Hormone therapy enhances anti-PD1 efficacy in premenopausal estrogen receptor-positive and HER2-negative advanced breast cancer

Summary

The efficacy of immunotherapy for estrogen receptor-positive/HER2-negative (ER+/HER2−) metastatic breast cancer (MBC) has not been proven. We conduct a phase 1b/2 trial to assess the efficacy of combining pembrolizumab (anti-PD1 antibody), exemestane (nonsteroidal aromatase inhibitor), and leuprolide (gonadotropin-releasing hormone agonist) for 15 patients with premenopausal ER+/HER2− MBC who had failed one to two lines of hormone therapy (HT) without chemotherapy. The primary endpoint of progression-free survival rate at 8 months (i.e., 64.3%) is achieved. Moreover, 5 of the 14 evaluable subjects exhibited partial responses (overall response rate = 35.7%). The combination of anti-PD1 antibody and anti-hormone therapy is associated with an enhanced immunoreactive microenvironment influencing treatment efficacy, as observed in pre- and post-treatment tumor samples through NanoString analysis. Post-treatment tumors are associated with increased immune response and immune cells. The findings indicate that combining HT with anti-PD1 antibody is a promising treatment strategy for patients with premenopausal ER+/HER2− MBC. This study was registered at ClinicalTrials.gov (NCT02990845).

Graphical abstract

Highlights

• •HT increases anti-PD1 antibody efficacy in premenopausal ER+/HER2− MBC
• •HT plus anti-PD1 antibody increases immune cell infiltration and activation
• •Estrogen suppression plus immunotherapy activates ER+/HER2− tumor microenvironment

Chen et al. demonstrate that the combination of hormone therapy and anti-PD1 antibody is effective in patients with premenopausal ER+/HER2− MBC. The combination treatment turns the tumor microenvironment into immunoactive status.

Introduction

Immunotherapy has revolutionized the treatment of most cancer types in the past decade.¹ Among various breast cancer (BC) types, immunotherapy is proven to be effective first in triple-negative BC (TNBC) either in the neoadjuvant setting for early BC (EBC)^2,3 or in the metastatic setting for programmed cell death-ligand 1 (PD-L1)-positive metastatic BC (MBC).^4,5 In estrogen receptor-positive/HER2-negative (ER+/HER2−) EBC, adding immunotherapy^6,7 to neoadjuvant chemotherapy can improve pathologic complete response. However, the efficacy of immunotherapy in ER+/HER2− MBC remains to be investigated. Predictors of immunotherapy efficacy include a high tumor mutational burden (TMB),⁸ substantial quantity of tumor-infiltrating lymphocytes (TILs), or high PD-L1 expression in the tumor.⁹ BC is characterized by a generally low TMB, less abundant TILs compared with other cancer types, and a low proportion of tumor cells with PD-L1 overexpression, and it can be considered as an immune-desert type of cancer, limiting the efficacy of immunotherapy.¹⁰ Therefore, the efficacy of immunotherapy for MBC must be enhanced, particularly for subtypes other than TNBC.

In estrogen receptor (ER)-positive (ER+) MBC, early development of immunotherapy is associated with a limited response.¹¹ One contributing factor to the reduced immunoreactivity of ER+ MBC compared with TNBC is the lower immune cell infiltration in ER+ MBC.^12,13 Additionally, the TMB in ER+ MBC is lower than that in TNBC,¹⁰ further influencing the overall response rate (ORR) to immunotherapy in patients with ER+ MBC.

In ER+ BC, estrogen facilitates tumor growth and spread¹⁴ by binding to ER with the transactivation of its downstream genes in BC cells.¹⁵ In addition to the role of estrogen on BC cells, the widely expressed ERα receptor on various immune cells also plays a crucial role.¹⁶ Estrogen exerts pleiotropic effects on polarization, cytokine production, proliferation, and effector function in immune cells.¹⁷ Estrogen can promote the polarization of M2 macrophages, thereby supporting BC tumor growth.¹⁸ Furthermore, estrogen inhibits antigen presentation by dendritic cells¹⁹ and upregulates immunosuppressive signals, including Foxp3 expression²⁰ on CD4⁺ T cells. Williams et al.²¹ reported the increased quantities of regulatory T cells, T helper (Th) cells, total macrophages, and PD-L1-positive macrophages in ESR1-mutant BC compared with wild-type ESR1 BC, suggesting an increase in TILs when ER+ BC becomes hormone resistant through an ESR1 mutation. The increase in tumor-infiltrating immune cells is indicative of tumor escape from immune surveillance; this is a fundamental process for sustaining embryonic and fetal development in the uterus, with the expression of allogeneic paternal antigens by the fetus.²² The immunological paradox of pregnancy remains largely unclear. The tolerance of a semi-allogeneic fetus by the maternal immune system implies the involvement of estrogen in fetal tolerance. Estrogen, the primary hormone produced during reproductive years, has been proposed to influence various immune cell populations in terms of their number and function, thereby playing a role in fetal tolerance.

