Inhibition of histone demethylase LSD1 suppresses CD47 expression and enhances efficacy of CD47 blockade in breast cancer

Abstract

Background

CD47 functions as a “don’t eat me” checkpoint, inhibiting macrophage-mediated phagocytosis in triple-negative breast cancer (TNBC). While anti-CD47 therapies can restore immune surveillance, their efficacy in TNBC is often limited by immune evasion and drug development challenges.

Methods

We investigated the crosstalk between the histone demethylase lysine-specific demethylase 1 (LSD1) and CD47 signaling in TNBC using in silico datasets, isogenic cell lines, conditional BRCA1 knockout models, and syngeneic mouse models. Techniques such as immunohistochemistry, multiplex immunofluorescence, immunoprecipitation, protein ubiquitination, chromatin immunoprecipitation, chemotaxis, flow cytometry, and phagocytosis assays were employed to examine the epigenetic regulation of CD47 by LSD1 and its impact on antitumor immunity. The efficacy of combining LSD1 inhibition with anti-CD47 therapy was evaluated in BALB/cJ mice bearing TNBC tumors.

Results

In TNBC tumors, CD47 expression is positively correlated with elevated LSD1 levels, which are associated with increased infiltration of M2 macrophages and a concomitant decrease in M1 macrophages and CD8+T cells. Inhibition of LSD1 led to downregulation of CD47 by suppressing the expression and activity of its key transcriptional regulators, NF-κB (p65) and STAT3, through distinct mechanisms. Loss of LSD1 reduced p65 transcription, which was linked to an accumulation of the repressive histone mark H3K9me2 at the p65 promoter. Conversely, LSD1 inhibition promoted polyubiquitination and subsequent destabilization of STAT3. LSD1 inhibition also enhanced phagocytosis and promoted M1 polarization of macrophages. CD47 downregulation induced the production of interferon-γ and Th1-type chemokines in TNBC cells, facilitating increased tumor infiltration of CD8+T cells. Furthermore, combining LSD1 inhibition with anti-CD47 therapy significantly improved antitumor efficacy compared with monotherapy in syngeneic tumor models, without inducing significant toxicity. This combination therapy also promoted a shift in macrophage polarization toward the M1 phenotype, further enhancing CD8+T cell infiltration within tumors. Depletion of CD8+T cells significantly diminished the antitumor efficacy of the combination therapy.

Conclusion

Targeting LSD1 enhances the efficacy of anti-CD47 therapy by overcoming immune evasion, offering a promising strategy to improve immunotherapy outcomes in TNBC.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• CD47 acts as a “don’t eat me” signal and immune checkpoint, inhibiting macrophage-mediated phagocytosis in triple-negative breast cancer (TNBC). While CD47 blockade can restore immune surveillance, anti-CD47 therapies face significant challenges, including on-target, off-tumor toxicities. Lysine-specific demethylase 1 (LSD1), a key histone demethylase, plays a role in regulating the tumor immune response, but its involvement in CD47-mediated immune evasion in TNBC remains poorly understood.

WHAT THIS STUDY ADDS

• This study provides novel evidence linking elevated LSD1 expression in TNBC tumors with increased CD47 expression, enhanced M2 macrophage infiltration, and decreased M1 macrophages and CD8+T cells. Mechanistically, LSD1 inhibition suppresses CD47 expression by downregulating NF-κB and STAT3 through distinct mechanisms. Functionally, LSD1 inhibition enhances macrophage phagocytosis, promotes M1 polarization, and increases CD8+T cell infiltration in TNBC. The combination of LSD1 inhibition with anti-CD47 blockade improves antitumor efficacy and demonstrates a favorable safety profile, suggesting a promising therapeutic strategy.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This study presents a novel approach targeting the crosstalk between LSD1 and immune components, offering a potential therapeutic strategy for breast cancer patients with suboptimal immune responses. As LSD1 inhibitors advance in clinical trials, their combination with anti-CD47 therapy could improve immunotherapy outcomes, providing new treatment options for patients who do not respond adequately to current therapies.

Introduction

Breast cancer (BC) is a leading cause of cancer-related mortality in women, characterized by significant molecular heterogeneity that influences therapeutic responses.^{1 2} A major clinical challenge is therapeutic resistance, both intrinsic and acquired, which limits treatment efficacy and contributes to relapse and metastasis.^{3 4} Resistance arises from factors such as genetic mutations, epigenetic changes, tumor microenvironment alterations, enhanced drug efflux, and therapy-induced senescence.^{5 6} Recent evidence highlights the critical role of the immune system in BC tumorigenesis, progression, and treatment resistance. Immune checkpoint blockade (ICB) has emerged as a promising strategy, particularly for refractory subtypes like triple-negative breast cancer (TNBC).⁷ However, the response to ICB in TNBC remains limited, with programmed cell death protein 1 (PD-1)/programmed death-ligand 1 (PD-L1) inhibitors as monotherapy in metastatic TNBC yielding an overall response rate of only 5%–25%.7,10

A hallmark of TNBC is its ability to manipulate the immune microenvironment, promoting tumor progression by recruiting immune cells that facilitate immune suppression, such as regulatory T cells, myeloid-derived suppressor cells, and M2-type tumor-associated macrophages (TAMs). These cells create an immunosuppressive microenvironment that limits the activity of cytotoxic T cells, M1 macrophages, and natural killer (NK) cells, which are critical for targeting and eliminating cancer cells.^{7 11} Macrophages exist in two polarization states: M1 (pro-inflammatory) and M2 (immune-suppressive). M1 macrophages are generally associated with promoting antitumor immunity, whereas M2 macrophages support tumor growth and immune suppression.¹² In TNBC, an increased presence of M2 macrophages correlates with poor prognosis and enhanced tumor progression.¹³ These macrophages are recruited to the tumor by cytokines, growth factors, and chemokines secreted by cancer cells. CD47, a transmembrane protein, serves as a ligand for the immune inhibitory receptor signal-regulatory protein alpha (SIRPα) on macrophages and dendritic cells (DCs), acting as a “don’t eat me” signal that inhibits phagocytosis and prevents adaptive immune activation.¹⁴ In cancer cells, CD47 is frequently upregulated, aiding immune evasion and promoting tumor tolerance.¹⁵ The expression of CD47 is regulated by key transcription factors such as c-Myc, HIF-1α, STAT3, and NF-κB.16,18 Targeting CD47 with monoclonal antibodies (mAbs) has shown promise in BC therapy.^{19 20} However, challenges remain, including off-target toxicity, where CD47 blockade may trigger immune activation against normal tissues, leading to adverse effects such as anemia.²¹ Additionally, resistance mechanisms may allow tumors to evade immune surveillance despite CD47 blockade, complicating treatment efficacy.²⁰ Therefore, more effective combinatorial strategies are needed to enhance CD47-targeted therapies.

