Iron diminishes immunosuppressive macrophages and enhances anti-PD-1 immunotherapy in breast cancer models
Qingfei Wang, Elizabeth L. Adams, Rylee A. Poole, Tahereh Soleimani, Grace Xiyu Wang, Hudie Li, Maegan L. Capitano, Ashiq Masood, Scott I. Abrams, Kelvin P. Lee, Siyuan Zhang, Mateusz Opyrchal
TL;DR
This study shows that iron can reduce immunosuppressive macrophages in breast cancer and improve the effectiveness of anti-PD-1 immunotherapy.
Contribution
The novel finding is that iron supplementation reprograms tumor-associated macrophages and enhances anti-PD-1 therapy in breast cancer models.
Findings
Iron treatment reduces immunosuppressive features in tumor-associated macrophages.
Iron supplementation increases CD8+ T cell infiltration and antitumor effects of PD-1 immunotherapy.
Iron modulates TAM metabolism and downregulates NF-κB pathways.
Abstract
Breast cancer remains a leading cause of cancer-related mortality among women globally, necessitating the development of innovative therapeutic strategies. The efficacy of immune checkpoint inhibitor-based immunotherapy in triple-negative breast cancer has provided a rationale for exploring its expansion to other breast cancer subtypes. Immunosuppressive tumor-associated macrophages (TAMs) within the tumor microenvironment have been demonstrated as a formidable barrier to the efficacy of approved immunotherapy. We aimed to identify and therapeutically modulate pathways that regulate the immunosuppressive properties of TAMs for more effective breast cancer immunotherapies. We integrated analyses of publicly available human breast cancer single-cell RNA sequencing (scRNA-seq) datasets with scRNA-seq profiling of murine mammary tumors to identify the signaling pathways associated with…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7- —IUSCCC Early Career Investigator Award EPAR3940
- —National Cancer Institute (NCI) Grant No. P30CA016056; Vera Bradley Foundation
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsImmune cells in cancer · Ferroptosis and cancer prognosis · Cancer Immunotherapy and Biomarkers
Background
Breast cancer continues to be the most frequently diagnosed cancer among women globally and remains a leading cause of cancer-related mortality [1], necessitating the development of innovative therapeutic strategies. Clinically, breast cancer treatment and prognosis depend on the subtype, as defined by the expression of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2), and disease stage. Patients with advanced triple-negative breast cancer (TNBC) have especially poor outcomes, largely due to a lack of effective targeted therapies [2]. Fortunately, the advent of pembrolizumab, an immune checkpoint inhibitor (ICI) targeting programmed cell death protein-1 (PD-1), along with chemotherapy in preselected TNBC patients has transformed the treatment landscape of this aggressive entity [3, 4]. The success seen in TNBC has provided a foundation and motivated research into expanding the ICI-based immunotherapy to other breast cancer subtypes [5]. And there is an urgent need to identify optimal combination strategies to further improve therapeutic outcomes for breast cancer patients.
It is now well-established that cancer development, progression, and treatment responses are influenced by bidirectional interactions between tumor cells and the non-neoplastic content within the cancer ecosystems, particularly, the tumor immune microenvironment (TIME). The TIME is broadly populated with T and B lymphocytes, natural killer (NK) cells, macrophages, and myriads of other myeloid cell types, which collectively shape the course of disease and serve as potential targets for cancer therapy [6]. Macrophages are universal and often the most abundant components of the tumor-infiltrating immune cells. Such tumor-associated macrophages (TAMs) possess remarkable diversity and plasticity and they can acquire multiple functional phenotypes depending on the microenvironmental cues they receive [7]. Notably, TAMs are instrumental in driving the immunosuppressive character of the tumor microenvironment (TME) by diverse mechanisms, including the production of anti-inflammatory cytokines and other mediators, secretion of pro-angiogenic factors, inhibition of cytotoxic T cell recruitment and activation, and promotion of regulatory T cells (Tregs), all of which have been demonstrated as major obstacles to the efficacy of various forms of immunotherapy, including ICIs [8–10]. Therefore, targeting pathways that up or downregulate different macrophage functional states, including their immunosuppressive activities within the TME could represent promising therapeutic strategies to enhance antitumor immune responses.
Numerous regulators, for example, signaling or transcriptional networks, cytokine exposure, and epigenetic modifiers, contribute to the diverse functional states or ‘programs’ of macrophages [10]. Based on these fundamental discoveries, an arsenal of approaches has been tested to reprogram TAMs, including in breast cancer [9, 10]. However, significant clinical benefits have yet to be realized. Another approach to potentially direct various macrophage functional states is through metabolic regulation. Iron is an essential micronutrient required for various enzymes and processes of living cells. Macrophages are critical for the regulation of iron availability and, reciprocally, iron availability can also influence functionality of macrophages and other immune cells [11]. It has been appreciated that iron regulates macrophage differentiation and their functional programs which could give rise to their biologic heterogeneity and plasticity. Therefore, modulating iron metabolism or controlling iron availability to functionally reprogram macrophages (i.e., TAMs) and enhance antitumor immunity has attracted considerable attention in recent years [11, 12]. However, results have varied considerably across the reported studies, possibly due to the variations in the model systems, cancer type, disease stage and treatment regimen. Additionally, and fundamentally, important gaps remain in our understanding of whether it is better to restrict iron utilization or to refuel the TME with iron. Therefore, in this study, we sought to revisit this critical unresolved area of iron metabolism in macrophage-breast cancer biology and therapy using an unbiased and comprehensive approach. We posited that such new insights may have important implications for the design of more precise, targeted, and effective cancer therapies.
In this study, through analyses of human breast cancer single-cell RNA sequencing (scRNA-seq) datasets, and single-cell transcriptional profiling of murine mammary tumors, we show that the activity of iron uptake and transport pathway in TAMs is strongly associated with their immunosuppressive features. Notably, we find that iron supplementation diminishes immunosuppressive phenotype of TAMs, increases CD8 + T-cell infiltration and their cytotoxic activity to enhance the efficacy of anti-PD-1-based immunotherapy. Our study advances the pivotal roles of iron in modulating immune function, including the functional states of TAMs and their impact on anti-tumor immune reactivity, and supports the notion that iron supplementation could be a promising combinatorial partner to enhance the efficacy of anti-PD-1 immunotherapy in breast cancer.
Methods
Animal model
The mouse strains were purchased from The Jackson Laboratory. The female MMTV-PyMT mice utilized for study were generated by breeding C57BL/6 females (JAX stock # 000664) with hemizygous B6-Tg(MMTV-PyMT) males (JAX stock #022974). Mice genotyping was performed according to the JAX genotyping protocol. The C57BL/6 syngeneic E0771 cell line was purchased from CH3 Biosystems (NY,940,001-Vial, Amherst). E0771 cells were cultured in DMEM High Glucose + 10% FBS + 1% Penicillin/Streptomycin. They were labeled with mCherry by lentiviral transduction. Cells were cultured in a humidified incubator in 5% CO2 at 37 °C, under sterile conditions. Cells were tested and found negative for mouse pathogens by IDEXX BioResearch Inc. E0771 tumor model was generated by injection of 2–3 × 10^5^ mCherry labeled E0771 cells in 100 µl phosphate buffered saline (PBS) into the abdominal (4th) mammary fat pad of 10 to 16-week-old female C57BL/6 mice.
