Neoadjuvant chemoimmunotherapy (NACI) has shown promise in the treatment of resectable esophageal squamous cell carcinoma (ESCC). The microbiomes of patients can impact therapy response, and previous studies have demonstrated that intestinal microbiota influences cancer immunotherapy by activating gut immunity. Here, we investigated the effects of intratumoral microbiota on the response of patients with ESCC to NACI. Intratumoral microbiota signatures of β-diversity were disparate and predicted the treatment efficiency of NACI. The enrichment of Streptococcus positively correlated with GrzB+ and CD8+ T-cell infiltration in tumor tissues. The abundance of Streptococcus could predict prolonged disease-free survival in ESCC. Single-cell RNA sequencing demonstrated that responders displayed a higher proportion of CD8+ effector memory T cells but a lower proportion of CD4+ regulatory T cells. Mice that underwent fecal microbial transplantation or intestinal colonization with Streptococcus from responders showed enrichment of Streptococcus in tumor tissues, elevated tumor-infiltrating CD8+ T cells, and a favorable response to anti-PD-1 treatment. Collectively, this study suggests that intratumoral Streptococcus signatures could predict NACI response and sheds light on the potential clinical utility of intratumoral microbiota for cancer immunotherapy.
Analysis of intratumoral microbiota in patients with esophageal cancer identifies a microbiota signature that is associated with chemoimmunotherapy response and reveals that Streptococcus induces a favorable response by stimulating CD8+ T-cell infiltration.
Esophageal squamous cell carcinoma (ESCC) accounts for approximately 90% of all esophageal cancer cases and is the sixth most common cause of cancer-related deaths worldwide (1, 2). ESCC remains a global challenge that requires a multidisciplinary approach with extensive treatment, including surgery, chemoradiotherapy, and chemotherapy (3, 4). ESCC has a relatively high tumor mutational burden (5, 6), suggesting that it could benefit from programmed cell death protein 1 (PD-1) blockade (7–9). Three phases III clinical trials recently reported that PD-1 blockade significantly improves progression-free survival (PFS) and overall survival (OS) when combined with chemotherapy as first-line therapy in advanced/metastatic ESCC (10–12). Moreover, as reported at the 2021 American Society of Clinical Oncology annual meeting, neoadjuvant chemoimmunotherapy (NACI) induces a noticeable improvement in the pathologic complete response for resectable ESCC (35. 3% and 42.5%, respectively). Several ongoing phase III clinical trials, such as NCT04807673 (13) and NCT04280822 (14), indicate that NACI could become a promising treatment for locally advanced ESCC. However, not all patients with ESCC respond to NACI (15, 16). This raises an important question regarding the predictive markers and potential mechanisms to determine the NACI response in patients with ESCC.
There is increasingly compelling that polymorphic variability in the microbiomes between individuals profoundly impacts the cancer phenotypes and therapy response (17–19). Most studies have focused on how the intestinal microbiota influences cancer immunotherapy by activating gut immunity (20–23). However, the role of intratumoral microbiota remains to be further explored (24). Riquelme and colleagues observed markedly different intratumoral microbiome signatures in patients of pancreatic ductal adenocarcinoma (PDAC) with short- and long-term survival, indicating its prognostic value (25). After performing a gradual and in-depth exploration of the mechanism, a recent study demonstrated that live bacteria that exist intracellularly in breast cancer cells could modulate the host-cell actin network and promote cell survival against fluid shear stress in circulation (26). These results suggest that the intratumoral microbiome may affect host cellular homeostasis, tumor biology, immune system activation, and treatment efficiency (27–30). The intratumoral microbiome has been detected within various tumor types and is inferred to be associated with the clinicopathologic features of patients (25, 31–33). Therefore, it is essential to investigate its predictive or prognostic significance in cancer immunotherapy and the specific underlying mechanism in patients with ESCC (32, 34, 35).
In this study, we demonstrated that the intratumoral microbiota signatures vary between NACI responders and nonresponders among patients with ESCC. The responders contained an abundance of tumor-resident Streptococcus compared with nonresponders. Streptococcus positively correlated with the infiltration of intratumoral GrzB+ T cells and CD8+ T cells. Streptococcus abundance predicted significantly prolonged disease-free survival (DFS) in patients with ESCC under NACI and surgical treatment. Moreover, fecal microbial transplantation (FMT) from NACI responders remodeled the composition of intratumoral microbiota that potentiated the immunotherapy response.