Epigenetic alterations, such as DNA hypermethylation and histone dysregulation, are critical in the initiation and progression of TNBC.^{22 23} Histone lysine-specific demethylase 1 (LSD1, also known as KDM1A or AOF2) was the first identified flavin adenine dinucleotide-dependent histone demethylase. LSD1 removes mono- and dimethyl groups from lysine 4 (H3K4) or lysine 9 (H3K9) of histone H3, leading to either repressive or activating transcriptional marks, depending on the cellular and chromatin context.24,27 The LSD1-CoREST complex also regulates macrophage polarization, with LSD1 inhibition promoting an antitumor M1-like macrophage phenotype.²⁸ However, the mechanisms by which LSD1 regulates the “don’t eat me” signal of CD47 and modulates macrophage-mediated phagocytosis in TNBC remain unknown. In this study, we investigate CD47-mediated immune evasion in TNBC and identify LSD1 as a key regulator of CD47 expression and immune modulation. We demonstrate that LSD1 inhibition suppresses CD47 expression by downregulating key transcription factors, including NF-κB and STAT3. This suppression of CD47 enhances the expression of Th1-type chemokines, promoting CD8+T cell infiltration into the tumor microenvironment. Furthermore, we provide preclinical evidence that combining the LSD1 inhibitor (LSD1i) with anti-CD47 blockade significantly improves therapeutic efficacy in TNBC.

Materials and methods

Reagents, plasmids, cell lines

SP2509 was obtained from Selleck Chemicals (Houston, Texas, USA). Tranylcypromine and MG132 were purchased from Thermo Fisher Scientific (Waltham, Massachusetts, USA). The CD47 and CD8 mAbs for animal studies were acquired from Bio X Cell (Lebanon, New Hampshire, USA). Plasmids pCMV3-N-GFPSpark, pCMV3-N-GFPSpark-RELA/NF-κB (p65) (mouse), and pCMV3-N-GFPSpark-STAT3 (mouse) were obtained from Sino Biological (Paoli, Pennsylvania, USA). pcDNA3-HA-ubiquitin was used as previously described.²⁹ The MDA-MB-231, 4T1, EMT6, and RAW 264.7 cell lines were sourced from American Type Culture Collection (ATCC) (Manassas, Virginia, USA). Mouse primary bone marrow-derived macrophages were obtained from iXCells Biotechnologies (San Diego, California, USA). Cells were cultured in growth medium as previously described.30,32

Immunohistochemistry

Formalin-fixed, paraffin-embedded tissue sections (5 mm thick) were deparaffinized and rehydrated through a graded ethanol series, then treated with 3% hydrogen peroxide to block endogenous peroxidase activity. Antigen retrieval was carried out by heating the sections in boiling citrate buffer (pH 6.0) for 10 min. The sections were subsequently incubated with primary antibodies, followed by a biotinylated secondary antibody. Signal amplification was achieved using the VECTASTAIN Elite ABC system, and staining was visualized with 3,3′-diaminobenzidine. Slides were counterstained with hematoxylin. Finally, samples were examined using a Leica inverted microscope, and staining intensity was quantified with ImageJ software.

Real-time qPCR

RNA extraction was performed using TRIzol reagent (Fisher Scientific, Pittsburgh, Pennsylvania, USA). Reverse transcription was carried out with moloney murine leukemia virus (MMLV) reverse transcriptase, as previously described.^{33 34} Quantitative real-time PCR was performed on the Applied Biosystems QS3 Real-Time PCR system, following established protocols.²⁹ Predesigned TaqMan Gene Expression Assays (Thermo Fisher Scientific) were used for gene expression analysis. Relative gene expression was calculated using the ΔΔCt method, with β-actin serving as the internal control.

Immunoblotting

Cell lysates were collected as previously described.^{30 35 36} Total protein concentration was measured using the BCA Protein Assay Kit (Thermo Scientific). Proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) on 12% polyacrylamide gels and transferred to polyvinylidene difluoride (PVDF) membranes. Membranes were developed using enhanced chemiluminescence. β-actin was used as a loading control. Blot images were captured using the Invitrogen iBRIGHT CL750 DSS imaging system. The antibodies used are listed in online supplemental table 1.

siRNA transfection

Pre-designed siRNA and non-targeting scramble siRNA (Santa Cruz Biotechnology, Dallas, Texas, USA) were transfected into cells according to the manufacturer’s protocol. Cells were collected post-transfection for subsequent analysis. The siRNAs used in this study are listed in online supplemental table 2.

Chromatin immunoprecipitation

Chromatin immunoprecipitation (ChIP) assays were performed using the MilliporeSigma ChIP Assay Kit, following protocols from our previous studies.^{29 31} 2 million cells were used per assay. Quantitative ChIP was conducted to assess changes in STAT3 and p65 binding at the CD47 promoter, as well as H3K4me2 and H3K9me2 marks at the p65 promoter, using quantitative PCR (qPCR) with SYBR Green Master Mix (Thermo Fisher Scientific). Primer sets are listed in online supplemental table 3. All primers were ordered and synthesized by Integrated DNA Technologies (Coralville, Iowa, USA).

Immunoprecipitation and ubiquitination assays

Cell lysates were prepared using immunoprecipitation (IP) lysis buffer, as previously described.²⁹ IP was performed by incubating lysates from EMT6 cells with LSD1, STAT3, p65, or control IgG antibodies, followed by the addition of protein G-plus agarose beads (MilliporeSigma, Burlington, MA, USA). The resulting complexes were subjected to immunoblotting using LSD1, STAT3, or p65 antibodies. For the ubiquitination assay, EMT6 cells were transfected with HA-Ubiquitin and treated with SP2509 for 24 hours, with co-treatment with 20 μM MG132 for 8 hours. Lysates were then collected and subjected to IP with Protein G-plus agarose beads. Immunoblots were performed using anti-HA, anti-STAT3, or anti-β-actin antibodies.

Enzyme-linked immunosorbent assay

ELISA kits for detecting mouse CXCL9 and CXCL10 (IP-10) were purchased from Abcam (ab203364/ab260067). Supernatants from EMT6 cells were collected 48 hours post-transfection with CD47 siRNAs. Experimental procedures were performed according to the manufacturer’s protocols, and absorbance was measured at 450 nm using the Agilent BioTek Synergy spectrophotometer.

Chemotaxis/migration assay

Human peripheral blood CD8+T cells (STEMCELL Technologies, Vancouver, Canada) were starved in serum-free Dulbecco's Modified Eagle Medium (DMEM) for 24 hours, followed by harvesting, washing, and counting. A total of 50,000 CD8+T cells were placed in the upper chambers of a Cell Migration/Chemotaxis Assay plate (Abcam, Cambridge, UK) and allowed to migrate towards the cell supernatants in the lower chamber, which were derived from cells transfected with siRNAs. After a 24 hours incubation period, 10 μL of cell dye was added to each well, and fluorescence was measured at Ex/Em=530/590 nm using the Agilent BioTek Synergy fluorescence spectrophotometer. A standard curve was generated by performing a twofold serial dilution of cells in wash buffer to correlate cell numbers with relative fluorescence units.

