You are here

Article Text

Article menu

Download PDFPDF

Basic and translational cancer immunology
Original research
Intratumoral microbiota predicts the response to neoadjuvant chemoimmunotherapy in triple-negative breast cancer
  1. Yilin Chen1,2,
  2. Lu Yang3,
  3. Yuhong Huang1,
  4. Teng Zhu1,
  5. Liulu Zhang1,
  6. Minyi Cheng1,
  7. Cangui Wu1,
  8. Peiyong Li1,4,
  9. Minting Liang1,5,
  10. Xiaoqi Zhang1,
  11. Hao Peng1 and
  12. Kun Wang1,2
  1. 1 Department of Breast Cancer, Guangdong Provincial People's Hospital, Guangdong Academy of Medical Sciences, Southern Medical University, 106 Zhongshan Er Road, Yuexiu District, Guangzhou 510080, People's Republic of China
  2. 2 School of Medicine, South China University of Technology, Guangzhou, Guangdong, China
  3. 3 Department of Radiation Oncology, Guangdong Provincial People’s Hospital, Guangzhou, Guangdong, China
  4. 4 Guangdong Medical University, Zhanjiang, China
  5. 5 Department of Breast Cancer, Shantou University, Shantou, Guangdong, China
  1. Correspondence to Dr Kun Wang; gzwangkun{at}126.com; Dr Hao Peng; phao{at}gdph.org.cn

Abstract

Background Neoadjuvant immunotherapy combined with chemotherapy (Chemo-IM) is associated with significantly improved pathological complete response (pCR) rates and long-term survival outcomes in patient with early-stage triple-negative breast cancer (TNBC). However, only a small proportion of patients benefit from the addition of immunotherapy. Here, we explored and confirmed the role of intratumoral microbiota in screening patients with TNBC who are likely to benefit from neoadjuvant Chemo-IM.

Methods Patients with previously untreated, non-metastatic TNBC receiving neoadjuvant Chemo-IM were enrolled. Differences in the intratumoral microbiota between the pCR and non-pCR groups were explored via 16S rDNA sequencing (16S-seq). Single-cell transcriptome sequencing (scRNA-seq) was employed to profile the tumor microenvironment (TME). Moreover, correlations between the intratumor microbiota and the TME were explored. Finally, machine-learning models based on the intratumoral microbiota were constructed to predict pCR.

Results A total of 89 female patients with early-stage TNBC treated by neoadjuvant Chemo-IM were enrolled. We found that the pCR group had greater diversity and a higher load of intratumoral microbiota than the non-pCR group. Intriguingly, scRNA-seq revealed significantly increased T cell infiltration and decreased tumor-associated macrophage infiltration into tumors in the pCR group. Moreover, intratumoral microbiota load was positively associated with CD4+CXCL13+ T cell infiltration and negatively associated with CD68+SPP1+ macrophage infiltration. Combined analysis of 16S-seq and scRNA-seq data revealed that intratumoral microbiota were present in both cancer and immune cells. Finally, we developed a model incorporating intratumoral microbiota and clinicopathological characteristics, and it showed strong power for predicting pCR to neoadjuvant Chemo-IM.

Conclusions Intratumoral microbiota may serve as a strong and specific predictor of the response of patients with early-stage TNBC to neoadjuvant Chemo-IM. Our findings could contribute to the development of individualized Chemo-IM strategies for treating TNBC.

  • Immunotherapy
  • Biomarker
  • Breast Cancer
  • Intratumoral

Data availability statement

Data are available upon reasonable request. Raw data and processed data of single-cell transcriptome sequencing are being submitted to GEO database (accession number is pending). All the other data should be requested to and will be fulfilled by the lead contact.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.

https://doi.org/10.1136/jitc-2024-010365

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    WHAT IS ALREADY KNOWN ON THIS TOPIC

    • While neoadjuvant chemoimmunotherapy (Chemo-IM) has been established as the standard care for early-stage triple-negative breast cancer (TNBC), there is a need to develop efficient biomarkers for individualized Chemo-IM.

    • Pathological complete response (pCR) after neoadjuvant treatment has been established as a strong predictor of long-term survival outcomes among patients with TNBC.

    • Intratumoral microbiota is associated with tumorigenesis and resistance to antitumor treatment; however, its role as a tool for predicting pCR to neoadjuvant Chemo-IM in patients with TNBC remains unknown.

    WHAT THIS STUDY ADDS

    • Intratumoral microbiota load is positively correlated with pCR to neoadjuvant Chemo-IM in patients with TNBC.

    • Tumors with pCR have significantly more T cells and less macrophages than tumors with non-pCR.

    • Intratumoral microbiota load is positively associated with T cell infiltration and negatively associated with macrophage infiltration.

    • We constructed and validated a machine-learning model based on intratumoral microbiota for predicting pCR to neoadjuvant Chemo-IM in patients with TNBC.

    HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

    • Intratumoral microbiota could serve as a strong and specific predictor of the response to neoadjuvant Chemo-IM in patients with early-stage TNBC. Our findings could contribute to the development of individualized Chemo-IM strategies for treating TNBC.

    Background

    Triple-negative breast cancer (TNBC) accounts for approximately 15% of all breast cancer cases and results in worse survival outcomes than other subtypes of breast cancer due to the lack of targeted therapies.1–4 Neoadjuvant chemotherapy is one of the most effective therapies for TNBC and pathological complete response (pCR) after neoadjuvant chemotherapy has been established as a strong predictor of long-term survival outcomes.5–8 Recent findings have shown that neoadjuvant chemotherapy combined with immunotherapy (Chemo-IM) achieves significantly increased pCR and progression-free survival rates compared with chemotherapy alone in patients with early stage TNBC,9 10 establishing neoadjuvant Chemo-IM as a preferred treatment for highly selected patients with early stage TNBC. Notably, additional immunotherapy only resulted in a 13% absolute improvement in the pCR rate, and nearly 40% of patients did not achieve pCR.10 Unfortunately, we still lack efficient strategies or biomarkers to identify patients who would benefit from neoadjuvant Chemo-IM. Therefore, identification and validation of powerful and specific biomarkers or strategies is crucial for developing individualized treatment strategies and avoiding overtreatment in patients with early-stage TNBC.