Materials and Methods
Ethics approval and consent to participate
Animal experiments were approved by the Institutional Animal Care and Use Committee of the Sichuan Cancer Hospital & Institute, Sichuan Cancer Center, School of Medicine, University of Electronic Science and Technology of China (Sichuan, P.R. China; no. SCCHEC-04-2022-004) and the Institutional Animal Care and Use Committee of the Tongji Hospital, Huazhong University of Science and Technology (no. 106-19-08-OH).
Human samples collection
Human tissues were collected from Sichuan Academy of Medical Sciences, Sichuan Provincial People's Hospital, University of Electronic Science and Technology of China, and Sichuan Cancer Hospital & Institute, Sichuan Cancer Center, School of Medicine, University of Electronic Science and Technology of China (Sichuan, P.R. China). We obtained written informed consent from all the patients, that the studies were conducted in accordance with recognized ethical guidelines of Sichuan Academy of Medical Sciences, Sichuan Provincial People's Hospital, University of Electronic Science and Technology of China, and Sichuan Cancer Hospital & Institute, Sichuan Cancer Center, School of Medicine, University of Electronic Science and Technology of China. Fresh tissues were immediately transferred to germ-free 15 mL conical tubes with a sterile DMEM culture medium. Samples were then processed in clean and sterile tubes with autoclaved dissection tools. The characteristics of all patients were provided in Supplementary Tables S1 and S2. The NACI response or no response was calculated according to the tumor regression grade score based on the American Society of Pathologists (CAP) and American Joint Committee on Cancer. Score 0: Complete remission, no residual tumor. Score 1: Nearly complete remission, single cells or rare small groups of cancer cells. Score 2: Partial remission, more than single cells or rare small groups of cancer cells with evident tumor regression. Score 3: Poor or no remission, extensive residual tumor or no regression. Score 0–2 is defined as responders, and Score 3 is defined as nonresponders. Supplementary Materials and Methods are provided in the Supplementary Data. These human samples were used to extract DNA for bacterial quantification and 16S library preparation.
Cell lines and cell culture
Mouse esophageal cancer cell line AKR cells (passages 8–10) were purchased from the ATCC. Mouse esophageal cancer cell line mEC25 (passages 10–12) was donated by Prof. Fu from the International Cancer Center of Shenzhen University (Shenzhen, P.R. China; PMID: 32510874; ref. 36). Cells were cultured in DMEM/H (HyClone, catalog no. SH30243.02) containing 10% FBS and 1% penicillin/streptomycin (Thermo Fisher Scientific, catalog no. 10378016) and incubated at 37°C in a humidified atmosphere with 5% CO2. All cell lines tested negative for Mycoplasma bacteria (December 9, 2022) as assessed by a Mycoplasma PCR Detection Kit (Beyotime, #C0301S).
Human tissue samples
Tissues were collected from patients with ESCC who underwent scheduled surgery. An initial cohort of 40 patients who received NACI (n = 25) or upfront surgery (n = 15) was included in this study. Each tissue was dissected into several pieces according to the tumor size under sterile conditions. One part of the tissue was fixed in formalin and embedded in paraffin, and the other was prepared for flow cytometry and bacterial culture. Tissues were fresh-frozen in liquid nitrogen or stored at −80°C before use. Patient information is provided in Supplementary Tables S1 and S2.
Fecal samples collection
Fecal samples were collected from patients with ESCC and healthy control donors. Feces were placed in sterile centrifugal tubes and immediately frozen at −80°C. Detailed information on feces providers for FMT is provided in Supplementary Table S3.
Multiplex immunofluorescence staining
Multiplex immunofluorescence (IF) staining for anti-CD163 (1:150, Cell Signaling Technology, catalog no. 93498S), anti-CD20 (1:200, Cell Signaling Technology, catalog no. 48750S), anti-CD4 (1:200, Cell Signaling Technology, catalog no. 25229S), anti-CD66b (1:200, Abcam, catalog no. Ab214175), anti-CD8α (1:200, Cell Signaling Technology, catalog no. 85336S), and anti-FoxP3 (1:150, Cell Signaling Technology, catalog no. 1f2653S) was performed using 5 μmol/L sections of formalin-fixed paraffin-embedded tumor samples by sequential staining after antigen retrieval in cell conditioning solution (pH 8.5) in a water bath at 98°C for 30 minutes. The Opal Polymer HRP Ms+Rb was used for the primary antibody detection and Opal 7-Color Manual IHC (Akoya Biosciences, catalog no. NEL811001KT), with six reactive fluorophores Opal 690, Opal 520, Opal 620, Opal 650, Opal570, Opal 540 plus 4′,6-diamidino-2-phenylindole nuclear counterstain, were added according to the manufacturer's instructions. The slides were imaged using Vectra 3.0 spectral imaging system (PerkinElmer) according to previously published instructions. The information on antibodies is provided in Supplementary Table S4.