In vivo treatment
Anti-PD-1 antibody (Cat# BE0146, clone RMP1-14) and rat IgG2a isotype control (Cat# BE0089) were purchased from BioXcell (West Lebanon, NH). Iron-dextran (Cat#D8517-25ML, Sigma-Aldrich, 100 mg/mL Fe) 0.5 mg/mouse, which is a clinically relevant dose equivalent to a human dose of 1.0 g per 60 kg, and anti-PD-1 antibody 150 ug/mouse were intraperitoneally (i.p.) administrated twice weekly for 2 weeks. MMTV-PyMT mice with spontaneous tumors were i.p. administrated with 1 dose of doxorubicin (HY-15142, MCE) at 2 mg/kg and cyclophosphamide (HY-17420, MCE) at 200 mg/kg 7 days before PD-1 antibody or PD-1 antibody plus iron treatment. The tumors were measured twice weekly using calipers. Tumor volume was calculated as length × width^2^/2. The relative tumor volume (RTV) was calculated as the ratio of the tumor volume on a given day of treatment to its volume at the start of treatment, normalizing the starting RTV to 1. In Fig. 4A left panel, individual RVT was normalized to the Ctrl, Fe, anti-PD-1, and combination treatment start date (Day 0). In Fig. 4B, individual RVT was normalized to anti-PD-1, and combination treatment start date (Day 8, when mice transitioned to PD-1-based immunotherapy after initial chemotherapy).
scRNA-seq by 10X Genomics and analysis
Cells from E0771 tumors for single-cell RNA-seq with 10X Genomics platform were prepared as described previously [13]. In brief, fresh tumors (2 weeks post MFP injection of mCherry-E0771 cancer cells) were resected and minced with sterile scissors into small pieces, then enzymatically digested in DMEM/F12 medium (10 ml/g tumor) containing 5% FBS, 2 mg/ml collagenase, 0.02 mg/ml hyaluronidase, and 0.01 mg/ml DNase I for 30 min at 37 °C with gentle agitation. Dissociated cells were centrifuged at 350 × g for 5 min with the brake on and supernatant was discarded. The pellet was re-suspended with 3–5 mL of pre-warmed TrypLE and incubated for 5 min. After adding 10 mL of DMEM/F12 medium supplemented with 2% FBS and passing through a 40 μm cell strainer, cells were centrifuged at 350 × g for 5 min and re-suspended in MACS buffer [PBS with 0.5% bovine serum albumin (BSA) and 2 mM EDTA]. The cell suspension was carefully layered on top of 15 ml Ficoll-Paque media solution in a 50-ml Falcon tube and centrifuged at 1000 g for 10 min at room temperature with the brake off. The buffy layer at the interface was transferred and washed with cold MACS buffer. Following Ficoll separation, dead cells were removed by using dead cell removal kit (Cat#130–090–101, Miltenyi Biotec) per manufacturer’s instruction. The live cell fraction was blocked by incubation with TruStain FcX in 50 µL cell staining buffer for 20 min on ice, then incubated with magnetically labeled CD45 antibody (Cat#130–052–301, Miltenyi Biotec) and passed through a LS magnetic column (Cat#130–042–401, Miltenyi Biotec). The cells retained in the column were eluted as CD45 + immune cells. The flow through fraction was collected as CD45- non-immune cells. The cells for 10X Genomics Chromium were prepared by following the CITE-seq and cell hashing protocol.
(https://cite-seq.com/wp-content/uploads/2019/02/cite-seq_and_hashing_protocol_190213.pdf).
Briefly, samples were stained individually with Total-seq rat anti-mouse hashtag 7–9 antibodies purchased from BioLegend (Cat# 155,813, 155,815, 155,817, Biolegend, M1/42; 30-F11). After 25 min of staining, samples were washed 4 times then pooled prior to loading onto the 10 × Chromium. Notre Dame Genomics and Bioinformatics Core Facility performed 10X Genomics Chromium single cell capture and cDNA and HTO libraries preparation according to the standard CITE-seq and 10X Genomics standard protocols. Libraries were validated by Qubit and Agilent Bioanalyzer DNA High Sensitivity Chip assays. After which, libraries were submitted to Indiana University School of Medicine Center for Medical Genomics for multiplexing into a single pool and Illumina sequencing on a NovaSeq6000. Raw FASTQ files were processed through CellRanger 6.0 (10X Genomics). Cells were demultiplexed to their original samples using the Cell Hashing tags (HTOs). Expression data normalization, quality control, dimensional reduction, clustering and differential expression analyses were performed using Seurat v5 package [14] in RStudio. Details about Seurat v5 can be found at: https://satijalab.org/seurat/.
Cell preparation for scRNA-seq by PIP-seq and analysis
Single-cell suspensions from E0771 tumors with indicated treatment were prepared by using mouse tumor dissociation kit (Cat#130–096–730, Miltenyi Biotec) in combination with the gentleMACS Octo Dissociator. Briefly, fresh E0771 tumors were resected and minced with sterile scissors, then transferred into gentleMACS C Tubes with 5 mL enzyme mix prepared as recommended in the protocol (5 mL enzyme mix for maximal 1 g tumor tissue). Program ‘37C_m_TDK_2’ for tough tumor was used with the heating function. After termination of the program, samples were resuspended after adding 10 mL of DMEM/F12 medium supplemented with 2% FBS and passed through a 40 µm cell strainer. Cell suspension was centrifuged at 350 g for 5 min and resuspended in PBS. Cells were then stained with zombie violet fixable viability dye (Cat# 423,114, BioLegend) on ice for 15 min followed by a 20-min incubation with TruStain FcX. Subsequently, cells were incubated with APC anti-mouse CD45 antibody (Cat#103,112, BioLegend) on ice for 30 min. After washing twice with cell staining buffer (Cat# 420,201, BioLegend), flow sorting was conducted on BD FACSAria Fusion Cell Sorter (BD Biosciences) to isolate live CD45^+^ cells. Gating strategies are shown in Supplementary Fig. S4A.
Single-cell RNA-seq for sorted CD45^+^ cells (pools of three to four tumors from each treatment group) were performed by PIP-seq [15]. Briefly, cells were suspended at a concentration of 5,000 live cells/µL with cell suspension buffer provided in the PIPseq T20 3ʹ Single Cell RNA Kit V4. A total of 40,000 cells plus 40 units of RNase inhibitor (Cat# M0314L, New England Biolabs) were added into pre-templated instant partitions (PIPs) according to the user’s guide (Doc ID:FB0002130, Fluent BioSciences). Samples were processed following the workflow of capture and lysis, mRNA isolation, complementary DNA (cDNA) synthesis, cDNA amplification, library preparation, library pooling, and sequencing accordingly. Sequencing of the cDNA libraries were performed on an Illumina NovaSeq X plus instrument at Indiana University School of Medicine Center for Medical Genomic, targeting a sequencing depth of ~ 400,000,000 reads per sample. The sequencing data was aligned using PIPseeker v3 according to the user’s guide (Doc ID:FB 0002787, Fluent BioSciences). Expression data normalization, quality control, dimensional reduction, clustering and differential expression analyses were performed using Seurat v5 package.
Gene set enrichment and correlation analysis
Gene set enrichment analysis was run using GSVA v2.2.0 [16] in R using single-cell expression matrix with UMI values. We applied mouse version hallmark gene sets (MH) and canonical pathways from the Reactome, Biocarta and WikiPathways databases from the M2 collection of Molecular Signatures Database (MSigDB) to each single cell to obtain the enrichment score for each signature. Immunosuppression signature has been described previously [17] and the genes are shown in Supplementary Fig. S1A for human and Supplementary Fig. S2B for mouse. Gene set enrichment data was merged with corresponding metadata and analyzed using the Seurat package. Signature scores of individual cells for TAM molecular signatures, iron uptake and transport pathway and immunosuppression were computed using the “AddModuleSocre” function of Seurat v5, by setting the number of control features selected from the same bin per analyzed feature as 5. Standard “FeaturePlot” and “VlnPlot” functions of Seurat were used to generate the feature plots or violin plots. For correlation analysis between gene signatures, Pearson correlation analyses and scatter plots were performed and generated using the “FeatureScatter” Seurat function in combination with ggplot2, by using the “geom_smooth” function with the method argument set to “lm” to plot a fitted linear regression line.
