Abstract Background Patients with liver metastases of triple-negative breast cancer (TNBC) show poor prognosis compared with other metastases. Chemotherapy is the primary treatment for advanced TNBC. Tumor cell diversity and the tumor microenvironment could affect therapeutic effect. However, whether liver metastases of TNBC exhibit differential chemotherapy efficacy compared with the primary tumors remains inadequately understood. The specific mechanisms that modulate chemotherapy efficacy in liver metastases need further investigation. Methods Single-cell RNA sequencing data from public databases were leveraged to contrast the immune profiles of liver metastases and primary tumors in TNBC. Murine models bearing liver tumors or primary tumors of TNBC were used to evaluate chemotherapy efficacy. Techniques such as immunohistochemistry, wound healing assays, and colony formation assays were employed to account for tumor heterogeneity. Intratumoral T lymphocytes and macrophages were quantified and characterized using RNA sequencing, immunohistochemistry, and flow cytometry. Antibody-mediated depletion of CD8+T cells or macrophages in mice substantiated their impact on chemotherapy responses. Results Single-cell RNA sequencing data showed the immune microenvironments of liver metastases and primary tumors exhibited significant differences, which may critically influence chemotherapy outcomes. Mouse models confirmed that chemotherapy was less effective against liver tumors compared with subcutaneous tumors. After excluding the influence of tumor cell heterogeneity, the weaker responsiveness in liver tumors was mediated by the impeded infiltration of CD8+T cells, attributed to the decreased activation of macrophages. Augmenting macrophage activation can improve the chemotherapeutic efficacy in liver tumors. Moreover, chemotherapy drove the immune microenvironment towards increased suppression through distinct mechanisms, with neutrophil extracellular traps (NETs) accumulating in liver tumors and impaired functionality of macrophages at the primary site. The combination of NET inhibitors or macrophage activators with chemotherapy enhanced treatment effectiveness. Conclusions These findings disclose the compromised chemotherapeutic efficacy in liver tumors of TNBC and elucidate the underlying immune-related mechanisms within the tumor microenvironment. Targeting the specific underpinnings of immune suppression at different tumor sites with selective drugs could optimize chemotherapeutic efficacy. Keywords: Breast Cancer, Tumor microenvironment - TME, Chemotherapy, Immunosuppression, Tumor infiltrating lymphocyte - TIL __________________________________________________________________ WHAT IS ALREADY KNOWN ON THIS TOPIC * Patients with liver metastases of triple-negative breast cancer (TNBC) show poor prognosis compared with other metastases. Chemotherapy is the primary treatment for advanced TNBC. The heterogeneity of tumor cells and the tumor microenvironment may impact the therapeutic effect. Chemotherapy drugs have been reported to induce shifts in the tumor immune microenvironment, typically facilitating tumor immune evasion and contributing to therapeutic resistance. WHAT THIS STUDY ADDS * This research primarily elucidated the comparatively weaker chemotherapy efficacy in liver tumors of TNBC compared with the primary tumors. The weaker responsiveness in liver tumors was mediated by the impeded infiltration of CD8+T cells, attributed to the hypoactive macrophages. Augmenting macrophage activation can improve the chemotherapy efficacy in liver tumors. Chemotherapy drove the immune microenvironment of both sites toward increased suppression through distinct mechanisms. Site-specific interventions can augment chemotherapy effectiveness. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY * Our study reveals diminished efficacy of chemotherapy in TNBC liver tumors and uncovers the associated immune-mediated mechanisms within the tumor microenvironment. Precision targeting of immune suppression across distinct tumor locales may enhance therapeutic outcomes. Background Breast cancer is the most frequently diagnosed malignancy in women globally and a significant driver of cancer-related mortality, mainly due to distant metastasis.[56]1 2 The 5-year survival rate for those with metastasis varies from 21% to 32%.[57]1 Triple-negative breast cancer (TNBC), defined by the absence of estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2, is associated with a high risk of distant metastasis, with an estimated 46% of patients experiencing such progression.[58]3–5 Patients with TNBC exhibit significant responses to immunotherapy, attributed to its heightened immunogenicity relative to other subtypes of breast cancer.[59]6 Nonetheless, immunotherapy exhibits efficacy in only a subset of patients with TNBC.[60]7 Chemotherapy is still the principal therapeutic intervention for metastatic TNBC.[61]8 Standard treatments for patients with metastatic TNBC often involve single-agent chemotherapy with drugs like taxanes, platinum compounds, and anthracyclines, which primarily disrupt tumor cell proliferation.[62]9 Nonetheless, chemoresistance remains a pivotal factor driving mortality in patients with metastatic breast cancer.[63]10 Chemotherapy efficacy is likely associated with the genetic diversity of tumor cells and the unique features of the tumor microenvironment (TME).[64]11–14 The TME, a complex ecosystem consisting of cancer and immune cells, fibroblasts, endothelial cells, and non-cellular elements, is significantly influenced by the anatomical location of the tumor.[65]15 Preclinical studies have revealed that orthotopic tumors, such as those in the kidney, colon, and prostate, are less sensitive to immunotherapy than subcutaneously implanted ones, likely due to the immunosuppressive nature of the orthotopic TME.[66]16 The liver, known for its immune tolerance, is central to managing autoimmune diseases, viral infections, and transplantation.[67]17 It has been observed that breast cancer liver metastases exhibit a unique immunosuppressive microenvironment characterized by the reprogramming of immune cell populations.[68]18 Moreover, liver metastases correlate with systemic immune dysfunction, potentially reducing the effectiveness of immunotherapy.[69]19 20 However, whether the anatomical location of liver metastases relative to the primary tumors affects chemotherapy efficacy in TNBC remains unclear. The precise impact of the immune characteristics of TNBC liver metastases on chemosensitivity warrants further investigation. In this study, single-cell RNA sequencing (scRNA-seq) data from liver metastases and primary tumors of TNBC patients identify immune microenvironment disparities, which may significantly affect chemotherapy outcomes. Here, we report that liver tumors of TNBC exhibit a weaker chemotherapy efficacy relative to primary tumors. Mechanically, impeded CD8+T cell infiltration in the liver tumors reduces chemosensitivity, driven by less activated upstream macrophages. Enhancing macrophage activation can improve chemotherapy efficacy in liver tumors. Moreover, chemotherapy induces an immunosuppressive state in both liver tumors and primary tumors, but with different mechanisms. Anatomically tailored strategies to target the immunosuppressive TME may enhance clinical responses. Methods Mice Six-week-old female BALB/c mice of the wild-type strain were housed in a specific pathogen-free environment at Nanjing Medical University’s Animal Core Facility. The mice were subjected to a 12-hour light/dark cycle with ad libitum access to food and water. Random allocation was used to distribute the animals into experimental groups. Cell lines The 4T1 mammary carcinoma cell line, native to BALB/c mice, was sourced from the American Type Culture Collection (USA). The 4T1 cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM, Hyclone, USA) enriched with 100 units/mL penicillin, 100 µg/mL streptomycin (Hyclone, USA), and 10% (v/v) fetal bovine serum (FBS, Hyclone, USA). Incubation occurred at 37°C under a 5% CO2 atmosphere. To generate the 4T1-luc cell line, 4T1 cells were