Hormone therapy potentiates chemoimmunotherapy by inducing SUMOylation-mediated immunogenic cell death

Abstract

Background

Combining immunotherapy with endocrine therapy and/or chemotherapy promotes prognosis in patients with estrogen receptor-positive/HER2-negative (ER⁺HER2⁻) breast cancer; however, the unclear underlying mechanisms limit their widespread application in clinical practice. This study aimed to investigate how ER signaling regulates antitumor immunity and influences responses to chemoimmunotherapy.

Methods

We evaluated pathological complete response rates between ER-positive and ER-negative breast cancer patients across internal and external cohorts treated with immunotherapy and neoadjuvant chemotherapy. Mechanistic studies were conducted to explore the crosstalk between ER and immunogenic cell death (ICD) through high mobility group box 1 (HMGB1) SUMOylation. The therapeutic potential of chemoimmunotherapy regimens was assessed using immunocompetent mouse models of ER⁺ breast cancer.

Results

Patients with ER⁻ breast cancer had higher pathological complete response rates after immunotherapy and neoadjuvant chemotherapy compared with those with ER⁺ breast cancer in both internal and external cohorts. Enhanced immune responses were observed in ER⁻ breast cancer after chemotherapy, which was attributed to the induction of ICD via the release of HMGB1. Mechanistically, ER inhibits HMGB1 release by directly interacting with the HMG Box B domain and suppressing its SUMOylation at the K114 site, thereby suppressing the antitumor immune response.

Conclusions

Therapeutically, combining ER antagonist with chemotherapy potentiated the efficacy of PD-1 blockade through HMGB1 in an immunocompetent mice model. These findings provide new mechanistic insights and clinical strategies for addressing immunotherapy resistance in ER⁺ breast cancer patients.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Most patients with estrogen receptor-positive/HER2-negative (ER⁺HER2⁻) breast cancer exhibit primary resistance to immunotherapy; however, the underlying mechanisms remain elusive.

WHAT THIS STUDY ADDS

• We demonstrate that ER signaling suppresses chemotherapy-induced immunogenic cell death by inhibiting the SUMOylation of high mobility group box 1. Conversely, ER degraders alleviate this inhibition, thereby reprogramming the postchemotherapy immune microenvironment.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• These findings provide a mechanistic rationale for implementing a triplet regimen combining endocrine therapy, chemotherapy, and immunotherapy in advanced ER-positive breast cancer. Furthermore, they establish a novel paradigm for immune combination strategies in other ER-high-expressing solid tumors, warranting validation through future pan-tumor clinical trials.

Introduction

Early-stage estrogen receptor-positive/HER2-negative (ER⁺HER2⁻) breast cancer, the most common breast cancer subtype, generally responds well to endocrine therapy. However, once recurrence or metastasis occurs, the efficacy of these treatments significantly decreases, highlighting the urgent need to explore new therapeutic approaches. Application of immunotherapeutic drugs, such as immune checkpoint inhibitors, is rapidly increasing and has major beneficial effects on oncology and human health.¹ Although breast cancer is referred to as an immunologically “cold” tumor, latest clinical trials showed that patients with ER⁻ breast cancer achieve significantly higher pathological complete response rates to immunotherapy compared with those with ER⁺ breast cancer.² Further research revealed that ER⁺/HER2⁻ tumors exhibit the lowest density of tumor-infiltrating lymphocytes, and the objective response rate in the hormonal receptor-positive (HR⁺) cohort was lower than that in patients with triple-negative breast cancer after immune therapy.^{3 4} However, it remains incompletely understood how to transform ER⁺HER2⁻ tumors into immunologically sensitive tumors and whether combining immunotherapy can benefit those with advanced ER⁺HER2⁻ breast cancer.

Recent clinical studies have shown that for patients with early-stage ER⁺/HER2⁻ breast cancer, combining immunotherapy with neoadjuvant chemotherapy (NAC) and endocrine therapy can significantly increase the pathological complete response rate (NCT04109066; NCT03725059).^5–7 Additionally, in metastatic ER⁺/HER2⁻ breast cancer, endocrine therapy combined with immunotherapy can also markedly enhance treatment efficacy (NCT02990845).⁸ These studies suggest that ER⁺/HER2⁻ breast cancer can reverse the local immunosuppressive microenvironment under different combination therapies. However, these findings are based on a limited number of clinical studies, further basic research is needed to confirm whether the immunosuppressive microenvironment can indeed be reshaped. The effectiveness of immunomodulatory strategies depends on the presence of a baseline immune response and the unleashing of pre-existing immunity.^{9 10} Cold tumors have been the focus of efforts to convert them into hot tumors.¹¹ However, there is heterogeneity in the methods used across different types of tumors.^{12 13} While combining immunotherapy with conventional cancer treatments can achieve better outcomes in some tumors that are otherwise insensitive to immunotherapy,^{12 13} the implementation of these approaches in clinical practice has been significantly slowed due to a lack of clear understanding of the underlying mechanisms.^{14 15} Therefore, understanding the mechanisms responsible for hot, altered, or cold immune tumor phenotypes is critical for boosting antitumor immunity.

Conventional chemotherapy drugs, such as doxorubicin and oxaliplatin, can elicit antitumor effects by inducing immunogenic cell death (ICD) of tumor cells.^15–19 ICD boosts antitumor immune responses through the release of damage-associated molecular patterns (DAMPs), such as high mobility group box 1 (HMGB1), calreticulin (CRT), ATP, and heat shock proteins.^19–22 This leads to the activation of the host immune system, which can act in combination with the direct cytotoxic effects of chemotherapy drugs on cancer cells to further improve their therapeutic effects.^17–23 Given the wide range of immunotherapeutic agents available, combining them with drugs that target other pathways in the immune regulatory cycle may help overcome treatment resistance.^{24 25}

In this study, through clinical data analysis, animal tumor-bearing models, and molecular biology techniques, we explored the remodeling of the immune microenvironment induced by endocrine therapy combined with chemotherapy in ER⁺/HER2⁻ breast cancer, as well as the therapeutic efficacy of combined immunotherapy in advanced breast cancer.

Results

NAC elicits an enhanced immune response in ER⁻HER2⁻ breast cancers

HR⁺ breast cancer accounts for around 75% of all cases and presents a significant challenge in clinical management due to its heterogeneous response to chemotherapy and immunotherapy.^26–28 To investigate the effect of HR status on therapeutic response to NAC and immunotherapy, we collected clinical data from 483 patients with breast cancer who had received NAC at Sun Yat-sen Memorial Hospital (table 1). Univariate analysis revealed that ER (OR, 0.98; 95% CI 0.97 to 0.99; p<0.001) and progesterone receptor (PR) were significantly associated with pathological complete response (table 2). In the multivariate model, ER was the only statistically significant variable that was associated with a lower probability of pathological complete response (OR, 0.98; 95% CI 0.97 to 0.99; p<0.01), which suggested that ER was a predictive factor for pathological complete response (table 2). Furthermore, patients with ER⁻ breast cancer tended to achieve pathological complete response after NAC and immunotherapy, with a lower pathological complete response rate (21.33% for NAC and 47.44% for neoadjuvant chemoimmunotherapy) than patients with ER⁺ breast cancer (7.10% for NAC and 25.00% for neoadjuvant chemoimmunotherapy) (online supplemental figures S1A,B, table 3). This observation was also validated in an external cohort of 161 patients with breast cancer who received NAC in the I-SPY 1 TRIAL (clinical data from GSE32603) and those (n=67) who received NAC plus pembrolizumab in the I-SPY 2 TRIAL evaluating combination immunotherapy in breast cancer (online supplemental figures S1A,B).

Table 1. Clinical characteristics of patients treated with NAC (N=483).

	N	%
Age
≤40	125	25.88
>40	358	74.12
Clinical stage
I–II	299	61.90
III	184	38.10
Tumor size
≤2 cm	45	9.32
>2 cm	438	90.68
Lymph nodes
Negative	158	32.71
Positive	325	67.29
ER
Negative	75	15.53
Positive	408	84.47
PR
Negative	184	38.10
Positive	299	61.90
HER2
Negative	289	59.83
Positive	194	40.17
pCR
No	438	90.68
Yes	45	9.32

Information was collected 483 patients who received NAC at Sun Yat-sen Memorial Hospital (Guangzhou, China) between 2010 and 2018. The NAC regimens comprised doxorubicin or epirubicin 60–80 mg/m2 every 3 weeks for four to eight cycles.

ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; NAC, neoadjuvant chemotherapy; pCR, pathological complete response; PR, progesterone receptor.

Table 2. Association of clinicopathologic factors with pathologic response in NAC setting.

Factor	Univariate analysis			Multivariate analysis
	OR	95% CI	P value	OR	95% CI	P value
Age (>40 years)	0.92	0.45 to 1.88	0.82
Clinical stage (III vs II or I)	1.85	1.00 to 4.09	0.05
Tumor size (>2 cm)	1.57	0.63 to 3.95	0.33
Lymph nodes (positive)	0.92	0.48 to 1.79	0.81
ER, continuous	0.98	0.97 to 0.99	<0.001***	0.98	0.97 to 0.99	<0.01**
PR, continuous	0.97	0.95 to 0.99	<0.001***	0.99	0.97 to 1.00	0.10
HER2 (positive)	1.38	0.72 to 2.64	0.33

P value was calculated with two-sided wald test.