    Approximately 13% of global cancer cases are correlated with infectious agents,11 and the microbiome has been widely proven to play pivotal roles in the tumorigenesis, progression and treatment resistance of various cancers.12–18 Gut and oral microbiota have gained widespread attention and been extensively investigated because of their abundance.19–22 In addition, microbiota have been detected within several tumors that were initially considered sterile,23 reinforcing the crucial role of intratumoral microbiota in cancer. One mechanism by which intratumoral microbiota promotes tumorigenesis is by influencing the functions of CD8+T cells and shaping the tumor microenvironment (TME) through their metabolites, facilitating immune escape by cancer cells.14 17 18 24 Indeed, analyses of nasopharyngeal carcinoma (NPC) tissues demonstrated that intratumoral microbiota load was negatively correlated with T-lymphocyte infiltration,16 indicating that intratumoral microbiota may serve as a potential biomarker for immunotherapy. However, little is known about whether intratumoral microbiota could serve as a predictor for patients with TNBC receiving neoadjuvant Chemo-IM.

    In our study, we enrolled patients with early-stage TNBC receiving neoadjuvant Chemo-IM and employed 16S rDNA sequencing (16S-seq) to characterize the intratumoral microbiota landscape of pCR and non-pCR tumors. We also performed single-cell transcriptome sequencing (scRNA-seq) to profile the TME of tumors from pCR and non-pCR groups, and investigated its relationship with intratumoral microbiota. Moreover, we generated a machine-learning model based on intratumoral microbiota to predict pCR in patients with early-stage TNBC.

    Methods

    Patient enrollment

    For the Chemo-IM cohort, patients treated at our center and the First People’s Hospital of Foshan between August 2020 and May 2024 were enrolled if they: (1) had newly diagnosed TNBC without distant metastasis; (2) were aged 18–70 years old; (3) did not receive previous anticancer treatments; (4) would undergo neoadjuvant Chemo-IM and radical surgery; (5) did not have other malignancies. Moreover, patients only receiving neoadjuvant chemotherapy in our previous clinical trial were also included as the chemotherapy (Chemo) cohort.25 The study protocols were approved by the Research Ethics Committee of our hospital (approval number: KY2023-211-01) and the First People’s Hospital of Foshan (without approval number). Written informed consent for the use of patient specimens was obtained from each participant.

    Patient tissues

    Pretreatment fresh tumor tissues of seven patients (four with pCR and three with non-pCR) from our hospital in the Chemo-IM cohort were subjected to scRNA-seq, and another eight fresh tissues (n=5 in pCR and n=3 in non-pCR) were subjected to flow cytometry analysis. Frozen tissues of 60 patients from our center and 29 patients from the First People’s Hospital of Foshan in the Chemo-IM cohort were collected for 16S-seq and real-time quantitative PCR (RT-qPCR) analysis of 16S rDNA. Moreover, formalin-fixed paraffin-embedded tissue slides of patients in Chemo-IM and Chemo cohorts were used for fluorescence in situ hybridization (FISH) staining of 16S rDNA and immunohistochemistry (IHC) staining of Lipopolysaccharide (LPS).

    Neoadjuvant treatment strategies

    Neoadjuvant Chemo-IM included albumin-bound paclitaxel (260 mg/m2, d1) plus carboplatin (AUC 5, d1) and sintilimab/camrelizumab/pembrolizumab (200 mg, d1) every 3 weeks for 4–6 cycles, or albumin-bound paclitaxel (260 mg/m2, d1) plus carboplatin (AUC 5, d1) and sintilimab/camrelizumab/pembrolizumab (200 mg, d1) every 3 weeks for 4 cycles followed by epirubicin (75 mg/m2, d1) plus cyclophosphamide (600 mg/m2, d1) and sintilimab/camrelizumab/pembrolizumab (200 mg, d1) every 3 weeks for 4 cycles. All the patients completed the scheduled cycles and received radical surgery treatment.

    Experimental methods

    Detailed information on experimental methods such as 16S-seq, scRNA-seq, FISH, IHC, multiplex immunofluorescence staining (mIF), flow cytometry, RT-qPCR and in vitro intratumoral microbiota culture were provided in online supplemental material.

    Supplemental material

    Statistical analysis

    Unpaired student’s t-test was applied for statistical comparisons. FISH and IHC data were quantified by Image-Pro Plus V.6.0 (Media Cybernetics, USA). Machine-learning models for predicting pathological tumor response were built using logistic regression after including intratumoral microbiota load, age, pathological grade, tumor stage and Ki67 in the training cohort and then validated in the validation cohort. Receiver operating curves were employed to compare their efficiencies in predicting pCR. Data in the histograms were expressed as mean±SD and the error bars represent the SD. Statistical p values were shown in the figure and sample sizes were provided in figure legends. p<0.05 was considered statistically significant. Unless indicated, only p<0.05 was shown in the figures.

    Results

    Patient characteristics

    A total of 60 female patients from Guangdong Provincial People’s Hospital (GDPH cohort) and 29 female patients from the First People’s Hospital of Foshan (FPHF cohort) with newly diagnosed, previously untreated, non-metastatic TNBC that was treated with neoadjuvant Chemo-IM between August 2020 and May 2024 were enrolled in our study, and their baseline characteristics were shown in table 1. Among these patients, about 63% (56/89) were diagnosed with stage I–II disease and only 6 (6.7%) patients had N3 disease. In the Chemo-IM cohort, 22 (24.7%) patients received albumin-bound paclitaxel plus carboplatin and programmed cell death protein-1 (PD-1) inhibitors, while 67 (75.2%) patients received albumin-bound paclitaxel plus carboplatin and PD-1 inhibitors followed by epirubicin plus cyclophosphamide and PD-1 inhibitors. T stage (p=0.265 and p=0.0547), N stage (p=0.444 and p=0.89), overall stage (p=0.835 and p=0.75) and Chemo-IM regimens (p=0.811 and p=1.0) were well balanced between pCR and non-pCR groups in the two cohorts, indicating that tumor stage and chemotherapy regimens were unlikely to be the key factors affecting pathological response. Surprisingly, the pCR group had a significantly greater percentage of patients with pathological grade III disease (p=0.02 and p=0.033) than the non-pCR group. Besides, Ki67 expression was significantly higher in the pCR group than that in the non-pCR group in the GDPH cohort (p=0.018). Moreover, we enrolled 39 patients from our previous clinical trial which compared neoadjuvant docetaxel plus carboplatin with epirubicin plus cyclophosphamide and docetaxel in early-stage TNBC (Chemo cohort, table 1). Although the non-pCR group had a higher percentage of patients with T3–4 and N1–3 disease than the pCR group, overall stages were well balanced (p=0.183). Similar to the pCR group in the Chemo-IM cohort, the pCR group in the Chemo-only cohort also had a significantly greater percentage of patients with pathological grade III disease (p=0.016) and higher Ki67 expression (p=0.04) than the non-pCR group. The flow chart of our study was summarized in online supplemental figure S1.

