumorigenesis. To measure effects of LPS stimulation on ESCC cell lines, we set PBS control and concentration gradients of 5 and 10 µg/mL. Our result clearly indicated that LPS stimulation promoted ESCC cell migration, invasion, and proliferation. Similar results were found in several studies. In breast cancer, LPS was believed to activate PI3K/Akt/GSK–3β/β-catenin through binding to TLR4, thus promoting angiogenesis in tumor tissues and leading to hypermetabolism of tumor cells (26). In gastric cancer, it is commonly recognized that Helicobacter pylori plays an important role in gastric cancer occurrence and development (27). LPS stimulation upregulated CXC chemokine receptor 7 which could promote proliferation and migration of tumor cells in gastric cancer (28). Also, research reported that LPS could induce immune escape through its O-antigen (29). LPS could promote m6A methylation modification of MIR155HG, which further upregulated METTL14 and PD-L1 in hepatocellular carcinoma (30). Some researchers have confirmed that LPS aggregates in colon cancer tissues and is significantly associated with poor response to PD-L1 monoclonal antibody. Effective control of G− bacteria or blocking TLR4 signaling pathway can significantly reverse the immunosuppressive TME and promote T-cell infiltration in tumor cells, thereby improving the therapeutic effect of PD-L1 mAb (31). Researchers proved that LPS stimulation induces the stemness of ESCC cells through activating TET3–HOXB2 signaling axis (9). It suggested that LPS stimulation may promote ESCC migration and invasion. Furthermore, lots of pieces of evidence indicated that LPS may not only cause inflammation but also induce immune evasion through activating PD-L1 expression on tumor cells. Previous research demonstrated that LPS was abundant in orthotopic colorectal cancer tissue and was associated with low responses to anti–PD-L1 mAb therapy (32). In gastric cancer, LPS stimulation could highly increase PD‐L1 expression in gastric cancer cells through NF‐κB activation (33). In our research, we found that G− bacterium had much higher abundance in ESCC tissues compared with adjacent tissues. LPS stimulation upregulated CCR1 and PD-L1 expression in ESCC. All above facts indicated that G− bacteria bulk deposition and LPS accumulation in TME may promote ESCC development through inducing immune evasion.

The inflammation immune microenvironment has been confirmed to be an important factor affecting immunotherapy, and effective regulation of the inflammation immune microenvironment has become an important means to improve the therapeutic effect of anti–PD-1/PD-L1 (34). In view of the immunosuppressive role of LPS–TLR4 regulatory signal axis in gastrointestinal tumors, we tried to explore whether LPS accumulation in esophageal cancer could effectively mediate the inflammation and immune microenvironment accompanied by esophageal flora imbalance and find out the potential molecular mechanism.

We conducted transcriptome sequencing and intersected the sequencing results with the sequencing data from The Cancer Genome Atlas (RRID: SCR_003193) and Gene Expression Omnibus (RRID: SCR_005012) databases and the highly correlated genes of TLR4. CCR1 was identified as the target. In our study, we observed that PD-L1 mRNA and protein abundantly expressed in KYSE150 and TE1 cells under LPS stimulation and was positively correlated to LPS concentration. In order to figure out whether CCR1 might be the regulator of PD-L1, we detected CCR1 and its ligands CCL3 and CCL5 mRNA and protein expression. As q-PCR and Western blot showed, CCR1, CCL3 and CCL5 expression were upregulated in LPS stimulation subsets which had the same trend with PD-L1. Based on our Kyoto Encyclopedia of Genes and Genomes analysis and multiple research studies, the MAPK pathway is a critical factor to CCR1–PD-L1 function. On one hand, MAPK pathways were one of the main drivers of PD-L1 expression (35). On the other hand, chemokines on tumor cells could directly activate MAPK pathway to promote tumorigenesis (36, 37). However, there remained unclear whether CCR1 may activate the MAPK pathway. In an effort to reveal the transcriptional regulatory mechanism of LPS-induced CCR1–PD-L1 expression, we perform Western blot to measure MAPK phosphorylation level, and the results indicated that there was high MAPK phosphorylation in LPS stimulation subsets where PD-L1 and CCR1 were also significantly upregulated. Flow cytometry indicated that PD-1 expression on T cells cocultured with ESCC cells significantly increased, which suggested that LPS stimulation could promote immune evasion through PD-L1/PD-1 combination. Our data clearly indicated that LPS-induced CCR1 overexpression could induce MAPK phosphorylation to regulate PD-L1 high expression to intensify immune evasion which promoted ESCC cells’ migration, invasion, and proliferation.