Bold entries indicate statistically significant P values.

^**p<0.01, ***p<0.001

ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; NAC, neoadjuvant chemotherapy; PR, progesterone receptor.

Table 3. Clinical characteristics of patients treated with NAC combined with immunotherapy (N=110).

	N	%
Age
≤40	40	36.36
>40	70	63.64
Tumor size
≤2 cm	12	10.91
>2 cm	98	89.09
Lymph nodes
Negative	39	35.45
Positive	71	64.55
ER
Negative	78	70.91
Positive	32	29.09
PR
Negative	96	87.27
Positive	14	12.73
HER2
Negative	80	72.73
Positive	30	27.27
pCR
No	65	59.09
Yes	45	40.91

Clinical data were collected from 110 patients who received NAC combined with immunotherapy at Sun Yat-sen Memorial Hospital (Guangzhou, China) between 2020 and 2025. The NAC regimens comprised doxorubicin or epirubicin 60–80 mg/m2 every 3 weeks for four to eight cycles concurrently with immunotherapy agents, including pembrolizumab, toripalimab, camrelizumab, and sintilimab.

ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; NAC, neoadjuvant chemotherapy; pCR, pathological complete response; PR, progesterone receptor.

Accumulating evidence has shown that anthracycline-based chemotherapy can modulate the antitumor immune response.^{17 29–31} Furthermore, a higher abundance of CD8⁺ cytotoxic T cells showed association with better clinical outcomes in patients with breast cancer, especially those with HER2⁺ and Triple-Negative Breast Cancer.³² Therefore, we analyzed the relationship between CD8a, Perforin, and ER in patients with locally advanced breast cancer who had received NAC, and then examined the number of CD8⁺ T cells in paired primary tumor samples before and after NAC in our cohort. In a subset of 87 patients with breast cancer, CD8A, PRF1, TBX21, and CD68 expression levels were negatively correlated to ESR1, the coding gene of ER-a (figure 1A, online supplemental figure S1C-G). Before NAC, the number of CD8⁺ T cells in tumor samples did not differ significantly between patients with ER⁻HER2⁻ and those with ER⁺HER2⁻ breast cancer. However, after NAC, patients with ER⁻HER2⁻ breast cancer showed higher CD8⁺ T-cell infiltration than those with ER⁺HER2⁻ breast cancer (figure 1B,C), with infiltration of CD8⁺ tumor infiltrating lymphocytes (TILs) being inversely correlated with the percentage of ER⁺ tumor cells (online supplemental figure S1H). Moreover, protein expression levels of HMGB1 and Granzyme B, along with CD8⁺ T cell infiltration, were markedly elevated in patients with breast cancer treated with triple therapy (fulvestrant, epirubicin, and PD-1 inhibitors) compared with those treated with single agent or dual combination (online supplemental figure S1I).

Figure 1. ER inhibits antitumor response to chemo-immunotherapy in breast cancer. (A) Expression of CD8A and ESR1 (ER-α) in tissue samples of breast cancer from the Gene Expression Omnibus (GEO) dataset. Patients were ranked in descending order of their ESR1 expression. (B) Representative immunohistochemical images of CD8⁺ T-cell infiltration in human primary breast cancer before or after neoadjuvant chemotherapy (NAC). (C) Representative immunofluorescent images of ER (red), CD8 (green), and DAPI (blue) in human primary breast cancer tissues. Scale bar, 20 μm. (D) Schematic overview of establishing an ER⁺ mice breast cancer cell line under treatment with MPA and DMBA (upper panel and lower left). Representative immunohistochemical images for ER-α show a difference in ER expression between DMMC-02 and EO771 cell lines. Scale bar, 50 µm (lower right). (E) Tumor growth curves of C57BL/6 mice receiving the indicated treatment after DMMC-02 tumors were palpable. Data are presented as mean±SD. (F) Production of IFN-γ by draining lymph node cells of C57BL/6 mice injected with DMMC-02 cells. (G, H) C57BL/6 mice were grafted with DMMC-02 cells and treated with fulvestrant (F) and epirubicin (E) in the presence or absence of neutralizing antibodies against PD-1 (a-PD1) after the tumors were palpable. (G) Tumor growth curves for the indicated treatment groups. (H) Representative immunofluorescent staining of CD8, TUNEL, and CK in slices of serial mouse tumor tissue sections. White boxes indicate the areas that are magnified in the insets below. Scale bar, 20 μm. (I) The tumor growth curves of recurrent DMMC-02 cells in C57BL/6 mice are shown. (J, K) Lung metastasis of DMMC-02 cells was checked by IVIS Lumina system, the representative images are shown in (J), and quantitative flux analysis is shown in (K). Data are presented as mean±SD; n=87 (A), 25 (B, C), 4 per group (E), 6 per group (F, G, I, K); Statistical significance was determined using a two-sided one-way ANOVA with Tukey’s test (E, F, G, I, K). *p<0.05, **p<0.01, ***p<0.001; ns, not significant. ANOVA, analysis of variance; ER, estrogen receptor; IVIS, in vivo imaging system.

The combination of ER antagonist and chemotherapy potentiates immune therapy in ER⁺ breast cancer

To evaluate the antitumor immune responses elicited by combination therapy comprising ER antagonist and chemotherapy against ER⁺ breast cancer in vivo, we established an ER⁺ mice breast cancer cell line—DMMC-02. This cell line was derived from transplantable 7,12-dimethylbenz[a]anthracene (DMBA)-induced mammary carcinoma of C57BL/6 mice (figure 1D, online supplemental figure S2A-D). Consistent with the clinical data, the combination therapy with fulvestrant and epirubicin inhibited tumor growth (figure 1E). Moreover, the combination treatment with fulvestrant and epirubicin significantly increased the concentration of the immune effector molecule IFN-γ in the draining lymph node cells, as demonstrated by injecting DMMC-02 tumor cells into C57BL/6 mice (figure 1F and online supplemental figure S2E). An increase in CD8⁺ T cells and IFN-γ secretion is a marker of immune activation. To investigate the relationship between ER and tumor immunotherapy in animal models, we inoculated mice with DMMC-02 cells or ER-knockdown DMMC-02 cells into the mammary fat pad. Chemotherapy combined with immunotherapy was administered when the tumor was palpable. We observed that in ER-knockdown DMMC-02 bearing mouse tumor models, chemotherapeutic agents significantly enhanced the efficacy of immunotherapy. However, in models with high ER expression, the synergistic effect of chemotherapeutic agents on immunotherapy was markedly attenuated (online supplemental figure S2F,G). These findings suggest that ER expression may serve as a critical factor modulating chemotherapy-induced immune activation.

Although ER⁺ breast cancer has far remained resistant to ICIs, endocrine therapy can enhance the efficacy of immunotherapy in clinical patients with ER⁺HER2⁻ breast cancer. However, related mechanistic studies are lacking, and further investigation is needed to determine whether their concurrent combination can also play a synergistic role in immunotherapy. To investigate whether the antitumor immune response elicited by epirubicin plus fulvestrant could enhance the efficacy of ICIs, we treated mice alone or in combination with epirubicin, fulvestrant, and anti-PD-1 when the tumor was palpable. Consistent with the above results, epirubicin plus fulvestrant treatment significantly attenuated the tumor growth; although anti-PD-1 therapy alone did not exhibit significant efficiency in tumor growth, the combination with epirubicin and fulvestrant apparently inhibited tumor growth (figure 1G). Immunofluorescence analysis of tumor specimens revealed that not only a significant increase in tumor cell death in the endocrine, chemotherapy, and immunotherapy groups, but also a marked enrichment of CD8⁺ T cells in the tumor microenvironment (figure 1H). Taken together, these findings suggest that the combination of endocrine and chemotherapy dramatically improves the efficacy of immunotherapy in ER⁺ breast cancer.

To further validate the therapeutic efficacy of these combination regimens in recurrent ER⁺ breast cancer, we collected recurrent tumor tissues from mice which had previously undergone endocrine therapy, chemotherapy, and surgical resection of the primary tumor. These tissues were re-inoculated into the mammary fat pad of immunocompetent mice. When tumors reached a maximum diameter of approximately 0.5 cm, we re-administered endocrine therapy combined with chemotherapy, or coadministered immunotherapy concurrently. We found that the mice model constructed with recurrent tumors did not show a significant response to endocrine therapy combined with chemotherapy (figure 1I). However, when immunotherapy was added, tumor growth was significantly suppressed (figure 1I). The above results indicate that combination therapy with ER antagonist and chemotherapy potentiates the responsiveness to immunotherapy in ER⁺ breast cancer. Additionally, to investigate the role of the above combination regimen in tumor metastasis, we established a tumor metastatic model via tail vein injection and administered the aforementioned combination therapies. We found that endocrine therapy combined with chemotherapy slightly reduced lung metastasis of tumors, but the addition of immunotherapy significantly decreased lung metastasis (figure 1J,K), demonstrating that the combination of endocrine therapy and chemotherapy markedly enhances the efficacy of immunotherapy in advanced ER⁺ breast cancer.