co-cultured with an appropriate concentration of lentivirus carrying the luciferase gene (GenePharma, Shanghai, China). Subsequent selection with puromycin at the optimal concentration for 10–15 days ensured the isolation of stably expressing 4T1-luc cells. Tumor models For subcutaneous tumor modeling, 4T1 cells (4×10^5 in 100 µL phosphate buffer saline, PBS) were implanted into the right fourth mammary fat pad of BALB/c mice. For experimental liver metastasis models, 4T1 cells were intrasplenically or intrahepatically injected as described. In mice in which intrasplenic injection (4×10^5 4T1 cells in 100 µL PBS) was used, the spleen was removed within 1 min after tumorous injection in the experimental groups and after PBS injection in the control groups. In experiments in which intrahepatic injection (2×10^5 4T1 cells in 20 µL PBS) was used, control mice underwent intrahepatic injection of PBS. Both sites were induced concurrently, and by day 15 postimplantation, subcutaneous tumors and tumors formed by subcapsular hepatic injection typically reached a volume of approximately 200 mm³. Mice were randomly allocated to experimental groups at this juncture. Tumor dimensions were measured every 3 days using calipers, applying the formula (π/6)×L×H to calculate volume, with L and H representing the longer and shorter diameters, respectively. Ethical guidelines dictated that mice were humanely euthanized via cervical dislocation once tumors reached a 1.5 cm diameter threshold. Bioluminescence imaging with the 4T1-luc cell line was used for in vivo tumor burden assessment (IVIS Spectrum, Perkin Elmer). The cross-planting model was developed through the following sterile procedure: Tumor tissues, subcutaneously and hepatically grown for 15 days, were excised, minced, and subjected to a digestion solution containing 100 mg collagenase, 5 mg DNase, and 5 mg hyaluronidase in 50 mL DMEM, enriched with 1 mL FBS. This mixture was incubated at 37°C for 45 min. Post-digestion, the suspension was filtered and centrifuged at 4°C to isolate the cells. The resulting subcutaneous and intrahepatic tumor cells were resuspended in complete DMEM and cultured in six-well plates. These cells were implanted orthotopically or ectopically to establish the cross-planting model upon reaching sufficient quantities. Mouse treatment For the chemotherapy regimen, each mouse received a dose of 10 mg/kg of paclitaxel, doxorubicin, or gemcitabine via intraperitoneal injection. The treatment was given five times per week for 2 weeks. The control group was given an equivalent volume of PBS. In the anti-CD8+PTX cohort, mice were administered an intraperitoneal injection of 200 μg of a mouse CD8 antagonist 1 day before commencing paclitaxel treatment. An additional 100 µg dose was administered every 3 days to ensure sustained depletion. Isotype controls were used to mitigate non-specific effects. Macrophage depletion was achieved on the same day as tumor inoculation via intravenous injection of 200 µL clodronate liposomes, administered before inoculation, followed by 150 µL booster injections every 3 days; control mice received PBS liposomes. Glufosinate was administered at a dosage of 10 mg/kg via gavage once daily for 14 days, with PBS serving as the control. Cl-amidine treatment involved daily intraperitoneal injections at 10 mg/kg dosage for 2 weeks, compared with PBS controls. Immunohistochemistry Paraffin sections of 4T1 subcutaneous tumors and liver tumors were stained with primary antibodies—anti-mouse CD8 antibody (CST, 98 941S), anti-mouse CD4 antibody (eBioscience, 4SM95), anti-mouse Ki67 antibody (Servicebio, [70]GB121141), or anti-mouse CD31 antibody (Servicebio, [71]GB120005). Subsequently, the tissue sections were incubated with horseradish peroxidase-conjugated goat anti-rat IgG secondary antibody (Santa Cruz Biotechnology, Santa Cruz, California, USA). This was followed by a chromogenic reaction with 3,3′-diaminobenzidine substrate (Beyotime, Nanjing, China) to develop the immunocomplexes. Wound healing assay Liver tumor and subcutaneous tumor cells were plated in six-well plates and cultured until confluent. A linear wound was introduced using a 200 µL pipette tip to disrupt the monolayer. Subsequently, the cells were maintained in a serum-free medium with DMSO for 24 hours to assess wound healing. The progression was tracked using an inverted microscope at 0 and 24 hours. Image J software was employed to quantify wound closure. All experiments were conducted thrice to confirm the reproducibility and reliability of the findings. Colony formation assay Liver tumor and subcutaneous tumor cells were plated in six-well plates and exposed to DMSO. Following a 10-day incubation, colonies were fixed using methanol and stained with a 0.1% crystal violet solution. Subsequent imaging and quantification of the colonies within each well were performed. To validate the findings, the experiment was independently replicated three times. Evans blue assay Mice with either subcutaneous tumors or liver tumors received an intraperitoneal injection of 200 µL of Evans Blue dye (30 mg/kg dissolved in PBS). Following a 30 min incubation period, the animals were humanely euthanized, after which the tumors were meticulously harvested and rinsed. To extract the dye, the tumors were incubated in dimethylformamide at 37°C overnight. A 100 µL aliquot of this solution was then transferred to a 96-well plate, with quintuplicate measurements per sample. The optical density at 620 nm was determined spectrophotometrically, employing a standard curve for dye concentration quantification. The concentration of Evans Blue within tumor tissue was determined by calculating the content per milligram of tissue based on absorbance measurements. Flow cytometry Under sterile conditions, excised subcutaneous and liver tumor tissues were minced and digested in a solution at 37°C for 45 min. Post-digestion, the suspension was filtered and centrifuged at 4°C, and the cell pellet was resuspended in PBS for flow cytometry. Mouse peripheral blood was processed with an erythrocyte-lysing reagent to eliminate red blood cells, yielding a single-cell suspension for cytometric analysis. Cells were first stained with the Zombie NIR Fixable Viability Kit (BioLegend, Cat# 423105) at 25°C for 15 min, followed by incubation with a CD16/32 blocking monoclonal antibody (Thermo Fisher Scientific, Cat# 14016181). For surface staining, the cells were incubated with fluorescence-conjugated antibodies in a flow cytometry staining buffer (BioLegend, USA) for 30 min at 4°C. The following fluorescence-conjugated antibodies (BioLegend and Thermo Fisher Scientific) were used: FITC anti-mouse CD45 Antibody (Biolegend, Cat# 103107), PerCP-Cyanine5.5 CD11b Monoclonal Antibody (Thermo Fisher Scientific, Cat# 45011280), PE F4/80 Monoclonal Antibody (Thermo Fisher Scientific, Cat# 12480180), PE-Cyanine7 CD86 (B7-2) Monoclonal Antibody (Thermo Fisher Scientific, Cat# 25086280), APC CD206 (MMR) Monoclonal Antibody (Thermo Fisher Scientific, Cat# 17206180), PE/Cyanine7 anti-mouse CD3 Antibody (Biolegend, Cat# 100219), Alexa Fluor 700 anti-mouse CD8a Antibody (Biolegend, Cat# 100729), PerCP anti-mouse CD4 Antibody (Biolegend, Cat# 100537), PE anti-mouse/human CD44 Antibody (Biolegend, Cat# 103007), Brilliant Violet 605 anti-mouse CD62L Antibody (Biolegend, Cat# 104438), PE/Cyanine7 anti-mouse I-A/I-E Antibody (Biolegend, Cat# 107629), APC CD4 Monoclonal Antibody (Thermo Fisher Scientific, Cat#17004182), PE-Cyanine7 CD45 Monoclonal Antibody (Thermo Fisher Scientific, Cat# 25045182), FITC CD3e Monoclonal Antibody (Thermo Fisher Scientific, Cat# 11003182), PE CD8 Monoclonal Antibody (Thermo Fisher Scientific, Cat# MA1145). Cell data were acquired using a BD FACSCanto cytometer (BD Biosciences) and analyzed with FlowJo V.10.8.1 software. Single-cell data analysis To compare the differences in the immune microenvironment between TNBC primary lesions and liver metastases before treatment, we integrated three single-cell sequencing datasets, comprising 5 TNBC primary lesions and 2 liver metastases from the GEO database (accession number [72]GSE169246),[73]21 12 TNBC primary tumors from the EGA database (accession number EGAS00001004809),[74]22 and 2 TNBC liver metastasis samples from the GEO database (accession number [75]GSE249361).