BMDM culture
Bone marrow-derived macrophages (BMDMs) were generated by harvesting femoral bone marrow from wild type C57BL/6 mice. Bone marrow was flushed out with ice-cold PBS and homogenized by pipetting. The cell suspension was then passed through a 70 μm cell strainer and red blood cells were lysed using RBC lysis buffer (Cat#420,301, BioLegend). After washing, cells were cultured in RPMI media containing 10% fetal bovine serum (FBS), 10 ng/ml M-CSF (Cat# 315–02-10UG, PeproTech), 1% penicillin/streptomycin for 5 days prior to treatment with or without iron-dextran (20 mM) as previously described [18].
T cell proliferation assay
Splenocytes were isolated from wild type C57BL/6 mouse spleens through mechanical disruption, RBSs lysis was performed, then the cells were washed twice with PBS, and filtered through 40µm strainer. Cells were labeled using CFSE cell division tracker kit (Cat#423,801, BioLegend) according to the protocol, followed by a wash with RPMI media containing 10% FBS and 1% penicillin/streptomycin. Primary BMDMs with or without iron treatment for 3 days were co-cultured with CFSE labeled splenocytes in the presence of anti-CD3/CD28 stimulation with Dynabeads Mouse T-Activator CD3/CD28 for T-Cell Expansion and Activation (Cat# 11452D; Thermo Fisher Scientific), with indicated ratio in round-bottom 96-well plates. After co-culturing for 72 h, cells were collected and stained for CD4 and CD8 with APC anti-mouse CD4 antibody (Cat# 100,412, BioLegend) and APC anti-mouse CD8 antibody (Cat# 100,712, BioLegend), respectively. CFSE dilution was measured using flow cytometry (BD LSRFortessa X-20). Proliferation index of CD4 + T-cell and CD8 + T-cell were calculated by proliferation modeling tool of FlowJo software. Proliferation index is the total number of divisions divided by the number of cells that went into division.
RNA isolation and quantitative PCR for gene expression
Total RNA samples were extracted from cells using the RNeasy Plus Mini Kit (Cat#74,136, Qiagen) according to the manufacturer’s instructions. One microgram of RNA was reverse transcribed into cDNA using qScript cDNA Synthesis Kit (Cat# 95,047–100, Quantabio) following the manufacturer’s protocol. Quantitative PCR was conducted on the QuantStudio 6 Pro Real-Time PCR System (Applied Biosystems) with PowerUp SYBR Green Master Mix (Cat# A25777, Applied Biosystems). Primers were synthesized by Sigma-Aldrich, and the sequences are listed in Supplementary Table 1. The expression level of target gene was normalized to the reference gene GAPDH by calculating (2^−∆Ct^) value.
Flow cytometry
Cryopreserved mammary tumors were digested with mouse tumor dissociation kit (Cat#130–096–730, Miltenyi Biotec) in combination with the gentleMACS Octo Dissociator. Cell suspensions of digested tumors or BMDMs were incubated with TruStain fcX (BioLegend, 101,319) in 100uL MACS buffer for 15 min at room temperature. Cells were then incubated on ice for 30 min with pre-conjugated antibodies including CD45-Pacific Blue (Cat# 103,125, BioLegend), F4/80-FITC (Cat# 123,107, BioLegend), CD274/PD-L1-PE/Cy7 (Cat# 124,313, BioLegend). Flow cytometry was performed on BD LSRFortessa X-20. FlowJo software was used for data analyses.
Serum iron and ferritin measurement
Serum was collected from mice at the end of experimentation and stored at −80°C. Iron concentration and total iron-binding capacity (TIBC) in mouse serum were measured using the Total Iron-Binding Capacity and Serum Iron Assay Kit (Cat# ab239715, Abcam). Each sample was prepared in duplicate, 10µL of serum was added per well, and the volume was brought to 25µL with TIBC assay buffer. For the serum iron assay, after adding the buffer, the samples were incubated for 10 min at 37°C. For TIBC assay, iron solution was added to the buffer before incubation under the same conditions. Subsequently, 25µL of TIBC detector was added, followed by another 10-min incubation at 37°C. Finally, TIBC developer solution was added and incubated for 10 min at 37°C. A standard curve was generated using 1 mM iron standard in each well to produce 0, 2, 4, 6, 8, and 10 nmol/well iron standards. Absorbance was measured at 570 nm for both standards and samples using the CYTATION 5 Reader (BioTek). Ferritin was measured using a mouse ferritin ELISA kit (Cat# ab157713, Abcam). Serum samples were diluted 40-fold in 1X diluent solution. Then, 100µL of the diluted sample was placed into predesignated wells. After incubation, washing, incubation with enzyme conjugated secondary detector antibody, incubation with chromogen substrate solution after washing and adding stop solution according to manufacturer’s instructions, absorbance was measured at 450 nm using the CYTATION 5 Reader (BioTek).
Immunohistochemistry
Tumor tissues were fixed in 10% neutral buffered formalin, processed routinely, and embedded in paraffin. After rehydration, blocking of endogenous peroxidases, paraffin-embedded sections (4 μm) were subjected to antigen retrieval in a pressure cooker with sodium citrate buffer (PH = 6.0) and incubated with R.T.U. Animal Free Blocker and Diluent (Cat# SP-5030–250, Vector Laboratories) for 1 h then antibodies specific for CD8α (Cat# 98941S, Cell Signaling Technology, 1:300) overnight at 4°C. HRP-conjugated secondary antibody was used. Remaining steps were performed using Pierce Peroxidase IHC Detection Kit (Cat# 36,000, Thermo Scientific). Slides were counterstained with hematoxylin. To quantify the positive staining cells, the numbers of infiltrating CD8 + cells were counted per field of view after examination of at least 10 fields of each section (200X). Images were acquired using a KEYENCE BZ-X microscope and software.
Immunofluorescence staining
Cryosections were used for CD8 and granzyme B double staining. For preparation of cryosections, dissected tissues were embedded in Tissue-tek O.C.T. (Cat# 62,550–12, Electron Microscopy Sciences) and frozen on dry ice. Frozen tissues were stored at −80 °C until they were sectioned at 7 μm. For staining, O.C.T. tumor cryosections were briefly air dried and fixed with 3% paraformaldehyde at room temperature for 15 min. Sections were then blocked with Animal Free Blocker for 1 h and incubated with primary antibodies for CD8 (Cat# 14–0808-82, Thermo Scientific, 1:150), granzyme B-FITC (Cat# 372,205, BioLegend, 1:50). Secondary antibody goat anti-Rat-Alexa Fluor 594 (Cat# A11007, Thermo Scientific, 1:500) was used for CD8 staining. After washing, sections were stained with DAPI to visualize nuclei. Immunofluorescence imaging was performed on a multicolor fluorescent microscope (Leica DM5500 B). Five random fields were acquired from each biological sample for quantification of positive stained cells.
Statistical analysis
Statistical tests were performed in GraphPad Prism (version 10.0) or in R (version 4.5.0). Differentially expressed genes between clusters in all datasets were defined using the Wilcoxon rank-sum test as implemented in the ‘FindAllMarkers’ Seurat function. Sample sizes, statistical test methods, and p-values are specified in the figure legends. P value < 0.05 was considered statistically significant.