Macrophage phagocytosis assays

The mouse macrophage cell line RAW 264.7 was labeled with CellTracker red fluorescent membrane dye (TargetMol PKH, Fisher Scientific) and seeded into 24-well culture plates. EMT6 cells treated with scramble or LSD1 siRNA were stained with Hoechst 33342 (blue, 5 μg/mL) to visualize nuclei and then co-cultured with labeled RAW 264.7 macrophages for 2 hours to allow macrophage-tumor cell interactions and phagocytosis. Phagocytosis was assessed by fluorescence microscopy, identifying RAW 264.7 macrophages engulfing EMT6 cells. Images were captured using a Leica inverted fluorescence microscope. For the flow cytometry-based phagocytosis assay, tumor cells were labeled with Hoechst 33342 for 5 min and co-cultured with RAW 264.7 cells or mouse primary bone marrow-derived macrophages (iXCells, 10MU-030) for 2 hour. Cells were then stained with 2 μL (per 200 μL sample) of PE-conjugated F4/80 rat anti-mouse antibody (BD Biosciences, BDB565410) at 4°C for 30 min in the dark. Samples were analyzed by flow cytometry using a Becton Dickinson LSR II at the Flow Cytometry Core Facility of the University of Iowa. Data were subsequently analyzed using FlowJo software (BD Biosciences).

Flow cytometry

Single-cell suspensions from mouse syngeneic tumors were prepared according to a previously established protocol.³⁷ Approximately 1×10⁶ cells were resuspended in 100 µL Live/Dead staining solution (Thermo Fisher Scientific). Cells were then incubated with anti-mouse CD16/32 antibody (Thermo Fisher Scientific) to block Fc receptors for 15 min on ice. Following Fc receptor blocking, cells were divided into several groups and stained with antibody mixtures containing fluorochrome-conjugated antibodies targeting specific surface markers (online supplemental table 4). The cells were incubated with the antibodies for 30 min at 4°C in the dark. Following incubation, the cells were washed two times with flow cytometry buffer and resuspended in 300–500 µL phosphate-buffered saline for analysis. Data were acquired using the Cytek Aurora Spectral Flow Cytometer at the University of Iowa Flow Cytometry Core Facility. Subsequent analysis was performed using FlowJo software. Dead cells and debris were excluded based on forward and side scatter parameters. Fluorescence compensation was performed using single-stained and unstained beads.

Animal studies

All animal studies were conducted at the Medical Lab animal facility of the University of Iowa in accordance with protocols approved by the University of Iowa Institutional Animal Care and Use Committee. Brca1-deficient mice (Trp53^tm1Brd Brca1^tm1Aash Tg(LGB-cre)74Acl/J, Stock No: 012620) were sourced from The Jackson Laboratory (Bar Harbor, Maine, USA). These mice carry the beta-lactoglobulin Cre (BLG-Cre) transgene, homozygous for floxed exons 22–24 of the Brca1 allele, and heterozygous for the p53 tumor suppressor gene deficiency. Mouse genotyping was performed using the Phire Tissue Direct PCR Master Mix kit (Fisher Scientific) as previously described.³⁸ After two rounds of pregnancy, tumors were allowed to develop and were then used for subsequent experiments. Syngeneic EMT6 tumors were established by injecting 1×10⁶ tumor cells into the mammary fat pads of 6–8 weeks old female BALB/cJ mice (The Jackson Laboratory). When tumor volume reached 50–100 mm³, the mice were randomized into experimental groups. SP2509 (50 mg/kg) was administered daily via intraperitoneal injection, while anti-CD47 antibodies (5 mg/kg) were given intraperitoneally every 3 days. CD8+T cells were depleted by intraperitoneal injection of anti-CD8 antibodies (250 μg per mouse) every 4 days. Vehicle and isotype-matched antibodies served as controls. Tumor volumes were measured with calipers and calculated as (length×width²)/2. The primary ethical endpoint for mouse tumors is a maximum diameter of 2 cm in any dimension.

Statistical analysis

Data are presented as mean±SD of three independent experiments. Quantitative variables were analyzed using a two-tailed Student’s t-test. Statistical analyses were performed with GraphPad Prism (GraphPad Software, La Jolla, California, USA). immunohistochemistry (IHC) staining was quantitatively assessed by calculating the average optical density using ImageJ software. For multiple groups comparison in animal studies, differences in tumor growth were analyzed using analysis of variance with Tukey’s post hoc multiple comparison test. P values<0.05 were considered statistically significant for all tests. Significance symbols are defined as *p<0.05, **p<0.01, ***p<0.001.

Results

LSD1 upregulates CD47 expression and is associated with M2-type macrophage polarization in TNBC

Our previous study showed that shRNA-mediated knockdown of LSD1 significantly suppressed the in vivo growth of syngeneic EMT6 tumors.³⁸ To determine whether LSD1 loss alters the tumor immune microenvironment, we analyzed immune-related gene expression in these tumors. qPCR revealed that LSD1 knockdown markedly increased Ifn-γ expression while concomitantly reducing Cd47 expression (figure 1A). Interferon (IFN)-γ is a key cytokine that promotes cellular immunity and drives antitumor immune responses.³⁹ In contrast, CD47 acts as an innate immune checkpoint by inhibiting macrophage-mediated phagocytosis.⁴⁰ By acting as a “don’t eat me” signal, CD47 suppresses phagocytic clearance by macrophages and DCs and is frequently overexpressed in cancers, where it correlates with poor clinical outcomes.⁴¹ In silico analysis of the Cancer Genome Atlas (TCGA) data revealed a positive correlation between LSD1 and CD47 expression in clinical BC samples (TIMER3 database) (figure 1B). Further analysis of the GSE62944 dataset showed elevated CD47 messenger RNA (mRNA) levels in basal-like breast tumors compared with other BC subtypes and normal-like tissues (figure 1C). However, the role of CD47 overexpression in breast tumorigenesis and its relationship with LSD1 remain poorly understood.