[76]23 To compare the differences in the immune microenvironment of TNBC primary lesions before and after treatment, single-cell sequencing data from primary tumor samples of five TNBC patients before and after paclitaxel treatment were retrieved from the GEO database (accession number [77]GSE169246).[78]21 This analysis involved samples from twenty-one TNBC patients, with their pertinent clinical characteristics and biopsied lesions detailed in [79]online supplemental table 1,2. The Seurat package (V.5.0.1) in R was employed to analyze the scRNA-seq data. The sctransform function was employed to preprocess the scRNA-seq data. Cells were excluded if they exhibited over 10% mitochondrial genes, had more than 8000 genes, or fewer than 200 genes. Gene expression normalization was achieved based on the raw UMI counts after normalizing total counts per cell (library size), which were then scaled by 1e6 and logarithmically transformed. Cell clusters were identified by examining the top 15 principal components across 4000 highly variable genes identified by the FindVariableFeatures function. The t-distributed stochastic neighbor embedding algorithm was used to present single-cell data. Differential gene expression analysis within cell clusters was performed using the COSine similarity-based marker gene identification function.[80]24 Supplementary data [81]jitc-13-3-s001.pdf^ (1.1MB, pdf) Evaluation of cellular functions based on single-cell data Gene Ontology (GO) analysis was performed using the clusterProfiler R package (V.4.8.3). The AddModuleScore function and single-sample Gene Set Enrichment Analysis (ssGSEA) performed by the GSVA R package (V.1.48.3) were employed to score individual cells for each gene signature. Gene sets from the MsigDB database ([82]http:// www.gsea-msigdb.org/) were used for enrichment analysis. Inference and analysis of cell-cell communication The R package CellChat (V.2.1.2) was used to quantitatively infer and analyze intercellular communication networks based on our scRNAseq data.[83]25 By leveraging network analysis and pattern recognition methods, CellChat predicts key signaling inputs and outputs for cells, shedding light on how cells and signals coordinate various functions. CellChat models the probability of cell–cell communication by integrating gene expression data with its database of known interactions between signaling ligands, receptors, and cofactors. Differences between groups were analyzed using Mann-Whitney U tests. RNA-seq analysis Total RNA was isolated from frozen tumor cells using Trizol reagent (Thermo Fisher, USA). cDNA library construction and sequencing were performed using the BGISEQ-500 platform (BGI Genomics, Shenzhen, China). The reads were then mapped to the Mus musculus genome using HISAT2. StringTie ([84]https://ccb.jhu.edu/software/stringtie) was employed to assemble the mapped reads for each sample. To estimate the expression levels of all transcripts, StringTie and Ballgown ([85]http://www.bioconductor.org/packages/release/bioc/html/ballgown.ht ml) were used following the generation of the final transcriptome. The expression abundance of mRNAs was determined by calculating the TPM (fragments per kilobase of transcript per million mapped reads) value using StringTie and Ballgown. Differentially expressed genes (DEGs) from RNA-seq data were identified using the R package limma (V.3.56.2). DEGs were selected based on a log fold change greater than 1 or less than −1, with adjusted p<0.05 using Bonferroni correction. These DEGs were subsequently subjected to enrichment analysis for GO and Kyoto Encyclopedia of Genes and Genomes (KEGG) functions using the R package clusterProfiler (V.4.8.3). The GSVA R package (V.1.48.3) was used to perform ssGSEA for each sample, which was applied to explore the different infiltration degrees of immune cell types, immune-related functions, and immune-related pathways based on the expression profile. Gene sets from the MsigDB database ([86]http:// www.gsea-msigdb.org/) were used for enrichment analysis. Statistical analysis Frequency tabulation and summary statistics were used to characterize the data distribution. Continuous variables are described by the means±SDs or medians and IQRs according to their distribution. Categorical variables were described using proportions and frequencies. Group comparisons were performed using the χ^2 or Fisher’s exact test for categorical variables. We incorporated a two-tailed unpaired t-test or the Mann-Whitney Wilcoxon test for continuous variables. Kaplan-Meier survival analysis was performed using the log-rank test. The p values were adjusted to the false discovery rate using the Benjamini-Hochberg procedure in multiple comparisons. A p<0.05 was considered to indicate statistical significance (ns, p≥0.05; *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001). Statistical analysis was conducted using R software. The results were visualized using the ggplot2 R package (V.3.4.4) and the pheatmap R package (V.1.0.12). Results Different immune microenvironments in liver metastases from primary tumors in TNBC To examine immune microenvironment differences between liver metastases and primary tumors in TNBC, scRNA-seq data from four TNBC liver metastasis samples and 17 treatment-naive TNBC primary tumor samples were integrated and analyzed ([87]figure 1a; [88]online supplemental table 1).[89]21–23 Following the exclusion of low-quality cells and adjustment for RNA dropouts, 53,050 immune cells were analyzed. Figure 1. [90]Figure 1 [91]Open in a new tab Different immune microenvironments in liver metastases from primary tumors in TNBC (a) Schematic overview of (b–f). Single cell sequencing data from 17 TNBC primary tumors and 4 liver metastases were collected and integrated for analysis. (b) T-distributed stochastic neighbor embedding (t-SNE) plots of immune cells colored by cell cluster. (c) Bar plots showing the distribution of immune cell proportions. (d) T-SNE plots of T cells colored by cell cluster. (e) Bar plots showing the distribution of T cell proportions. (f) Bar plots showing the mean scores of GO pathways enriched in upregulated genes in CD8+T cells. Significance determined by χ^2 test or Fisher’s exact test (c, e). Significance was determined as p<0.05. ns, p≥0.05; ***p<0.001, ****p<0.0001. Parts of (a) were drawn using pictures from Servier Medical Art. Servier Medical Art by Servier is licensed under a Creative Commons Attribution 3.0 Unported License ([92]https://creativecommons.org/licenses/by/3.0/). DC, dendritic cell; TNBC, triple-negative breast cancer. Bioinformatics analysis identified six unique immune cell subtypes: T cells, B cells, innate lymphoid cells (ILCs), dendritic cells (DCs), neutrophils, and macrophages ([93]figure 1b; [94]online supplemental fig 1a). Significant differences in immune cell infiltration between liver metastases and primary tumors were observed. Metastatic samples had a markedly lower T cell, DC, neutrophil, and B cell presence but showed increased ILC and macrophage infiltration compared with primary tumors ([95]figure 1c). Given the critical roles of T cells in antitumor immunity, 11 T cell clusters were identified ([96]figure 1d; [97]online supplemental figure 1b,c). During T cell clustering, functional states were considered, distinguishing naive (Tn) and proliferating (Tprf) clusters that consist of both CD8+and CD4+ T cells. Furthermore, several canonical clusters of CD8+or CD4+ T cells were delineated, encompassing regulatory T cells (Treg), central memory T cells (Tcm), effector T cells (Teff), effector memory T cells (Tem), tissue-resident memory T cells (Trm), and mucosal-associated invariant T (MAIT) cell subsets ([98]figure 1d; [99]online supplemental figure 1b). In contrast to the primary tumors, liver metastases exhibited an increased presence of specific T cell subsets, including CD4-CXCL13, Tprf-MKI67, CD8_Trm-ZNF683, and CD8_MAIT-KLRB1 ([100]figure 1e). Conversely, at the primary tumor site, there was an increased prevalence of