Chemotherapy enhances the antigen-presenting capacity of ER⁻ but not ER⁺ breast cancer

Effective tumor immunity relies on functional antigen-presenting cells (APCs), among which dendritic cells (DCs) are the major mediators of tumor antigen presentation.^{33 34} To investigate the relationship between ER expression and tumor antigen presentation in the context of chemotherapy-induced immune potentiation, we first treated ER⁻ (MDA-MB-231, MDA-MB-468) and ER⁺ (MCF-7, BT-474) breast cancer cells with epirubicin, one of the commonly used anthracyclines in breast cancer. We first evaluated the IC50 of the chemotherapeutic agent epirubicin in breast cancer cells with varying levels of ER expression (online supplemental figure S3A). Based on these IC50 values, we treated breast cancer cells at the corresponding concentrations and performed apoptosis analysis, which confirmed no significant difference in dead cells between ER⁺ and ER⁻ groups (online supplemental figure S3B,C). Next, we cocultured monocyte-derived DCs with the ER⁻ and ER⁺ tumor cells. Coculture of ER⁻ breast cancer cells treated with epirubicin resulted in elevated DC maturation, as indicated by increasing expression of CD80 and CD86 (figure 2A,B). To further assess the immunostimulatory activity of ER⁻ breast cancer cells, we cocultured T cells with DCs exposed to epirubicin-treated cancer cells. T-cell proliferation was induced by DCs treated with ER⁻ tumor cells (figure 2C,D). Moreover, the ER⁻ tumor cell-treated group showed increased levels of perforin and granzyme B (figure 2E,F). Cytotoxicity assays were performed at different effector-to-target (E:T) ratios, and the optimal killing effect was observed at an E:T ratio of 10:1 (online supplemental figure S3D), which was used for subsequent analyses. The ER^- tumor cell-treated group showed enhanced cytotoxicity of cytotoxic T lymphocytes (CTLs) (figure 2G,H). We also evaluated CD8⁺ T cell death following coculture with ER⁺ and ER⁻ tumor cell types using TUNEL staining (online supplemental figure S3E). Although coculture with tumor cells indeed led to an increased proportion of dead CD8⁺ T cells compared with T-cell monoculture, no significant differences were observed among the different coculture groups. To exclude the possibility that tumor cell death resulted from direct chemotoxicity, we included a control in which CellTracker-labeled tumor cells were cultured under identical conditions without T cells (target tumor cells only). Minimal tumor cell death was observed in the absence of CD8⁺ T cells, confirming that the observed cytotoxicity was primarily mediated by CTLs (online supplemental figure S3D). The above results indicate that ER⁻ breast cancer enhances the antigen-presenting capacity of immune cells and further killing of tumor cells after chemotherapy.

Figure 2. ER⁻ breast cancer elicits a vigorous immune response after chemotherapy. (A, B) Epirubicin-treated breast cancer cells were treated for their effects on DCs. Maturation status of DCs was monitored by examining the surface expression of CD80 and CD86. The mean fluorescence intensity (MFI) of each marker and statistics are presented. (C, D)Proliferated T cells labeled with carboxyfluorescein succinimidyl ester (CFSE) following coculture with DCs under the indicated treatment were analyzed using flow cytometry. The percentages of gated CD8⁺ T cells stained with CFSE are shown. (E, F) CTLs receiving the indicated treatments were stained for intracellular perforin and granzyme B and analyzed using flow cytometry. The percentages of gated CD8⁺ T cells with perforin and granzyme B staining are shown in (E), and the statistics are shown in F. (G, H) CTLs were cocultured with CellTracker (Deep Red)-labeled breast tumor cells for 18 hours. Dead tumor cells are defined as PI⁺ CellTracker⁺ cells using flow cytometry. The percentages of PI⁺ tumor cells and the statistics are shown. Data are presented as mean±SD; n=3 per group (B, D, F, H). Statistical significance was determined using a two-sided one-way ANOVA with Tukey’s test (B, D, F, H), ***p<0.001; ns, not significant. ANOVA, analysis of variance; ER⁻, estrogen receptor-negative.

Chemotherapy induces ICD in ER⁻ breast cancer cells by promoting HMGB1 release

The above results indicate that tumor cell death induced by epirubicin triggered a T cell-mediated immune response in ER⁻ breast cancer. Based on literature demonstrating that combinational chemotherapy induces ICD,^{16 31 35} we hypothesized that epirubicin triggers ICD in ER⁻ breast cancer cells. ICD is characterized by the release or surface expression of DAMPs from dying tumor cells following specific treatments (anthracyclines, oxaliplatin, radiation therapy, etc). To determine whether epirubicin-treated ER⁻ breast cancer cells underwent ICD, we treated breast cancer cells with epirubicin in vitro and then determined the levels of CRT, ATP, and HMGB1. CRT exposure (online supplemental figure S4A) and ATP secretion (online supplemental figure S4B) showed no difference in ER⁻ and ER⁺ breast cancer cells. However, elevated HMGB1 release from the nucleus was observed only in the ER⁻ breast cancer cells, as confirmed using ELISA, immunofluorescence, and western blotting (figures 3A–C and online supplemental figure S4C).

Figure 3. ER⁻ breast cancer releases HMGB1 after chemotherapy. (A–H) Breast cancer cells were treated with or without epirubicin (IC50) for 24 hours. (A) HMGB1 levels in the extracellular medium (supernatant) of breast cancer cells were measured using ELISA. (B) Representative immunofluorescent images showing intracellular localization of HMGB1. Green, HMGB1; blue, DAPI. Scale bar, 5 µm. (C) Nuclear (HMGB1-n) and cytoplasmic (HMGB1-c) proteins were separated and blotted using anti-HMGB1 antibody. (D) MCF-7 breast cancer cells were transduced with shGFP or ESR1 shRNAs (shER1, shER2). HMGB1 levels in nuclear (HMGB1-n) and cytoplasmic (HMGB1-c) fractions of MCF-7 cells. (E) Analysis of HMGB1 expression in MCF-7 cells using ELISA. (F) Representative immunofluorescent images of BT474 cells treated with epirubicin, fulvestrant, or a combination of both for 24 hours. HMGB1 (red), ER (green), and DAPI (blue). Line-scan analysis was performed to analyze the colocalization. Scale bar, 5 µm. (G) Representative immunoblots for MCF-7 breast tumor cells treated with epirubicin, fulvestrant, or a combination of both for 24 hours. Nuclear (HMGB1-n) and cytoplasmic (HMGB1-c) proteins were separated and blotted using anti-HMGB1 antibodies. (H) DC-activated CTLs subjected to the indicated treatments were co-cultured with CellTracker (Deep Red)-labeled MCF-7 breast cancer cells for 18 hours. Dead tumor cells are defined as PI⁺ CellTracker⁺ cells using flow cytometry. The percentages of PI⁺ tumor cells are shown. Data are presented as mean±SD (A, E, H), n=3 per group (A, E, H), and three biologically independent experiments in western blotting. Statistical significance was determined using a two-sided one-way ANOVA with Tukey’s test (E, H) and a two-sided Student’s t-test (A). ***p<0.001, ns, not significant. ANOVA, analysis of variance; HMGB1, high mobility group box 1.

To identify whether impaired release of HMGB1 in the nucleus of ER⁺ breast cancer cells after chemotherapy was related to ER levels, we overexpressed ER in ER⁻ breast cancer cells, or knocked down ER in ER⁺ breast cancer cells. ER overexpression significantly inhibited HMGB1 translocation in epirubicin-treated MDA-MB-231 cells (online supplemental figure S4D,E). In contrast, ER knockdown increased HMGB1 release in MCF-7 cells (figure 3D, online supplemental figure S4F,G). The combination of epirubicin and ER degradation agent fulvestrant significantly promoted HMGB1 release (figure 3E–G, online supplemental figure S4H–J), while estrogen deprivation therapy or tamoxifen did not induce HMGB1 secretion from the nucleus (figure 3E). Moreover, combination treatment with fulvestrant improved the immunogenicity of tumor cells, as indicated by the elevated cytotoxicity of CTLs (figure 3H). These data indicated that chemotherapy induced ICD in ER⁻ breast cancer but failed to stimulate the release of HMGB1 in ER⁺ breast cancer.