We hypothesize that combining immunotherapy with anti-estrogen treatment can improve immunotherapy efficacy in patients with ER+ MBC. This effect may be more pronounced in premenopausal patients given the substantially higher estrogen levels in this demographic. In this study, we assessed the efficacy of incorporating a gonadotropin-releasing agonist (leuprolide) and an aromatase inhibitor (exemestane) into immunotherapy (pembrolizumab) for treating premenopausal ER+ MBC.

Results

Patients

A total of 39 patients were screened between September 2017 and December 2020 (Figure 1). For this study, 16 patients were enrolled, with 15 receiving at least one dose of treatment. The most common reason for screening failure was hepatitis B virus carrier status. Fourteen patients were evaluable, and as of July 2024, the median follow-up time was 39.6 months. Patient characteristics are provided in detail in Table 1. The median age of the patients was 47.8 years (38.2–59.3). Two-thirds of the patients presented with de novo stage IV disease. The median disease duration, defined as the duration between the diagnosis of breast cancer and clinical trial enrollment, was 20.6 months (2.6–99.3 months). All patients had received tamoxifen before enrolling into this study. Prior GnRH agonist usage was noted in 73% of the patients, whereas only 27% had received a CDK4/6 inhibitor before participating in this study. The most prevalent site of metastatic disease was the bone (80%), followed by the lungs (27%).

Figure 1.

CONSORT study flow diagram: Total number of patients screened, excluded, enrolled, and evaluable in the study

Exclusion reasons were specified at different stages. HBsAg, hepatitis B surface antigen; AST, aspartate aminotransferase; ALT, alanine aminotransferase; ULN, upper limits of normal range; TNBC, triple-negative breast cancer.

Table 1.

Patient characteristics

	All patients	Evaluable patients PR	Evaluable patients SD	Evaluable patients PD	p value
N	15	5	7	2	–
Median age (range)	47.8 (38.2–59.3)	50.2 (40.8–59.3)	47.6 (38.2–52.1)	47.6 (41.4–53.7)	0.39
Disease status	–	–	–	–	0.12
De novo stage IV no. (%)	10 (66.6)	4 (80.0)	5 (71.4)	0 (0.0)	–
Recurrent disease no. (%)	5 (33.3)	1 (20.0)	2 (28.6)	2 (100.0)	–
Previous treatment in stage IV setting	–	–	–	–	–
Tamoxifen (%)	15 (100.0)	5 (100.0)	7 (100.0)	2 (100.0)	1.00
Letrozole (%)	9 (60.0)	1 (20.0)	6 (85.7)	2 (100.0)	0.03
GnRH agonist (%)	11 (73.0)	1 (20.0)	7 (100.0)	2 (100.0)	<0.01
CDK4, 6 inhibitor (%)	4 (27.0)	0 (0.0)	1 (14.3)	2 (100.0)	0.01
Tumor sites at enrollment	–	–	–	–	–
Liver (%)	3 (20.0)	1 (20.0)	1 (14.3)	1 (50.0)	0.55
Lung (%)	4 (26.7)	0 (0.0)	2 (28.6)	2 (100.0)	0.03
Bone (%)	12 (80.0)	5 (100.0)	5 (71.4)	1 (50.0)	0.28
Lymph node (%)	2 (13.3)	0 (0.0)	0 (0.0)	2 (100.0)	<0.01
Others (%)	2 (13.3)	1 (20.0)	1 (14.3)	0 (0.0)	0.79
PAM50	–	–	–	–	0.66
Luminal A	6 (40.0)	2 (40.0)	3 (42.9)	0 (0.0)	–
Luminal B	7 (46.7)	3 (60.0)	3 (42.9)	1 (50.0)	–
HER2 enriched	1 (6.7)	0 (0.0)	0 (0.0)	1 (50.0)	–
Basal	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	–
NA	1 (6.7)	0 (0.0)	1 (14.3)	0 (0.0)	–

Treatment

At the time of data collection, all of the patients had discontinued study treatment. Dose reduction or delay of exemestane or leuprolide was not allowed in this study. Pembrolizumab was given from 150 mg every 2 weeks to 100 mg every 2 weeks after the protocol amendment in 2018. Ten and five patients received pembrolizumab at 150 and 100 mg every 2 weeks in the study. There were 8 dose delays, 3 dose discontinuations, and 5 dose reductions. All of these dose delays, discontinuations, and reductions had resulted from adverse effects (AEs). There was no obvious association between the best overall response and the dose level received by the patients (Table S1).