    View this table:
    Table 1

    Baseline characteristics of patients enrolled in this study

    Intratumoral microbiota was positively associated with pathological tumor response

    To elucidate the landscape of intratumoral microbiota in TNBC, we first performed 16S-seq on pretreatment tissues from 57 patients (35 in pCR group and 22 in non-pCR group) from the GDPH cohort in the Chemo-IM cohort. Contamination was removed following the standard procedure (online supplemental figure S2). After removing contaminant amplicon sequence variants (online supplemental figure S3), we identified a wide range of bacterial communities at the class, order, family and genus levels (online supplemental figure S4A-D). Our analysis demonstrated that Gammaproteobacteria was the most abundant intratumoral microbiota at the class level, comprizing 45.4% of the microbiota in the pCR group and 40.3% in the non-pCR group, whereas Acinetobacter accounted for the highest proportion at the genus level (15.4% in pCR and 15.1% in non-pCR; Online supplemental figure S4A and S4D). We also performed 16S-seq analysis of 29 samples in the FPHF cohort to validate the above findings. As expected, about 50–80% of intratumoral bacteria identified in the FPHF cohort were consistent with those in the GDPH cohort (online supplemental figures S4E–G). Next, we performed scRNA-seq analysis of pretreatment tissues from seven patients receiving neoadjuvant Chemo-IM in the GDPH cohort (four in pCR group and three in non-pCR group) to further explore the distribution of intratumoral microbiota in TNBC. As shown, Escherichia was identified as the most abundant intratumoral microbiota at both genus and species levels (online supplemental figures S4H and S4I). After incorporating the 16S-seq and scRNA-seq data, we generated a schematic phylogenetic tree (figure 1A). These high-throughput data suggested the presence of microbiota within TNBC tissues. To further validate our high-throughput sequencing findings, we performed an in vitro culture assay and showed that intratumoral microbiota isolated from TNBC tissues were successfully cultivated in vitro under both aerobic and anaerobic conditions (online supplemental figure S4J). Moreover, 16S-seq analysis of these colonies identified many aerobic and anaerobic bacteria which were quite consistent with those from tumor tissues (online supplemental figure S4D–S4K), further supporting the notion that bacteria cultured in vitro originated from tissues. Collectively, our data verified that intratumoral microbiota were rich and diverse in TNBC tissues.

    Figure 1
    Figure 1

    pCR tumors had higher intratumoral microbiota load. (A) Schematic phylogenetic tree depicting the representative bacterial genera of TNBC tissues based on 16S-seq and scRNA-seq data. Various colors and shades within the circles denote the classifications of bacteria at the order (inner circle) and phylum (middle circle) levels. (B) RT-qPCR analysis of intratumoral microbiota load of pCR and non-pCR tumors (n=10 in control, n=22 in non-pCR and n=35 in pCR groups). (C) Representative images of FISH staining of 16S rDNA in pCR and non-pCR tumors from the GDPH cohort. Scale bars, left panels, 400 µm; right panels, 40 µm. (D) Statistical quantification of 16S rDNA in pCR and non-pCR tumors from the GDPH cohort (n=23 in non-pCR and n=37 in pCR). (E) Representative images of IHC staining of LPS in pCR and non-pCR tumors from the GDPH cohort. Scale bars, left panels, 375 µm; right panels, 35 µm. (F) Statistical quantification of LPS in pCR and non-pCR tumors from the GDPH cohort (n=22 in non-pCR and n=37 in pCR). (G) Representative images of FISH staining of 16S rDNA in tumors from the FPHF cohort. Scale bars, left panels, 400 µm; right panels, 40 µm. (H) Statistical quantification of 16S rDNA in pCR and non-pCR tumors from the FPHF cohort (n=11 in non-pCR and n=18 in pCR). (I) Representative images of IHC staining of LPS in tumors from the FPHF cohort. Scale bars, left panels, 375 µm; right panels, 35 µm. (J) Statistical quantification of LPS in pCR and non-pCR tumors from the FPHF cohort (n=11 in non-pCR and n=18 in pCR). (K) Statistical quantification of 16S rDNA in tumors with radiology-defined CR and non-CR (n=20 in non-CR and n=40 in CR). (L) Statistical quantification of LPS in tumors with radiology-defined CR and non-CR (n=19 in non-CR and n=40 in CR). 16S-seq, 16S rDNA sequencing; CR, complete response; FISH, fluorescence in situ hybridization; FPHF, First People’s Hospital of Foshan; GDPH, Guangdong Provincial People’s Hospital; IHC, immunohistochemistry; pCR, pathological complete response; RT-qPCR, real-time quantitative PCR; ScRNA-seq, single-cell transcriptome sequencing.