For qRT-PCR quantification, briefly, 10 μL reaction mix containing TB Green Premix Ex Taq II (2X; Takara, catalog no. RR820Q), 400 nmol/L of forward primer, 400 nmol/L of reverse primer, and 1 μL sample DNA was loaded on the CFX96 Real-Time PCR Detection System (Bio-Rad). The qRT-PCR reaction was programmed as follows: denaturation at 95°C for 2 minutes, followed by 40 cycles of 95°C for 5 seconds, 55°C for 30 seconds, and 72°C for 1 minute and completed with a dissociation curve. Bacteria load was estimated by comparing threshold cycles (Ct) values to a bacteria standard curve produced with Escherichia coli DNA. All samples conducted at least three biological replicates. The primers are provided in Supplementary Table S5.
Tumor-resident bacterial culture and identification
To isolate anaerobic or aerobic bacteria, ESCC tissue samples (around 0.1 g) from responders and nonresponders were minced into small pieces and then homogenized with a glass homogenizer in 1 mL ice-cold DMEM High Glucose (DMEM/H; HyClone) under sterile conditions. Another tube of DMEM/H that underwent the same workflow of tissue samples, including exposure to the surgery room environment, laboratory environment, and tumor tissue shredding process, was used as a control for bacterial culture. For aerobic culture, 100 μL sample homogenate was plated on Columbia blood agar (CBA; JX601) + 5% sheep blood (Solarbio, catalog no. TX0030) aerobically with 5% CO2. For anaerobic culture, 100 μL sample homogenate was plated on CBA (JX601) + 5% sheep blood (Solarbio, catalog no. TX0030) and placed in an anaerobic bag (AnaeroPouch). The plates were incubated at 37°C for 3 days in aerobic or anaerobic conditions. Bacterial colonies were collected on day 3. The DMEM/H control groups were plated and set simultaneously. The collected colonies were prepared for 16S rRNA sequencing and transmission electron microscopy (TEM) according to the detailed methods provided in the Supplementary Data.
Antibiotics treatment and FMT
C57BL/6 mice, ages 6 weeks, were purchased from Beijing HFK Bioscience CO., LTD. An antibiotic cocktail was consecutively orally administered to the mice for 4 weeks to ablate the gut microbiome. The antibiotics (ATBx) were dissolved in a sterile medium and included vancomycin (0.5 mg/mL), streptomycin (5 mg/mL), and lincomycin (3 mg/mL). After 4 weeks, the mice were recolonized by FMT, receiving PBS-diluted fecal samples every 4 days by oral gavage at three dosages.
Isolation and identification of intratumoral Streptococcus
For the isolation and identification of intratumoral Streptococcus in the NACI responders, colonies were obtained and streaked in CBA (JX601) + 5% sheep blood (Solarbio, catalog no. TX0030) aerobically with 5% CO2, and conditioned for 1–3 days to form single colony. Colonies were selected to run colony PCR separately; primers (27F:5′-GAGAGTTTGATCCTGGCTCAG-3′; 1492R: 5′-TACGGCTACCTTGTTACGAC-3′) were used. The reaction was programmed as follows: 98°C for 3 minutes, 39 cycles of 98°C for 10 seconds, 55°C for 15 seconds, and 72°C for 30 seconds. Final extension reactions were carried out for 5 minutes at 72°C. The PCR product was sent for sequencing, and results were aligned to the 16S rRNA sequences (Bacteria and Archaea) database on the NCBI BLAST website.
Intestinal colonization with Streptococcus or E. coli and in vivo treatment efficiency of anti-PD-1
Streptococcus was isolated from the NACI responders and grown under aerobic conditions. E. coli (ATCCBAA-1429, 01261K) was purchased from the China Center of Industrial Culture Collection. ATBx-treated mice were given oral gavage of PBS-diluted Streptococcus [1 × 108 colony-forming unit (CFU) in a total of 200 μL], PBS-diluted E. coli (1 × 108 CFU in a total of 200 μL) or PBS (200 μL) every 2 days for six dosages. The colonization of Streptococcus or E. coli was confirmed by culturing the feces on CBA (JX601) + 5% sheep blood (Solarbio, catalog no. TX0030) aerobically with 5% CO2 and real-time PCR using primers specific for Streptococcus or E. coli. The mice were randomly assigned to indicated groups. Furthermore, mice in the CD8α+ T-cell exhaustion group were treated with an anti-mouse CD8α antibody (200 μg, BE0004-1, Bio-X Cell) according to the manufacturer's instructions. All the mice were implanted subcutaneously with mEC25 tumor cells (5 × 106 cells) at their flanks. An indicated group of mice was fed with the second round of an antibiotic cocktail containing vancomycin (0.5 mg/mL), tetracycline (0.2 mg/mL), and azithromycin (0.5 mg/mL) to ablate the intestinal and intratumoral Streptococcus for 10 days. Other mice were fed sterile water without ATBx for the same period. Mice were administrated anti-PD-1 (250 μg per mouse) for a total of three times by intraperitoneal injection. Tumor growth was monitored every 3 days following tumor implantation.