Primary endpoints and secondary endpoints

Of the 14 evaluable patients, the disease in nine remained under control at 8 months. The progression-free survival (PFS) rate at 8 months was 64.3%, with a median PFS of 10.1 months in the evaluable patients (Figure 2A). Five patients achieved partial responses (PRs), resulting in an ORR of 35.7%. Seven patients achieved stable disease (SD), yielding a clinical benefit rate of 85.7% (Figure 2B). Analysis of the response groups revealed that the PFS of the PR, SD, and progressive disease (PD) groups was 13.7, 8.7, and 2.4 months, respectively (Figure S1A).

Figure 2.

Clinical outcomes of treated patients

(A) PFS estimates for evaluable subjects (N = 14) from the commencement of the study, with dotted lines representing the 95% confidence intervals.

(B) Waterfall plot summarizing the maximal change in tumor size from baseline among patients with evaluable disease (N = 14). Bar colors depict different best overall responses based on RECIST 1.1. Blue bars represent partial response (PR), yellow bars indicate stable disease, and red bars represent progressive disease. Additional indicators include PD-L1 (PD-L1 score %), TIL% (tumor-infiltrating lymphocyte %), Mt per Mb (mutations per megabyte), and PD@8M (progressive disease at 8 months), denoted by white (not progressed at 8 months) and gray (progressed at 8 months) boxes. PAM50 represents PAM50 subtypes; A, luminal A; B, luminal B; H, HER2− enriched; NA, not available.

(C) OS estimates for evaluable subjects (N = 14) from the commencement of the study, with dotted lines representing the 95% confidence intervals.

(D) Swimmer plot summarizing the duration of response for evaluable subjects (N = 14). Blue circle represents partial response (PR), yellow triangle indicates stable disease, and red square represents progressive disease. Arrow represents still alive at the time of follow-up.

One patient in the PR group and four patients in the SD group were still alive at the time of follow-up (Figure 2D). The median overall survival (OS) was 39.6 months (Figure 2C). The median OS of the PR, SD, and PD groups was 39.6, 51.8, and 17.6 months, respectively (Figure S1B).

Among the 5 patients with PR, the median duration of response was 10.4 months (4.6–27.1 months, Figure 2D).

AEs

We have conducted the study according to the 3 + 3 design to monitor if unacceptable dose-limiting toxicity (DLT) had occurred. There was one grade 3 hepatitis among the first 3 patients. There was no other DLT noted in the subsequent 3 patients. Therefore, pembrolizumab was given at 150 mg every 2 weeks. However, additional two patients with grade 3 hepatitis were noted afterward. After discussing with the safety monitoring committee, pembrolizumab was given at 100 mg every 2 weeks after the protocol amendment in 2018.

Table 2 outlines the nonimmunologic AEs and immune-related AEs (irAEs) observed in the study. Among the nonimmunologic AEs, fever and urinary tract infection were the most common (21.4%; all grades), followed by flushing, hypokalemia, nausea, rash, constipation, cough, and anemia (14.3%; all grades). The majority of these nonimmunologic AEs were grade 1 or grade 2. Only two patients developed the nonimmunologic grade 3 AEs of skin infection and bacteremia, which resolved smoothly, allowing the patients to continue treatment.

Table 2.