    Next, we evaluated the difference in intratumoral microbiota load between pCR and non-pCR tumors. Analysis of 16S-seq data revealed that patients who achieved pCR presented significantly increased α diversity (p=0.015) although β diversity did not significantly differ between the two groups (online supplemental figures S5A and S5B). Our RT-qPCR analysis of tumors in the GDPH cohort using the relative quantification method demonstrated that bacterial load was significantly greater in tissues from the pCR group than in tissues from the non-pCR group (figure 1B). Then, we developed a standard curve for absolute quantification and still found that the pCR group had a significantly greater bacterial load than the non-pCR group (online supplemental figures S5C and S5D). FISH staining of 16S rDNA and IHC staining of LPS in the GDPH cohort also demonstrated significantly increased intratumoral bacterial load in pCR tissues, which was further confirmed by the results from the FPHF cohort (figure 1C–J). These data indicated that intratumoral microbiota load was positively correlated with pathological tumor response. We then investigated the relationship between intratumoral microbiota and tumor response evaluated by radiological methods. As shown, tumors with radiologically defined complete response (CR) had significantly greater intratumoral bacterial loads than those without CR (figure 1K and L and online supplemental figure S5E). Furthermore, we analyzed the intratumoral microbiota load in pCR and non-pCR groups within the Chemo cohort. Intriguingly, intratumoral bacterial load was comparable between these two groups (online supplemental figures S5F and S5I). Taken together, these findings suggested that intratumoral microbiota could serve as a strong and specific predictor of the response to Chemo-IM but not of the response to Chemotherapy only in TNBC.

    T cells and macrophage infiltration were associated with pathological tumor response

    To elucidate the mechanisms underlying different tumor responses, we employed scRNA-seq to decipher the landscape of the TME of tumors in pCR and non-pCR groups in the Chemo-IM cohort. We generated scRNA-seq profiles of pretreatment tumors from seven patients (four in pCR group and three in non-pCR group; figure 2A). After quality control, we obtained a total of 76,369 single cells for subsequent analysis, with an average of 1,162 genes and 3,884 unique molecular identifiers. We subsequently annotated 7-cell clusters by curated gene sets, including 36,076 epithelial cells, 1,258 endothelial cells (ECs), 1,202 mural cells, 3,687 B cells, 18,325 T cells, 4,935 plasma cells and 7,224 mononuclear phagocytes (MPs) (figure 2B and online supplemental figure S6A; online supplemental table S1). We further performed normalization, dimensionality reduction, and clustering of each major cell type to identify subclusters, and we obtained a total of 38 subclusters containing 3 epithelial cell subclusters, 5 EC subclusters, 6 mural cell subclusters and 24 immune cell subclusters (figure 2C–2H and online supplemental figures S6B and S6G). Overall, patients who achieved pCR had more T cell (30.75% vs 20.54%, online supplemental table S2), B cell (6.75% vs 2.56%, online supplemental table S3) and plasma cell (11.15% vs 2.34%, online supplemental table S4) infiltration and less MP (7.75% vs 13.32%, online supplemental table S5) infiltration in tumors than patients in the non-pCR group. In terms of immune cell subclusters, patients who achieved pCR had more CD4+CXCL13+ follicular helper T cells (13.55% vs 9.0%), CD4+FOXP3+ regulatory T cells (19.0% vs 11.41%), naive B cells (54.7% vs 49.74%), first subcluster plasma cells (51.35% vs 12.16%) and FOLR2+ macrophages (35.76% vs 21.8%) infiltrating tumors (figure 2C–2H; online supplemental tables S2-S6). Surprisingly, CD8+ T cell infiltration did not significantly differ between pCR and non-pCR groups, whereas tumor-associated macrophages (TAMs), including antitumor FOLR2+ macrophages and tumor-promoting SPP1+ macrophages, were differentially distributed between the two groups (figure 2G, online supplemental table S6), suggesting that macrophages may play a pivotal role in driving immunotherapy resistance in TNBC. We subsequently performed mIF staining to validate the findings of scRNA-seq. CD4+CXCL13+ T cells were significantly increased in tumors from the pCR group, while CD68+SPP1+ macrophages were more abundant in tumors from the non-pCR group (figure 2I). Moreover, flow cytometry analysis also demonstrated that tumors from the pCR group had significantly lower percentages of CD68+SPP1+ macrophages and greater percentages of CD4+CXCL13+ T cell infiltration (figure 2J and K, online supplemental figure S7). Collectively, our analysis demonstrated the distinctive TME between pCR and non-pCR groups.

    Figure 2
    Figure 2

    scRNA-seq identified the distinctive tumor microenvironment of pCR and non-pCR tumors. (A) Workflow diagram of scRNA-seq of pCR and non-pCR tumors. (B) UMAP plots (left panel) and statistical histograms (right panel) of cells from seven samples, with each cell color coded to indicate the associated cell types. (C–H) UMAP plots (left panels) and statistical histograms (right panels) of subclusters of cancer and immune cells. (I) mIF staining of markers of CD4+CXCL13+ T cells and CD68+SPP1+ macrophages (n=6 in pCR and n=5 in non-pCR). (J) Statistical analysis of CD68+SPP1+ macrophages by flow cytometry assay (n=5 in pCR and n=3 in non-pCR). (K) Statistical analysis of CD4+CXCL13+ T cells by flow cytometry assay (n=5 in pCR and n=3 in non-pCR). cDC, classical dendritic cell; DAPI, 4′,6-diamidino-2-phenylindole; ECs, endothelial cells; FACS, Fluorescence-Activated Cell Sorting; mIF, multiplex immunofluorescence staining; MPs, mononuclear phagocytes; pCR, pathological complete response; pDC, plasmacytoid dendritic cell; ScRNA-seq, single-cell transcriptome sequencing; TNBC, triple-negative breast cancer; UMAP, Uniform Manifold Approximation and Projection.

    Intratumoral microbiota is positively correlated with an activated TME

    Having elucidated the distinctive landscape of intratumoral microbiota and TME between pCR and non-pCR groups, we next explored the relationship between intratumoral microbiota and TME. We performed mIF analysis on 11 patient samples with 6 in the pCR group and 5 in the non-pCR group. Our analysis demonstrated that intratumoral microbiota load was positively correlated with the infiltration of CD4+CXCL13+ T cells but negatively correlated with the infiltration of CD68+SPP1+ macrophages (figure 3A–3D), indicating that intratumoral microbiota may serve as an inducer of antitumor immunity. A previous study revealed that intratumoral microbiota was present in both cancer and immune cells.23 To further evaluate how intratumoral microbiota affects the TME in TNBC, we performed mIF and FISH staining on consecutive slides to analyze the distribution of intratumoral microbiota. Consistent with previous findings, intratumoral microbiota were found in both cancer and immune cells, with a higher abundance in cancer cells (figure 3E and F). Moreover, analysis of our scRNA-seq data confirmed that cancer cells contained significantly more bacterial genome transcripts than immune cells at both genus and species levels (figure 3G and H). These results suggested that intratumoral microbiota may regulate the TME of TNBC through both cancer and immune cells.