HMGB1 release from the nucleus is regulated by SUMOylation

The release of intracellular proteins is regulated by various factors, including not only induction of cell permeability but also expression levels and post-transcriptional regulation.^{36 37} To investigate why HMGB1 release differs between ER⁻ and ER⁺ breast cancer cells after chemotherapy, we first examined the expression of HMGB1 in these cells. Western blot analysis revealed no significant difference in HMGB1 expression between the cell types (online supplemental figure S5A), indicating that the differential release is not due to variations in expression levels between the cells. Post-translational modifications, such as phosphorylation, methylation, acetylation, and SUMOylation, dynamically alter the compartmentalization and trafficking of proteins that regulate immunological processes.³⁸ Although previous studies have shown that acetylation of lysine residues in HMGB1 promotes its release,³⁹ our detection revealed no difference in acetylation; similarly, no difference was observed in phosphorylation (online supplemental figure S5B,C). Interestingly, we observed that ER⁻ breast cancer cells exhibited stronger HMGB1 SUMOylation than ER⁺ breast cancer cells (figure 4A). We used Subasumstat to inhibit cellular SUMOylation and found that Subasumstat significantly suppressed fulvestrant-induced and epirubicin-induced HMGB1 release, but did not suppress basal HMGB1 release in untreated MCF-7 cells (figure 4B). These results indicate that HMGB1 SUMOylation plays an important role in fulvestrant-induced and epirubicin-induced HMGB1 release in ER⁺ breast cancer.

Figure 4. ER-α hinders the SUMOylation of HMGB1. (A) Representative immunoblot analysis of SUMO1-modified HMGB1 immunoprecipitated using anti-HMGB1 antibodies from lysates of breast tumor cell lines. (B) HMGB1 expression in the cytoplasm of MCF-7 cells, which were stimulated with Subasumstat (Suba) for 24 hours, was assessed by immunoblotting. (C) Representative immunoblots analysis of SUMO1-modified HMGB1 immunoprecipitated using anti-HMGB1 antibodies from lysates of MCF-7 cells treated with or without fulvestrant for 24 hours. (D) HMGB1 SUMOylation was evaluated using an in vitro SUMOylation assay in the presence of SAE1/2 (SUMO E1), UBC9 (SUMO E2), SUMO1/2, ATP, HMGB1, and recombinant ER proteins. (E) Seven potential SUMOylated sites were predicted using online SUMOylation site analysis systems. Immunoblotting of SUMO1 in MDA-MB 231 cells were transfected with full-length or mutated His-containing HMGB1 plasmids. (F) Mapping of SUMOylation sites of HMGB1. (G-I) MDA-MB 231 cells were transfected with plasmids expressing HMGB1 WT, K86, K114, or K177 expression plasmids. (G) HMGB1 levels in the culture medium (HMGB1-m) and total HMGB1 inside cells (HMGB1-t) were detected using immunoblots. (H) Proliferating T cells labeled with carboxyfluorescein succinimidyl ester (CFSE) were coculture with DCs and analyzed using flow cytometry. The percentages of gated CD8⁺ T cells stained with CFSE are shown. (I) Percentages of PI⁺ tumor cells are shown. Data are presented as mean±SD (H, I), n=3 per group (H, I). Statistical significance was determined using a two-sided one-way ANOVA with Tukey’s test (H, I), **p<0.01***p<0.001. ANOVA, analysis of variance; ER, estrogen receptor; HMGB1, high mobility group box 1.

To further investigate the effect of ER on HMGB1 SUMOylation, we used fulvestrant to degrade ER protein and found that ER degradation improved HMGB1 SUMOylation in the ER⁺ breast cancer cell (figure 4C). Besides, overexpression of ER inhibited SUMOylation in ER⁻ breast cancer cell (online supplemental figure S5D), while silencing ER in ER⁺ breast cancer cell reversed its inhibitory effect on HMGB1 SUMOylation (online supplemental figure S5E). In vitro SUMOylation analysis indicated the attachment of SUMO-1, rather than SUMO-2/3, to HMGB1 (figure 4D, online supplemental figure S5F,G). Moreover, in vitro attachment of SUMO-1 to HMGB1 required the presence of both SUMO E1 and E2, and ER impeded the SUMOylation process (figure 4D).

SUMOylation usually occurs in a highly conserved recognition motif. Based on the online SUMOylation site analysis system (PhosphoSitePlus, SUMOplot Analysis Program, and JASSA), we chose seven putative SUMOylation sites, including lysine residues at positions 82, 86, 90, 114, 147, 167, and 177 of human HMGB1 (online supplemental figure S5H). To identify the lysine residue modified by SUMO, we constructed wild-type (WT) and lysine mutants of HMGB1 (K82R, K86R, K90R, K114R, K147R, K167R, and K177R). The K82, K90, K147, and K167 variants showed SUMOylation at levels similar to those of WT, whereas the K86, K114, and K177 mutants showed decreased SUMOylation of HMGB1 (figure 4E). K86 located between HMG-Box A and HMG-Box B, K114 in HMG-Box B, and K177 between HMG-Box B and the C-tail (figure 4F). To further determine whether HMGB1 was multi-modified at several lysine residues, we evaluated HMGB1 release using Western blotting, which showed that only the K114 mutant suppressed HMGB1 release from the nucleus (figure 4G). Furthermore, epirubicin-treated ER⁻ breast cancer cells with HMGB1 deletion and overexpressing the K114 mutant HMGB1 showed defects in stimulating the immune response, as indicated by decreased T-cell proliferation (figure 4H) and cytotoxicity of CTLs (figure 4I and online supplemental figure S5I).

ER interacts with HMGB1 and inhibits its SUMOylation

Our above results shown that ER inhibits the SUMOylation of HMGB1; however, the precise mechanism by which ER abrogates the process of HMGB1 SUMOylation is still unknown. To elucidate this mechanism, we first examined whether ER physically interacts with HMGB1. Coimmunoprecipitation assays in MCF-7 and BT-474 cells revealed that HMGB1 coprecipitated with ER, providing strong evidence for a direct interaction between ER and HMGB1 (figure 5A and online supplemental figure S6A). This observation suggests that ER may influence HMGB1 function by directly binding to it. To further delineate the structural basis of this interaction, we performed domain-mapping experiments using a series of deletion mutants of the HMGB1 domain, and found that ER interacted with the HMG Box B domain of HMGB1 (figure 5B,C). This finding indicates that the binding of ER to the HMG Box B domain may sterically hinder or allosterically alter the accessibility of K114, thereby preventing its SUMOylation.

Figure 5. ER-α hinders the SUMOylation of HMGB1 by direct binding. (A) Immunoprecipitation assay for the endogenous interaction between HMGB1 and ER-α in MCF-7 cells. (B) His-tagged HMGB1 or its deletion mutants were expressed in MCF-7 cells, immunoprecipitated with anti-His antibody, and immunobloted with anti-ER-α antibody. The numbers indicate amino acids included in the constructs. (C) Schematic of HMGB1 deletion mutants. (D) Schematic of ER-α deletion mutants. (E) Lysates of MDA-MB 231 cells were transfected with plasmids encoding His-tagged ER-α or its deletion mutants. Cell lysates were immunoprecipitated (IP) using anti-His antibody and analyzed using immunoblot with anti-HMGB1 antibody. The numbers indicate amino acids included in the constructs. (F) The interaction between HA-tagged full-length or truncated ER-α and HMGB1 was analyzed by precipitation of HMGB1 followed by immunoblotting using anti-HA antibody. (G) Effect of ER on HMGB1 SUMOylation was evaluated using an in vitro SUMOylation assay in the presence of SAE1/2 (SUMO E1), UBC9 (SUMO E2), SUMO1, ATP, and HMGB1 with a truncated ER recombinant protein. Following the SUMOylation reaction, the samples were analyzed by immunoblotting using anti-SUMO1 antibodies. (H) Predicted AlphaFold structure of HMGB1. Red arrows indicate the location of the modification site (K114). (I) Cell lysates treated as indicated were pulled down using an anti-CRM1 antibody and analyzed by Western blot for HMGB1. Changes in HMGB1 protein levels in the IP fraction indicate variations in the CRM1-HMGB1 interaction resulting from the treatments. (J−M) Nuclear HMGB1 protein was enriched via coimmunoprecipitation (J) The representative image is shown. (K−L) The nuclear HMGB1 DNA-binding capacity following chemotherapy, ER inhibitor monotherapy, or combination therapy was detected by DNA gel electrophoresis. Representative image is shown in K and statistics are shown in L. (M) DNA concentration was detected by nanodrop. (N−P) MCF-7 cancer cells were transfected with plasmids expressing HMGB1 mutation (K86V, K114V). The nuclear HMGB1 DNA-binding capacity was evaluated by DNA gel electrophoresis (N, O) or nanodrop (P). Data in (A, B, E–G) are representative of three biologically independent experiments. Mean±SD (L, M, O, P), n=3 per group (L, M, O, P). Statistical significance was determined using a two-sided one-way ANOVA with Tukey’s test (L, M, O, P), **p<0.01, ***p<0.001. ANOVA, analysis of variance; ER, estrogen receptor; HMGB1, high mobility group box 1.

To further validate the role of ER in this process, we investigated which domain of ER mediates its interaction with HMGB1. By using ER-deletion mutants, we identified the DNA-binding domain (DBD, 180–282 amino acids) as the critical region responsible for binding HMGB1 (figure 5D–F). To assess whether this interaction is functionally relevant to HMGB1 SUMOylation, we performed rescue experiments. Furthermore, restoring the expression of full-length ER or its DBD (180–282 amino acids) reversed the effect of ER deletion on HMGB1 SUMOylation (figure 5G), suggesting that the DBD of ER interacts with HMGB1 and inhibits its SUMOylation (figure 5H). Taken together, these findings reveal a previously unrecognized mechanism by which ER, through its DNA-binding domain, directly interacts with HMGB1 to inhibit its SUMOylation at K114, thereby modulating its post-translational modification and potentially influencing its functional activity.