Adverse effects

Non-immunologic adverse effects	G1 (N)	G1 (%)	G2 (N)	G2 (%)	G3 (N)	G3 (%)	G4 (N)	G4 (%)	All (N)	All grades (%)
Fever	3	21.4	0	0.0	0	0.0	0	0.0	3	21.4
UTI	0	0.0	3	21.4	0	0.0	0	0.0	3	21.4
Flushing	2	14.3	0	0.0	0	0.0	0	0.0	2	14.3
Hypokalemia	2	14.3	0	0.0	0	0.0	0	0.0	2	14.3
Nausea	2	14.3	0	0.0	0	0.0	0	0.0	2	14.3
Rash	2	14.3	0	0.0	0	0.0	0	0.0	2	14.3
Constipation	1	7.1	1	7.1	0	0.0	0	0.0	2	14.3
Cough	1	7.1	1	7.1	0	0.0	0	0.0	2	14.3
Anemia	0	0.0	2	14.3	0	0.0	0	0.0	2	14.3
Alopecia	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1
Anorexia	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1
CMV viremia	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1
Dysuria	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1
Ear local erythema	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1
Eczema	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1
Eosinophilia	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1
Hyperuricemia	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1
Malaise	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1
Neutrophil count decrease	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1
Proteinuria	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1
Pruritus	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1
Rhinorrhea	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1
Scalp pain	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1
Stomach pain	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1
Abdominal distention	0	0.0	1	7.1	0	0.0	0	0.0	1	7.1
Corneal opacity	0	0.0	1	7.1	0	0.0	0	0.0	1	7.1
Episcleritis	0	0.0	1	7.1	0	0.0	0	0.0	1	7.1
Hemorrhoid	0	0.0	1	7.1	0	0.0	0	0.0	1	7.1
Hypocalcemia	0	0.0	1	7.1	0	0.0	0	0.0	1	7.1
Insomnia	0	0.0	1	7.1	0	0.0	0	0.0	1	7.1
Leukopenia	0	0.0	1	7.1	0	0.0	0	0.0	1	7.1
MRONJ	0	0.0	1	7.1	0	0.0	0	0.0	1	7.1
Mucositis	0	0.0	1	7.1	0	0.0	0	0.0	1	7.1
Skin infection	0	0.0	0	0.0	1	7.1	0	0.0	1	7.1
Staphylococcus bacteremia	0	0.0	0	0.0	1	7.1	0	0.0	1	7.1

Immune-related adverse events	G1 (N)	G1 (%)	G2 (N)	G2 (%)	G3 (N)	G3 (%)	G4 (N)	G4 (%)	All (N)	All grades (%)
AST elevation	1	7.1	1	7.1	4	28.6	0	0.0	6	42.9
ALT elevation	0	0.0	1	7.1	3	21.4	1	7.1	5	35.7
Hypothyroidism	2	14.3	2	14.3	0	0.0	0	0.0	4	28.6
Maculo-papular rash	1	7.1	2	14.3	0	0.0	0	0.0	3	21.4
Pneumonitis	0	0.0	1	7.1	1	7.1	1	7.1	3	21.4
Diarrhea	2	14.3	0	0.0	0	0.0	0	0.0	2	14.3
Herpes zoster	0	0.0	1	7.1	1	7.1	0	0.0	2	14.3
Hyperthyroidism	0	0.0	2	14.3	0	0.0	0	0.0	2	14.3
Cushing syndrome	0	0.0	1	7.1	0	0.0	0	0.0	1	7.1
Hypophysitis	1	7.1	0	0.0	0	0.0	0	0.0	1	7.1

The most prevalent irAEs were aspartate aminotransferase (AST) elevation (42.9%), followed by alanine aminotransferase (ALT) elevation (35.7%), hypothyroidism (28.6%), maculo-papular skin rash (21.4%), and pneumonitis (21.4%). Regarding severe irAEs (grade 3 or 4), four patients had AST elevation (28.6%), four had ALT elevation (28.6%), two had pneumonitis (14.3%), and one had herpes zoster infection (7.1%). All patients recovered from severe irAEs with adequate steroid treatment, and no patient developed grade 5 irAEs.

Correlative studies

Clinical/pathologic factors

Before enrollment into this study, the patients in the PD group (Table 1) exhibited a statistically higher likelihood of receiving letrozole (p = 0.034), gonadotropin-releasing hormone agonist (p = 0.006), and CDK4/6 inhibitors (p = 0.012) than those in the PR group. Compared with the PR group, the PD group also showed a statistical prevalence of lung metastases (p = 0.03) and lymph node metastases (p = 0.001).

Traditional predictive biomarkers of the immunotherapy response, such as PD-L1, TMB, and TILs, were also analyzed in this study (Figure 2B). Two patients did not have samples for PD-L1 or TIL evaluation. The PD-L1 score was 2% in 1 patient and <1% in all other 11 patients, and no correlation was found between the PD-L1 score and treatment response. In the TIL analysis, the patients in the PR group scored between 3% and 40%, and the patients in the SD and PD groups scored between <1% and 8% (Figure 2C). In eight patients with sufficient tumor samples for TMB evaluation, the TMB ranged between 0.40 and 1.49 mutations per megabyte, and no correlation was found between the TMB level and treatment response.