Chemokines influence cancer development processes in many aspects. They can affect angiogenesis, metastasis, cancer cell proliferation, stemness, and invasiveness of cancer cells. Therefore, they play key roles in patient prognosis and response to cancer therapy (38). Lots of research studies have proved that CCR1 is highly overexpressed in multiple cancers including prostate cancer cells, ovarian cancer, and hematolymphoid neoplasia (39–41). Multiple preclinical studies demonstrated that blocking CCR1 has significant effects for inhibiting cancers such as colon cancer and breast cancer. Research proved that CCR1 antagonists associated with PD-L1 inhibitor had great effect on breast cancer (42, 43). CCR1 expressed by cancer cells could interact with tumor-associated cells to activate PI3K/AKT and ERK 1/2, which can promote cancer cell proliferation (44). Research proved that CCR1 is aberrantly overexpressed in human hepatocellular carcinoma tissues (45). Another finding also suggested that CCR1 activation may promote hepatocellular carcinoma progression (46). Besides the effect on tumor cells, CC chemokine family also induce tumor immune invasion through immunosuppressive cell infiltration and metastasis (47, 48). However, it remains unclear of the role of CCR1 on immune check points like PD-L1. CCR1 inhibiting therapeutic has drawn lots of attention recently. CCR1 inhibition can reduce myeloid-derived suppressor cell infiltration, which had the ability to significantly inhibit immune cell response and negatively regulate immune response. In murine models of multiple myeloma bone disease, CCX721, a classical CCR1 antagonist, was proved to decrease tumor burden and osteolytic lesion by osteoclasts inhibiting (49). In breast cancer, anti–PD-L1 antibody combined with CCX9588 was proven to be a promising therapeutic approach in an orthotopic breast cancer mouse model (43). In colon cancer, the use of BL5923 suppressed the recruitment of immature myeloid cells and reduced metastatic colonization, which prolonged OS time of mice with hepatic metastasis of colon cancer (42). In the current study, BX471, a typical CCR1 antagonist, was used to inhibit CCR1 expression in ESCC cell lines. CCR1 mRNA and protein expressions were significantly decreased after BX471 stimulation. Meanwhile, CCR1 silencing inhibited MPAK phosphorylation and PD-L1 expression in in KYSE150 and TE1 cells. Flow cytometry proved the results too. Our data clearly indicated that BX471-induced CCR1 inhibition could suppress MAPK phosphorylation to regulate PD-L1 low expression and ESCC cells’ migration, invasion, and proliferation.

Another key finding of our research is that, besides promoting PD-L1/PD-1 combination, LPS stimulation induced immune evasion through promoting T-cell exhaustion and suppressing T-cell cytotoxicity and proliferation. Research believed that T-cell membrane and intranuclear molecules can also be used by tumors which make them resistant to T-cell attack (50). In T-cell cytotoxicity aspect, LAG3 expression, also called CD107a, is elevated in tumor-infiltrating lymphocytes in colon cancers and ovarian cancers (51, 52). Persistent antigen stimulation in cancer or chronic infection leads to chronic LAG3 high expression to promote T-cell exhaustion expression (53). Exocytosecretion of dissolved particles enables cytotoxic T lymphocytes and NK cells to participate in cytotoxic activities via perforin and granase, which are stored in lipid bilayer vesicles containing CD107a. During exocytosis, this vesicle fuses with the plasma membrane, thereby mobilizing CD107a to the cell surface to perform cytotoxicity (54). In our current research, LAG3 expression elevated, whereas CD107a expression descended in LPS-induced T cells cocultured with ESCC cells. The results clearly indicated that LPS stimulation promoted T-cell exhaustion and suppressed T-cell cytotoxicity. CCR1 inhibition reversed the trends of LAG3 and CD107a. CCR1 inhibition avoided T-cell exhaustion and elevated T-cell cytotoxicity.

There were some limitations in our current research. First, we used Jurkat cells to perform coculture assays. Jurkat cells expressed stable CD3-positive markers. We did not detect CD4/CD8 markers on it. It remained uncertain how CD4/CD8-positive T cells might affect LPS-induced tumorigenesis in ESCC cells. Second, inhibitory marker of T cells, besides PD-1 and LAG3, had multiple other molecules including TIM3, TIGIT, and CTLA-4. It was better to detect these markers to present a more integrity process of T-cell exhaustion. Finally, we did not detect more detailed effects of microbial metabolites on immune cell functional pathways and mechanisms of how LPS induced CCR1 upregulation. We did not perform these assays because of experimental condition limits.

Our study showed that there was a high G− bacterium abundance in ESCC. G− bacterium deposition may induce immune escape in ESCC. Accumulation of LPS had a promotion on the ESCC cell lines KYSE150 and TE1 malignant biological behavior. As a CC chemokine receptor, CCR1 directly regulated the expression of PD-L1. The promoting effect of LPS on ESCC might be mediated by CCL3/CCL5–CCR1–MAPK–PD-L1 signaling. CCR1 overexpression suppresses T-cell cytotoxicity and proliferation, whereas it promoted T-cell exhaustion.

No disclosures were reported.

H. Yang: Conceptualization, resources, data curation, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. J. Cai: Conceptualization, software, investigation, methodology. X. Huang: Data curation, software, methodology. C. Zhan: Conceptualization, supervision. C. Lu: Validation, methodology. J. Gu: Data curation, software. T. Ma: Data curation, software. H. Zhang: Data curation, software. T. Cheng: Conceptualization, data curation, software, methodology. F. Xu: Conceptualization, supervision, funding acquisition, writing–review and editing. D. Ge: Conceptualization, resources, supervision, funding acquisition.

The work was supported by the National Natural Science Foundation 81872291 and the National Natural Science Foundation 82003280.