Figure 1. LSD1 inhibition attenuates CD47 expression in TNBC. (A) qRT-PCR analysis of immune regulatory factors in EMT6 tumors expressing control or LSD1 shRNA (n=6 per group). (B) Analysis of TCGA data showing a positive correlation between LSD1 and CD47 expression. (C) CD47 mRNA expression across breast cancer subtypes (TCGA database: https://www.ncbi.nlm.nih.gov/pubmed/26209429). (D) Immunohistochemical staining of CD47, F4/80, iNOS, arginase 1 and CD8 in mouse BRCA1-deficient mammary gland tumors and matched adjacent mammary tissues. Scale bar, 200 μm. Quantification of staining intensity was performed by AOD analysis using ImageJ software (n=3 per group). (E) qRT-PCR analysis of Lsd1 and Cd47 mRNA expression in EMT6, 4T1 and MDA-MB-231 cells transfected with control or LSD1-targeting siRNA for 24 hours. β-actin was used as an internal control. (F) Immunoblot analysis of LSD1 and CD47 in EMT6, 4T1 and MDA-MB-231 cells transfected with control or LSD1-targeting siRNA for 48 hours. (G) Immunoblot analysis of CD47 in EMT6 and MDA-MB-231 cells treated with increasing concentrations of SP2509 for 24 hours. CD47 antibodies: Santa Cruz, sc53050 for EMT6 and 4T1 cells; Santa Cruz, sc12730 for MDA-MB-231 cells. Data are mean±SD from three independent experiments unless otherwise indicated. *p<0.05, **p<0.01, ***p<0.001. AOD, average optical density; IFN, interferon; IHC,immunohistochemistry; IL, interleukin; KD, knockdown; LSD1, lysine-specific demethylase 1; mRNA, messenger RNA; NS, not significant; PD-L1, programmed death-ligand 1; qRT-PCR, quantitative reverse transcription polymerase chain reaction; SCR, scramble; TCGA, the Cancer Genome Atlas; TNBC, triple-negative breast cancer; TNF, tumor necrosis factor.

To investigate this, we employed a genetically engineered BRCA1-deficient mouse TNBC model (Brca1^{f22−24/p53±/BLG-Cre}; The Jackson Laboratory), which harbors a conditional knockout of BRCA1 and a heterozygous deletion of p53 specifically in mammary epithelial cells. Cre recombinase expression is driven by the BLG-Cre promoter during lactation (online supplemental figure 1A). Following two rounds of pregnancy to activate Cre recombinase, BRCA1-deficient tumors were induced in the mouse mammary glands, and both tumor and adjacent normal tissues were subsequently harvested (online supplemental figure 1B,C). These tumors exhibited increased levels of LSD1 and Ki-67 compared with adjacent mammary tissues (online supplemental figure 2A,B). IHC analysis revealed significantly higher CD47 expression in BRCA1-deficient tumors than in adjacent normal tissue (figure 1D). While the total macrophage population (F4/80+) remained statistically unchanged between tumors and adjacent normal tissues, M1-type macrophage infiltration (iNOS+) was significantly reduced, whereas M2-type macrophages (arginase 1+) were markedly elevated in the tumors (figure 1D). Furthermore, CD8+T cell infiltration was substantially diminished in tumors compared with normal tissue, while the CD4+T cell population remained unaffected (figure 1D; online supplemental figure 2C). Together, these findings suggest a positive association between LSD1 and CD47 in BRCA1-deficient mammary tumors and indicate that the immune microenvironment in these tumors is skewed toward a protumorigenic M2-like TAM phenotype. This is coupled with impaired cytotoxic T lymphocyte (CTL) infiltration during tumor progression.

To further investigate the role of LSD1 in regulating CD47 expression and its impact on macrophage function, we performed siRNA-mediated knockdown of LSD1 in multiple human and murine TNBC cell lines. This knockdown resulted in a significant reduction in CD47 expression at both the mRNA and protein levels (figure 1E,F). Consistent with these findings, treatment with LSD1 inhibitors (SP2509, tranylcypromine) also led to a downregulation of CD47 protein levels (figure 1G; online supplemental figure 3A). These results suggest that LSD1 inhibition suppresses CD47 expression in TNBC cells.

LSD1 inhibition suppresses NF-κB and STAT3 through distinct mechanisms in TNBC

Among the known transcriptional regulators of CD47, c-Myc, HIF-1α, NF-κB, and STAT3 bind to promoter and intronic regions to drive its transcription.⁴² To evaluate the effect of LSD1 depletion on these factors, we examined their expression following LSD1 knockdown. siRNA-mediated depletion of LSD1 markedly reduced NF-κB (p65) and STAT3 protein levels, with minimal effects on c-Myc and HIF-1α in EMT6, 4T1, and MDA-MB-231 TNBC cells (figure 2A). Consistently, treatment with the LSD1 inhibitor SP2509 decreased NF-κB (p65) and STAT3 expression (figure 2B; online supplemental figure 3B), while only marginally affecting c-Myc and HIF-1α. These findings indicate that LSD1 preferentially regulates NF-κB and STAT3 signaling in TNBC. TCGA analysis further revealed a positive correlation between LSD1 expression and both p65 and STAT3 levels in BC and basal-like tumors (online supplemental figure 3C). Reverse transcription polymerase chain reaction (RT-PCR) analysis showed that LSD1 knockdown reduced p65 mRNA levels, whereas STAT3 mRNA remained unchanged (figure 2C,D), suggesting transcriptional regulation of p65 and post-transcriptional regulation of STAT3.

Figure 2. Downregulation of NF-κB and STAT3 by LSD1 inhibition. (A) Immunoblot analysis of c-Myc, p65, STAT3 and HIF-1α in MDA-MB-231, EMT6 and 4T1 cells transfected with control or LSD1-targeting siRNA for 48 hours. β-actin was used as loading control. (B) Immunoblot analysis of c-Myc, p65, STAT3 and HIF-1α in MDA-MB-231 and EMT6 cells treated with increasing concentrations of SP2509 for 24 hours. (C, D) qRT-PCR analysis of p65 and Stat3 mRNA expression in EMT6 and MDA-MB-231 cells transfected with control or LSD1 siRNA for 48 hours. LSD1, lysine-specific demethylase 1; mRNA, messenger RNA.

STAT3 is known to be regulated by various post-translational modifications, including phosphorylation and methylation.⁴³ Previous studies have shown that STAT3 is methylated at K140 by SET9 and subsequently demethylated by LSD1 on binding to a subset of its target gene promoters.⁴⁴ In addition, phosphorylation of LSD1 by PKCα is required for p65 binding, facilitating its demethylation and enhancing its stability.⁴⁵ To further investigate how LSD1 regulates STAT3 and p65 in TNBC cells, we performed reciprocal co-IP assays to assess potential physical interactions. LSD1 was found to interact with STAT3, but not with p65 (figure 3A), suggesting direct regulation of STAT3. Consistently, ubiquitination assays revealed that treatment with an LSD1 inhibitor increased STAT3 polyubiquitination (figure 3B), indicating that LSD1 regulates STAT3 via a post-translational mechanism, potentially by modulating its stability or degradation.