NanoString analysis

We successfully collected pre-treatment tumor samples from 12 subjects and post-treatment (i.e., before cycle 3) tumor samples from five subjects for NanoString BC360 and IO360 panel analysis. The IO360 biological signatures across all samples are illustrated in Figure 3A. In pre-treatment tumors, only the signatures of Th1 cells and natural killer (NK) CD56dim cells were significantly elevated in non-responders compared to responders (Figure S2A; Table S2). After the combination treatment of leuprolide, exemestane, and pembrolizumab, the expression of CD45 was upregulated (Figures 3E and 3F), suggesting an increase in infiltrating lymphocytes following treatment. Additionally, the signatures of specific immune cells, such as NK CD56dim (Figure 3G) and Th1 cells (Figure 3H), were upregulated in post-treatment samples compared to pre-treatment samples. Gene signatures associated with immune cell function or population were also upregulated after treatment (Figure 3B). Major histocompatibility complex (MHC) class II-related (Figure 3C) and T cell-related (Figure 3D) genes were upregulated after treatment. In the IO360 panel analysis, immune cell abundance was indicated by representative genes. The most upregulated immune cells after treatment were CD56dim NK cells (Figure 3G) and Th1 cells (Figure 3H).

Figure 3.

Pre-treatment and post-treatment RNA profiling derived from NanoString BC360 and PanCancer IO360 panels

(A) Immune-related gene signatures.

(B) Immune cell-related signatures.

(C) HLA-related gene expression levels.

(D) CD8 T cell-related gene signatures from samples collected before treatment (Pre, n = 12) and after treatment before cycle 3 (C3, n = 5). Paired comparisons of samples collected after 2 cycles of treatment (C3, n = 5) and before treatment (baseline, n = 5) are summarized in (E–H).

(E–H) Genes upregulated in samples collected before cycle 3 and at baseline; (F) CD45 expression levels; (G) NK45 dim expression levels; (H) TH1 cells expression levels.

RNA sequencing

Transcriptomic profiling was performed on pre-treatment tumor samples from 12 subjects, of which 11 patients were evaluable in terms of responses. One sample was excluded from further analysis due to poor sequencing quality. Principal component analysis revealed that PR samples clustered together, while PD and SD samples also formed a cluster but exhibited high variance among them (Figure S2C).

Differentially expressed gene analysis between responders (PR, n = 5) and non-responders (PD/SD, n = 5) was performed, accounting for potential confounding factors such as age, molecular subtype, stage, and prior treatment. Using a p value of <0.05 and absolute value of log2 fold change as threshold to determine significantly differentially expressed genes, 557 genes were highly expressed in responders, while 662 genes showed higher expression in non-responders (Figure 4A). CCDC74A, SYT1, VSTM2A, and CCDC74B were the most significantly upregulated genes in responder tumor samples, whereas AREG, RND3, FAIM2, and MET were the most significantly downregulated genes (Table S4).

Figure 4.

RNA-seq analysis from samples collected before treatment based on treatment response

(A) Volcano plot of DEGs between responder and non-responder groups. With log2 (fold change) as the x axis and log10 (p.adj) as the y axis, the volcano plot was constructed according to the gene expression level. The red dots indicate upregulated genes in patients with SD/PD (N = 5), blue dots indicate upregulated genes in patients with PR (N = 5), and black dots indicate non-significant differentially expressed genes. Gene set enrichment analysis of DEGs.

(B) Estrogen-responsive genes.

(C) Immune response gene. DEGs, differentially expressed genes; PD, progressive disease; PR, partial response; SD, stable disease.

Gene set enrichment analyses were performed to identify differentially activated pathways between responders and non-responders (Table S5). The significantly enriched MSigDB hallmark gene sets are summarized in Figure S2D. Genes involved in cell proliferation and cell cycle, such as E2F targets, G2M checkpoint, c-MYC targets, and mitotic spindle, were relatively highly expressed in responder tumors. Additionally, genes related to estrogen response were upregulated in responders compared to non-responders (Figure 4B), consistent with results from BC360 data (Figure S2E). Genes involved in immune response and T/B cell functions were highly expressed in responder tumors (Figure 4C). However, immune cell composition analysis from pre-treatment samples (Figure S2B), in line with the results of NanoString IO360 analysis (Figure S2A), did not identify remarkable immune cell composition alterations. These results suggest that the quality of immune cells, rather than their quantity before treatment, plays a critical role in the response to combination therapy of leuprolide, exemestane, and pembrolizumab.

Discussion

In this study, incorporating a GnRH agonist (leuprolide) and an aromatase inhibitor (exemestane) into immunotherapy (pembrolizumab) substantially enhanced the ORR compared with historical controls among patients with premenopausal ER+/HER2− MBC.