T cell subsets such as CD8-CXCL13, Tn-IL7R, CD8_Teff-GNLY, CD8_Tem-GZMK, and Tn-LEF1 ([101]figure 1e). The CD8-CXCL13 subset, which has been reported as a tumor-reactive CD8 subset, represented a significant proportion of T cell subsets in primary tumors[102]21 26 ([103]figure 1e). For ILCs, five clusters were identified ([104]online supplemental figure 1d). The primary lesion was mainly enriched with the ILC1−IL32, while liver metastases were enriched with ILC1−CCL20 and ILC2−XCL1 ([105]online supplemental figure 1e). To delineate the comprehensive differences in CD8+T cells, CD4+T cells, and ILCs between liver metastases and primary tumors, GO enrichment analysis was conducted. This analysis revealed that CD8+T cells and CD4+T cells in primary tumors, prior to chemotherapy, were enriched in pathways related to immune activation, such as positive regulation of leukocyte cell-cell adhesion and positive regulation of T cell activation ([106]figure 1f; [107]online supplemental figure 1f). In contrast, ILCs in liver metastases demonstrated enhanced immune function relative to the primary tumor site, with GO enrichment analysis showing involvement in key pathways including leukocyte cell–cell adhesion, leukocyte-mediated cytotoxicity, and cell killing ([108]online supplemental figure 1g). Considering B cells an important component of the immune microenvironment, further clustering and subgroup analysis of B cells revealed enrichment of Bfoc−MKI67, Bmem−IFIT3, and Bn−CCR7 subpopulations in the primary tumors, while the pB−IGHG1 subpopulation was enriched in liver metastases ([109]online supplemental figure 2a,b). GO pathway enrichment analysis revealed that the primary tumor was mainly characterized by antigen presentation-related functions, as evidenced by the enrichment of pathways such as MHC protein complex assembly and MHC class II protein complex assembly ([110]online supplemental figure 2c). In contrast, liver metastases showed enrichment in protein synthesis-related pathways and B cell receptor signaling pathways ([111]Online supplemental figure 2c). These results highlight the immune microenvironment heterogeneity between liver metastases and primary tumors in TNBC, where liver metastases are marked by heightened innate immune activation and primary tumors by a more vigorous adaptive immune response. Weaker chemotherapeutic efficacy in liver tumors compared with subcutaneous tumors in preclinical models The diversity in the immune microenvironment may critically influence chemotherapy outcomes.[112]27 Murine models harboring 4T1 liver tumors or subcutaneous tumors, simulating liver metastases and primary tumors of TNBC, were established to assess the effects of chemotherapy. TNBC experimental liver metastasis models were constructed by intrasplenically injecting 4T1 cells ([113]figure 2a). Compared with the primary tumors, which exhibited a size reduction after 14 days of paclitaxel treatment as assessed by in vivo bioluminescence imaging, liver tumors in mice showed no significant change in tumor burden pre-treatment and post-treatment ([114]figure 2b). Mouse survival and HE staining further confirmed reduced chemotherapy efficacy in liver tumors ([115]figure 2c; [116]online supplemental figure 3a.b). Additionally, the metastatic model demonstrated lower CD8+T cell infiltration in the TME prior to treatment compared with the primary tumor, consistent with our single-cell analysis results ([117]figures 1c, 2d). Figure 2. [118]Figure 2 [119]Open in a new tab Weaker chemotherapeutic efficacy in liver tumors compared with subcutaneous tumors in preclinical models (a) Schematic overview of (b–d). The mice bearing 4T1 liver tumors (Liver) induced by intrasplenic inoculation or subcutaneous tumors (SC) were treated with PTX from day 1 to day 14. The control group was given an equivalent volume of PBS. (b) Representative bioluminescent imaging of mice in (a) on day 15, with n=3 for visualization and quantification. (c) Survival curves of mice in (a), monitored until the time of death, with n=6. (d) Quantification of CD8+T cells in Liver and SC on day 0, with n=4. (e) Schematic overview of (f–i). The mice bearing 4T1 liver tumors (Liver) induced by subcapsular hepatic inoculation or SC were treated with PTX from day 1 to day 14. The control group was given an equivalent volume of PBS. (f) Tumor volume curves of mice in (e), with n=6. (g) Representative images of tumors harvested from mice in (e) on day 15, with n=6. (h) Survival curves of mice in (e), monitored until the time of death, with n=6. (i) Representative bioluminescent imaging of mice in (e) on day 15, with n=3 for visualization and quantification. The experiments were independently replicated on multiple occasions. Data are mean±SE (b, d, f, i). Significance determined by two-tailed unpaired t-test (b, d, f, i) and log-rank test (c, h). Significance was determined as p<0.05. ns, p≥0.05; *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. Parts of (a, e) were drawn using pictures from Servier Medical Art. However, the chemotherapy efficacy of splenic inoculation-induced liver metastases may be influenced by multiple factors, including tumor heterogeneity and anatomical variations in the immune microenvironment. Ki-67 expression, a marker for tumor proliferation, was assessed using immunohistochemistry (IHC). The higher expression of Ki67 in liver tumors suggests a stronger proliferative capacity before treatment ([120]online supplemental figure 3c,d). Blood vessel density, crucial for drug delivery and treatment response, was examined in liver tumors and subcutaneous tumors using CD31 antigen staining before chemotherapy.[121]28 The evaluation revealed a higher expression of CD31 in liver tumors ([122]online supplemental figure 3c,e). These results suggest that splenic injection-induced experimental liver metastases may exhibit heterogeneity compared with the primary tumors. Although tumor heterogeneity’s impact on chemotherapy efficacy has been extensively researched,[123]29 30 the role of anatomical differences in the immune microenvironment remains under-reported. Furthermore, the liver metastasis model induced by splenic inoculation exhibits multifocal growth ([124]online supplemental figure 3a,b), which is not conducive to a direct comparison of the chemotherapy efficacy between primary tumors and liver tumors. Consequently, we opted for subcapsular hepatic injection of 4T1 breast cancer cells to construct liver tumors.[125]19 20 When liver tumors and subcutaneous tumors reached similar sizes on the 15th day postimplantation, paclitaxel treatment was initiated ([126]figure 2e). Liver tumors exhibited weaker responsiveness to paclitaxel treatment compared with subcutaneous tumors, with higher tumor volumes observed ([127]figure 2f,g). Mice with liver tumors also had significantly lower survival rates than those with subcutaneous tumors ([128]figure 2h). Consistently, in vivo bioluminescence imaging confirmed a more significant tumor burden in mice with liver tumors following chemotherapy compared with those with subcutaneous tumors ([129]figure 2i). These findings indicate that the efficacy of paclitaxel is comparatively weaker in liver tumors compared with subcutaneous tumors. Additionally, tumor heterogeneity was assessed in subcutaneous tumors and liver tumors induced by subcapsular hepatic inoculation before paclitaxel treatment, with no statistically significant difference found in Ki-67 and CD31 expression between these two types of tumors ([130]Online supplemental figure 4a,b,c). Wound healing assays indicated equivalent migratory capacities between the two tumor cell types before chemotherapy ([131]online supplemental figure 4d). Consistently, colony formation assays demonstrated that the proliferative and colony-forming capabilities of liver tumor cells were comparable to those of subcutaneous tumor cells before chemotherapy ([132]online supplemental figure 4e). Moreover, the Evans Blue assay, a standard method for measuring drug permeation, was employed to assess drug diffusion in both sites. The assay showed no substantial difference in the diffusion of Evans Blue dye, suggesting similar