Figure 3. LSD1 regulates STAT3 and p65 through distinct mechanisms. (A) Co-immunoprecipitation analysis in EMT6 cells using antibodies against STAT3, p65 or LSD1, followed by immunoblot detection of the indicated proteins. IgG served as a control. (B) Ubiquitination assay in EMT6 cells transfected with HA-ubiquitin and treated with SP2509 (10 μM, 24 hours) and MG132 (20 μM, 8 hours), followed by STAT3 immunoprecipitation and immunoblot analysis. (C) Schematic of tiling ChIP primers spanning −392 to +254 bp relative to the p65 TSS. (D) ChIP-qPCR analysis of H3K4me2 and H3K9me2 occupancy at the p65 promoter in EMT6 cells transfected with control or LSD1 siRNA. (E) TCGA analysis showing positive correlations between CD47 and p65 (RELA), and between CD47 and STAT3 expression in all breast tumors as well as in basal-like tumors. ChIP, chromatin immunoprecipitation; LSD1, lysine-specific demethylase 1; qPCR, quantitative PCR; TCGA, the Cancer Genome Atlas;TSS, transcription start site.

Next, we examined whether LSD1 knockdown affects the transcriptional activity of p65 using ChIP analysis. LSD1 can function as either a transcriptional repressor or activator by demethylating mono- or dimethylated H3K4 or H3K9, respectively, depending on cellular context and interacting partners.⁴⁶ To investigate how LSD1 inhibition modulates p65 transcription through alterations in the histone methylation landscape at its promoter, we conducted tiling ChIP assays using primers spanning approximately −392 to +254 base pairs relative to the p65 transcriptional start site (TSS) (figure 3C). H3K4 methylation was detected at the P1, P2, P3, and P5 regions, whereas H3K9 methylation was enriched at P1, P3, and P5 (online supplemental figure 4A,B). qPCR of ChIP-enriched DNA showed that LSD1 depletion significantly increased H3K9me2 across multiple promoter regions (P3, P5), while H3K4me2 levels were unchanged or reduced (P3) (figure 3D). These results indicate that LSD1 maintains a permissive chromatin state at the p65 promoter. Its depletion leads to selective enrichment of the repressive H3K9me2 mark, promoting transcriptional suppression. The lack of induction in H3K4me2 further suggests that LSD1 regulates p65 primarily through modulation of repressive rather than activating histone marks.

LSD1 promotes CD47 expression through activating NF-κB and STAT3

In silico analysis revealed a significant positive correlation between CD47 expression and both NF-κB and STAT3 in breast tumor and basal-like cohorts (figure 3E). To determine whether p65 and STAT3 directly regulate CD47 and whether LSD1 acts upstream of these pathways, EMT6 cells were transfected with plasmids encoding mouse p65 (pCMV3-N-GFP-RELA/NF-κB) or STAT3 (pCMV3-N-GFP-STAT3). Overexpression of either factor significantly increased CD47 mRNA and protein levels (figure 4A,B), indicating that both act as key activators of CD47 in TNBC. Consistently, rescue experiments showed that enforced expression of p65 or STAT3 restored CD47 expression in LSD1-depleted EMT6 cells (figure 4C,D). Collectively, these findings suggest that LSD1 promotes CD47 expression, at least in part, through activating NF-κB and STAT3 signaling pathways.

Figure 4. LSD1 suppresses CD47 expression through downregulation of p65 and STAT3. (A) qRT-PCR analysis of p65, Stat3 and Cd47 expression in EMT6 cells transfected with control (pCMV3-N-GFPSpark), RELA/NF-κB (p65) or STAT3 expression plasmids for 48 hours. (B) Immunoblot analysis of p65, STAT3 and CD47 in EMT6 cells transfected as in (A) (CD47 antibody: Cell Signaling Technology/36096). (C, D) Immunoblot analysis of CD47 in EMT6 cells co-transfected with control or LSD1 siRNA together with control or p65 expression plasmids (C) or STAT3 expression plasmids (D) for 48 hours (CD47 Antibody: Cell Signaling Technology, #36096). (E) Schematic of tiling ChIP primers spanning −1305 to +21 bp relative to the CD47 TSS. (F) ChIP-qPCR analysis of p65 and STAT3 occupancy at the CD47 promoter in EMT6 cells transfected with scramble control or LSD1-targeting siRNA. Data are representative of three independent experiments with similar results. ChIP, chromatin immunoprecipitation; LSD1, lysine-specific demethylase 1; mRNA, messenger RNA; qPCR, quantitative PCR; qRT-PCR,quantitative reverse transcription polymerase chain reaction; TSS, transcription start site.

To elucidate the mechanism by which LSD1 facilitates NF-κB and STAT3 recruitment to the CD47 promoter, we performed tiling ChIP assays spanning ~1.3 kb upstream of the TSS (figure 4E). STAT3 binding was localized to −654 to −944 bp region (P5), whereas p65 occupied a broader region spanning −576 to −1186 bp (P4-P6) (online supplemental figure 5A,B). Quantitative ChIP analysis further demonstrated that siRNA-mediated knockdown of LSD1 markedly reduced p65 enrichment at the P4 and P6 regions and significantly diminished STAT3 binding at P5 (figure 4F). These findings indicate that LSD1 is required for efficient recruitment of both NF-kB and STAT3 to the CD47 promoter, functioning as a key chromatin regulator that integrates these signaling pathways to drive CD47 transcription.

Downregulation of CD47 by LSD1 inhibition promotes M1-type macrophage phagocytosis and antitumor immunity

Macrophages play a pivotal role in tumor immune surveillance, and their ability to phagocytose tumor cells is crucial for eliminating malignant cells. However, many cancers evade this defense by upregulating CD47, which inhibits macrophage recognition and suppresses phagocytic clearance.⁴⁷ To assess whether LSD1 inhibition enhances macrophage-mediated phagocytosis, EMT6 cells transfected with control or LSD1-targeting siRNA were co-cultured with RAW 264.7 murine macrophages, followed by evaluation of phagocytic activity. LSD1 depletion in EMT6 cells significantly increased macrophage engulfment, as demonstrated by enhanced uptake observed via microscopy (figure 5A; online supplemental figure 6A). This effect was further confirmed by flow cytometry using EMT6 cells expressing control or LSD1 shRNA. Tumor cells were labeled with Hoechst 33342, co-cultured with RAW 264.7 macrophages for 2 hours, stained with a PE-conjugated F4/80 antibody, and analyzed by flow cytometry. LSD1 inhibition significantly increased macrophage-mediated phagocytosis (figure 5B; online supplemental figure 6B). To validate these findings in a more physiologically relevant system, primary mouse bone marrow-derived macrophages (mBMDMs) were used. Consistent with results obtained in RAW 264.7 cells, LSD1 depletion markedly enhanced phagocytosis of EMT6 tumor cells by mBMDMs (figure 5C; online supplemental figure 6C). RT-PCR analyses showed that treatment with SP2509 significantly upregulated the mRNA expression of pro-inflammatory cytokines and chemokines, including Tnf-α, Il-6, Cxcl9, and Cxcl10, in RAW 264.7 cells (figure 5D). Correspondingly, protein levels of these factors were also elevated following SP2509 treatment (figure 5E). The induction of these immune mediators aligns with changes in macrophage polarization markers toward an M1 phenotype.⁴⁸ Moreover, immunoblot analysis demonstrated that SP2509 treatment (10 μM) decreased arginase 1 expression while increasing iNOS levels in RAW 264.7 cells (figure 5E), further supporting the conclusion that LSD1 inhibition promotes the switch from an M2-like to an M1-like macrophage phenotype.