Normally, when cells undergo immunogenic cell death, the nuclear membrane and plasma membrane rupture, allowing the loosely bound nuclear protein HMGB1 to diffuse into the extracellular space.⁴⁰ The loss of HMGB1 from the nucleus is directly linked to the instability of chromatin, making the DNA more accessible and contributing to the release of this protein into the extracellular space.⁴¹ To investigate the mechanism by which SUMOylation at K114 regulates HMGB1 translocation, we examined both its interaction with nuclear export machinery and its chromatin binding capacity. Coimmunoprecipitation assays using CRM1 (Exportin-1/XPO1), the best-characterized nuclear export receptor,⁴² showed no significant difference in HMGB1 association between the indicated conditions (figure 5I), suggesting that SUMOylation at K114 does not regulate HMGB1 nuclear export through enhanced interaction with CRM1.

We then assessed HMGB1-DNA interactions and found that combined treatment with epirubicin and fulvestrant markedly reduced chromatin-bound HMGB1 (figure 5J–M). Importantly, mutation of K114 significantly decreased HMGB1–DNA association compared with WT (figure 5N–P), indicating that this residue plays a critical role in chromatin binding. Together, these results suggest that SUMOylation at K114 promotes HMGB1 release primarily by weakening its chromatin binding rather than by modulating its interaction with nuclear export machinery.

ER antagonist and chemotherapy potentiate immunotherapy via HMGB1 release in vivo

To further evaluate the antitumor immune responses elicited by combination therapy of ER antagonist against ER⁺ breast cancer in vivo, we transplanted the ER⁺ breast cancer mice cell into immunocompetent C57BL/6 mice. Consistent with the ex vivo data, we observed that the combination therapy with fulvestrant and epirubicin promoted the release of HMGB1 and enhanced the infiltration of CD8⁺ T cells into the tumor microenvironment in vivo (figure 6A,B). Moreover, coculture of DCs with HMGB1-knockout tumor cells treated with chemotherapy markedly attenuated the upregulation of costimulatory molecules on DCs (online supplemental figure S7A,B). Additionally, HMGB1 deficiency significantly impaired CD8⁺ T cell proliferation and cytotoxicity, and correspondingly reduced tumor cell apoptosis within the coculture system (online supplemental figure S7C–E). Collectively, these findings indicate that HMGB1 promotes antitumor immune responses during immunogenic cell death.

Figure 6. The combination therapy of chemotherapy and immunotherapy plus fulvestrant inhibits the growth of established tumors by strengthening immune responses in immunocompetent mice. (A) Representative immunohistochemical images for HMGB1 and CD8 staining of mouse samples from DMMC-02 cells treated with epirubicin, fulvestrant, and the combination of epirubicin and fulvestrant (n=3 per group). Scale bar, 20 µm. (B) Representative immunoblots for DMMC-02 cells treated with epirubicin, fulvestrant, and the combination of epirubicin and fulvestrant for 24 hours (upper panel). The medium (m) protein was extracted and blotted for HMGB1. Western blot analysis of HMGB1 protein expression (lower panel). Results are normalized to GAPDH protein expression and presented as fold increase. (C–F) C57BL/6 mice were grafted with DMMC-02 cells and treated with fulvestrant (F) and epirubicin (E) in the presence or absence of neutralizing antibodies against PD1 (a-PD1), CD8 (a-CD8), or HMGB (a-HMGB) after tumors were palpable. (C) Tumor growth curves of C57BL/6 mice injected with fulvestrant and epirubicin, or together with α-PD1 and either α-CD8 or α-HMGB1 after DMMC-02 tumors were palpable. (D) Tumor size is estimated using PET/CT; representative images are shown, and circles indicate tumors. Percent ID/g, percentage of injected dose per gram. (E) Representative immunofluorescent images for ER and CD8; the nuclei were counterstained with DAPI. Scale bar, 20 µm. (F) ELISpot analysis of IFN-γ. (G, H) DMMC02 cells with HMGB1 knockout were implanted into C57BL/6 mice, which were treated with fulvestrant (F) and epirubicin (E) in the presence or absence of neutralization antibodies against PD-1 (a-PD1) after the tumors were palpable. The tumor growth (G) and CD8⁺ T cells infiltration (H) are shown. (I) Schematics highlight the major findings of this study. Data are presented as mean±SD (B, C, G); n=3 (B) and 6 (C, G) per group. Statistical significance was determined using a two-sided one-way ANOVA with Tukey’s test. ***p<0.001. ANOVA, analysis of variance; ER, estrogen receptor; HMGB1, high mobility group box 1; ICD, immunogenic cell death; NAC, neoadjuvant chemotherapy; PET/CT, positron emission tomography/CT.

To further elucidate the role of HMGB1 in mediating the antitumor effects of the combination therapy in vivo, we administered α-HMGB1 neutralizing antibodies to tumor-bearing mice subjected to the combination therapy. The blockade of HMGB1 abolished the synergistic effect of epirubicin and fulvestrant plus anti-PD-1 in tumor suppression, as evidenced by an increase in tumor burden (figure 6C). Notably, the inhibitory effect of HMGB1 blockade closely resembled that of CD8⁺ T cell depletion, indicating that HMGB1 may exert its tumor-suppressive effects through CD8⁺ T cell activation (figure 6C). This was further corroborated by positron emission tomography/CT (PET-CT), immunofluorescence, and enzyme-linked immunospot (ELISpot) assays, all of which confirmed a reduction in CD8⁺ T cell infiltration and activity on HMGB1 blockade (figure 6D–F). Collectively, these findings suggest that ER antagonist potentiate combined immunotherapy and chemotherapy through HMGB1-dependent antitumor CD8⁺ T cell response. To further confirm these findings, we generated HMGB1-knockout ER⁺ cells (online supplemental figure S8A) and orthotopically inoculated them into the mammary fat pads of C57BL/6 mice. Our results indicated that the triple therapy was significantly more effective than dual therapy (chemotherapy plus ER inhibition); however, on HMGB1 knockout, the efficacy of the triple therapy was markedly diminished (figure 6G and online supplemental figure S8B). Immunofluorescence analysis of CD8-positive cells at the tumor site revealed that the triple therapy significantly promoted CD8⁺ T cell infiltration into tumor tissues, whereas HMGB1 knockout substantially inhibited this infiltration (figure 6H).

Moreover, we employed a lentiviral vector to overexpress ER in EO771 cells to further validate our results (online supplemental figure S8C,D). Consistent with the above findings, the combination of chemotherapy and ER inhibition therapy significantly enhanced CD8⁺ T cell infiltration into the tumor microenvironment (online supplemental figure S8E) and inhibited orthotopic tumor growth (online supplemental figure S8F,G); however, CD8⁺T cells depletion or HMGB1 inhibition significantly interrupted the functions of chemotherapy and ER inhibition (online supplemental figure S8F,G). RNA-seq and pathway analyses on in vivo tumor samples showed that relevant signaling pathways such as inflammatory response, TNF/NF-KB signaling, and apoptosis-related pathways were significantly upregulated following chemotherapy and ER inhibition therapy, indicating that these treatments enhanced both local immune responses and tumor cell death (online supplemental figure S8H).

Discussion

In this study, we demonstrated that ER interrupts HMGB1 SUMOylation-mediated ICD to promote immune evasion; ER inhibition remodels the tumor immune suppressive microenvironment, enhancing the efficacy of both chemotherapy and immunotherapy (figure 6I). ER⁺ breast cancer is highly sensitive to endocrine therapy, but effective therapeutic strategies remain lacking for recurrent or metastatic tumors. Immunotherapy, a critical modality for treating various solid tumors, has demonstrated clinical benefit across multiple tumor types. Although breast cancer is typically characterized as an immunosuppressive “cold tumor”, successes in triple-negative breast cancer treatment have challenged this notion. Studies have shown that tumors with low or absent ER expression respond significantly better to immunotherapy than those from patients with high ER expression.^{7 43} However, why ER expression correlates so strongly with divergent tumor immune responses remains poorly understood. ER degradation agent fulvestrant could modulate macrophage and neutrophil infiltration in previous reports⁴⁴; however, the regulation of cancer immune responses remains unknown.⁴⁵ Our findings show that fulvestrant elicits antitumor immune responses by enhancing the maturity and antigen presentation of DC cells and infiltration of CTLs in ER⁺ breast cancer. Furthermore, this effect was significantly augmented following combination chemotherapy. As is well established, chemotherapy demonstrates suboptimal efficacy in patients with ER⁺ breast cancer; thus, its role in ER⁺ breast cancer extends beyond mere tumor cytotoxicity.