FMT and tumor volume paired subcutaneous tumor implantation
To evaluate the influence of tumor volume on the bacterial content, FMT-administered C57BL/6 mice were subcutaneously implanted with AKR cells (5 × 106) at both flanks. We paired the tumor volume between FMT from the responders and nonresponders and sacrificed the mice at different time points. Tumors were collected for detecting the hypoxia burden and bacterial content.
Analysis of 16S rRNA sequencing data was performed on https://cloud.majorbio.com/. Samples with abundances of amplicon sequence variant (ASV) ≥ 1% were considered for analysis and those with < 1% were defined in the “others” category. The analysis of α-diversity, including richness estimators and diversity estimators, was performed using the Wilcoxon signed-rank test. The β-diversity principal coordinate analysis (PCoA) plots were analyzed between adjacent normal and tumor tissues, upfront surgery group and NACI group, and responders and nonresponders using weighted UniFrac. The different microbiota compositions between different samples were computed using the Wilcoxon signed-rank test. Continuous variables were calculated using the t test in GraphPad Prism (v7.0; GraphPad). A comparison of more than two groups was performed with one-way ANOVA (v7.0; GraphPad). Kaplan–Meier curves were estimated for survival distributions. The log-rank test was used to test the difference in survival distributions between the subgroups. HRs and 95% confidence intervals (CI) were calculated. P ≤ 0.05 was considered statistically significant. A detailed description of the other methods used in this study can be found in the Supplementary Data.
All the data relevant to the study are included in the article or uploaded as Supplementary Data, and are available upon request from the corresponding author. The raw data for single-cell RNA sequencing (scRNA-seq) reported in this publication can be accessed under the Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi, accession number GSE203115, Gene Expression Omnibus, RRID: SCR_005012). The 16S rRNA sequencing data performed in this study can be accessed under the Sequence Read Archive database (https://www.ncbi.nlm.nih.gov/sra; accession number PRJNA839584).
Differences in intratumoral microbiota between NACI responders and nonresponders
To explore the role of intratumoral microbiota composition in mediating the treatment response to NACI in patients with ESCC, we collected tumor tissues, adjacent normal tissues, and feces from 25 patients with locally advanced ESCC who received NACI and 15 who were treated with upfront surgery (Fig. 1A). We conducted TEM, RNA FISH, IHC staining, and qRT-PCR to validate the presence of bacteria in the tumor and adjacent normal tissues. High-resolution electron microscopy analysis demonstrated that bacteria-like structures primarily existed in the cytosol rather than in the extracellular space (Fig 1B; Supplementary Fig. S1A). Positive and negative control of FISH probes were shown in Supplementary Fig. S1B. 16S RNA fish probe, lipopolysaccharides (LPS), and lipoteichoic acid (LTA) costaining demonstrated the presence of bacteria in tumor tissues and adjacent normal tissues (Fig. 1C; Supplementary Fig. S1C). No significant differences in LPS and LTA expression were observed between NACI responders and nonresponders or between tumor and adjacent normal tissues (Fig. 1D; Supplementary Fig. S1D and S1E). We subsequently introduced the environmental background control group (EBC) and the non-template control (NTC) group to demonstrate the existence of intratumoral bacteria by qRT-PCR. The results demonstrated that the tumor and adjacent normal tissues had significantly higher bacterial content than the EBC and NTC groups (Supplementary Fig. S1F). The bacteria content between the tumor and adjacent normal tissues or between NACI responders and nonresponders had no significant difference, which was consistent with the IHC results (Supplementary Fig. S1F and S1G). We assessed the general microbial landscape of all the tissues and the EBC control samples using 16S rRNA sequencing (Fig. 1E; Supplementary Fig. S1H). The majority of the microbes in the EBC control samples were Proteobacteria, whereas the tissue samples were enriched for Firmicutes (Supplementary Fig. S1I). The apparent clustering of β-diversity in the PCoA analysis demonstrated that the control samples had significantly distinct microbial communities compared with the tumor and adjacent normal tissues (Supplementary Fig. S1J). Further analysis using different methodologies revealed that the α- and β-diversity were similar between the tumor and adjacent normal tissues and between the upfront surgery and NACI treatment groups (Supplementary Fig. S2A–S2L).