Figure 5. CD47 suppression by LSD1 inhibition enhances macrophage phagocytosis and tumor immunogenicity. (A) Representative images of RAW 264.7 macrophages engulfing EMT6 cells transfected with scramble or Lsd1 siRNA. Macrophages were labeled with PKH26 (red) and tumor cells with Hoechst 33342 (blue). (B, C) Phagocytosis of Hoechst 33342-labeled EMT6 cells expressing scramble or Lsd1 shRNA after 2 hours co-culture with RAW 264.7 macrophages (B) or mBMDMs (C), assessed by PE-F4/80 staining and flow cytometry (n=3). (D) qRT-PCR analysis of Cxcl9, Cxcl10, Tnf-α and Il6 expression in RAW 264.7 macrophages treated with SP2509 (5 or 10 μM) for 24 hours. β-actin was used as an internal control. (E) Immunoblot analysis of CXCL9, CXCL10, TNF-α, IL-6, arginase 1 and iNOS in RAW 264.7 macrophages treated with SP2509 (5 or 10 μM) for 24 hours. (F) qRT-PCR analysis of indicated immune-related genes in EMT6 cells transfected with scramble or Cd47 siRNA for 48 hours. (G) Immunoblot analysis of CXCL9 and CXCL10 in EMT6 cells transfected with scramble or Cd47 siRNA for 48 hours. (H) ELISA quantification of CXCL9 and CXCL10 in supernatants from EMT6 cells transfected with scramble or Cd47 siRNA for 48 hours. (I) Chemotaxis assay assessing CD8+T cell migration in response to conditioned media from MDA-MB-231 cells transfected with scramble or CD47 siRNA for 48 hours. Data are mean±SD of three independent experiments. CXCL,chemokine (C-X-C motif) ligand; IL, interleukin; LSD1, lysine-specific demethylase 1; mBMDMs, mouse bone marrow-derived macrophages; mRNA, messenger RNA; qRT-PCR, quantitative reverse transcription polymerase chain reaction; TNF, tumor necrosis factor.

Next, we investigated how CD47 inhibition reshapes the immune landscape by profiling key immune-regulatory factors in EMT6 cells. qPCR analysis revealed that treatment with CD47-specific siRNA led to a marked upregulation of Th1-associated chemokines (Ccl4, Cxcl9, Cxcl10) and the pro-inflammatory cytokine Ifn-γ (figure 5F), potent mediators of monocyte activation and chemoattractants for NK cells and Th1 T cells. In contrast, CD47 silencing significantly reduced the expression of the immunosuppressive chemokine Ccl2 (figure 5F). CCL2 and its receptor CCR2 play critical roles in promoting tumor cell survival and invasion, pathological angiogenesis, and the recruitment of immune-suppressive cells.⁴⁹ Western blot confirmed that CD47 depletion increased CXCL9 and CXCL10 expression (figure 5G). ELISA of culture supernatants from EMT6 cells treated with CD47 siRNA revealed elevated extracellular CXCL9 and CXCL10 levels (figure 5H), indicating enhanced secretion of Th1-type chemokines that promote CTL recruitment. Ex vivo chemotaxis assays further demonstrated that human peripheral blood-derived CD8+T cells migrated more efficiently toward conditioned medium from CD47-depleted MDA-MB-231 cells (figure 5I). Collectively, these results demonstrate that CD47 inhibition, potentiated by LSD1 depletion, enhances both innate and adaptive antitumor immunity in TNBC.

LSD1 inhibition potentiates in vivo response of TNBC tumors to CD47 blockade

To determine whether LSD1 inhibition enhances the therapeutic efficacy of anti-CD47 treatment in TNBC, BALB/cJ mice bearing EMT6 tumors were treated with SP2509 in combination with an anti-CD47 mAb (online supplemental figure 7A). Monotherapy with either agent reduced tumor growth by approximately 30% relative to controls. In contrast, combination therapy elicited a substantially greater antitumor effect, achieving more than 70% tumor growth inhibition (figure 6A,B). At the study endpoint, tumors were harvested and weighed. In line with tumor volume measurements, tumors from the combination group were significantly lighter than those from control or single-agent groups (figure 6C). No evidence of systemic toxicity was observed during treatment, as indicated by stable body weights and the absence of treatment-related mortality across all groups (online supplemental figure 7B). To further assess in vivo treatment safety, liver and kidney tissues were examined histologically using H&E staining. No overt histopathological abnormalities were observed following either monotherapy or combination treatment (online supplemental figure 8).

Figure 6. LSD1 inhibitor enhances antitumor efficacy of CD47 antibody. (A) EMT6 cells were orthotopically implanted into the mammary fat pads of female BALB/cJ mice. Once tumors became palpable, mice (n=7 per group) received intraperitoneal injections of vehicle/IgG (DMSO and IgG2a isotype control), SP2509 (50 mg/kg, daily), anti-CD47 mAb (5 mg/kg, every 3 days), or the combination. Tumor volumes are shown as mean±SD (mm³). SP2509 versus combination, ***p<0.001; CD47 versus combination, **p<0.01; Veh/IgG versus combination, ***p<0.001; Veh/IgG versus SP2509, *p<0.05. ANOVA with Tukey’s post hoc multiple comparison test. (B) Tumors were excised and imaged at the experimental endpoint. (C) Tumor weights were recorded at necropsy and are shown as mean±SD. (D) Representative immunohistochemical staining of CD47, F4/80, iNOS, arginase 1, and CD8 in paraffin-embedded tumor sections from mice treated as in (A). (E) AOD values quantify the mean staining intensity of the indicated markers in tumors (n=3). ANOVA, analysis of variance; AOD, average optical density; DMSO, dimethyl sulfoxide; IHC, immunohistochemistry; LSD1, lysine-specific demethylase 1; mAb, monoclonal antibody.