The BOLERO-2 study²³ recruited patients with ER+/HER2− MBC who progressed after multiple lines of hormone therapy, with ORRs of 0.4% for exemestane alone and 9.5% for exemestane with or without everolimus. In the CONFIRM study²⁴ involving patients with ER+/HER2− MBC with disease progression after prior endocrine therapy, the ORR for fulvestrant monthly treatment at 500 mg was 9.1%. The JAVELIN-1b study²⁵ assessed the therapeutic effects of avelumab in unselected patients with MBC refractory to or progressing after standard-of-care therapy, yielding an ORR of 3.0%. The KEYNOTE-028 (KN-028) study,¹¹ focusing on heavily treated (median lines of prior therapies = 9) patients with PD-L1-positive ER+/HER2− MBC, reported an ORR of 12.0% for pembrolizumab alone.

Specifically, we compared the demographic characteristics, clinical characteristics, and treatment outcomes between our study and the BOLERO-2, JAVELIN-1b, and KN-028 trials, respectively (Table S6). Our study focused on premenopausal Asian patients with a statistically younger mean age than almost all the other three (p = 0.0003, p = 0.0219, and p = 0.1006, respectively). All of our patients were chemotherapy naive with advanced BC, but all patients in JAVELIN-1b and KN-028 had received chemotherapy before receiving immunotherapy for advanced BC (p < 0.0001). Only 6.7% of our patients were PD-L1 positive, while all patients in KN-028 and 62.5% of patients in JAVELIN-1b were PD-L1 positive (p < 0.0001). However, the ORR of 33.3% observed in this study was significantly higher than the ORR of 0.4%, 3%, and 12% in BOLERO-2, JAVELIN-1b, and KN-028 (p < 0.0001, p = 0.0004, and p = 0.1256, respectively). Therefore, the combination of exemestane, leuprolide, and pembrolizumab in this study demonstrated a numerically higher ORR compared with most historical controls in patients with ER+/HER2− MBC failing to respond to hormone therapy. However, all of the comparisons need to be interpreted with caution because these analyses were not randomized comparisons, and the patient groups were different.

In the MORPHEUS-HR+ BC study,²⁶ patients who failed to respond to one to two lines of hormone therapy without chemotherapy were randomized to receive either fulvestrant alone or fulvestrant, ipatasertib, plus atezolizumab, resulting in ORRs of 0% and 23% and median PFS of 1.9 and 4.4 months, respectively. Pembrolizumab combined with letrozole and palbociclib in patients with ER+/HER2− MBC who failed first-line treatment led to a complete response rate of 31%.²⁷ Rugo et al. reported ORRs of 23.1% and 28.6% in treatment-naive and pretreated patients, respectively, who received pembrolizumab/abemaciclib.²⁸ Severe AST or ALT elevation was noted (AST elevation ≥ grade 3: 17.9%–34.6%, ALT elevation ≥ grade 3: 10.7%–42.3%) in the study.²⁸ The PACE trial²⁹ revealed that compared with fulvestrant alone, the addition of avelumab to palbociclib and fulvestrant for patients who had not responded to palbociclib and aromatase inhibitor treatment resulted in ORRs of 13.0% and 7.3%, respectively. The median PFS was 8.1 and 4.8 months, respectively. The combination of palbociclib and fulvestrant with avelumab denotes the incorporation of endocrine therapy into immunotherapy for treating patients with ER+/HER2− MBC. The ORR and PFS results in our current study not only reflected substantial efficacy compared with hormone therapy or immunotherapy in the second- or third-line setting in ER+/HER2− MBC but also surpassed the results of recent early-phase studies of immunotherapy in ER+/HER2− MBC.

In advanced cancer treatment, PD-L1 has been recognized as a predictive marker of immunotherapy efficacy^30,31,32; this is particularly true in triple-negative MBC, where the PD-L1 expression level is correlated with positive outcomes in trials of various immunotherapy agents.^4,33 In the KEYNOTE-355 study,³³ adding pembrolizumab to chemotherapy provided significant OS benefits for patients with PD-L1-positive triple-negative MBC, with a combined positive score higher than 10%. However, PD-L1-positive immune cells or tumor cells are typically rare in ER+ BC, although these cells are slightly more in the ER+/basal subtype.³⁴ In a previous study of samples from the SEARCH and NEAT trial, approximately 2,400 subjects were luminal subtypes.³⁴ The PD-L1 expression levels of >1% on immune cells and tumor cells among the 2,400 subjects were as low as 3% and 0.4%, respectively.³⁴ In the present study, all 12 patients, except 1 patient, had PD-L1 scores of <1%. In the patient whose PD-L1 score was >2%, the best overall response was a PR with a PFS of 10.1 months. Notably, PD-L1 status did not predict the immunotherapy response. TMB is considered a predictive marker of immunotherapy efficacy in most cancer types.³⁵ In a study evaluating data from 3,969 patients in publicly available genomic studies, the median TMB level in ER+ MBC (1.1 mutation/Mb) was significantly lower than that in TNBC (1.8 mutations/Mb).³⁶ In our study, the TMB level was similarly low and did not correlate with immunotherapy efficacy. TILs are considered another potential indicator of the immunotherapy response.¹² The PCD4989g phase 1 study of atezolizumab in advanced solid tumors included a total of 116 patients with TNBC. Higher ORR, longer PFS, and longer OS were noted in patients with higher TILs at baseline.³⁷ The KEYNOTE 086 study of cohorts A and B conducted combined biomarker analysis, and the results revealed that higher stromal TILs were associated with higher ORR and OS³⁸ in patients with TNBC treated with pembrolizumab monotherapy. In our exploratory biomarker evaluation, the average stromal TILs were higher in patients with PR than in patients with no PR. However, the sample size in our study was not large enough for attaining statistical significance.