Recently, some researchers discovered chemotherapy triggers ICD in multiple tumors, which releases potent danger signals known as DAMPs and improves tumor immune killing.^{14 28 30} We further found that ER inhibited chemotherapy-induced HMGB1 release, which partially accounts for the poor efficacy of immunotherapy and chemotherapy in ER⁺ breast cancer. However, how ER modulates chemotherapy-induced HMGB1 release remains unclear. HMGB1 release is subject to multiple layers of regulation, but the most critical step involves its translocation from the nucleus to the cytoplasm; among these regulatory processes, post-transcriptional regulation is a key mediator. We demonstrated that ER inhibits HMGB1 SUMOylation by interacting with the HMG Box domain of HMGB1, thereby preventing its release from the nucleus and contributing to the immune tolerance observed in ER⁺ breast cancer. These mechanisms explain why chemotherapy or endocrine therapy enhances the efficacy of immunotherapy in ER⁺HER2⁻ breast cancer. However, ER signaling operates through three key regulatory modalities, including ligand-dependent, ligand-independent, and direct protein binding. We have only identified the direct protein-binding modality in its regulation of ICD; other pathways may contribute, warranting further investigation. Furthermore, future research could focus on evaluating these findings in other cancers, such as endometrial, ovarian, and prostate cancers, to explore the broader applicability of ER inhibition in cancer therapy.

Collectively, our findings shed light on why ER⁺ breast cancer exhibits the characteristics of cold tumors and evaluated the efficacy of ER inhibition combined with chemotherapies and immunotherapies, such as anti-PD-1/PD-L1. This study provides new insights into the mechanisms and therapeutic strategies for ER⁺ breast cancer, highlighting the potential of ER inhibition as a novel therapeutic strategy to enhance immune response in ER⁺ breast cancer and provide a rationale for future clinical trials combining ER inhibitors and immunotherapies with chemotherapies.

Materials and methods

Sex as a biological variable

Our study only examined female patients with breast cancer and used female mice for the animal model because breast cancer predominantly affects females, and the disease model is specifically relevant to females.

Patients and tissue samples

Tumor samples were obtained from patients with invasive breast carcinoma at Sun Yat-sen Memorial Hospital, Sun Yat-sen University (Guangzhou, China) between 2010 and 2025.

To investigate the effects of HR (ER, PR) status on therapeutic response to NAC (pCR according to the FDA guidance), clinical data were collected from 483 patients who received NAC at Sun Yat-sen Memorial Hospital (Guangzhou, China) between 2010 and 2018 (table 1). The NAC regimens comprised doxorubicin or epirubicin 60–80 mg/m² every 3 weeks for four to eight cycles. An external validation cohort consisting of 161 patients was obtained from the I-SPY 1 Trial (GSE32603). All patients adopted doxorubicin and cyclophosphamide as initial chemotherapy (54% received biweekly dose-dense treatment, 34% followed triweekly standard cycles, and 12% received modified regimens). Patients with missing ER status or pCR information were excluded from the analysis. The internal and external cohorts were analyzed independently to evaluate the association between HR status and response to NAC.

To investigate the effects of HR (ER, PR) status on therapeutic response to NAC plus immunotherapy (pCR according to the FDA guidance), clinical data were collected from 110 patients who received NAC combined with immunotherapy at Sun Yat-sen Memorial Hospital (Guangzhou, China) between 2020 and 2025 (table 3). The NAC regimens comprised doxorubicin or epirubicin 60–80 mg/m² every 3 weeks for four to eight cycles concurrently with immunotherapy agents including pembrolizumab, toripalimab, camrelizumab and sintilimab. An external validation cohort consisting of 67 patients was obtained from the I-SPY 2 Trial. Participants underwent standard neoadjuvant chemotherapy (NACT): weekly intravenous paclitaxel at 80 mg/m² for 12 weeks, followed by four cycles of AC regimen consisting of 60 mg/m² doxorubicin combined with 600 mg/m² intravenous cyclophosphamide every 2 to 3 weeks.

For gene expression analysis of ESR1 and CD8A, PRF1, TBX21, RNA sequencing (RNAseq) profiles from paired prechemotherapy and postchemotherapy tumor samples of 87 patients with breast cancer were obtained from the GEO database (GSE192341). Immunohistochemistry staining was performed for paired biopsy and surgically resected samples from 25 patients before and after NAC at Sun Yat-sen Memorial Hospital between 2018 and 2023. In addition, immunofluorescence staining for ER and CD8 was performed on 25 female patients who underwent NAC at the Sun Yat-sen Memorial Hospital between 2018 and 2023.

Cell culture and induction of cell death

Human breast cancer cell lines, MDA-MB 231, MDA-MB 468, MCF-7, and BT-474, were obtained from American Type Culture Collection (ATCC) and cultured according to standard protocols. Cells were maintained at 37°C and 5% CO₂ and tested for mycoplasma contamination. In some experiments, cells were stimulated with 100 nM Subasumstat for 24 hours. MDA-MB 231, MDA-MB 468, MCF-7, and BT-474 were treated with IC50 concentration of epirubicin measured using the MTS assay.

Cell viability assays

To test cell viability, 1.5×10³ cells suspended in 50 μL culture medium were seeded into flat-bottom 96-well plates. After 24 hours of incubation at 37 °C, 50 µL of drug-containing medium or medium-only vehicle controls were added to designated wells in duplicate. Cell viability was quantified at 24 hours and 48 hours post-treatment using the MTS-CellTiter 96 aqueous non-radioactive cell proliferation assay (Cat# G3582, Promega) according to the manufacturer’s instructions. MTS reagent was added to each well, and the plates were incubated at 37 °C in the dark for 1–4 hours. The absorbance was measured at 490 nm using an Infinite F500 plate reader (Tecan). Each analysis included six replicate wells, and assays were repeated in a minimum of three independent experiments.

In vitro cellular apoptosis assay

Cells exposed to indicated chemotherapeutic agents for the indicated time periods were dissociated using 0.25% trypsin-EDTA and collected via centrifugation. Apoptosis was quantified with the Annexin V Apoptosis Detection Kit (Cat# 640911, BioLegend). Cells were incubated for 15 min at room temperature with 100 µL of binding buffer containing 5 µL of 647-conjugated antibody against annexin V. After washing, cells were resuspended in 200 µL binding buffer containing 5 µL of propidium iodide (PI) staining solution and promptly subjected to flow cytometric analysis.

Isolation of primary cells from peripheral blood

Primary cells were isolated from peripheral blood following a previously published protocol. Mononuclear cells were harvested from the interface cell layer formed between 45% and 60% Percoll. Magnetic-activated cell sorting using the CD8⁺ T Cell Isolation Kit (Cat# 130–096-495, Miltenyi Biotec) or the CD14⁺ Cell Isolation Kit (Cat# 130-050-201, Miltenyi Biotec) was used to obtain pure CD8⁺ T cell or CD14⁺ populations from mononuclear cells.

Generation of tumor-specific DCs and T cells

DCs were generated following previously published protocols with modifications.²⁹ CD14⁺ mononuclear cells were isolated from the peripheral blood samples collected from HLA-A2⁺ healthy donors and breast cancer patients, followed by 6-day culture in DMEM supplemented with 50 ng/mL GM-CSF (Cat# NBP2-34893, Novus Biologicals), 20 ng/mL IL-4 (NBP2-34896, Novus Biologicals), and 10% heat-inactivated autologous serum. Culture medium and cytokines were refreshed every other day, and cell differentiation was evaluated using light microscopy. The resulting DCs were cocultured for 24 hours with tumor cells by five freeze–thaw cycles (200 µg protein per 1×10⁶ cells/mL) or with epirubicin-treated tumor cells. To generate tumor-specific CTLs, CD8⁺ T cells isolated from peripheral blood were incubated with antigen-specific DCs (5:1) in RPMI 1640 medium supplemented with 25 U/mL IL-2 (Cat# 200-02-50, PeproTech) and 10% heat-inactivated autologous serum for 6 days.

Flow cytometry

For flow cytometric analysis, cells were incubated with the antibodies listed below: anti-CD80-PE (Cat# 12-0809-42, eBioscience), anti-CD83-FITC (Cat# 11-0839-42, eBioscience), anti-CD86-APC (Cat# 17-0862-82, eBioscience), anti-CD11c-ef450 (Cat# 48-0116-42, eBioscience), anti-CD8a-APC (Cat# 17-0086-42, eBioscience), anti-Perforin-eFluor 450 (Cat# 48-9994-42, eBioscience), and anti-granzyme B-PE (Cat# 12-8896-42, eBioscience). For intracellular staining, cells were first treated using an Intracellular Fixation and Permeabilization Kit following the manufacturer’s protocols (Cat# 88-8824, eBioscience). Samples were subsequently analyzed using the CytoFlex Flow cytometer (Beckman Coulter). The data were analyzed using the FlowJo software (version 10, TreeStar).

Fluorescence detection of cell surface CRT

First, breast cancer cells were washed with FACS buffer (1×PBS supplemented with 5% fetal bovine serum) and subsequently incubated for 30 min with rabbit anti-CRT antibodies (1:100, Cat# ab92516, Abcam) at 4°C in FACS buffer. Cells were then incubated for 30 min with Alexa Fluor secondary antibodies (1:300, Cat# A32731, Thermo Fisher Scientific) at 4°C in FACS buffer. Following three washes with FACS buffer, cell-surface CRT was quantified using a CytoFlex Flow cytometer (Beckman Coulter).