IHC analysis revealed that both SP2509 and anti-CD47 mAb monotherapies reduced Ki-67 expression, whereas combination therapy elicited a more pronounced decrease, indicating enhanced suppression of tumor cell proliferation (online supplemental figure 9A). CD47 expression was significantly reduced in EMT6 tumors treated with SP2509 or combination therapy. Macrophage infiltration (F4/80+) was increased in these groups, with the greatest enrichment observed following combination treatment (figure 6D,E). Notably, the combination regimen upregulated the M1 macrophage marker iNOS while downregulating the M2 marker arginase 1 relative to control and monotherapy groups, suggesting enhanced innate immune activation. We next assessed the impact of combination therapy on T-cell infiltration within the tumor microenvironment. SP2509 monotherapy increased CD8+T cell infiltration, which was further enhanced by combination therapy (figure 6D,E). No significant changes in CD4+T cell infiltration were observed across treatment groups (online supplemental figure 9B). To validate these findings, multiplex immunofluorescence staining was conducted on tumor samples to profile immune cell subsets, including macrophages (CD68, iNOS, arginase 1) and T cells (CD3, CD8, CD4). Our results indicate that the combination treatment promoted macrophage polarization toward an M1-like phenotype, which was closely associated with enhanced CD8+ T-cell infiltration in EMT6 tumors (online supplemental figures 10 and 11).

Furthermore, we examined the in vivo regulation of NF-κB and STAT3 signaling in response to therapy. Both SP2509 monotherapy and combination treatment with anti-CD47 mAb significantly reduced NF-κB and STAT3 expression in tumors (figure 7A). Combination therapy also robustly upregulated CXCL9/10 and granzyme B, a cytotoxic serine protease produced by CTLs and NK cells that induces apoptosis in target cells (online supplemental figure 12A). Consistent with the effects of CD47 siRNA in EMT6 cells, CD47 blockade in tumors suppressed Ccl2 mRNA expression while increasing Ccl4 and Ifn-γ levels (online supplemental figure 12B). To assess potential causality and feedback regulation between LSD1 and CD47, we examined LSD1 protein expression following CD47 knockdown via siRNA, as well as in tumors treated with anti-CD47 monotherapy. CD47 depletion, achieved by either siRNA or antibody treatment, did not alter LSD1 protein levels (online supplemental figure 12C,D), indicating that although LSD1 regulates CD47 expression, this relationship is not bidirectional and CD47 does not provide feedback regulation of LSD1.

Figure 7. Depletion of CD8 T cells compromises the antitumor effect of combination therapy. (A) Representative immunohistochemical staining of NF-κB (p65) and STAT3 in EMT6 tumors treated with vehicle/isotype IgG control, SP2509, CD47 mAb, or combination therapy. AOD values represent the mean staining intensity of NF-κB (p65) and STAT3 in tumors (n=3). (B) Female BALB/cJ mice (n=7 per group) bearing EMT6 tumors were treated with vehicle/IgG (DMSO and IgG2a or IgG2b isotype controls), anti-CD8 mAb (250 μg per mouse, every 4 days), or the combination of SP2509 (50 mg/kg, daily) and anti-CD47 mAb (5 mg/kg, every 3 days), with or without CD8 mAb. Tumor growth was monitored by caliper measurements and the volumes are presented as mean±SD (mm³). Veh/IgG versus combination, ***p<0.001; combination versus CD8 mAb+combination, **p<0.01; CD8 mAb versus CD8 mAb+combination, ***p<0.001. ANOVA with Tukey’s post hoc multiple comparison test. (C) Tumors were excised and imaged at the experimental endpoint. (D) Primary tumors were collected, and tumor weights were measured for each animal at necropsy. (E) Tumor-infiltrating macrophages were analyzed by flow cytometry. Single-cell suspensions were stained with antibodies against CD45, CD11b, F4/80, CD11c, and CD206. M1-like macrophages were defined as CD45+CD11b+F4/80+CD11c+, and M2-like macrophages as CD45+CD11b+F4/80+CD206+. (F) Quantification of M1-like and M2-like macrophages and the M1/M2 ratio in tumors. Combo: combination. (G) Proposed model illustrating the role of LSD1 in regulating CD47 signaling and the response to anti-CD47 therapy. ANOVA, analysis of variance; AOD, average optical density; DMSO, dimethyl sulfoxide; IHC, immunohistochemistry; LSD1, lysine-specific demethylase 1; mAb, monoclonal antibody.

To investigate the role of CD8+T cells in the antitumor efficacy of combination therapy, EMT6 cells were orthotopically implanted into the mammary glands of mice, followed by treatment with vehicle/isotype IgG, anti-CD8 mAb, or SP2509 plus anti-CD47 mAb, in the presence or absence of a CD8-depleting mAb (online supplemental figure 13A). Combination therapy significantly inhibited tumor growth; however, this effect was largely abrogated by CD8+T cell depletion (figure 7B,C). Tumor weights were consistent with volume measurements (figure 7D), and all treatments were well tolerated, with no significant changes in body weight (online supplemental figure 13B). Immunofluorescence staining confirmed that anti-CD8 mAb effectively blocked CD8 T-cell infiltration induced by combination therapy (online supplemental figure 13C,D). Flow cytometric analysis showed that combination therapy increased the infiltration of M1 macrophages (F4/80+CD11c+) while decreasing M2 macrophages (F4/80+CD206+), resulting in a significantly elevated M1/M2 ratio (figure 7E,F), indicative of enhanced macrophage-mediated antitumor immunity. CD8+T cell depletion reversed these effects, reducing M1 macrophages, increasing M2 macrophages, and lowering the M1/M2 ratio (figure 7E,F). These findings demonstrate that CD8+T cell activation enhances the therapeutic efficacy of the combination treatment by promoting M1 polarization of TAMs and strengthening the antitumor immune response to CD47 blockade.

In summary, these results demonstrate that LSD1 inhibition downregulates CD47 in TNBC, enhancing the antitumor efficacy of anti-CD47 therapy. Combination treatment not only suppresses tumor proliferation but also promotes innate and adaptive immune activation, including macrophage polarization toward the M1 phenotype and CD8+T cell tumor infiltration. These findings support a model in which LSD1 regulates CD47-associated transcriptional pathways, including NF-κB and STAT3, thereby modulating tumor immunity and response to combination therapy (figure 7G).

Discussion

Elevated CD47 expression correlates with several key pathological features of TNBC, such as EMT, advanced tumor stages, lymph node metastasis, and increased recurrence risk.⁵⁰ CD47 upregulation promotes immune evasion by enabling tumor cells to escape macrophage-mediated phagocytosis through interaction with SIRPα, a receptor expressed on macrophages. Despite these associations, the molecular mechanisms regulating CD47 expression and function in TNBC remain incompletely understood. Using an engineered BRCA1-deficient mouse model, we observed that CD47 expression was significantly upregulated during tumorigenesis, coinciding with an increased proportion of immunosuppressive M2 macrophages and a concomitant reduction in antitumor M1 macrophages. These findings provide the first direct evidence of CD47 dysregulation in the pathogenesis of TNBC resulting from BRCA1 loss. Elevated CD47 expression has also been associated with homologous recombination deficiency and pathogenic BRCA mutations in ovarian cancer.⁵¹ Recent studies further show that CD47 overexpression promotes immune evasion in BRCA1-deficient ovarian tumors and that anti-CD47 therapies can enhance the efficacy of PARP inhibitors.⁵² Nevertheless, the precise role of CD47 in driving BRCA1-mutant tumor initiation and progression remains unclear.