Because of the insufficient predictive power of PD-L1, TMB, and TIL levels for immunotherapy efficacy, RNA sequencing (RNA-seq) and NanoString analysis were performed to explore potential biomarkers. Xu et al. have reported both CCDC74A and VSTM2A in two gene elements associated with luminal A and luminal B BC subtypes. CCDC74A and VSTM2A-related gene elements were also associated with NK cell killing in this study.³⁹ Our study also identified an increase in NK cells and upregulations of CCDC74A and VSTM2A gene expression in responders. CCDC74B was reported to be one of the immune-related genes associated with BC prognosis.⁴⁰ Amphiregulin (AREG) was a downstream effector of estrogen signaling in ER+ BC⁴¹ and was secreted by suppressive immune cells such as mast cells, basal cells, and regulatory T cells.^41,42 RND3 can enhance proinflammatory activation in macrophage via its regulation in NOTCH.⁴³ Therefore, the downregulation of AREG and RND3 would be suggestive of an immunoactive microenvironment from responders in our study. These differentially expressed top upregulated or downregulated genes suggest that these genes affect immunoactive function.

BC has been characterized as a cold tumor due to its limited immune cell infiltration compared with hot tumors with a substantial amount of immune cell infiltration.⁴⁴ The cold tumor microenvironment in BC has been considered a barrier to active immunotherapy.⁴⁴ In this study, a comparison of pre-treatment and post-treatment tumor samples revealed that pembrolizumab, leuprolide, and exemestane treatment led to the upregulation of genes related to immune cell populations, T cells, and MHC class II. The abundance of immune cells was also increased in post-treatment tumor samples. These findings suggest that the combination of immunotherapy and anti-hormone therapy has the potential to activate the immune microenvironment, leading to an increased immunotherapy response.

Limitations of the study

Despite the remarkable findings, this study has a few limitations. First, the study was terminated prematurely before enrolling all subjects due to the reimbursement policy of CDK4/6 inhibitors solely for patients with postmenopausal ER+/HER2− MBC in Taiwan, resulting in increased surgical menopause procedures in patients with premenopausal ER+/HER2− MBC. Second, this pilot study did not have a control arm. The phase 1b/2 design aimed to investigate the potential signals upregulated after the combination of anti-hormone therapy and immunotherapy for patients with premenopausal ER+/HER2− MBC. The sample size was insufficient for producing a conclusive outcome, but the exceptional response of the treatment compared with historical controls in the same population was robust enough for further developing the combination of anti-hormone therapy and immunotherapy in patients with ER+/HER2− MBC. Third, pre-treatment tumor samples were sufficient for both RNA-seq and NanoString profiling, but post-treatment tumor samples were only available for NanoString profiling. The factors associated with treatment effects were identified from comparisons of pre-treatment and post-treatment samples through NanoString profiling. Although post-treatment samples were insufficient for RNA-seq analysis, genes and pathways associated with the response to the combination of immunotherapy and anti-hormone therapy were evaluated.

In conclusion, our study identified that the combination of immunotherapy and anti-hormone therapy is an effective strategy for patients with premenopausal ER+/HER2− MBC. Further confirmatory study using the combination of immunotherapy and anti-hormone therapy should be warranted. The combination of immunotherapy and anti-hormone therapy induces an immunoactive microenvironment in the tumor, and pre-existing upregulation of immune-related pathways in pre-treatment tumors was associated with response to the combination therapy.

Resource availability

Lead contact

For further information and requests regarding this manuscript, please contact the lead author, Yen-Shen Lu, at (yslu@ntu.edu.tw).