ELISA

Tumor cells were cultured in DMEM supplemented with 10% human serum until reaching 80% confluence. These cells were then treated with the indicated treatment and cultured in fresh, serum-free media. Conditioned media were collected at 24 hours and subjected to ELISA. ELISA kit for HMGB1 was purchased from Novus Biologicals (Cat# NBP2-62781). The experiments were performed according to the manufacturer’s protocols.

Immunofluorescence and immunohistochemistry

Paraffin-embedded tissue blocks were sectioned at a thickness of 4–6 μm. Deparaffinized slides were heated in an EDTA buffer (pH 8.0) using a pressure cooker for 3 min for antigen retrieval. The samples were blocked in PBS with 10% bovine serum albumin (BSA) for 10 min at room temperature.

For immunofluorescence, sections were incubated overnight at 4℃ with primary antibodies against CD8 (1: 100, Cat# sc-7970, Santa Cruz), ER-α (1: 100, Cat# 21244-1 AP, Proteintech), or HMGB1 (1: 100, Cat# ab18256, Abcam). Subsequently, samples were incubated with Alexa fluor–conjugated secondary antibodies (1:300, Cat# A31572 and Cat# A21202, Thermo Fisher Scientific) for 1 hour at room temperature. Cells were counterstained with 4’,6-diamidino-2-phenylindole (DAPI, 5 mg/mL, Cat# D3571, Thermo Fisher Scientific). The images were obtained using the Zeiss LSM800 confocal microscope. Line-scan analysis of the representative images was performed using ImageJ and created by GraphPadsoftware.

For the TUNEL assay, the aforementioned tissue was incubated overnight with a primary antibody, washed three times, and stained with an In Situ Cell Death Detection Kit (Cat# 11684817910, Roche). TUNEL images were acquired by an LSM880 confocal microscope. TUNEL positivity was expressed as the number of TUNEL⁺ cells divided by the total number of DAPI⁺ nuclei per field.⁴⁶

For immunohistochemistry (IHC), samples were incubated overnight at 4℃ with primary antibodies including rabbit anti-human ER-α (1: 100, Cat# ab138492, Abcam), rabbit anti-human CD8 (1:100, Cat#, ZA-0508, ZSGB-BIO), anti-ER polyclonal (1:100, Cat# 21 244-1 AP, Proteintech), anti-HMGB1 (1:100, Cat# ab79823, Abcam), or anti-CD8 (1: 100, Cat# ab7778, Abcam) antibodies. After three PBS washes, sections were incubated with IHC secondary antibody (Cat# GK500705, Gene Tech) following the manufacturer’s protocols.

T cell proliferation assay

Autologous CD8⁺ T cells were stained with 0.5 mM 5,6-carboxyfluorescein succinimidyl ester (CFSE) (Cat# C34554, Thermo Fisher) for 15 min at room temperature, then co-cultured with the previously generated mature DCs at a 5:1 ratio. Briefly, CD8⁺ T cells were incubated with antigen-loaded DCs at a 5:1 ratio in RPMI 1640 medium supplemented with 25 U/mL IL-2 (Cat# 200-02-50, PeproTech) and 10% heat-inactivated autologous serum. After 4 days of coculture, CD8⁺ T cell proliferation was quantified from CFSE dilution profiles obtained by flow cytometry (Beckman Coulter).

Cytotoxicity assays

Breast cancer cells were stained for 15 min at 37 °C using the CellTracker Deep Red Dye (Cat# C34565, Thermo Fisher Scientific). Tumor antigen–activated CD8⁺ T cells, generated by incubation with tumor antigen–pulsed DCs as described above, were collected using CD8 MicroBeads (Cat# 130-045-201, Miltenyi Biotec). Effector and labeled target breast cancer cells were combined at a 10:1 ratio. Following the 18-hour incubation, the entire culture was collected, stained with PI (3.75 mM stock diluted 1:500) immediately before analysis on a CytoFlex Flow cytometer (Beckman Coulter).

Determination of ATP content

Tumor cells, cultured in 24-well plates to 70%–80% confluence was obtained, washed, and incubated with or without epirubicin for 24 hours. Supernatants were collected, and ATP concentration was measured immediately using the ATP Determination Kit (Cat# A22066; Invitrogen). ATP-driven chemiluminescence was recorded using the Infinite F500 instrument (Tecan, Männedorf, Switzerland).

Transfection with short hairpin RNA

The lentiviral vector pLKO.1-EGFP-puro was used to generate stable cell clones. Lentiviral short hairpin RNA (shRNAs) targeting ESR1 were purchased from IGE bio, and the corresponding target sequences are listed in online supplemental table S1.

Nuclear and cytoplasmic extraction

Nuclear and cytoplasmic proteins were extracted using NE-PER Nuclear and Cytoplasmic Extraction Reagents (Cat# 78833, Thermo Fisher Scientific) following the manufacturer’s protocols. Briefly, cytoplasmic protein extraction was performed by adding CER I and CER II subsequently to fresh cells. After centrifugation, the cytoplasmic extract was transferred, and the insoluble pellet fraction containing nuclei was suspended with NER for nuclear extract. Nuclear/cytoplasmic protein concentrations were quantified with the Pierce BCA protein assay kit (Cat# A55860, Thermo Fisher Scientific).

Immunoblotting

Protein lysates were fractionated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and electrotransferred onto polyvinylidene fluoride membranes. Membranes were incubated with antibodies against HMGB1 (Cat# ab79823, Abcam), histone H3 (Cat# 9715S, CST), GAPDH (Cat# HRP-60004, Proteintech), SUMO-1 (Cat# 4930s, CST), SUMO-2/3 (Cat# 4971, CST), ER-α (Cat# 21244-1 AP, Proteintech), His tag (Cat# 66005-1 Ig, Proteintech), and HA tag (Cat# 3724T, CST) antibodies. Peroxidase-conjugated secondary antibodies (Cat# 7074, 7076, CST) were used, and protein bands were visualized using enhanced chemiluminescence system (ECL, Thermo Fisher Scientific). Blot images were cropped for presentation based on molecular weight markers (kDa).

Animal experiments

6–8-week-old female C57BL/6 mice were obtained from the Animal Experiment Center of Sun Yat-sen University and maintained under specific pathogen-free facility. Female mice were used in this study, as the modeled disease is clinically relevant in females.

To construct mice ER⁺ breast cancer model in mice, 6-week-old female mice were anesthetized using ketamine–xylazine. Each animal received subcutaneous medroxyprogesterone acetate (MPA) pellets designed to release 50 mg over 90 days. DMBA (5 mg/mL) was delivered via oral gavage at 200 µL per dose during week 1, 2, 4, 5, 7, and 8. Palpable lesions were removed and divided into 0.5 mm³ tumor fragments. Fresh tumor fragments were implanted into the inguinal mammary gland fat pads of NSG mice. Once the tumor mass became prominent, it was removed aseptically and implanted into NSG mice. When a tumor transplant developed a palpable tumor, single tumor cells were isolated and cultured. Fibroblast contamination was removed by selective trypsinization to develop a Mice ER⁺ breast cancer cell line, namely the DMMC-02 cell line.

DMMC-02 cells (1×10⁶) were injected into the fat pads of C57BL/6 mice. Treatment was initiated when the tumor volume reached 100–200 mm³. PBS, epirubicin (7 mg/kg, weekly, intraperitoneally), fulvestrant (150 mg/kg, once a week, subcutaneously), or a combination treatment of epirubicin and fulvestrant was administered to the C57BL/6 mice of the respective groups.

To evaluate the function of HMGB1 and CD8 in our experiments, mice were intraperitoneally administered (twice a week) with either an α-HMGB1 (1 mg/kg; 3E8, BioLegend) or α-CD8 (200 µg; 2.43, BioXcell) antibody 1 week before the initiation of epirubicin (7 mg/kg, weekly) and fulvestrant (150 mg/kg, once a week) treatments. In some experiments, twice-weekly intraperitoneal injection of α-PD-1 (200 µg; RMP1-14, BioXcell) was used to check the efficiency of immunotherapy. Tumors were not permitted to exceed the ethics-approved limit of 2000 mm³. Caliper measurements were obtained every 3 days and volume was calculated as (length×width²)/2, as previously described.⁴⁷

To assess the local immune response, 5×10⁵ DMMC-02 cells were injected into the footpad of C57BL/6 mice. After 5 days, popliteal lymph node cells were harvested and homogenized; single-cell suspensions were prepared by filtering tissue homogenates through a sterile cell strainer (70 µm; Becton Dickinson). Isolated lymph node cells (1×10⁵) were cultured in round-bottom 96-well plates containing 200 µL complete culture medium, with or without lysates of DMMC-02 cells killed with a freeze–thaw cycles. After 3 days, culture supernatants were collected, and the secretion of IFN-γ was determined using ELISA (Cat# 1210002, DAKAWE).