Our previous work identified LSD1 as a critical regulator of mammary tumor initiation and progression following BRCA1 loss.³⁸ In this study, we show that CD47 upregulation in BRCA1-deficient tumors correlates with increased LSD1 expression. However, the mechanisms through which LSD1 controls CD47 expression in BC have remained unclear. We demonstrate that LSD1 inhibition selectively reduces the expression of NF-κB and STAT3, two key transcription factors involved in CD47 regulation. Notably, LSD1 regulates these factors through distinct mechanisms. We further identified a physical interaction between LSD1 and STAT3 and found that LSD1 inhibition promotes STAT3 polyubiquitination, suggesting a post-translational mode of regulation. Prior studies have shown that LSD1 demethylates STAT3 at specific target promoters, thereby enhancing STAT3 activation and downstream transcriptional activity.⁴⁴ Consistent with this, a recent study by Ladaika et al revealed that LSD1, in collaboration with CoREST, facilitates STAT3 demethylation, promoting chromatin binding and enteroendocrine cell differentiation in mucinous colorectal cancer.⁵³ Our findings expand the understanding of LSD1’s multifaceted role in regulating CD47 expression by coordinating NF-κB and STAT3 activity, revealing previously unrecognized regulatory mechanisms underlying CD47 overexpression. Importantly, CD47 depletion did not alter LSD1 protein levels, indicating that while LSD1 regulates CD47, the relationship is not reciprocal. Whether a causal relationship or feedback loop exists during early tumorigenesis remains unresolved, and additional studies are needed to define the broader molecular landscape through which LSD1 influences CD47 regulation.

The M1 macrophage phenotype is well recognized for its capacity to enhance antitumor immunity by activating other immune cells and promoting tumor cell killing.⁴⁸ Our data show that LSD1 inhibition increases macrophage phagocytic activity and drives M2-to-M1 repolarization, suggesting that this shift may potentiate the antitumor functions of TAMs. The concurrent suppression of NF-κB and STAT3 signaling following LSD1 inhibition further indicates that targeting these pathways may help alleviate the immunosuppressive tumor microenvironment that often limits effective immune responses. Moreover, we observed that CD47 inhibition elevates IFN-γ and Th1-type chemokine expression and enhances CD8+T cell migration. These findings support a model in which macrophages, through tumor cell phagocytosis, may prime CD8+T cells and thereby amplify antitumor immunity. This is consistent with prior evidence showing that M1-like TAMs stimulate and recruit CD8+T cells, whereas M2-like TAMs suppress T-cell activity.⁵⁴ Conversely, signals from CD8+T cells can reprogram TAMs from a protumor M2 phenotype toward an antitumor M1 state, highlighting a dynamic and reciprocal interaction essential for effective immune responses.⁵⁵ Our results also underscore the role of Th1-type chemokines, such as CXCL9 and CXCL10, in recruiting CD8+T cells into tumors. This chemokine-mediated recruitment appears to be a key mechanism by which CD47 inhibition enhances immune-cell infiltration and strengthens antitumor immunity.

CD47-targeting monoclonal antibodies have emerged as a promising therapeutic strategy in BC due to CD47’s central role as an immune checkpoint that enables tumor immune evasion.²⁰ By blocking CD47, these antibodies disrupt the “don’t-eat-me” signal, thereby enhancing immune recognition and clearance of malignant cells. Despite this therapeutic potential, several challenges continue to limit the clinical translation of CD47-directed mAbs.^{56 57} A major barrier is on-target toxicity, as CD47 is broadly expressed on both tumor and normal cells, including erythrocytes, leading to adverse effects such as anemia and thrombocytopenia.⁵⁶ Another major challenge is the limited selectivity of CD47-targeted agents. Achieving robust tumor-specific activity while preserving normal tissue integrity remains a significant obstacle for the clinical use of these therapies. These clinical challenges highlight the need for more refined therapeutic approaches, including rational combination strategies, to improve efficacy and safety.

Our in vivo studies demonstrate that combining LSD1 inhibition with CD47 mAb therapy significantly enhances the therapeutic index in syngeneic TNBC models while maintaining acceptable toxicity in mice. These findings suggest that LSD1 inhibition lowers CD47 expression on tumor cells, and that pairing this effect with an anti-CD47 antibody blocks the remaining CD47, thereby amplifying macrophage-mediated tumor clearance. This dual approach may also enable reduced antibody dosing to mitigate toxicity and may provide more sustained suppression of CD47, which can re-emerge between treatment cycles. Immune profiling revealed that this combination strengthens antitumor immunity by promoting M1 macrophage polarization and alleviating the immunosuppressive tumor microenvironment. Mechanistically, this is achieved through suppression of NF-κB and STAT3 signaling, along with increased expression of antitumor cytokines and chemokines, including IFN-γ, CXCL9, and CXCL10, that facilitate CD8+T cell recruitment into tumors. As a result, the combination enhances macrophage-mediated phagocytosis and primes CD8+T cells, culminating in a more robust antitumor immune response. Recent studies have shown that bispecific antibodies (BisAbs) targeting both CD47 and PD-L1 can improve macrophage-CTL crosstalk while reducing toxicity compared with single-agent checkpoint blockade.⁵⁸ Several PD-L1/CD47 BisAbs are currently undergoing clinical evaluation.⁵⁹ However, challenges remain, including immune-related toxicity, tumor heterogeneity, resistance mechanisms, and pharmacokinetic limitations.^{56 60} Our previous work demonstrated that LSD1 inhibition enhances T-cell infiltration and synergizes with PD-1 blockade to promote TNBC eradication.³¹ These findings suggest that combining LSD1i with BisAbs targeting CD47/PD-L1 may help overcome the obstacles associated with this dual ICB, warranting further investigation.

In conclusion, our study demonstrates that LSD1 inhibition reduces CD47 expression by downregulating NF-κB and STAT3 signaling, thereby enhancing macrophage-mediated antitumor immunity through the promotion of an M2-to-M1 shift in TAMs. This strategy targets the critical crosstalk between epigenetic modulators and immune components, offering a novel therapeutic approach for BC patients with poor immune responses. As LSD1 inhibitors advance in clinical trials, combining them with anti-CD47 therapy holds considerable promise for improving the efficacy of BC immunotherapy.

Data availability statement

Data are available upon reasonable request.