In addition to similar α-diversity, an apparent clustering of β-diversity in the PCoA between the NACI responders and nonresponders was generated in genus and ASVs (Fig. 1F; Supplementary Fig. S3A–S3F). To further investigate these findings, we conducted high-dimensional class comparisons using linear discriminant analysis of effect size (LEfSe), which detected marked differences in the predominance of bacterial communities between the NACI responders and nonresponders (Fig. 1G; Supplementary Fig. S3G): The tumors of the NACI responders exhibited a predominance of Firmicutes and Bacteroidota at the phylum level, Negativicutes and Bacteroidia at the class level, and Veillonellales-Selenomonadales and Actinomycetales at the order level. The major genera with distinct features between the NACI responders and nonresponders were shown in Fig. 1H. Collectively, the results demonstrated the presence of intratumoral microbiota, which has a significantly different signature in the NACI responders and nonresponders among patients with ESCC.
Association between tumor-resident Streptococcus, NACI response, and DFS in patients with ESCC
We comprehensively profiled the distinct tumor-resident microbiota between the NACI responders and nonresponders at the genus level. The tumor tissues of the responders exhibited a predominance of Streptococcus, Actinomyces, Granulicatella, and Abiotrophia. Nonresponders demonstrated a greater abundance of Tardiphaga and Pajaroellobacter (Fig. 2A). Streptococcus had the most significant change upon the comparison between the responders and nonresponders (Fig. 2A). We further verified the abundance of Streptococcus in the tumor tissues of responders and nonresponders using qRT-PCR, Streptococcus probe, and Streptococcus antibody. The results demonstrated that the NACI responders had a significantly higher Streptococcus abundance than the nonresponders (Fig. 2B–D; Supplementary Fig. S4A–S4C). We cultured the intratumoral bacteria by plating the dissociated tumor cells in CBA and found that live bacteria grew (Fig. 2E; Supplementary Fig. S4D). Live Streptococcus was observed in the culturable bacteria of ESCC tissues by scanning electron microscopy (Fig. 2F).
We further performed 16S rRNA gene sequencing of the culturable bacteria. The results revealed that the major genera detected in the culturable bacteria were highly consistent with those in the tissues, and Streptococcus was more abundant in the responders than that in the nonresponders (Fig. 2G). We used differentially expressed Streptococcus and several other bacteria, including Prevotella, Bifidobacterium, Moraxella, and Dialister, to conduct an AUC-ROC analysis. The results revealed that the AUC value of Streptococcus was 0.84 (95% CI: 0.6985–0.9874), which was higher than the others, indicating an excellent discrimination power for NACI responders among patients with ESCC (Fig. 2H; Supplementary Fig. S4E). Furthermore, the abundance of Streptococcus predicted prolonged DFS compared with those with lower Streptococcus (P = 0.0341; Fig. 2I). Other bacteria, including Prevotella, Bifidobacterium, Moraxella, and Dialister, demonstrated no significant roles in the prediction of DFS (Supplementary Fig. S4F–S4I). These data suggested that the increased presence of Streptococcus might enhance the treatment efficiency of NACI in patients with ESCC.
Association between tumor-resident Streptococcus and the infiltration of specific immune cells
Previous studies have revealed that the gut microbiota composition could shape the tumor immune microenvironment (TIME) by modulating the specific infiltration of immune cells (27, 31, 37, 38). We speculated whether the tumor-resident microbiota has the same function. Multiplex IF staining was used to characterize the intratumoral infiltration of immune cells. A significantly higher number of CD8+ T cells, CD20+ B cells, and CD66b+ myeloid cells were found in the TIME of NACI responders (Fig. 3A; Supplementary Fig. S5A). We used IHC to delineate the distinct infiltration of immune cells between NACI responders and nonresponders. The results demonstrated that the density of tumor-infiltrating GrzB+ and CD8+ T cells was higher in the responders. The nonresponders had greater infiltration of FOXP3+ or CD4+ T cells in both pre-NACI and post-NACI tissue slices (Fig. 3B and C). Moreover, our results demonstrated that the responders had higher expression of cytokines related to CD8+ T-cell trafficking, including CCL3, CCL4, CCL5, CCL20, CXCL10, and CXCL11 (Fig. 3D).