    Figure 3
    Figure 3

    Intratumoral microbiota was positively associated with activated TME. (A–B) Correlation analysis of LPS and CD4+CXCL13+ T cells and CD68+SPP1+ macrophages (n=11). (C–D) Correlation analysis of 16S rDNA and CD4+CXCL13+ T cells and CD68+SPP1+ macrophages (n=11). (E) Colocalization analysis of 16S rDNA in cancer (pan-CK) and immune (CD45) cells by FISH and mIF. Scale bars, 20 µm. (F) Statistical analysis of pan-CK+16S rDNA+ and CD45+16S rDNA+ cells in pCR and non-pCR tumors (n=8). (G–H) Histograms of bacterial gene fragments detected in different cell types. (I) Bar plots showing the adjusted Wilcoxon p value comparing the transcriptional diversity of bacterium-associated cells with unassociated cells for each cell type. (J) KEGG pathway enrichment analysis of 16S-seq data in pCR and non-pCR tumors. (K) COG pathway enrichment analysis of 16S-seq data in pCR and non-pCR tumors. COG, Clusters of Orthologous Groups; ECs, endothelial cells; FISH, fluorescence in situ hybridization; IHC, immunohistochemistry; KEGG, Kyoto Encyclopedia of Genes and Genomes; LPS, Lipopolysaccharide; mIF, multiplex immunofluorescence staining; MPs, mononuclear phagocytes; pCR, pathological complete response; TME, tumor microenvironment.

    To understand how intratumoral microbiota regulated the TME, we performed a comprehensive analysis of scRNA-seq data and identified distinct bacterial profiles across epithelial and immune cells. As shown, a wide spectrum of intratumoral bacteria, including Escherichia, Streptomyces, Clostridium, Ralstonia and Salmonella, preferred to reside in epithelial cells of tumors in the non-pCR group, whereas their distributions in T cells and epithelial cells were almost comparable in tumors in the pCR group (online supplemental figures S8A and S8B). Then, we focused on the intratumoral bacteria with differential distributions between cancer and immune cells. Intriguingly, Acinetobacter was predominantly distributed in XCL1+ NK, MKI67+ proliferating and CXCL13+CD4+ T cells of pCR tumors; however, Streptococcus was located mainly in the epithelial cells of tumors in the non-pCR group (online supplemental figures S8C and S8F). Next, we determined whether the presence of cell-associated bacteria was correlated with the diversity of somatic cell states by calculating the Shannon diversity of each cell’s transcriptional profile.26 As shown, the non-pCR group exhibited significantly increased diversity in epithelial cells, whereas the pCR group had increased diversity, mainly in immune cells (figure 3I and online supplemental figures S9A). These results suggested that intratumoral microbiota within cancer cells may have distinctive functions from those within immune cells. Acinetobacter may activate immune cells to perform antitumor functions, whereas Streptococcus promotes the resistance of cancer cells to immunotherapy, which was validated by previous findings.18 27 Moreover, our analysis revealed that the functions of intratumoral microbiota were predominantly enriched in metabolism-related pathways (figure 3J and K). Further investigation of metabolic pathways revealed diverse metabolites, such as carbohydrates, amino acids and lipids (online supplemental figures S9B and S9D), highlighting the complexity of metabolism-mediated regulation of TME. Collectively, those high-throughput data suggested that intratumoral microbiota may activate the antitumor immunity through their metabolites in TNBC.

    Intratumoral microbiota predicts pCR to neoadjuvant Chemo-IM

    As the combined positive score (CPS), defined as the sum of programmed death-ligand 1 (PD-L1)-positive tumor cells and PD-L1-positive immune cells counts per 100 total cells, has been reported to be a strong biomarker of immunotherapy in TNBC,9 10 we first determined the CPS of tumors from pCR and non-pCR groups. However, CPS was comparable between the two groups (p=0.0867, figure 4A), indicating that CPS alone may not be a sufficiently strong predictor of the response to neoadjuvant Chemo-IM in patients with early-stage TNBC. Next, we evaluated whether intratumoral microbiota could serve as a strong biomarker for neoadjuvant Chemo-IM in patients with early-stage TNBC. To this end, we generated machine learning models after including clinicopathological factors (age, pathological grade, tumor stage, Ki67 expression, and CPS) and intratumoral microbiota load (quantitative FISH and IHC staining data). After random stratification, 40 patients in the GDPH cohort served as the training set, 17 patients served as the internal validation set (online supplemental table S7) and 29 patients in the FPHF cohort served as the external validation set. As shown, models incorporating either LPS (area under curve (AUC)=0.883 in training set, AUC=0.800 in internal validation set, AUC=0.773 in external validation set) or FISH data (AUC=0.891 in training set, AUC=0.814 in internal validation set, AUC=0.768 in external validation set) demonstrated significantly stronger power than the model constructed alone with clinicopathological factors (AUC=0.789 in training set, AUC=0.686 in internal validation set, AUC=0.672 in external validation set) in both training and validation sets (figure 4B–D). Moreover, a combination of LPS and FISH data achieved the strongest predictive performance for pCR (AUC=0.955 in training set, AUC=0.843 in internal validation set, AUC=0.813 in external validation set; figure 4B–D). Taken together, these findings suggested that intratumoral microbiota could serve as a powerful predictor of response to neoadjuvant Chemo-IM in early-stage TNBC.

    Figure 4
    Figure 4

    Intratumoral microbiota predicts tumor response to neoadjuvant Chemo-IM in early-stage TNBC. (A) CPS of pCR and non-pCR tumors (n=22 in non-pCR and n=35 in pCR). (B–D) ROCs of different predictive models with or without intratumoral microbiota in training (B), internal validation (C) and external validation sets (D). The clinical model was constructed using the factors including age, pathological grade, tumor stage, Ki67 expression and CPS. AUC, area under curve; Chemo-IM, chemotherapy combined with immunotherapy; CPS, combined positive score; FISH, fluorescence in situ hybridization; FRP, False Positive Rate; LPS, Lipopolysaccharide; pCR, pathological complete response; ROCs, receiver operating curves; TNBC, triple-negative breast cancer; TRP, True Positive Rate.