Tumor metastasis and in vivo imaging system

To evaluate the potential for lung metastasis, a luciferase-expressing DMMC-02 cell line was established by pretransduction with a lentiviral vector carrying the luciferase gene. These engineered cells were then injected into the animal models via tail vein and accepted relative treatment. After the designated experimental period, the mice were euthanized, and their lungs were carefully excised. The harvested lung tissues were immediately immersed in a solution of d-luciferin, the substrate for luciferase, for a duration of 10 min to permit sufficient penetration and reaction. Following this incubation, the tissues were imaged ex vivo with the IVIS Lumina imaging system (Caliper Life Sciences) to capture the bioluminescent signals indicative of metastatic foci. Finally, a quantitative bioluminescent flux analysis was performed on the acquired images using the Living Image software V.3.0 (Caliper Life Sciences) to objectively measure the metastatic burden.

PET/CT Imaging

Tumor progression and therapeutic efficacy were evaluated using Fluorine-18(18F)-fluorodeoxyglucose (¹⁸F-FDG) PET/CT following drug administration. Experiments were performed as we previously described. Briefly, mice were fasted overnight, anesthetized with pentobarbital, and then injected intravenously via the tail vein with ¹⁸F-FDG (5 µCi/g) in 100 µL of saline. A 15-minute static scan was acquired using an Inveon microPET/CT scanner. Volumes-of-interest were analyzed in Inveon Research Workplace software. Tissue uptake was calculated by normalizing the measured radioactivity concentration (Bq/cm³) to the decay-corrected activity concentration (Bq) and multiplying the resulting value by 100, reported as percentage injected dose per gram (%ID/g).

Quantitative reverse transcription real-time PCR

For total RNA extraction, the culture medium was discarded, and cells were rinsed with PBS and lysed with TRIzol (Cat# A33251, Thermo Fisher Scientific). RNA was extracted following the manufacturer’s protocols. The extracted RNA was reverse-transcribed into cDNA using PrimeScript RT Master Mix (Cat# RR036A, Takara Bio). Quantitative PCR was subsequently performed with a LightCycler 480 instrument (Roche) using SYBR Premix Ex Taq II (Cat# RR820A, TaKaRa Bio). The primer sequences are provided in online supplemental table S2.

IFN-γ ELISpot assay

IFN-γ-based ELISpot assay was conducted as previously described.⁴⁸ Briefly, DMMC-02 tumor-bearing mice were treated with different drugs when the grafts became palpable. Five weeks post-treatment, lymph node samples were harvested to prepare a single-cell suspension. T cells were enriched using CD3⁺ T-cell isolation kits (Cat #70903-100, BEAVER). Purified T cells were co-cultured with irradiated DMMC-02 tumor cells (3×10⁴) to induce tumor-specific T-cell activation. Following 48 hours of culture, the ELISpot assay was carried out with the IFN-γ ELISpot kit (Cat# 2210002, Dakewe Biotech) according to the manufacturer’s protocols. IFN-γ-positive spots were counted using an Immunospot S6 Analyzer (Cellular Technology).

Immunoprecipitation

Cells with the indicated treatments were lysed in IP lysis buffer (Cat# 87787, Thermo Fisher Scientific) supplemented with protease inhibitor cocktail (Cat# 78446, Thermo Fisher Scientific). Lysates were mixed with the indicated antibodies at room temperature for 1 hour. Dynabeads Protein A (50 μL; Cat# 10 001D, Thermo Fisher Scientific) were then added for another 1 hour incubation. Immunocomplexes were washed four times with immunoprecipitation buffer and released by boiling for 5 min in 1X loading buffer, and separated using 10% SDS-polyacrylamide gels and probed with the corresponding antibodies.

Mammalian expression vectors

SUMOylation-site mutant constructs were generated by replacing a lysine with an arginine at the indicated mutation site. cDNAs encoding full-length human HMGB1, mutant constructs, HMGB1 Box A (amino acids 1–90), and HMGB1 Box B (amino acids 90–215) were subcloned into the pcDNA4.1 vector with an N-terminal 6×His tag. cDNAs encoding full-length human ER and its deletion mutants (amino acids 1–185, amino acids 180–282, and amino acids 282–595) were subcloned into the pcDNA4.1 vector containing an N-terminal HA tag.

Recombinant protein production and purification

Full-length and truncated ER-α constructs were cloned into pET28a expression vector (Novagen) with a C-terminal His tag. and introduced into Escherichia coli BL21(DE3). Transformants were expanded at 37 °C in kanamycin-supplemented LB medium to an OD600 of 0.7. Expression was initiated with a 0.4 mM isopropyl-β-D-thiogalactopyranoside, and cultures were maintained for a further 18 h at 20°C. Bacterial cells were pelleted and suspended in lysis buffer containing 20 mM sodium phosphate (pH 7.5), 10 mM imidazole, 500 mM NaCl and EDTA-free protease inhibitors. After sonication, insoluble debris was removed via centrifugation. His-tagged proteins were purified from the supernatants using a HisPur Ni-NTA Purification Kit (Cat# 88229, Thermo Fisher Scientific) and eluted with elution buffer containing 20 mM sodium phosphate (pH 7.6), 500 mM NaCl, 500 mM imidazole. The eluates were further concentrated and subjected to buffer exchange using Amicon Ultra-0.5 spin columns (Cat# Z677094; Merck Millipore). Concentrations were derived from A₂₈₀ measurements obtained with an ND-2000C NanoDrop spectrophotometer (NanoDrop Technologies), and purity was examined by Coomassie blue staining. Purified proteins were transferred to storage buffer comprising 50 mM Tris-HCl, 500 mM KCl, 1 mM DTT, and 5% glycerol. Purified recombinant proteins were rapidly frozen in liquid nitrogen and stored at −80°C.

In vitro SUMOylation assay

Reactions were assembled with the SUMOylation assay kit (Cat# BML-UW8955, Enzo Life Sciences) in a final reaction volume of 30 μL. The reaction buffer contained 5 mM MgCl2, 0.1% Tween-20, 50 mM NaCl, and 20 mM HEPES (pH 7.6), together with 0.78 μM E1, 8.4 μM peptide, 0.35 μM UBC9, 22.7 μM SUMO-1 or GST-fusion-SUMO-2, and 4 mM ATP. In some experiments, truncated recombinant ER protein was supplemented to the reaction mixture. After 15 min at 37 °C, each reaction was quenched with 30 μL of 360 mM DTT loading buffer, and proteins were resolved by SDS-PAGE.

In vivo coimmunoprecipitation assay

After the indicated treatments, cells were solubilized in immunoprecipitation buffer (Cat# 87787, Thermo Fisher Scientific) supplemented with protease inhibitor cocktail (Cat# 78446, Thermo Fisher Scientific). Precleared cell lysates were incubated overnight at 4 °C with an anti-His (Cat# 66 005–1 Ig, Proteintech) or anti-HA mAb (Cat# 3724T, CST), then bound to protein A beads (Cat# 10 001D, Thermo Fisher Scientific). Immunocomplexes were washed four times with buffer, separated by SDS-PAGE, and examined by immunoblotting with the corresponding antibodies mentioned above.

CRISPR-mediated gene knockout

The HMGB1-targeting gRNA sequences were as follows: human HMGB1 gRNA, 5′- GGAAGAGTCGACTCGCTT-3′ and mouse HMGB1 gRNA, 5′-TCATACTCACGGAGGCCTCT-3′. Cas9 and sgRNA lentivirus were obtained from GenePharma and transduced into human tumor cell lines (MDA-MB-231, MDA-MB-468, MCF-7, and BT-474) and the murine tumor cell line DMMC02. Following transduction, cells were selected with 2.5 μg/mL puromycin for 1 week to generate stable HMGB1 knockout tumor cell lines.

Generation of ER-overexpressing E0771 cells

To establish stable ER overexpression in E0771 murine mammary carcinoma cells, a cDNA encoding full-length murine ER was subcloned into the pcDNA4.1 vector. ER overexpression was validated by quantitative reverse-transcription PCR and immunoblot analysis. Cells with stable ER overexpression were expanded and used for all subsequent experiments.

Analysis of HMGB1–DNA binding

To assess DNA binding to HMGB1, nuclear proteins were first extracted from each treatment group. HMGB1-containing complexes were then enriched by immunoprecipitation. The immunoprecipitate was run on a 1% agarose gel and visualized under UV illumination. Band intensity was quantified using ImageJ software. For relative quantification of nucleic acid content associated with HMGB1 complexes, immunoprecipitated samples were measured using a NanoDrop spectrophotometer.

Quantification and statistical analysis

Statistical analyses and data visualization were performed using SPSS for Windows V.13.0 (SPSS) and GraphPad Prism V.7.0 (Dotmatics). Sample sizes and detailed statistical parameters are provided in the corresponding figure legends. Data are presented as the mean±SD. Univariable and multivariable binary logistic regression analyses were performed to evaluate associations between clinicopathological factors and pathological complete response (pCR). Two-tailed Student’s t-tests were used for comparisons between the two groups. One-way analysis of variance with Tukey’s multiple comparison tests was adopted for analyses involving three or more groups. Pearson’s correlation analysis was conducted to evaluate correlations among the expression of ESR1, CD8A, and PRF1. p < 0.05 was defined as statistically significant.

Data availability statement

Data are available on reasonable request.