We evaluated the correlation between tumor-resident Streptococcus abundance and immune cells infiltration in patients with ESCC. Intriguingly, we observed a positive correlation between the tissue densities of GrzB+ and CD8+ T cells with Streptococcus and a negative correlation between FOXP3+ and CD4+ T cells with Streptococcus in both pre-NACI and post-NACI tissues (Fig. 3E and F), which was further confirmed by flow cytometry (Fig. 3G). Our results also demonstrated a positive correlation between Granulicatella and Abiotrophia abundance and infiltration of GrzB+ and CD8+ T cells in post-NACI patients (Supplementary Fig. S5B). We also confirmed that a higher abundance of tumor-resident Streptococcus was commonly associated with more infiltration of CD8+ T cells (Fig. 3H and I). Taken together, these findings demonstrated that Streptococcus abundance in ESCC tissues typically indicated an activated TIME that could enhance the therapeutic efficiency of NACI.
scRNA-seq reveals the tissue-resident immune cell landscape of NACI responders and nonresponders
To better understand the heterogeneity within and across NACI responders and nonresponders, we conducted scRNA-seq on 13,280 single cells from three patients with ESCC, including two responders and one nonresponder. Detailed clinical and pathologic information about the patients is provided in Supplementary Table S2. Each cell cluster [cancer-associated fibroblasts, epithelial cells including malignant and nonmalignant cells, immune cells including myeloid, natural killer (NK), and T and B cells] was annotated by the average expression of the curated gene sets. This approach revealed a complex TIME containing nine cell subclusters (Fig. 4A; Supplementary Fig. S6A). We subsequently identified signature genes for each immune subcluster and successfully developed ten T/NK subtype-specific signatures (Supplementary Fig. S6B). The mean proportion of CD4+ regulatory T cells (Treg)/CD4+ T cells (46.03% vs. 83.08%), central memory CD4+ T/ CD4+ T cells (53.97% vs. 16.92%), CD8+ exhausted T cells/CD8+ T cells (43.06% vs. 71.16%), and effector memory CD8+ T cells/CD8+ T cells (56.94% vs. 28.84%) between responders and nonresponders are shown in Fig. 4B. The cell clusters demonstrated varying enrichment in the NACI responders and nonresponders (Fig. 4C and D). The tumor tissue of nonresponders was infiltrated by a decreased proportion of effector T cells (expressing cytotoxicity-related markers, such as GZMK). An increased ratio of exhausted T cells (expressing exhaustion-related markers, e.g., FOXP3 and CTLA4) indicated an immunosuppressive microenvironment within NACI nonresponders (Fig. 4E; Supplementary Fig. S6C).
Our study further demonstrated that the nonresponders had more infiltration of ICOS+FOXP3+ Treg and CXCL13+CD8+ exhausted or cycling T cells (Fig. 4F; Supplementary Fig. S6D). CXCL13+CD8+ T cells express higher immune checkpoint levels, including HAVCR2, CTLA4, PRDM1, and Ki-67 (39). Upon monocle trajectory analysis of the CD8+ T cells, the differentiation trajectory exhibited a branched structure, starting with effector memory CD8+ T cells that bifurcated into either exhausted T or cycling T cells (Fig. 4G; Supplementary Fig. S6E). Conventional analysis of CD4+ T cells also inferred a differentiation trajectory primarily organized into two main branches and self-assorted according to the NACI response (Fig. 4H). Collectively, scRNA-seq data revealed a comprehensive differential and tissue-specific tumor-resident immune cell landscape between the NACI responders and nonresponders.
Streptococcus orchestrates the infiltration of intratumoral immune cells to enhance antitumor immunity
Gut microbiota can modulate the intratumoral microbial landscape, as demonstrated by direct translocation into tumor tissues in pancreatic cancer (25, 40). Upon comparing the taxonomic composition of different samples, we determined that the human gut microbiota represented approximately 5% of the human ESCC microbiota, which was also present in normal adjacent tissues (Fig. 5A). Therefore, we performed FMT from patients with ESCC and healthy control donors to mice that had been previously treated with ATBx (Fig. 5B). FMT increased the stool bacterial content that was eliminated by ATBx treatment (Fig. 5C). Tumor growth decreased in mice receiving FMT from responders compared with that in the nonresponders or healthy control donors under anti-PD-1 treatment (Fig. 5D and E). Bacterial origin investigation revealed that many bacteria of human stool origin were observed in the murine gut microbiota in the recipient mice post-FMT. Intriguingly, human donor bacteria were present in the murine AKR tumors post-FMT; however, they were absent from the mice stool and tumors before FMT (Fig. 5F). We further investigated whether FMT could influence immune system activation in vivo. Polychromatic flow cytometry analysis revealed that mice receiving FMT from responders had a significantly higher percentage of CD8+ T and effector T cells (CD44+CD62L−CD8+ T cells) as compared with stools transferred from nonresponders or healthy control donors (Fig. 5G and H). IHC staining further confirmed that infiltration of CD8+ and GrzB+ T cells was upregulated in mice that received FMT from responders compared with nonresponders (Fig. 5I). We further analyzed the blood concentrations of cytokines, including IFNγ and IL1β, which were critical for innate and adaptive immunity against viral, bacterial, and protozoan infections (41). The data revealed that the concentrations of IFNγ and IL1β were higher in mice that received FMT from responders (Supplementary Fig. S7A). We used the murine ESCC cell line mEC25 to confirm the in vivo results, which further established that FMT from NACI responders could remodel the intratumoral immune microenvironment by increasing the infiltration of cytotoxic T cells and reducing the infiltration of FOXP3+CD4+ T cells and decrease tumor growth under anti-PD-1 treatment (Supplementary Fig. S7B–S7F). 16S rRNA sequencing revealed that Streptococcus was enriched in the tumor tissues of mice that received FMT from responders compared with the healthy control donors and nonresponders (Fig. 5J). Streptococcus abundance positively correlated with GrzB+ and CD8+ expression in tumor xenograft-bearing mice (Fig. 5K).