chemotherapy drug distribution in liver tumors and subcutaneous tumors before treatment ([133]online supplemental figure 4f). To further explore the influence of cell line genotype on drug response, animal models with reinjected liver tumor and subcutaneous tumor cells into identical or contrasting anatomical locations were established ([134]online supplemental figure 4g). Regardless of whether the tumor cells originated from the hepatic or subcutaneous tissue, the previously observed inferior therapeutic efficacy in the liver metastases compared with subcutaneous tumors persisted ([135]online supplemental figure 4h). Moreover, within the same treatment groups, no statistically significant differences in efficacy were observed between tumors of different origins, indicating that the efficacy of chemotherapy was related to the site of implantation rather than the genotype of the implanted cells ([136]online supplemental figure 4h). Therefore, the liver tumor model induced by subcapsular injection of tumor cells can exclude the influence of tumor heterogeneity on the efficacy of chemotherapy, allowing for a more focused investigation on how anatomical differences in the immune microenvironment impact chemotherapy efficacy. For subsequent experiments, the liver tumor model induced by subcapsular injection of tumor cells was used. To assess the therapeutic effects of alternative chemotherapeutic agents on tumor models, doxorubicin (DOX) or gemcitabine (GEM) was administered when liver tumors and subcutaneous tumors were of similar size on the 15th day postimplantation ([137]online supplemental figure 5a). Liver tumors showed weaker responsiveness to these drugs compared with subcutaneous tumors ([138]online supplemental figure5b). While both drugs improved survival rates in tumor-bearing mice compared with the control group, the survival benefit in mice with liver tumors was significantly less than that observed in mice with subcutaneous tumors ([139]online supplemental 5c,d). These results suggest that chemotherapy appears to be less effective in liver tumors compared with subcutaneous tumors, warranting further investigation into the underlying reasons. The differences in chemotherapeutic efficacy were mediated by CD8+ T cells Excluding the intrinsic properties of tumors, the differences in the immune microenvironment between liver tumors and primary tumors before chemotherapy may contribute to the variability in chemotherapy responses. To substantiate these differences in the immune microenvironment, RNA sequencing (RNA-seq) was performed to analyze pre-chemotherapy gene expression profiles in mouse models of liver tumors and subcutaneous tumors ([140]figure 3a). GO analysis revealed that the predominant upregulated gene categories in subcutaneous tumors before chemotherapy were related to T cell activation, lymphocyte differentiation, and immune response pathways ([141]figure 3b). The findings imply a reduced activation of the adaptive immune system in liver tumors, highlighting the need to focus more on adaptive immunity rather than innate response. IHC was conducted to evaluate T cell infiltration in mouse tumors before paclitaxel treatment, revealing a reduced presence of CD4+and CD8+ T cells in liver tumors compared with subcutaneous tumors ([142]figure 3c). Flow cytometry confirmed these findings, indicating reduced levels of both CD4+and CD8+ T cells in liver tumors before chemotherapy ([143]figure 3d). Given the potential impact of systemic immune status on local tumor infiltration, peripheral blood T cells were also quantified by flow cytometry. No significant differences in peripheral blood CD8+and CD4+ T cell counts were observed among mice with tumors from different sites before treatment ([144]figure 3e), indicating that the disparity in T cell infiltration is tumor-specific rather than a reflection of systemic immune status. Figure 3. [145]Figure 3 [146]Open in a new tab The differences in chemotherapeutic efficacy were mediated by CD8+T cells (a) Schematic overview of (b–e). 4T1 subcutaneous tumors (SC), liver tumors (Liver) induced by subcapsular hepatic inoculation, and peripheral blood from the mouse models were collected before paclitaxel treatment for further analysis, including RNA-seq, IHC, and flow cytometry analysis. (b) Bar plots showing the mean scores of GO pathways enriched in differentially expressed genes between Liver (n=5) and SC (n=3). (c) Representative IHC staining for CD8 and CD4 in Liver and SC, with quantitative analysis, n=3. The scale bar represents 30 µm. (d) Representative flow cytometry plots and quantification of CD8+T and CD4+T cells in Liver and SC, with n=6. (e) Representative flow cytometry plots and quantification of CD8+T and CD4+T cells in peripheral blood, with n=6. (f) Schematic overview of (g, h). Mice bearing SC and Liver were intraperitoneally injected with aCD8 or isotype control from day 0 to day 15, with paclitaxel administration from day 1 to day 14. (g) Representative images and quantification of tumor burden from mice in (f) on day 15, with n=4. (h) Survival curves of mice in (f), monitored until the time of death, with n=6. The experiments were independently replicated on multiple occasions. Data are mean±SE (c, d, e, g). Significance determined by two-tailed unpaired t-test (c, d, e, g) and log-rank test (h). Significance was determined as p<0.05. ns, p≥0.05; *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. Parts of (a, f) were drawn using pictures from Servier Medical Art. GO, Gene Ontology; IHC, immunohistochemistry. To determine whether the lack of CD8+T cells in liver tumors contributes to reduced potency of chemotherapy, CD8+T cells were systemically depleted in mice using a CD8α monoclonal antibody (aCD8), beginning 1 day before paclitaxel treatment and continuing for the entire experimental period ([147]figure 3f). Following this treatment, no significant difference in tumor size was observed between liver tumors and subcutaneous tumors ([148]figure 3g). Additionally, the survival disparity between mice with tumors from different sites disappeared following CD8+T cell depletion ([149]figure 3h). These results suggest that the variance in chemotherapy response between liver tumors and subcutaneous tumors is primarily attributed to the presence of intratumoral CD8+T cells. Hypoactive macrophages impede CD8+ T cell infiltration in liver metastases To identify the regulatory mechanisms that modulate CD8+T cell infiltration in liver metastases and primary tumors pre-chemotherapy, the ‘CellChat’ (V.2.1.2) package was used to analyze cell-to-cell communication in the human scRNAseq dataset ([150]figure 1a).[151]25 Among the diverse immune cell subsets interacting with CD8+T cells, macrophages displayed the most significant variation in their interactions between tumors from different sites ([152]figure 4a). GO analysis was conducted to explore the biological processes associated with macrophages in liver metastases and primary tumors. The analysis indicated that macrophages in primary tumors were more enriched in immune response processes, such as response to type II interferon, myeloid leukocyte migration, and cytokine-mediated signaling pathways ([153]figure 4b). Further analysis of ligand-receptor interactions between macrophages and CD8+T cells revealed a decrease in antigen presentation-related interactions in the liver metastases, suggesting a reduced immune-activating capacity of macrophages in liver metastases ([154]online supplemental figure 6a). GSEA on these macrophages revealed suppressed antigen processing and presentation function in liver metastases ([155]online supplemental figure 6b). Macrophage populations were categorized into seven clusters, with liver metastases mainly harboring the Macro-CCL20 subset ([156]online supplemental figure 6c,d,e). GSEA indicated that this subset exhibited reduced activity in pathways associated with APC co-stimulation, IFN gamma response, and phagocytosis, consistent with the generally subdued activation status of liver macrophages ([157]online supplemental figure 6f). To further confirm the distinct roles of macrophages at different sites, RNA-seq in mouse models was conducted and revealed a reduced antigen presentation score in liver tumors before treatment ([158]figure 4 c,d). Flow cytometry analysis of macrophages confirmed that the expression of CD86 and MHC-II, key to antigen presentation, was reduced in liver tumors compared with subcutaneous tumors before treatment ([159]figure 4e). This suggests that macrophages in liver tumors exhibited a reduced capacity for antigen presentation and co-stimulation. Figure 4. Figure 4 [160]Open in a new tab Hypoactive macrophages impede CD8+T cell infiltration in liver metastases. (a) Heatmap of cell-to-cell communication among various cell subpopulations in TNBC liver metastases and primary tumors, derived from human scRNA-seq. (b) Bar plots showing the mean scores of GO pathways enriched in differentially expressed genes in macrophages, derived from human scRNA-seq. (c) Schematic overview of (d-, e). 