Results
Iron metabolic gene signature is correlated with immunosuppressive phenotype of TAMs in human breast cancer independent of subtype
To identify immunosuppressive components within the breast cancer TME at the single-cell resolution, we sought to analyze a publicly available single-cell RNA sequencing (scRNA-seq) dataset containing primary tumors collected from five TNBC patients [19]. After performing single-cell gene set variation analysis (GSVA) [16] using an immunosuppression signature published previously (Supplementary Fig. S1A), we found that myeloid cells, as annotated by the source study, exhibited highest signature scores for immunosuppression among all other identified cell populations (Fig. 1A). Next, to explore the signaling pathways associated with this immunosuppressive state, we calculated the Pearson’s correlation between immunosuppression signature score and the enrichment score of individual canonical pathway genesets. This analysis showed that iron uptake and transport was among the top pathways enriched in myeloid cells (Fig. 1B) and correlated strongly with the immunosuppression signature score across all cells in the dataset (Supplementary Fig. S1B). Accordingly, myeloid cells expressed markedly higher levels of genes involved in iron and heme metabolism, such as iron storage protein ferritin (FTH1 and FTL1), transferrin receptor (TFRC/CD71), heme oxygenase enzyme HMOX1 for heme degradation, heme receptor Lrp1, iron transporter natural resistance-associated macrophage protein 1 (SLC11A1), and iron transporter ferroportin (SLC40A1), relative to other immune cells (Supplementary Fig. S1C). To investigate the complexity of myeloid cells in more detail, we sub-clustered these cells and identified seven macrophage (TAM) subsets, two neutrophil/monocyte (Neut/Mono) subsets (clusters 3 and 4) and one dendritic cell (DC) subset (cluster 9) based on the expression of marker genes (Supplementary Fig. S1D and Fig. 1C). Notably, TAM subsets possessed significantly higher enrichment scores for both immunosuppression signature and iron uptake and transport pathway compared with Neut/Mono and DC subsets (Fig. 1D and E). Furthermore, the positive correlation between these two signatures was stronger in TAMs relative to Neut/Mono and DC subsets (Fig. 1F). On top of this, we asked whether the above phenomena and correlation were specific or limited to TAMs within TNBC. We then performed gene expression and pathway analysis across breast cancer clinical subtypes in another scRNA-seq dataset published by Wu et al. [20]. Consistently, within each subtype, TAM subpopulations displayed greater enrichment scores of iron uptake and transport pathway and immunosuppression signature compared to all other cell types (Supplementary Fig. S1E and F). Despite the expression of major iron and heme metabolism genes, including FTH1, FTL, TFRC/CD71, HMOX1, LRP1, SLC11A1 and SLC40A1 exhibited highest in TAMs from TNBC (Fig. 1G), the average enrichment scores for both signatures were greater in TAMs from HER2 + tumors relative to those of ER + and TNBC subtypes (Fig. 1H). The positive association between these two signatures were significant across all cells within each subtype (Supplementary Fig. S1G), and importantly, this association in TAM subsets seemed to be independent of clinical subtype (Fig. 1I). Together, these analyses revealed a shared and close association between iron metabolism and immunosuppression in TAMs across breast cancer patients, suggesting that iron metabolism signaling may be involved in regulating the immunosuppressive activity of TAMs within the breast cancer TME.Fig. 1. The association between iron metabolism and immunosuppression in TAMs of human breast cancer. A and B** Violin plots showing enrichment scores for immunosuppression (A) and iron uptake and transport (B) signatures among cell types. ***, P < 1 × 10^–10^, myeloid cells vs other cell types by two-tailed Student’s t-test. The red dashed line shows the mean of enrichment score. C Re-clustering, uniform manifold approximation and projection (UMAP) visualization and subtype identification of myeloid cells. D and E UMAP (left) and violin plots (right) showing enrichment scores for immunosuppression signature (D) and iron uptake and transport pathway (E) among cell types identified in C. ***, P < 1 × 10^–10^, TAM vs Neut/Mono and vs DC by two-tailed Student’s t-test. F Correlation between enrichment scores for iron uptake and transport and immunosuppression signatures across TAM, neutrophil/monocytes and DC as identified in C. Each point represents a single cell. Dot color indicates cell type. r represents Pearson correlation coefficient. G Bubble plot showing the expression levels of major iron and heme metabolic genes in TAMs from the indicated clinical subtype of breast cancer. Cell type classification and detailed patient information can be found in the source study. H Violin plots showing enrichment scores for iron uptake and transport and immunosuppression signatures among TAMs from the indicated clinical subtype of breast cancer. ns, not significant. ***, P < 1 × 10^–10^, vs TNBC subtype by two-tailed Student’s t-test. The red dashed line shows the mean of enrichment scores in TNBC group. I Correlation between enrichment scores for iron uptake and transport and immunosuppression signatures in TAMs from the indicated clinical subtype of breast cancer. Each point represents a single cell. r represents Pearson correlation coefficient
Single-cell RNA-Seq identifies iron metabolic gene-enriched TAMs with elevated immunosuppressive features in mouse breast cancer
To corroborate and delve further into these findings from humans, we performed scRNA-seq by using the 10 × Genomics platform with sample multiplexing/cell hashing, on a mixture of magnetic-activated cell sorting (MACS)-isolated immune (CD45 +) and non-immune (CD45-) cells from mammary fat pad (MFP) injected E0771 tumors. Parental E0771 cells were derived from a spontaneous medullary breast cancer in C57BL/6 mouse and readily form aggressive tumors when implanted into the MFP [21, 22]. It has been characterized as a luminal B or triple-negative subtype of breast cancer with vast amounts of tumor infiltrating lymphoid and myeloid cells in previous studies [23, 24]. After sample demultiplexing, quality-control filtering, unsupervised clustering and marker gene expression analysis, we identified 12 clusters of main cell types (Fig. 2A and B, Supplementary Fig. S2A), including macrophages (TAMs, clusters 0, 7, 6, 4 and 5, cluster 5 representing proliferative macrophages), marked by high expression of C1qa, C1qc, Mrc1 and Trem2, monocytes (Mono, cluster 2) by expression of Ccr2, Hp and Chil3, dendritic cells (DCs, cluster 9) by expression of Kmo, Cd209a and Zbtb46, CD8 + T cells (cluster 8), CD4 + T cells (cluster 10), tumor cells (clusters 1 and 2) and fibroblasts (Fibro, cluster 11). We next converted the immunosuppression signature to the mouse analog (Supplementary Fig. S2) and performed geneset enrichment analysis. We observed that TAMs had higher overall immunosuppression signature scores compared to other cell types, and specifically, clusters 0, 4, 6, and 7 were among the top subsets when compared to other clusters (Fig. 2C), suggesting that TAMs, particularly clusters 0, 4, 6, and 7 subpopulations, may act as important immunosuppressors in the TME. Similarly, compared with other cell types, the TAMs showed significantly elevated enrichment of the iron uptake and transport pathway; clusters 0, 4, and 6 TAMs exhibited higher levels, while cluster 7 had the highest enrichment score among all clusters (Fig. 2D). Indeed, TAM clusters exhibited increased expression of major iron and heme metabolic transcripts, such as ferritin (Fth1 and Ftl1), transferrin (Trf), heme oxygenase enzymes Hmox1/2, heme receptor Lrp1, iron transporter Slc11a1 and ferroportin (Slc40a1) (Fig. 2E). Notably, and consistent with the results from patients, we found that the extent of the iron uptake and transport pathway signature was positively correlated with the enrichment score of immunosuppression signature across all cells (Supplementary Fig. S2C), and this correlation was stronger in TAMs relative to monocytes and DCs clusters (Fig. 2F).Fig. 2. Single-cell RNA-Seq identifies iron metabolic gene-enriched TAMs with elevated immunosuppressive features in E0771 breast cancer model. A Unsupervised clustering and uniform manifold approximation and projection (UMAP) visualization of single-cell transcriptional profiling of cells isolated from E0771 tumors. B Bubble plot showing the expression of marker genes for each cell type as annotated in A. Dot color indicates expression level and size indicates the proportion of cells expressing each gene. C UMAP (left) and violin plots showing immunosuppression signature score among cell types (middle) and clusters (right) identified in A. The red dashed line shows the mean of enrichment score. D UMAP (left) and violin plots showing iron uptake and transport signature score among cell types (middle) and clusters (right) identified in A. The red dashed line shows the mean of enrichment score. E Bubble plot showing the expression levels of major iron and heme metabolic genes for the indicated clusters and cell types. F Correlation between enrichment scores for iron uptake and transport and immunosuppression signatures across TAMs, monocytes and DCs. Each point represents a single cell. G Bubble plot showing the annotation of TAM subsets based on the enrichment of molecular signatures. H Bubble plot showing normalized enrichment of top hallmark genesets within TAMs. ns, not significant. *, P < 0.05, **, P < 1 × 10^–5^ and ***, P < 1 × 10^–11^ vs cluster 7 or TAM cells by two-tailed Student’s t-test
To gain more functional implications of TAM clusters, we sought to characterize these cells according to recent nomenclature that is based on their signature genes, enriched pathways, and predicated function from single-cell omics studies [7]. Molecular signature enrichment analysis of individual clusters confirmed the classification of cluster 5 as proliferating TAM (Prolif-TAM), and cluster 6 was predominated by signature genes of interferon-primed TAM (IFN-TAM), while clusters 0 and 4 appeared to resemble lipid-associated TAM (LA-TAM) (Fig. 2G). Most notably, cluster 7 was enriched for signatures related to immune regulatory TAM (Reg-TAM), inflammatory cytokine-enriched TAM (Inflam-TAM), and pro-angiogenic TAM (Angio-TAM) (Fig. 2G). Meanwhile, hallmark pathway analysis revealed high enrichment of the inflammatory response signature in TAMs broadly (clusters 5, 0, 7, 6 and 4) (Fig. 2H). Particularly, cluster 7 had highest enrichment for angiogenesis, TGF-β signaling, reactive oxygen species pathway, hypoxia, TNF-α signaling via NFκB, as well as the heme metabolism pathways among the TAM clusters (Fig. 2H). This functional and transcriptomic diversity highlight the heterogenous nature of TAMs. Nevertheless, our single-cell RNA profiling and analysis of the mouse breast cancer model confirmed the findings from patients that the activity of iron metabolism pathway in TAMs is positively associated with their immunosuppressive properties, suggesting that this strong association may be a common feature within diverse TAM subpopulations of breast cancer, and the iron metabolic gene-enriched TAMs represent potential targets for new therapeutic strategies.