We further conducted animal experiments to determine the presence of a causal relationship between Streptococcus and immune cells’ infiltration and immunotherapy response (Fig. 6A). The constructed ATBx mice were intestinally cloned using Streptococcus isolated from the tumor tissues of NACI responders (Supplementary Fig. S8A and S8B). After the successful intestinal cloning of Streptococcus, mice were divided into different groups and administered the indicated treatments. Furthermore, they were inoculated with subcutaneous xenografts followed by anti-PD-1 immunotherapy. The results demonstrated that the intestinal cloning and intratumoral abundance of Streptococcus significantly increased the infiltration of intratumoral CD8+ and GrzB+ T cells and enhanced the response to anti-PD-1 immunotherapy. However, depletion of Streptococcus using ATBx decreased the intratumoral Streptococcus abundance, reduced infiltration of the CD8+ T cells, and decreased the response to anti-PD-1 immunotherapy. Moreover, when the CD8+ T cells were exhausted by the anti-CD8α antibodies, the increased intestinal Streptococcus clones and intratumoral Streptococcus abundance did not enhance the efficacy of anti-PD-1 immunotherapy (Fig. 6B–F; Supplementary Fig. S8C–S8F). We further intestinally cloned the ATBx mice by Streptococcus and E. coli, separately (Fig. 6G; Supplementary Fig. S8G–S8I). Results showed that compared with E. coli, Streptococcus could enhance the infiltration of CD8+ T cells and the efficiency of anti-PD-1 treatment (Fig. 6H–K). These data indicate a causal relationship between intratumoral Streptococcus abundance and CD8+ T-cell infiltration. Thus, Streptococcus can promote the infiltration of CD8+ T cells and enhance the efficacy of anti-PD-1 immunotherapy.
To determine whether the hypoxic burden can affect bacterial content, especially intratumoral Streptococcus abundance, we established xenograft mouse models that received FMT from responders and nonresponders. The mice were sacrificed at different time points to pair the tumor volume between the two groups (Supplementary Fig. S8J–S8L). The results demonstrated that the expression of HIF1α increased along with tumor growth, while the average intratumoral bacterial load remained constant (Supplementary Fig. S8M and S8N). Importantly, our results demonstrated that Streptococcus was enriched in the tumor tissues of the mice that received FMT from responders compared with nonresponders; however, there was no growth advantage concomitant with tumor growth (Supplementary Fig. S8O). Collectively, these data suggest that gut microbiota may colonize specific intratumoral sites of ESCC, modify both the gut and intratumoral bacterial composition, remodel the TIME, and ultimately affect the response to cancer immunotherapy.
The intratumoral microbiota constitutes an essential component of the TIME, and its significant role in regulating immune activation, cancer-associated inflammation, and treatment response has received extensive attention (17, 18, 31). This study investigated the composition and diversity of intratumoral microbiota and its effect on NACI response in patients with locally advanced ESCC. We performed a comprehensive analysis to distinguish the microbiota signatures in tumor tissues and paired adjacent normal tissues from 40 patients with ESCC treated with NACI or upfront surgery. We detected a substantial abundance of microbiota in the intratumoral and adjacent normal tissues of patients with ESCC. No differences in either α- or β-diversity were found between the tumor and adjacent normal tissues. However, apparent clustering between the genus and ASVs from the NACI responders and nonresponders was generated by β-diversity. The NACI responders and nonresponders had a distinctive tumor microbiota signature with specific bacterial genera that could predict the DFS of patients with locally advanced ESCC who underwent NACI treatment.