    Discussion

    Neoadjuvant Chemo-IM has emerged as one of the standard strategies for treating early-stage TNBC; however, precise selection of patients who may benefit from additional immunotherapy remains a critical challenge for clinical practice. On the basis of this premise, our current study explored and validated intratumoral microbiota as a strong predictor of the response to neoadjuvant Chemo-IM in early-stage TNBC. To the best of our knowledge, this is the first study to investigate the role of intratumoral microbiota in predicting pathological tumor response after neoadjuvant Chemo-IM in patients with TNBC. We found that patients who achieved pCR had significantly greater intratumoral microbiota loads, greater T cell infiltration and lower TAM infiltration in tumor tissues than patients who achieved non-pCR. Moreover, combined analysis of 16S-seq and scRNA-seq data demonstrated that intratumoral microbiota may promote an antitumor immune phenotype via their metabolites. Finally, we demonstrated that intratumoral microbiota could serve as a powerful predictor of pathological tumor response after neoadjuvant Chemo-IM. Thus, our findings provided a novel strategy for delivering individualized neoadjuvant Chemo-IM in early-stage TNBC.

    Since intratumoral microbiota have been found in seven human cancer types including breast cancer,23 they have been widely investigated and reported to play pivotal roles in prognosis prediction and treatment resistance.28–30 A study by Qiao et al demonstrated that intratumoral microbiota served as a robust prognostic tool in NPC and was negatively associated with T lymphocyte infiltration.16 Intratumoral microbiota also served as an independent predictive marker of the response of locally advanced rectal cancer to neoadjuvant chemoradiotherapy31 and regulated neoadjuvant Chemo-IM in patients with resectable esophageal squamous cell carcinoma.32 These studies highlighted the crucial roles of intratumoral microbiota in neoadjuvant cancer treatment. Similar to previous findings, our study also revealed that intratumoral microbiota could predict response to neoadjuvant Chemo-IM in TNBC. In contrast to previous results showing that Streptococcus was positively correlated with GrzB+ and CD8+ T cell infiltration in esophageal carcinoma tissues32 and that a high intratumoral load was associated with reduced CD8+ T-cell infiltration in NPC,16 our findings showed that Streptococcus was mainly distributed in cancer cells of patients who did not achieve pCR and that a high intratumoral microbiota load was positively associated with an activated TME in TNBC. A recent study demonstrated a positive correlation between α diversity and favorable outcomes in patients with pancreatic cancer,33 which was similar to our results. These previous studies, together with our reports, suggested that host-microbiota interactions determined the distinctive functions of intratumoral microbiota in different cancers.

    To understand the mechanisms by which intratumoral microbiota regulated the TME, we performed a combined analysis of 16S-seq and scRNA-seq data. Intriguingly, pathway enrichment analysis demonstrated that intratumoral microbiota were enriched in metabolic pathways. Notably, metabolites from gut microbiota have been reported to be associated with tumorigenesis, diagnosis and treatment of colorectal cancer.34 Moreover, the metabolite trimethylamine N-oxide derived from Clostridiales promoted the efficacy of immunotherapy in TNBC.17 Thus, these literatures, together with our findings, suggested that metabolites from intratumoral microbiota may regulate TME and could serve as promising biomarkers for predicting the efficacy of neoadjuvant Chemo-IM in TNBC. However, this speculation needs to be validated by future cause-effect experiments. Notably, intratumoral microbiota were present in both cancer and immune cells, as revealed by mIF and scRNA-seq data. Therefore, how intratumoral microbiota metabolites regulate cancer and immune cells separately to promote the resistance of immunotherapy needs to be elucidated in future studies.

    Despite these profound results, the limitations of our study should be acknowledged. Because the neoadjuvant Chemo-IM strategy has not been approved for a long time in China, only a small proportion of patients have received this treatment. Therefore, the sample size of our study may not have been large enough. To attenuate the impact of small sample size, we performed scRNA-seq to further support the results of 16S-seq, as scRNA-seq has been reported to provide supplemental information.35 However, future studies with large cohorts are needed to validate our findings. Additionally, although our analysis of high-throughput data suggested that intratumoral microbiota may influence cancer and immune cells separately through their metabolites to regulate the efficacy of neoadjuvant Chemo-IM, further experimental studies are required to clarify the detailed mechanism. Finally, we did not identify the specific intratumoral bacteria that were differentially presented between pCR and non-pCR tumors because the 16S-seq used in our study could only qualitatively analyze the type and constituent of bacteria but not quantitatively compare the difference. The 16S-seq with quantitative analysis functions is needed to address this limitation in future studies.

    In conclusion, we explored and validated intratumoral microbiota as a strong predictor of response to neoadjuvant Chemo-IM in early-stage TNBC. Our findings could contribute to the development of individualized Chemo-IM strategies for TNBC.

    Data availability statement

    Data are available upon reasonable request. Raw data and processed data of single-cell transcriptome sequencing are being submitted to GEO database (accession number is pending). All the other data should be requested to and will be fulfilled by the lead contact.

    Ethics statements

    Patient consent for publication

    Ethics approval

    This study involves human participants and was approved by the Research Ethics Committee of Guangdong Provincial People’s Hospital (Guangdong Academy of Medical Sciences), (approval number: KY2023-211-01). Participants gave informed consent to participate in the study before taking part.

    Acknowledgments

    We sincerely thank Dr. Wei Li (The First People’s Hospital of Foshan) for his assistance in obtaining data from an external validation cohort at The First People’s Hospital of Foshan. We also acknowledge Dr. Binbin Chen and Dr. Jing Han (Sun Yat-sen University Cancer Center) for their valuable support in conducting experiments, including flow cytometry, fluorescence in situ hybridization, and immunohistochemistry.