Iron treatment on primary macrophages enhances the proliferation of activated T cells
Motivated by the above findings, we sought to investigate the effect of iron on the functional state of macrophages. We reasoned that if iron treatment could alter the functional state of macrophages, this would affect T cell activity when they are co-cultured in vitro. To test this hypothesis, primary bone marrow-derived macrophages (BMDMs) treated with or without iron were subsequently co-cultured with splenocytes in the presence of anti-CD3/CD28 stimulation. After co-culturing for 3 days, we observed that the proliferation of activated CD4 + and CD8 + T cells co-cultured with iron-treated BMDMs was significantly increased compared with those of control BMDMs (P < 0.001) in a density dependent manner (Fig. 3A and B). Nonetheless, in vitro iron treatment had no or limited effect on viability of human TNBC cells Hs578T and MDA-MB-468 (Supplementary Fig. S3). Markedly, we found that iron treatment significantly reduced the expression of genes associated with immunosuppressive function in BMDMs, including Spp1, Arg1, Ptgs2, Cxcl2, CD274/PD-L1 and Tgf-β1 (Fig. 3C). We also observed that transferrin receptor (Tfrc/CD71) expression was decreased in iron-treated BMDMs (Fig. 3C), indicative of increased iron availability in the culture. In addition, through flow cytometry analyses, we found that iron treatment significantly decreased PD-L1 protein levels in BMDMs compared to the control (Fig. 3D). These data suggests that iron treatment decreases the immunosuppressive activity of primary macrophages.Fig. 3. Iron treatment on BMDMs enhances proliferation of activated T cells in vitro. A Flow cytometry histograms showing representative carboxyfluorescein succinimidyl ester (CFSE) dilution in activated splenic CD4 + T and CD8 + T cells 3 days after co-culture with BMDMs, respectively. B Quantification of CD4 + T and CD8 + T proliferation across experimental replicates (n = 4) as done in A. Results shown were normalized proliferation index calculated by Flowjo. Proliferation index is the total number of divisions divided by the number of cells that went into division. Error bars represent SD. **, P < 0.001, iron treated BMDMs vs control BMDMs co-culture at indicated ratio, by two-tailed Student’s t-test. C Expression of genes associated with immunosuppressive function in BMDMs with or without iron treatment determined by RT-qPCR. mRNA levels were normalized to Gapdh. Error bars represent SD. ***, P < 0.0001, by two-tailed Student’s t-test. D Analysis of PD-L1 levels in BMDMs with or without iron treatment by flow cytometry. Left, representative histograms; Right, quantification of PD-L1 expression by mean fluorescence intensity (MFI). Error bars represent SD. #, P < 0.01, by two-tailed Student’s t-test
Iron supplementation enhances therapeutic efficacy of anti-PD-1 immunotherapy in preclinical breast cancer models
It has been well recognized that the immunosuppressive TME impedes anti-tumor immune responses and contributes to the ineffectiveness of immunotherapy. The above results prompted us to test whether addition of iron would improve responses to anti-PD-1 ICI therapy in preclinical animal models. We first evaluated this combination to treat E0771 tumors. Iron or PD-1 antibody monotherapy had a moderate effect on tumor growth inhibition compared to the control group, whereas combination treatment substantially decreased tumor growth compared with control, iron, or PD-1 antibody monotherapy (P < 0.01) (Fig. 4A). Next, we extended our studies in MMTV-PyMT spontaneous mammary tumor model. As disease progresses, PyMT tumors display loss of ER and PR [25], overexpression of androgen receptor (AR) [26], high rate of pulmonary metastasis [27, 28], and have robust infiltration of macrophages and T cells [29–31]. Mice with PyMT tumors were treated with a standard frontline neoadjuvant chemotherapy, doxorubicin (Adriamycin) combined with cyclophosphamide (AC) for a week, then switched to anti-PD-1 antibody with or without iron (Fig. 4B). Consistent with the results seen in the E0771 model, combination treatment led to significant inhibition of tumor growth compared with PD-1 antibody monotherapy (P < 0.01, Fig. 4B). Additionally, we observed no obvious change in circulating serum iron concentration with or without iron treatment in MMTV-PyMT mice, while there was a decreased total iron-binding capacity (TIBC) and elevated serum ferritin levels after iron supplementation (Fig. 4C-E). These results demonstrated that iron supplementation improved clinical indicators of cancer-associated anemia. Together, results from these two breast cancer models showed that iron supplementation enhances the antitumor effect of PD-1 antibody treatment, suggesting a potential strategy for achieving optimal therapeutic activity of PD-1 based immunotherapy.Fig. 4. Iron supplementation enhances anti-tumor effect of PD-1-based immunotherapy in two syngeneic breast cancer models and diminishes iron deficiency-like phenotype. A Tumor growth curves in E0771-cell mammary fat pad injection model. Iron-dextran was administrated twice a week with 0.5 mg/mouse i.p. for 2 weeks, PD-1 mAb 150 ug/mouse i.p. twice weekly for 2 weeks. Left, tumor volumes and right, relative tumor volumes, presented as the mean ± SEM (n = 6–8). **, P < 0.01, by 2-side Wald test for the combination treatment group versus the control or the single-agent treatment groups for the tumor growth curves. B Tumor growth curves in MMTV-PyMT mice with spontaneous tumors. Mice were first treated with 1 dose of doxorubicin and cyclophosphamide (left), then switched to PD-1 mAb or PD-1 mAb + iron treatment and tumor size were normalized at day 8 (middle), the average of relative tumor volumes, presented as the mean ± SEM (right). **, P < 0.01, by 2-side Wald test for the combination treatment group versus the control or the single-agent treatment groups for the tumor growth curves. C-E Measurement of serum iron levels (C), plasma ferritin levels (D) and total iron binding capacity (TIBC) (E) in MMTV-PyMT model. **, P < 0.01, by two-tailed unpaired Student’s t-test
Iron supplementation in combination with anti-PD-1 immunotherapy increases CD8 + T-cell infiltration and cytotoxicity