4T1 liver tumors (Liver) induced by subcapsular hepatic inoculation and subcutaneous tumors (SC) from mouse models were collected before paclitaxel treatment for further analysis, including RNA-seq and flow cytometry analysis. (d) Box plots of gene set enrichment scores for antigen processing and presentation in SC (n=3) and Liver (n=5), calculated using ssGSEA. (e) Representative flow cytometry plots and quantification of MHC-II+ (n=3) and CD86+ (n=5) macrophages. (f) Schematic overview of (g-–j). Mice bearing LliverlLiver and SC were intraperitoneally injected with clodronate (Clo) or PBS liposomes from day −15 to day 15, with paclitaxel administration from day 1 to day 14. (g) Representative images and quantification of tumor burden from mice in (f) on day 15, with n=4. (h) Survival curves of mice in (f), monitored until the time of death, with n=6. (i) Representative IHC staining for CD8 on day 0, with quantitative analysis, n=3. ScaleThe scale bar represents 30 μm µm. (j) Representative flow cytometry plots and quantification of CD8+T cells on day 0, with n=5. (k) Schematic overview of (l). Mice bearing LiverlLiver were treated with glufosinate (Glu) or PBS via gavage from day −14 to day −1, with paclitaxel administration from day 1 to day 14. (l) Representative images and quantification of tumor burden from mice in (k) on day 15, with n=5. The experiments were independently replicated on multiple occasions. Data are mean±standard errorSE (d, e, g, i, j, l). Significance determined by two-tailed unpaired t-test (d, e, g, i, j, l) and log-rank test (h). Significance was determined as Pp<0.05. ns, pp≥0.05; *, pp<0.05; **, pp<0.01; ***, pp<0.001; ****, pp<0.0001*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. Parts of (c, f, k) were drawn using pictures from Servier Medical Art. GO, Gene Ontology; IHC, immunohistochemistry; ssGSEA, single-sample gene set enrichment analysis; TNBC, triple-negative breast cancer. To investigate the impact of macrophage activation on CD8+T cell infiltration and the efficacy of chemotherapy in liver tumors vs subcutaneous tumors, clodronate liposomes to systemically deplete macrophages in animal models were employed ([161]figure 4f). This depletion, initiated before tumor inoculation and maintained throughout the experimental period, eliminated the differential therapeutic responses to paclitaxel between liver tumors and subcutaneous tumors on the 15th day post-treatment ([162]figure 4g). Additionally, the targeting of macrophages using clodronate liposomes abrogated the survival differences between mice bearing liver tumors and those with subcutaneous tumors after paclitaxel treatment ([163]figure 4h). The findings indicated that macrophages were implicated in the modulation of chemotherapeutic efficacy. Given the impact of CD8+T cell infiltration on chemotherapeutic efficacy, CD8+T cell levels in tumors were assessed following macrophage depletion before paclitaxel treatment. No significant difference in CD8+T cell infiltration was detected between liver tumors and subcutaneous tumors by IHC ([164]figure 4i). Flow cytometry corroborated these results, indicating equivalent CD8+T cell levels in both sites following macrophage depletion ([165]figure 4j). This suggests that the reduced activity of macrophages in liver tumors leads to decreased CD8+T cell infiltration. Glufosinate, a glutamine synthetase inhibitor, has demonstrated the ability to reprogram macrophages toward an active antitumor phenotype, potentially inhibiting metastasis.[166]31 To explore the possibility that macrophage activation in the pre-treatment liver TME could enhance the effectiveness of chemotherapy, mice with liver tumors were pre-treated with glufosinate or PBS before undergoing paclitaxel treatment ([167]figure 4k). Mice administered with glufosinate demonstrated a significant reduction in liver tumor volume by day 15 following chemotherapy compared with the control group ([168]figure 4l). In summary, the findings indicate that macrophages with reduced activity in liver metastases impede the infiltration of CD8+T cells, representing a potential therapeutic target for improving the efficacy of chemotherapy in liver metastases of TNBC. Chemotherapy induces an immunosuppressive TME in primary tumors Chemotherapy drugs have been reported to induce shifts in the tumor immune microenvironment, typically resulting in a state of immunosuppression that facilitates tumor immune evasion and contributes to therapeutic resistance.[169]32 33 In pursuit of strategies to augment chemotherapeutic efficacy, a comprehensive examination of the microenvironmental changes induced by chemotherapeutic agents was conducted. Considering the pre-existing infiltration and activation of macrophages and CD8+T cells in primary tumors, subsequent analyses were conducted to assess the alterations in these cellular subsets post-treatment. The scRNA-seq dataset, sourced from the GEO database, included primary tumors from five patients with advanced TNBC before and after paclitaxel monotherapy[170]21 ([171]figure 5a; [172]online supplemental table 2). After stringent quality control, 50,418 immune cells from ten samples were analyzed. Six immune cell clusters were identified, including T cells, B cells, ILCs, DCs, neutrophils, and macrophages ([173]online supplemental figure 7a,b). The expression of HLA-A, HLA-B, and HLA-C within macrophages decreased after treatment, indicative of impaired antigen presentation capabilities ([174]figure 5b). Concurrently, there was a reduction in the levels of inflammatory markers TNF, CD86, and CXCL9 following paclitaxel treatment ([175]figure 5b). These observations collectively suggest a suppression of the macrophage-mediated antitumor immune response following chemotherapy. GSEA on CD8+T cells revealed impaired activation and inflammatory function post-treatment ([176]figure 5c). Additionally, 10 T cell clusters were distinguished ([177]online supplemental figure 7c,d). The CD8-CXCL13 subset, known for its tumor reactivity, experienced a substantial decrease post-treatment[178]21 26 ([179]online supplemental figure 7e). These changes suggest a compromise in T cell functionality following treatment. RNA-seq of subcutaneous tumors in mouse models before and after paclitaxel treatment provided further insights into the dynamics of the tumor immune microenvironment ([180]figure 5d). GSEA further indicated downregulation of pathways related to pro-inflammatory macrophage signaling, T cell cytotoxicity, costimulatory signals, and cytokine receptor functions after treatment ([181]figure 5e). Conversely, pathways associated with the anti-inflammatory signaling of macrophages showed no significant change ([182]figure 5e). These findings imply that chemotherapy induces an immunosuppressive microenvironment in the primary site, marked by reduced functionality in both macrophages and T cells. Figure 5. Figure 5 [183]Open in a new tab Chemotherapy induces an immunosuppressive tumor microenvironment in TNBC primary tumors (a) Schematic overview of (b, c). Single-cell sequencing data from primary tumor samples of five TNBC patients before and after paclitaxel treatment were collected and analyzed. (b) Violin plots of gene expression on immune-related genes of macrophages. (c) Violin plots of activation and inflammation scores of CD8+T cells, calculated using ssGSEA. (d) Schematic overview of (e–g). 