    References

    1. ↵
      2. Waks AG ,
      3. Winer EP
      . Breast Cancer Treatment: A Review. JAMA 2019;321:288–300. doi:10.1001/jama.2018.19323
      OpenUrlCrossRefPubMed
    2. ↵
      2. Leon-Ferre RA ,
      3. Goetz MP
      . Advances in systemic therapies for triple negative breast cancer. BMJ 2023;381:e071674. doi:10.1136/bmj-2022-071674
    3. ↵
      2. Denkert C ,
      3. Liedtke C ,
      4. Tutt A , et al
      . Molecular alterations in triple-negative breast cancer-the road to new treatment strategies. Lancet 2017;389:2430–42. doi:10.1016/S0140-6736(16)32454-0
      OpenUrlCrossRefPubMed
    4. ↵
      2. Adams S ,
      3. Gatti-Mays ME ,
      4. Kalinsky K , et al
      . Current Landscape of Immunotherapy in Breast Cancer: A Review. JAMA Oncol 2019;5:1205–14. doi:10.1001/jamaoncol.2018.7147
      OpenUrl
    5. ↵
      2. Broglio KR ,
      3. Quintana M ,
      4. Foster M , et al
      . Association of Pathologic Complete Response to Neoadjuvant Therapy in HER2-Positive Breast Cancer With Long-Term Outcomes: A Meta-Analysis. JAMA Oncol 2016;2:751–60. doi:10.1001/jamaoncol.2015.6113
      OpenUrl
    6. ↵
      2. Conforti F ,
      3. Pala L ,
      4. Sala I , et al
      . Evaluation of pathological complete response as surrogate endpoint in neoadjuvant randomised clinical trials of early stage breast cancer: systematic review and meta-analysis. BMJ 2021;375:e066381. doi:10.1136/bmj-2021-066381
    7. ↵
      2. von Minckwitz G ,
      3. Untch M ,
      4. Blohmer J-U , et al
      . Definition and Impact of Pathologic Complete Response on Prognosis After Neoadjuvant Chemotherapy in Various Intrinsic Breast Cancer Subtypes. JCO 2012;30:1796–804. doi:10.1200/JCO.2011.38.8595
      OpenUrlAbstract/FREE Full Text
    8. ↵
      2. Yoshioka T ,
      3. Hosoda M ,
      4. Yamamoto M , et al
      . Prognostic significance of pathologic complete response and Ki67 expression after neoadjuvant chemotherapy in breast cancer. Breast Cancer 2015;22:185–91. doi:10.1007/s12282-013-0474-2
      OpenUrlCrossRefPubMed
    9. ↵
      2. Schmid P ,
      3. Cortes J ,
      4. Dent R , et al
      . Event-free Survival with Pembrolizumab in Early Triple-Negative Breast Cancer. N Engl J Med 2022;386:556–67. doi:10.1056/NEJMoa2112651
      OpenUrlCrossRefPubMed
    10. ↵
      2. Schmid P ,
      3. Cortes J ,
      4. Pusztai L , et al
      . Pembrolizumab for Early Triple-Negative Breast Cancer. N Engl J Med 2020;382:810–21. doi:10.1056/NEJMoa1910549
      OpenUrlCrossRefPubMed
    11. ↵
      2. de Martel C ,
      3. Georges D ,
      4. Bray F , et al
      . Global burden of cancer attributable to infections in 2018: a worldwide incidence analysis. Lancet Glob Health 2020;8:e180–90. doi:10.1016/S2214-109X(19)30488-7
      OpenUrlCrossRefPubMed
    12. ↵
      2. Dai J-H ,
      3. Tan X-R ,
      4. Qiao H , et al
      . Emerging clinical relevance of microbiome in cancer: promising biomarkers and therapeutic targets. Protein Cell 2024;15:239–60. doi:10.1093/procel/pwad052
      OpenUrl
    13. ↵
      2. Fu A ,
      3. Yao B ,
      4. Dong T , et al
      . Tumor-resident intracellular microbiota promotes metastatic colonization in breast cancer. Cell 2022;185:1356–72. doi:10.1016/j.cell.2022.02.027
      OpenUrlCrossRefPubMed
    14. ↵
      2. Jia D ,
      3. Wang Q ,
      4. Qi Y , et al
      . Microbial metabolite enhances immunotherapy efficacy by modulating T cell stemness in pan-cancer. Cell 2024;187:1651–65. doi:10.1016/j.cell.2024.02.022
      OpenUrlCrossRefPubMed
    15. ↵
      2. Park EM ,
      3. Chelvanambi M ,
      4. Bhutiani N , et al
      . Targeting the gut and tumor microbiota in cancer. Nat Med 2022;28:690–703. doi:10.1038/s41591-022-01779-2
      OpenUrlCrossRefPubMed
    16. ↵
      2. Qiao H ,
      3. Tan X-R ,
      4. Li H , et al
      . Association of Intratumoral Microbiota With Prognosis in Patients With Nasopharyngeal Carcinoma From 2 Hospitals in China. JAMA Oncol 2022;8:1301–9. doi:10.1001/jamaoncol.2022.2810
      OpenUrl
    17. ↵
      2. Wang H ,
      3. Rong X ,
      4. Zhao G , et al
      . The microbial metabolite trimethylamine N-oxide promotes antitumor immunity in triple-negative breast cancer. Cell Metab 2022;34:581–94. doi:10.1016/j.cmet.2022.02.010
      OpenUrlCrossRefPubMed
    18. ↵
      2. Fu K ,
      3. Cheung AHK ,
      4. Wong CC , et al
      . Streptococcus anginosus promotes gastric inflammation, atrophy, and tumorigenesis in mice. Cell 2024;187:882–96. doi:10.1016/j.cell.2024.01.004
      OpenUrlCrossRefPubMed
    19. ↵
      2. Nagata N ,
      3. Nishijima S ,
      4. Kojima Y , et al
      . Metagenomic Identification of Microbial Signatures Predicting Pancreatic Cancer From a Multinational Study. Gastroenterology 2022;163:222–38. doi:10.1053/j.gastro.2022.03.054
      OpenUrlCrossRefPubMed
    20. ↵
      2. Nakatsu G ,
      3. Zhou H ,
      4. Wu WKK , et al
      . Alterations in Enteric Virome Are Associated With Colorectal Cancer and Survival Outcomes. Gastroenterology 2018;155:529–41. doi:10.1053/j.gastro.2018.04.018
      OpenUrlCrossRefPubMed
    21. ↵
      2. Terrisse S ,
      3. Derosa L ,
      4. Iebba V , et al
      . Intestinal microbiota influences clinical outcome and side effects of early breast cancer treatment. Cell Death Differ 2021;28:2778–96. doi:10.1038/s41418-021-00784-1
      OpenUrlPubMed
    22. ↵
      2. Zhang X ,
      3. Zhang Y ,
      4. Gui X , et al
      . Salivary Fusobacterium nucleatum serves as a potential biomarker for colorectal cancer. iScience 2022;25:104203. doi:10.1016/j.isci.2022.104203
      OpenUrl
    23. ↵
      2. Nejman D ,
      3. Livyatan I ,
      4. Fuks G , et al
      . The human tumor microbiome is composed of tumor type-specific intracellular bacteria. Science 2020;368:973–80. doi:10.1126/science.aay9189
      OpenUrlAbstract/FREE Full Text
    24. ↵
      2. Zhang Q ,
      3. Zhao Q ,
      4. Li T , et al
      . Lactobacillus plantarum-derived indole-3-lactic acid ameliorates colorectal tumorigenesis via epigenetic regulation of CD8+ T cell immunity. Cell Metab 2023;35:943–60. doi:10.1016/j.cmet.2023.04.015
      OpenUrlCrossRefPubMed
    25. ↵
      2. Zhang L ,
      3. Wu Z-Y ,
      4. Li J , et al
      . Neoadjuvant docetaxel plus carboplatin vs epirubicin plus cyclophosphamide followed by docetaxel in triple-negative, early-stage breast cancer (NeoCART): Results from a multicenter, randomized controlled, open-label phase II trial. Int J Cancer 2022;150:654–62. doi:10.1002/ijc.33830
      OpenUrlCrossRefPubMed
    26. ↵
      2. Ghaddar B ,
      3. Biswas A ,
      4. Harris C , et al
      . Tumor microbiome links cellular programs and immunity in pancreatic cancer. Cancer Cell 1240;40. doi:10.1016/j.ccell.2022.09.009
    27. ↵
      2. Zhu G ,
      3. Su H ,
      4. Johnson CH , et al
      . Intratumour microbiome associated with the infiltration of cytotoxic CD8+ T cells and patient survival in cutaneous melanoma. Eur J Cancer 2021;151:25–34. doi:10.1016/j.ejca.2021.03.053
      OpenUrlCrossRefPubMed
    28. ↵
      2. Cao Y ,
      3. Xia H ,
      4. Tan X , et al
      . Intratumoural microbiota: a new frontier in cancer development and therapy. Signal Transduct Target Ther 2024;9:15. doi:10.1038/s41392-023-01693-0
      OpenUrl
    29. ↵
      2. Bullman S
      . The intratumoral microbiota: From microniches to single cells. Cell 2023;186:1532–4. doi:10.1016/j.cell.2023.03.012
      OpenUrlCrossRefPubMed
    30. ↵
      2. Yang L ,
      3. Li A ,
      4. Wang Y , et al
      . Intratumoral microbiota: roles in cancer initiation, development and therapeutic efficacy. Signal Transduct Target Ther 2023;8:35. doi:10.1038/s41392-022-01304-4
    31. ↵
      2. Sun L ,
      3. Qu J ,
      4. Ke X , et al
      . Interaction between intratumoral microbiota and tumor mediates the response of neoadjuvant therapy for rectal cancer. Front Microbiol 2023;14. doi:10.3389/fmicb.2023.1229888
    32. ↵
      2. Wu H ,
      3. Leng X ,
      4. Liu Q , et al
      . Intratumoral Microbiota Composition Regulates Chemoimmunotherapy Response in Esophageal Squamous Cell Carcinoma. Cancer Res 2023;83:3131–44. doi:10.1158/0008-5472.CAN-22-2593
      OpenUrl
    33. ↵
      2. Riquelme E ,
      3. Zhang Y ,
      4. Zhang L , et al
      . Tumor Microbiome Diversity and Composition Influence Pancreatic Cancer Outcomes. Cell 2019;178:795–806. doi:10.1016/j.cell.2019.07.008
      OpenUrlCrossRefPubMed
    34. ↵
      2. Qu R ,
      3. Zhang Y ,
      4. Ma Y , et al
      . Role of the Gut Microbiota and Its Metabolites in Tumorigenesis or Development of Colorectal Cancer. Adv Sci (Weinh) 2023;10:2205563. doi:10.1002/advs.202205563
    35. ↵
      2. Galeano Niño JL ,
      3. Wu H ,
      4. LaCourse KD , et al
      . Effect of the intratumoral microbiota on spatial and cellular heterogeneity in cancer. Nature New Biol 2022;611:810–7. doi:10.1038/s41586-022-05435-0
      OpenUrl