To gain a better understanding into whether and how the addition of iron with PD-1 antibody treatment influences the tumor immune microenvironment to exert better therapeutic activity, we next performed scRNA-seq with application of the PIP-seq technology [15] to analyze CD45 + immune cells sorted via fluorescence-activated sorting from E0771 tumors (Supplementary Fig. S4A). After filtering, we continued with the analysis of 97,543 cells derived from control, iron or PD-1 antibody alone, and the combination treated samples. Unsupervised clustering divided those cells into 23 clusters (Fig. 5A and Supplementary Fig. S4B). Using differentially expressed marker genes, we identified four populations of myeloid cells: macrophages (TAMs, including clusters 1–4, 6, 9, 10, 17 and 18), monocytes (cluster 0), neutrophils (cluster 22) and dendritic cells (DCs, clusters 7, 13, 16 and 19); four types of lymphocytes: CD8 + T cells (clusters 5, 11 and 20), CD4 + T cells (clusters 8 and 14), NK cells (cluster 12), and B cells (cluster 21), as well as fibroblasts (cluster 15) (Fig. 5A and B). Comparing cell type proportions (Fe vs. Ctrl or PD-1 + Fe vs. PD-1), we observed a clear increase in CD8 + T-cell frequency with iron treatment while the frequency of CD4 + T-cell decreased (Fig. 5C). The frequencies of TAMs, the most abundant cell type and other types of cells, such as monocytes, DCs, neutrophils, NK and B cells, were either not apparently altered or not altered consistently when comparing combination treatment with PD-1 antibody only, iron alone and control treatment (Fig. 5C). We then focused on further looking into CD8 + T and CD4 + T-cell populations. First, immunohistochemistry staining showed a noticeable increase of CD8 + T-cell present in the combination treated tumors in both E0771 and PyMT models compared to monotherapy-treated tumors (Fig. 5D-F). Next, CD8 + T, CD4 + T, and NK cells were further subsetted for clustering and differential gene expression analysis, which enabled us to distinguish Treg (subclusters 1 and 3) and T helper (subcluster 7) subtypes among CD4 + T cells (Supplementary Fig. S5A and B). Upon iron administration, the proportion of Tregs was decreased while the proportion of CD4 + T helpers and cytotoxic CD8 + T cells were increased (Fig. 5G). Strikingly, the CD8 + T to Treg ratio (CD8/Treg) in the combination treatment group was more than doubled compared to iron or PD-1 antibody alone (Fig. 5H), suggesting a stronger anti-tumor response. Notably, gene set enrichment analysis comparing cytotoxic gene signature (composed of Cd8a, Ccl5, Gzma*, Gzmb, *Gzmk, Klrd1, Nkg7, Prf1 and Xcl1) (Supplementary Fig. S5C) revealed significantly higher enrichment scores in CD8 + T cells with combination treatment relative to PD-1 antibody alone (Fig. 5I). Along with these results, in PyMT model, tumors with the combination treatment had significantly more CD8 and granzyme B double-positive cells than those treated with PD-1 antibody alone (Supplementary Fig. S5D). Collectively, these results indicate that addition of iron could increase the infiltration and cytotoxic activity of CD8 + T cells, while decreasing intratumoral Treg frequencies to enhance efficacy of PD-1-based immunotherapy.Fig. 5. Iron supplementation increases CD8 + T-cell infiltration and cytotoxic activity of PD-1-based immunotherapy. A Unsupervised clustering, UMAP plot and cell type annotation of single-cell transcriptional profiling of CD45 + cells isolated from control, iron or PD-1 antibody alone and combination treated E0771 tumors. B Bubble plot showing the expression of marker genes for each cell type as annotated in A. Dot color indicates expression level and size indicates the proportion of cells expressing each gene. C Frequency of different main immune cell types as clustered and annotated in A, with indicated treatment. D and E Representative images of CD8 immunohistochemistry staining for E0771 tumors (D) and MMTV-PyMT tumors (E) with indicated treatment. Scale bar, 50 μm. Error bars represent SEM. F Quantification of CD8 + cells infiltrating in E0771 tumors (top) and MMTV-PyMT tumors (bottom) with indicated treatment. P-value determined by one-way ANOVA (top) and by Student’s t-test (bottom). G Percent of indicated cell subtypes determined by sub-clustering of T- and NK-cell clusters in E0771 tumors with indicated treatment. H Determining of CD8 + T to Treg ratio (CD8/Treg) as sub-clustered and annotated in G. I Violin plots showing enrichment scores for cytotoxic signature across CD8 + T cells grouped by different treatment. ***, P < 1 × 10^–10^, compared to PD-1 antibody treatment group by two-tailed Student’s t-test. The red dashed line shows the mean of cytotoxic signature score in PD-1 antibody monotherapy group
Iron supplementation diminishes immunosuppressive phenotype of TAMs
Since the overall proportion of TAMs showed no significant change upon treatment with iron, anti-PD-1 antibody, or combination therapy (Fig. 5C), we sought to investigate the subpopulation composition and functional states of TAMs following treatment with these agents. Interestingly, when we analyzed the differences in gene expression between treatment groups, we found lower levels of iron and heme metabolic genes transcripts (Supplementary Fig. S6A), such as Fth1 and Ftl1 (encode ferritin), Hmox1 and Hmox2 heme oxygenase enzymes*,* heme receptor Lrp1 and Tfrc (encodes transferrin receptor or CD71) in TAMs with combination or iron treatment compared to PD-1 antibody alone or control treatment, respectively. We next evaluated whether iron could affect the functional phenotypes of TAM clusters. Gene set enrichment analysis enabled us to identify and annotate TAM subsets: Prolif-TAM (cluster 17), IFN-TAM (cluster 3), LA-TAM (clusters 1 and 4), Inflam-TAM (cluster 6), Reg-TAM (clusters 9 and 18) and Angio-TAM (cluster 10), while cluster 2 showed no clear bias to a specific subtype (Fig. 6A). Notably, when examining the frequencies of TAM subpopulations, we found that Inflam-TAM, Reg-TAM and Angio-TAM subtypes were decreased, whereas LA-TAMs were increased upon iron treatment (Fig. 6B). Negligible changes were observed regarding Prolif-TAM, IFN-TAM and cluster 2 TAMs when comparing combination treatment to PD-1 antibody alone, or iron to control, respectively (Fig. 6B). Furthermore, combining iron with PD-1 antibody treatment decreased immunosuppression, Angio-TAM and Inflam-TAM signatures, while increased LA-TAM signature in TAMs as a whole compared to those of PD-1 antibody monotherapy (Fig. 6C and D), with immunosuppressive signature being notably reduced in Inflam-TAM, Reg-TAM and Angio-TAM subpopulations (clusters 6, 9, 18, and 10) (Supplementary Fig. S6B), suggesting that iron treatment diminishes the immunosuppressive phenotype of TAMs. Additionally, flow cytometry analyses showed that iron, PD-1 antibody alone and the combination treatment significantly decreased PD-L1 levels within TAMs (Supplementary Fig. S6D), while increasing them in the residual tumor cells **(Supplementary Fig. S6E), compared to the control tumors. The above findings suggest that iron supplementation could remodel the population structure and functional state of TAMs by polarizing them to a less immunosuppressive phenotype, ultimately mitigating immunosuppression within the TME and promoting anti-tumor activity of anti-PD-1 based immunotherapy.Fig. 6. Iron supplementation diminishes immunosuppressive phenotype of TAMs. A Bubble plot showing the enrichment of the molecular signature of TAM subsets across myeloid clusters. B Identification and quantification of TAM subsets based on enrichment of the molecular signature as shown in A. C Bubble plot showing the enrichment scores for the immunosuppression and subtype molecular signature in TAMs with indicated treatment. D Violin plots showing the enrichment scores for immunosuppression signature in TAMs with indicated treatment. , P < 1 × 10^–5^, ***, P < 1 × 10^–11^, compared to PD-1 antibody monotherapy by two-tailed Student’s t-test. The red dashed line shows the mean of enrichment score for immunosuppression signature
Iron may attenuate NF-κB inflammatory pathways and induce metabolic shifts to reprogram TAMs to a less immunosuppressive phenotype