Furthermore, we successfully isolated live bacteria that grew on CBA by plating dissociated ESCC tumor cells. This notion is supported by several recent studies on comprehensive intratumoral microbiota analysis, which report a close relationship between the microbiota and TIME (41–43). Nejman and colleagues found that the diversity and composition of the gut microbiome were different between responders and nonresponders to anti-PD-1 treatment in patients with melanoma (31). Germ-free mice receiving fecal transplants from responders demonstrated enhanced systemic and antitumor immunity (27). Riquelme and colleagues reported that PDAC tumor microbiome composition and diversity could affect immune cell infiltration, ultimately influencing patient survival (25). Our results demonstrated that the abundance of intratumoral Streptococcus could influence the infiltration of CD8+ T cells, ultimately affecting NACI treatment efficiency in locally advanced ESCC. The abundance of Streptococcus was a vital predictive factor for NACI response in patients with ESCC, which provided insights into the future clinical application of the intratumoral microbiota.
We performed scRNA-seq on available tumor tissues to unravel the infiltration of intratumoral immune cells and cross-talk between the intratumoral microbiota and tissue-resident immune cells. The infiltration of CXCL13+CD8+ exhausted T cells was more abundant in the NACI nonresponders with a higher level of immune checkpoints, such as HAVCR2 and CTLA4, PRDM1, and Ki-67. Despite the close relationship between immune cell infiltration, treatment response/prognosis, and intratumoral microbiota signatures, we did not detect a more specific or direct influence. The underlying mechanisms need to be further evaluated (44–46).
Our preclinical data demonstrated that modifying the intratumoral microbiota via FMT from the NACI responders induced an antitumor response and activated the immune system in tumor-bearing mice. We demonstrated the abundance of Streptococcus and infiltration of CD8+ T cells and observed the partial existence of Streptococcus in CD8+ T cells using the FISH probe. Our study suggests that FMT from NACI responders or intestinal cloning of Streptococcus could potentiate the anti-PD-1 treatment response in tumor-bearing mice by recruiting antitumor immune cell infiltration, supporting the cross-talk between the gut and intratumoral microbiota and their role in remodeling the TIME. Further studies should focus more on exploring the mechanism by which Streptococcus may infect CD8+ T cells and the potential factors playing critical roles in the evaluation of the cross-reactivity between the T cells recognizing tumor neoantigens and the tumor-resident microbiome. This may help understand the mechanisms by which bacteria exert immuno-activating effects and promote research into novel therapeutic strategies for ESCC (47, 48).
In summary, our study elucidates the presence of intratumoral microbiota associated with the presurgical chemoimmunotherapy response in patients with ESCC. The enrichment of Streptococcus signatures results in a favorable response to cancer immunotherapy. FMT may remodel the composition of tumor-resident microbiota that potentiates immunotherapy response, indicating potential clinical implications of intratumoral microbiota for cancer immunotherapy. Future investigations are warranted to elucidate the detailed mechanisms related to the tumor-resident microbiota and dynamic alterations.
No disclosures were reported.
H. Wu: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. X. Leng: Resources, supervision, investigation. Q. Liu: Data curation, formal analysis, investigation, visualization, methodology, writing–original draft. T. Mao: Data curation, formal analysis, investigation, methodology, writing–original draft. T. Jiang: Data curation. Y. Liu: Data curation, formal analysis, investigation. F. Li: Investigation, visualization. C. Cao: Data curation. J. Fan: Investigation, methodology. L. Chen: Data curation. Y. Chen: Data curation. Q. Yao: Data curation. S. Lu: Supervision. R. Liang: Data curation, investigation, methodology. L. Hu: Data curation. M. Liu: Data curation, methodology. Y. Wan: Investigation, visualization. Z. Li: Data curation, software. J. Peng: Data curation. Q. Luo: Data curation. H. Zhou: Data curation. J. Yin: Data curation. K. Xu: Data curation. M. Lan: Data curation. X. Peng: Data curation, investigation. H. Lan: Data curation. G. Li: Data curation, investigation. Y. Han: Data curation, supervision. X. Zhang: Supervision. Z.-X.J. Xiao: Supervision. J. Lang: Supervision. G. Wang: Data curation, supervision, validation, investigation. C. Xu: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, visualization, writing–original draft, project administration, writing–review and editing.
The authors want to thank all the members of Prof. Xu's lab who continued to provide kind help and suggestions. This work was supported by the National Natural Science Foundation of China (nos. 92259102 and 81873048).
The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).