    Footnotes

    • Collaborators Not applicable.

    • Contributors KW, HP, and YC designed the study. YC, LY, YH, TZ, and LZ carried out experiments. YC and HP performed the 16S rDNA sequencing data analysis. YC, HP and LY performed scRNA sequencing data analysis. XZ contributed to data collection. PL, LY, MC, CW, XZ and ML collected clinical data. YC and HP wrote the manuscript. HP and KW revised the manuscript. KW is the guarantor.

    • Funding This study is supported by grants from the Ministry of Science and Technology of the People's Republic of China National Natural Science Foundation of China (82171898, 82372726), Deng Feng project of high-level hospital construction (DFJHBF202109), Guangdong Basic and Applied Basic Research Foundation (2022A1515012277, 2023A1515010222), Guangzhou Science and Technology Project (202002030236, 2025A04J7230), Macao Science and Technology Development Fund (20210701181316106/AKP), Beijing Medical Award Foundation (YXJL-2020-0941-0758), Beijing Science and Technology Innovation Medical Development Foundation (KC2022-ZZ-0091-5), Development Cancer for Medical Science and Technology National Health Commission of the People’s Republic of China (WKZX2023CX110002) and Beijing Life Oasis Public Service Center (cphcf-2022-058).

    • Competing interests No, there are no competing interests.

    • Provenance and peer review Not commissioned; internally peer reviewed.

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.