To explore the mechanisms by which iron may reprogram TAMs and diminish their immunosuppressive function, we next leveraged our scRNA-seq data to compare the signaling pathway changes in TAMs derived from combination and anti-PD-1 monotherapy. This analysis revealed that TNF-α signaling via NF-κB, inflammatory response, hypoxia and glycolysis, heme signaling and wound healing were among the top down-regulated hallmark and canonical pathways in combination treated TAMs compared to PD-1 antibody monotherapy (Fig. 7A-F). NF-κB signaling is a master regulator of inflammation and immunosuppression. It is a key driver of TAM polarization towards an alternatively activated immunosuppressive phenotype by the production of inflammatory cytokines, promoting tumor progression and angiogenesis. Hypoxia in the TME is known to induce a metabolic shift towards glycolysis in TAMs. Heme signaling has critical immune modulatory functions in macrophages, which can trigger a functional switch toward immunosuppressive phenotypes [32]. Growing evidence has suggested that glycolysis contributes to immunosuppression by promoting alternatively activated (M2-like) polarization of TAMs and producing metabolites like lactate, which can dampen immune responses [33, 34]. Meanwhile, immunosuppressive TAMs are often associated with wound healing and tissue repair, exhibiting increased glycolysis. Indeed, many genes that are involved in NF-κB signaling, inflammatory response and immunosuppression, as well as hypoxia, tissue repair and wound healing were downregulated upon iron treatment, including toll-like receptors Tlr2, Tlr4 and co-receptor Cd14 that activate NF-κB signaling, NF-κB regulators Nfkbia and Nfkbiz, inflammatory cytokines like Tnf and its induced proteins (Tnfaip2 and 3), Il-1β, Il-6 and chemokines Cxcl2, Ccl2, Pf4, inflammasome component Nlrp3, factors associated with immunosuppression like Cd274/PD-L1, Tgfβ1, Ptgs2, Arg1 and Spp1, hypoxia-inducible factor 1-alpha (Hif1a) that regulates the cellular response to hypoxia, mannose receptor C-type1(Mrc1/CD206) and macrophage metalloelastase Mmp12 that play roles in tissue repair, wound healing, and inflammation (Fig. 7G). Of note, these results were validated in iron treated BMDMs (Fig. 7H). In addition to reducing the expression of genes associated with immunosuppression (Fig. 3C), iron treatment caused a drastic downregulation of chemokines and inflammatory mediators that can promote immunosuppressive phenotype, such as Ccl2, Ccl4 and Ccl7; transcription factor Hif1a and inflammasome component Nlrp3; as well as NF-κB signaling components, including Tnf, Il-1β, Il-6, chemokines Cxcl1 and Cxcl2; and hemoglobin subunit Hba-a1 (Fig. 7H). These systemic and interconnected alterations indicate that iron may reprogram TAMs and diminish their immunosuppressive function potentially by modulating NF-κB inflammatory pathways and metabolic activity.Fig. 7. Iron may attenuate NF-κB inflammatory pathways and induce metabolic shifts to reprogram TAMs. A Volcano plots comparing enrichment score of hallmark gene sets and canonical pathways in TAMs from E0771 tumor with PD-1 antibody alone or PD-1 antibody plus iron combination treatment. Each point represents one pathway or gene set. Each point represents one pathway/gene set. X-axis, mean difference of single cell enrichment score (the score of PD-1 antibody treated TAMs – that of combination treated TAMs); Y-axis, -log10 (P-value). P-value determined by Student’s t-test. B-F Violin plots showing the relative enrichment score of the indicated top down-regulated hallmark gene sets and canonical pathways in combination treated TAMs compared to PD-1 antibody monotherapy. ***, P < 1 × 10^–10^, compared to PD-1 antibody monotherapy by Student’s t-test. G Bubble plot showing the expression levels for genes involved in NF-κB signaling, inflammatory response, immunosuppression, hypoxia and wound healing. Dot color indicates expression level and size indicates the proportion of cells expressing each gene. H Expression of genes associated with inflammatory response and NF-kB signaling in BMDMs with or without iron treatment determined by RT-qPCR. mRNA levels were normalized to Gapdh. Error bars represent SD. ***, P < 0.0001, by two-tailed Student’s t-test
Discussion
Macrophages are often abundant in tumors, and accumulating evidence indicates that TAMs enable and sustain multiple aspects of the hallmarks of cancer, including evading growth suppression and immune destruction, tumor promoting inflammation, activating invasion and metastasis, angiogenesis and conferring resistance to therapies [8, 10, 29, 30]. Growing interest has been focused on targeting TAMs to improve current cancer treatment. In this study, we show that TAMs with high expression for iron metabolic genes also harbor robust immunosuppressive transcriptional features in both murine and human breast cancers. This finding is well in line with a recent study that iron-rich macrophages identified in mouse and human tumor microenvironments have immunosuppressive and angiogenic properties that enhanced tumor growth [35]. We also demonstrate that systemic iron supplementation can reprogram TAMs and diminish their immunosuppressive function, as well as increase infiltration and cytotoxic activity of CD8 + T cells, which results in enhanced efficacy of anti-PD-1 immunotherapy in two preclinical breast cancer models. Hence, targeting iron metabolic pathway may represent a promising approach to combine with clinical immunotherapies to improve patient outcomes.
Our results, along with findings by others, underscore the critical role of iron in regulating polarization and function of TAMs, which supports the notion that macrophages undergo polarization towards a classically activated, pro-inflammatory state with antitumor activities under iron sufficient conditions, while a deficiency in iron usually leads to more pro-tumor, anti-inflammatory phenotypes [12], although this dichotomy has been considered oversimplified considering TAMs could show overlapping pro-inflammatory/anti-inflammatory features as evidenced by single cell studies [7]. The NF-κB pathway is a master regulator of innate immune responses and macrophage polarization. Depending on the specific stimuli and context, NF-κB activation in TAMs has been associated with the acquisition of both immunosuppressive and immunostimulatory phenotypes, with opposite effects on tumor progression and treatment response [36, 37]. We show that iron treatment reduces the suppressive capacity of primary macrophages and promotes T cell proliferation in vitro. Moreover, iron supplementation reduces immunosuppressive TAM subsets and antitumor responses of CD8 + T cells in vivo. Both our in vitro and in vivo data show downregulation of inflammatory cytokines and chemokines in macrophages accompanied by decreased NF-κB signaling upon iron administration, which indicate that increased iron availability in the breast cancer TME may shift TAMs away from an immunosuppressive phenotype, coordinated with an enhancement of antitumor immune response, possibly through downregulation of the NF-κB inflammatory program. Further investigation of the underlying mechanisms and the interplay of iron metabolism, NF-κB signaling and TAM function in different scenarios may provide insights for the development of novel cancer immunotherapies targeting TAMs.