4T1 subcutaneous tumors in mouse models were collected before and after paclitaxel treatment for further analysis, including RNA-seq and flow cytometry analysis. (e) Heatmap of gene set enrichment scores, with n=3, calculated using ssGSEA. (f) Representative flow cytometry plots and quantification of CD86+ and CD206+ macrophages, with n=6. (g) Representative flow cytometry plots and quantification of CD8+ and CD44+CD62L− CD8+T cells, with n=4. (h) Schematic overview of (i–l). Mice bearing subcutaneous tumors were treated with paclitaxel and glufosinate (Glu) from day 1 to day 14. The control group was given an equivalent volume of PBS. (i) Representative images and quantification of tumor burden from mice in (h) on day 15, with n=4. (j) Survival curves of mice in (h), monitored until the time of death, with n=6. (k) Representative flow cytometry plots and quantification of CD86+ and CD206+ macrophages, with n=4. (l) Representative flow cytometry plots and quantification of CD8+and CD44+CD62L- CD8+T cells, with n=4. The experiments were independently replicated on multiple occasions. Data are mean±SE (f, g, i, k, l). Significance determined by two-tailed unpaired t-test (e, f, g, i, k, l), Wilcoxon test (b, c), and log-rank test (j). Significance was determined as p<0.05. ns, p≥0.05; *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. Parts of (a, d, h) were drawn using pictures from Servier Medical Art. ssGSEA, single-sample gene set enrichment analysis; TNBC, triple-negative breast cancer. To validate the findings, flow cytometry was used to evaluate the functional phenotype of macrophages and CD8+T cells in subcutaneous tumors before and after paclitaxel treatment. Analysis of post-treatment samples revealed a marked reduction in the frequency of pro-inflammatory macrophages characterized by CD86 expression, alongside an elevation in the presence of anti-inflammatory macrophages marked by CD206 ([184]figure 5f). This indicates a phenotypic transition induced by paclitaxel, favoring the emergence of immunosuppressive macrophages. Additionally, paclitaxel treatment led to a notable reduction in CD8+T cell infiltration and the frequency of CD44+CD62L- effector memory CD8+T cells ([185]figure 5g). These flow cytometric results align with the immunosuppressive microenvironment indicated by sequencing data. Given the crucial role of macrophages in regulating CD8+T cell infiltration, combination therapies that activate antitumor macrophages may augment the efficacy of chemotherapy. In mouse models with 4T1 subcutaneous tumors, a combined treatment of paclitaxel and glufosinate significantly reduced tumor volume compared with paclitaxel alone ([186]figure 5h,i). While glufosinate monotherapy did not improve survival, the combination with paclitaxel markedly enhanced the survival outcomes compared with paclitaxel monotherapy ([187]figure 5j). Flow cytometry confirmed that, compared with paclitaxel monotherapy, the combination treatment induced a shift in macrophages to a pro-inflammatory phenotype from the anti-inflammatory phenotype accompanied by increased infiltration of CD8+T cells and a higher frequency of CD44+CD62L- effector memory CD8+T cells ([188]figure 5k,l). These findings indicate that targeting dysfunctional macrophages induced by chemotherapy could be a promising strategy to enhance the effectiveness of paclitaxel treatment for TNBC primary tumors. Chemotherapy induces an immunosuppressive TME in liver tumors To explore the changes in the immune microenvironment of liver tumors before and after chemotherapy, RNA-seq on liver tumors in mouse models was conducted ([189]figure 6a). GSEA showed no significant shifts in CD8+T cell infiltration, T cell activation, or macrophage inflammatory functions post-treatment ([190]figure 6b,c). However, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis pointed to an upregulation of pathways involved in the formation of neutrophil extracellular traps (NETs), with genes like MPO, CAMP, NCF1, ELANE, and CTSG exhibiting increased expression in treated liver tumors ([191]figure 6d,e). Further GSEA substantiated the upregulation of NETs-related pathways following paclitaxel treatment in liver tumors ([192]figure 6f). Nonetheless, no significant difference in NETs scores was found in pre-treatment liver metastases and primary lesions, as well as primary tumors before and after chemotherapy ([193]online supplemental figure 8a,b). Previous research has indicated that NETs can exert immunosuppressive effects, impeding tumor immunotherapy and facilitating tumor metastasis.[194]34 This implies that the NET formation following chemotherapy in liver tumors may further dampen the immune microenvironment. Combining paclitaxel with a NET inhibitor could boost the chemotherapeutic efficacy against liver tumors. Cl-amidine, a broad-spectrum peptidyl arginine deiminase four inhibitor, can prevent NET formation by blocking neutrophil histone citrullination.[195]35 36 In mouse models, the co-administration of Cl-amidine with paclitaxel significantly reduced the size of liver tumors compared with paclitaxel alone by the 15th day post-chemotherapy ([196]figure 6g,h). While Cl-amidine alone did not enhance mouse survival, its combination with paclitaxel significantly improved survival outcomes over paclitaxel monotherapy ([197]figure 6i). These findings indicate that inhibiting the NETs formation could potentiate the chemotherapeutic efficacy in TNBC liver tumors. Figure 6. [198]Figure 6 [199]Open in a new tab Chemotherapy induces an immunosuppressive tumor microenvironment in TNBC liver tumors (a) Schematic overview of (b–f). 4T1 liver tumors induced by subcapsular hepatic inoculation in mouse models were collected before and after paclitaxel treatment for RNA-seq. (b) Violin plots of gene set enrichment scores for CD8 gene signature, with n=5, calculated using ssGSEA. (c) Heatmap of gene set enrichment scores, with n=5, calculated using ssGSEA. (d) Bar plots showing the mean scores of KEGG pathways enriched in differentially expressed genes, with n=5. (e) Volcano plots of differentially expressed genes, with n=5. (f) Violin plots of gene set enrichment scores for NETs, with n=5, calculated using ssGSEA. (g) Schematic overview of (h, i). Mice bearing 4T1 liver tumors induced by subcapsular hepatic inoculation were intraperitoneally injected with paclitaxel and Cl-amidine from day 1 to day 14. The control group was given an equivalent volume of PBS. (h) Representative images and quantification of tumor burden from mice in (g) on day 15, with n=4. (i) Survival curves of mice in (g), monitored until the time of death, with n=6. The experiments were independently replicated on multiple occasions. Data are mean±SE. (h) Significance determined by two-tailed unpaired t-test (b, c, f, h) and log-rank test (i). Significance was determined as p<0.05. ns, p≥0.05; **p<0.01, ***p<0.001, ****p<0.0001. Parts of (a, g) were drawn using pictures from Servier Medical Art. ssGSEA, single-sample gene set enrichment analysis; TNBC, triple-negative breast cancer. In summary, chemotherapy drives both liver tumors and primary tumors toward a more immunosuppressive state, although via different mechanisms. Targeting these site-specific immunosuppressive effects with tailored therapeutic strategies may enhance treatment efficacy. Discussion The liver is a prevalent site for breast cancer metastases, often resulting in a poor prognosis for patients with untreated liver metastases, who typically have a median survival of merely 3–6 months.[200]37 38 Chemotherapy serves as the primary first-line treatment for patients with liver metastases of breast cancer, especially those with the subtype of TNBC.[201]6 8 9 However, the interplay between metastatic heterogeneity and chemotherapy sensitivity in TNBC remains not fully understood, prompting research into the underlying mechanisms and the development of effective combination therapies.[202]39 The environmental conditions in different organs can significantly influence drug responses.