Prior reports have proposed both iron supplementation and iron restriction/chelation as strategies to modulate TAM function and antitumor immunity. A study in non-small cell lung cancer observed that the infiltration of iron-loaded TAMs correlated with tumor regression in patients. In this study, ex vivo hemolytic red blood cells exposed TAMs or in vivo iron oxide nanoparticles treatment could repolarize TAMs to exert antitumor function, suggesting the delivery of iron to TAMs could be an applicable adjuvant therapeutic strategy to promote antitumor immune responses [38]. Another study showed that the FDA-approved iron supplement, ferumoxytol, inhibits the growth of early-stage breast cancer and metastasis of lung cancer by increasing tumor infiltrating pro-inflammatory macrophages [39]. While iron deficiency induced by an iron-deficient diet promotes 4T1 mouse mammary tumor and MDA-MB-231 human breast cancer growth and metastasis, importantly, correcting iron deficiency by iron therapy with intraperitoneal injection of iron dextran reduces primary tumor growth and lung metastasis [40]. Intriguingly, iron supplementation by a high iron diet promotes antitumor responses by increasing IFNγ production in T cells and improves the efficacy of anti-PD-1 cancer immunotherapy in mouse models of colorectal and lung cancer [41]. However, a study showed that intravenous iron isomaltoside administration increases tumor growth, inhibits the efficacy of chemotherapy with IL-2/doxorubicin and moderately impairs anti-PD-L1 immunotherapy in the E0771 mouse breast cancer model [42]. Another study reported that deferiprone, an FDA-approved iron chelator, synergizes with chemotherapy and delays metastatic disease progression of mice with ovarian cancer by bolstering type-I IFN responses that activates NK cells to restrain metastatic disease [43]. In leptomeningeal metastasis, inflammatory cytokines derived from macrophages within cerebrospinal fluid could induce iron-binding protein lipocalin-2 expression in cancer cells, enabling them to capture limiting iron for survival and growth [44]. This study also demonstrated that iron chelation therapy with deferoxamine impairs metastatic cancer cell growth in mice [44]. Our work highlights the benefits of iron supplementation on the efficacy of anti-PD-1 immunotherapy in the context of breast cancer. The effects of iron on the tumor microenvironment and antitumor immunity are likely context dependent, and the contrasting results from different studies may be explained by variations in the model system, cancer type, disease stage, iron formulation, and treatment regimen. Nevertheless, our data suggests that within the breast cancer tumor microenvironment, iron supplementation can synergize with anti-PD-1 therapies to decrease the immunosuppressive features of TAMs and improve outcomes in preclinical models.
Epidemiological studies show a high prevalence of iron deficiency anemia in breast cancer patients and other malignant diseases [45–47]. A study reported that iron deficiency is significantly associated with lymph node invasion and iron overload appears beneficial for survival and limits tumor recurrence in young breast cancer patients [40]. Another investigation in lung cancer patients found that the quality and efficacy of the antitumor response following anti-PD-1 immunotherapy with nivolumab may be modulated by plasma ferritin levels, indicating a positive relationship between nivolumab therapy and iron supplementation [41]. Importantly, in our models, iron supplementation also ameliorated laboratory indicators of iron deficiency-associated anemia, further supporting its potential clinical relevance (Fig. 4D and E). These data would suggest the necessity to correct cancer-associated anemia in breast cancer, even when asymptomatic. The effects of iron application on reducing immunosuppressive TAMs within the TME in our study support combination strategies that integrate iron-based interventions with anti-PD-1 immunotherapy (and potentially other immunotherapies), especially in those with iron-deficient anemia. Given the current clinical use of multiple iron formulations with favorable safety profiles, translation to clinical trials is feasible. However, further investigation is required to determine optimal dosing, timing of treatment, patient selection and biomarkers/gene signatures, and caution is warranted in light of the dual roles of iron reported in tumor growth and anticancer immune response.
This study has several limitations. First, our mechanistic findings are largely based on unbiased transcriptomic analysis, while the actual function of metabolic and signaling pathways is a complex process, coordinately determined by multiple factors, including protein abundance, enzymatic activities, and post-transcriptional regulations, which were not fully explored here. Additional complementary strategies, such as protein level validation or metabolic proteomic profiling, and gain- and loss-of-function studies, are needed to better characterize TAMs biology following iron and immunotherapy in future studies. Second, considering the multifaceted and systemic role of iron, its potential direct or indirect effects on other immune cells could not be excluded; for instance, we observed changes in CD4 + T helpers and Tregs following iron and immunotherapy. Further investigation is required to determine how iron affects the phenotype, trafficking, and function of different T-cell subpopulations within the TME and systemically, and how these changes ultimately influence anti-tumor immune responses.
Conclusion
Our study reveals iron metabolic activity is closely associated with the functional phenotype of TAMs, and demonstrates the pivotal roles of iron in reprograming TAMs away from an immunosuppressive state and enhancing the efficacy of anti-PD-1 immunotherapy in preclinical breast cancer models. Therefore, iron supplementation may have clinical potential to benefit breast cancer patients receiving anti-PD-1 immunotherapy, which warrants further investigation.
Supplementary Information
Supplementary Material 1: Figure S1. The association between iron metabolism and immunosuppression in TAMs of human breast cancer. A, Bubble plot showing the expression levels of signature genes associated with immunosuppression. B, Correlation between enrichment scores for iron uptake and transport, and immunosuppression signatures across all cells colored by cell type based on the study by Wu et al, 2020. Each point represents a single cell. Cell type classification and detailed patient information can be found in the source study. C, Bubble plot showing the expression levels of major iron and heme metabolic genes in the indicated cell types. D, Bubble plot showing the expression of marker genes for identifying the subtypes of myeloid cells. Dot color indicates expression level and size indicates the proportion of cells expressing each gene. E and F, Violin plots showing enrichment scores for immunosuppression signature (E), and iron uptake and transport pathway signature (F) among the identified cell types from the indicated clinical subtypes of breast cancer based on the study by Wu et al, 2021. The red dashed line shows the mean of the enrichment scores. G, Correlations between enrichment scores for iron uptake and transport and immunosuppression signatures across all cells (colored by cell type) from the indicated clinical subtypes of breast cancer based on the study by Wu et al, 2021. Figure S2. Analysis of single-cell RNA-Seq data from E0771 breast cancer model. A, Clustering and UMAP plots showing all retained cells merged (left) and retained cells from each E0771 tumor/sample (right 3 panels) for analysis after scRNA-seq. B, Bubble plot showing the expression levels of signature genes associated with immunosuppression. C, Correlations between enrichment scores for iron uptake and transport and immunosuppression signatures across all cells colored by cluster (left) and colored by cell type (right). Each point represents a single cell. Figure S3. In vitro iron treatment had no significant effect on viability of TNBC cells with clinically relevant doses. A and B, Effect of iron treatment (with doses from 0.5 to 50 mM) on viability of TNBC cells HS578T (A) and MDA-MB-468 (B). Representative cell growth plots where each point represents the mean ± SD (n = 5-6 per condition). Experiments were repeated at three independent times with similar results. **, P< 0.01 versus other treatment at day 3, by two-tailed unpaired Student’s t-test. Figure S4. Single-cell RNA-Seq profiling to characterize tumor infiltrated immune cells after iron supplementation in E0771 model. A, Gating strategy for flow sorting of viable tumor infiltrating CD45+ cells. B, Clustering and UMAP plots of scRNA-seq data showing cells isolated from E0771 tumors with indicated treatment. Figure S5. Characterization of T and NK cells after iron treatment in E0771 and PyMT tumors. A, UMAP plot and clusters of unbiased sub-clustering of T and NK cells in E0771 tumors. B, Bubble plot showing the expression levels of marker genes for the subtype annotation of T and NK cells as clustered in A. C, Bubble plot showing the expression levels of cytotoxic signature genes in each subclusters as clustered in A. D, Representative images and quantification of CD8 and granzyme B immunofluorescence staining in MMTV-PyMT tumors with the indicated treatments. Arrows indicate CD8 and granzyme B double-positive cells. Scale bar, 50 mm. Error bars represent SEM. P-value determined by Student’s t-test. Figure S6. Iron supplementation diminishes immunosuppressive phenotype of TAMs. A, Bubble plot showing the expression levels of major iron metabolic genes in TAMs of E0771 tumors with indicated treatment. Dot color indicates expression level and size indicates the proportion of cells expressing each gene. B, Violin plots showing enrichment scores of immunosuppression gene signature among TAM, monocyte and neutrophil clusters with indicated treatment. ns, not significant. *, P < 0.01 and **, P < 1 x 10^-5^ vs PD-1 by two-tailed Student’s t-test. The red dashed line shows the mean of enrichment score. C-E, Analysis of PD-L1 levels in TAMs and tumor cells of E0771 model after the indicated treatments by flow cytometry. C, Gating strategy. D and E, representative histograms and quantification of PD-L1 expression by mean fluorescence intensity (MFI) in TAMs (D) and tumor cells (E). Error bars represent SD. P-value determined by one-way ANOVA with Tukey’s test.