[203]40 For instance, Alsaggar et al observed that melanoma tumors in the lungs exhibited a better response to chemotherapy than those in the liver and kidneys.[204]41 Similarly, Erstad et al discovered that the anatomical location of pancreatic cancer influences the TME and the sensitivity to chemotherapy.[205]42 However, current preclinical assessments of therapeutic efficacy often rely on subcutaneous tumor models, which hinders direct comparisons between primary and metastatic tumors in distinct anatomical settings. This study employed murine models of TNBC liver metastases to show that liver tumors are less responsive to chemotherapy than identical subcutaneously grown tumors. The experimental liver metastasis model, established through intrahepatic injection of 4T1 cells, enables single-focus tumor growth under the liver capsule, which makes liver tumors comparable in size to subcutaneous tumors. Moreover, this model bypasses the variability of tumor cell heterogeneity, vascular density, and drug distribution in treatment efficacy assessments, confirmed by various experiments including IHC, wound healing assays, colony formation assays, Evans Blue assays, and tumor cross-transplantation experiments. This approach facilitates a focused investigation into the influence of the TME at different anatomical sites on chemotherapeutic efficacy, compared with the experimental liver metastasis model constructed by intrasplenically injecting. Chemotherapeutic drugs are traditionally thought to exert their antitumor effects by directly killing cancer cells.[206]43 However, recent studies suggest a strong link between their antitumor efficacy and immune system engagement. The abundance of macrophages and T cells in breast cancer correlates with patients’ responses to chemotherapy and serves as a prognostic marker for clinical outcomes.[207]27 Additionally, higher levels of stromal tumor-infiltrating lymphocytes are associated with improved patient survival following adjuvant chemotherapy.[208]44 45 In this study, significant disparities were observed in the immune microenvironment between liver metastases and primary tumors of TNBC before chemotherapy. Primary tumors showed a more robust adaptive immune response, while liver metastases exhibited increased innate immune activity. The weaker efficacy of chemotherapy in liver tumors, as seen in animal models, suggests that adaptive immunity, notably CD8+T cell activity, plays a critical role in chemotherapy outcomes. This notion was supported by the absence of differential chemosensitivity following CD8+T cell depletion with an antagonist. Macrophages, recognized for their critical interactions with CD8+T cells, demonstrated reduced activation within liver metastases. Depletion of macrophages normalized pre-treatment CD8+T cell infiltration and eliminated the initial therapeutic differences between tumors from different sites, indicating that macrophages are crucial for the efficacy of chemotherapy by regulating CD8+T cell presence. Therefore, stimulating macrophages to a state of heightened activation before chemotherapy could potentially enhance treatment efficacy in liver tumors. The administration of glufosinate before chemotherapy, which activates macrophages, improved chemotherapy outcomes in liver tumors. In summary, the relatively suppressed immune microenvironment in liver metastases limits the efficacy of chemotherapy. Targeting macrophages represents a potential strategy to enhance the effectiveness of chemotherapeutic interventions. Chemotherapy can induce a suppressive phenotype in the tumor immune microenvironment, potentially aiding tumor progression and metastasis.[209]46 For example, Hughes et al noted that chemotherapy could lead to the alternative activation of tumor-associated macrophages around blood vessels, which may foster tumor vascularization and relapse.[210]47 Monteran et al discovered that doxorubicin chemotherapy upregulates complement factors in pulmonary fibroblasts, creating an immunosuppressive microenvironment conducive to lung metastasis of TNBC.[211]48 Our studies further show that the immune microenvironments of subcutaneous tumors and liver tumors shift toward inhibitory states post-chemotherapy through distinct mechanisms. Before treatment, the superior response of subcutaneous tumors to chemotherapy was linked to the infiltration and activation of macrophages and CD8+T cells. The initially activated immune microenvironment in subcutaneous tumors transforms post-chemotherapy, with macrophages adopting an immunosuppressive phenotype and a concurrent decrease in CD8+T cell infiltration. This suggests that chemotherapy may dampen the active immune state, highlighting macrophages as targets for intervention to boost chemotherapeutic efficacy in subcutaneous tumors. Adding the macrophage-activating agent glufosinate to chemotherapy regimens has indeed improved treatment outcomes. The combined therapeutic approach induces a reactivation of macrophages and an increase in CD8+T cell infiltration, underscoring the role of macrophages in modulating the efficacy of chemotherapy by regulating CD8+T cell infiltration, in line with our previous observations regarding pre-treatment immune profiles in liver tumors and subcutaneous tumors. However, no phenotypic changes in macrophages and CD8+T cells were observed in liver tumors following chemotherapy, likely due to their pre-existing immunosuppressive profile. Further research has revealed that liver tumors treated with paclitaxel showed an enrichment of the NETs formation pathway, a known contributor to weaker chemotherapeutic efficacy in lung metastases of breast cancer.[212]49 This suggested that NETs may serve as a potential target for improving the chemotherapeutic efficacy in liver tumors, which was a distinctly different approach from the macrophage-targeted interventions in subcutaneous tumors. Combining the NETs inhibitor Cl-amidine with paclitaxel demonstrated enhanced efficacy against liver tumors. In conclusion, chemotherapy may induce an immunosuppressive microenvironment in both liver tumors and primary tumors, potentially leading to chemotherapy resistance. These effects appear to be mediated by different mechanisms, underscoring the potential of site-specific interventions to augment chemotherapy effectiveness. This study acknowledges several limitations. First, the small clinical sample size restricts the statistical power of our findings. Specifically, the limited liver metastasis samples hindered a comprehensive single-cell analysis pre-treatment and post-treatment. Second, the study did not delve into the specific interaction receptors and ligands between macrophages and CD8+T cells, nor did it explore the potential contributions of other immune cells to chemotherapeutic efficacy. Third, the mechanisms by which chemotherapy induces the formation of NETs in liver tumors and how NETs affect the effectiveness of chemotherapy are not reported here. Fourth, while the liver metastasis models used in this study were not spontaneously metastatic to the liver, these models may not fully replicate the hematogenous spread observed in clinical liver metastases. Moreover, further research is required to ascertain whether the observed chemotherapy differences across different sites apply to other tumor types beyond TNBC. In summary, chemotherapy shows weaker effectiveness against TNBC liver tumors compared with primary tumors, likely due to the less active state of macrophages and lower CD8+T cell infiltration in the microenvironment. Enhancing macrophage activation could improve chemotherapy efficacy in liver tumors. Moreover, chemotherapy induces an immunosuppressive shift in the immune microenvironment of both liver tumors and primary tumors, but through different pathways. Therapeutic strategies targeting these unique mechanisms may augment the efficacy of chemotherapy in site-specific tumor contexts, though further research is required to validate these observations. Understanding how distinct tumor immune microenvironments affect chemotherapy outcomes is crucial for enhancing chemotherapeutic efficacy across different tumor sites. Acknowledgments