Materials availability

The authors declare that all results supporting the findings of this study are available within the paper and its supplemental materials.

Data and code availability

• •All requests for raw and analyzed data and materials will be promptly reviewed by the corresponding author and appropriate National Taiwan University Hospital committees to verify if the request is subject to any intellectual property or confidentiality obligations. Requests may be made to yslu@ntu.edu.tw; response time will be within approximately 30 business days. Release of individual-level data may be restricted due to patient confidentiality considerations. Any data and materials that can be shared will be de-identified and released via a data or material transfer agreement. The raw data of RNA-seq and IO360 are deposited on the Gene Expression Omnibus (GEO) under the accession number GEO: GSE261380 and GEO: GSE261815.
• •Scripts to conduct the RNA-seq analyses are publicly available on Zenodo website, https://doi.org/10.5281/zenodo.10816773.
• •All code used in this manuscript is available as of the date of publication on Zenodo website, https://doi.org/10.5281/zenodo.10816773.
• •Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request.

Acknowledgments

We thank all the participants in the trial and their families as well as all the study investigators. Medical writing editing was provided by Wallace Academic Editing. This study was partially supported by Merck Sharp & Dohme, a subsidiary of Merck. This work was supported by the Ministry of Health and Welfare, Taiwan (MOHW109-TDU-B-211-114002 to Y.-S.L.), the National Science and Technology Council (MOST 110-2314-B-002-218 to Y.-S.L.; MOST 109-2314-B-002-230 and 110-2628-B-002-040 to I.-C.C.), and National Taiwan University Cancer Center (NTUCCS-111-04 to I.-C.C.). We thank Hedy Chiu from Taiwan Genomic Industry Alliance Inc. for the support in RNA-seq analysis and Sarah Church from the NanoString Technologies for the support in PanCancer BC360/IO360 analysis. We also thank the National Center for High-Performance Computing for computer time and facilities.

Author contributions

The study was designed by I.-C.C. and Y.-S.L. The authors enrolled patients into the study and the study team was responsible for data collection. C.-L.H. performed bioinformatic analyses for RNA-seq and PanCancer BC360/IO360 datasets. Data analyses were conducted by I.-C.C. and Y.-S.L. The first draft of the manuscript was prepared by I.-C.C. and Y.-S.L., who vouch for the completeness and accuracy of the data and for the fidelity of the study to the protocol. All the authors contributed to the editing of the manuscript and made the decision to submit the manuscript for publication.

Declaration of interests

I.-C.C. received honoraria from AstraZeneca, Daiichi Sankyo, Gilead, Merck Sharp & Dohme, Novartis, Pfizer, and Sanofi and received research grants from AstraZeneca, Merck Sharp & Dohme, and Novartis. D.-Y.C. received honoraria from Amgen, AstraZeneca, Daiichi Sankyo, Eisai, Eli Lilly, MSD, Novartis, ONO Pharma, Pierre Faber, Pfizer, Roche, Sanofi, and TTY Biopharm. T.W.-W.C. received honoraria from Roche, Novartis, Eli Lilly, Eisai, Pfizer, Daiichi Sankyo, and AstraZeneca. Y.-S.L. received honoraria from Merck Sharp & Dohme and Pfizer and received research grants from Merck Sharp & Dohme and Pfizer.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
PD-L1 IHC 22C3	Dako/Agilent	SK-006, RRID: AB_2889976
Biological samples
Human breast tumor samples	This study	NA
Deposited data
RNAseq data	Gene Expression Omnibus	GEO: GSE261380
IO360& BC360 data	Gene Expression Omnibus	GEO: GSE261815
RNAseq analysis script	Zenodo	https://doi.org/10.5281/zenodo.10816773
Software and algorithms
GraphPad Prism 10.0.3	Graphpad Software, Inc.	https://www.graphpad.com/features
R (v. 4.2.2, with Bioconductor 3.16)	The R Foundation	https://www.r-project.org
IBM SPSS Statistics 23	IBM	https://www.ibm.com/products/spss-statistics

Experimental model and study participant details

This study included human participants from Taiwan. Data were collected in accordance with the Declaration of Helsinki. The study was approved by the institutional research ethics committee (201610003MIFA) and was registered on clinicaltrials.gov (NCT02990845).

Patients

The patients enrolled in this study are all female adults older than 20 years old. Before enrolling into this study, they had received endocrine therapy for their metastatic or locally advanced ER+/HER2- BC. None of the participants have autoimmune deficiency. The patients all received written informed consent before enrolling into the study.

Method details

Study design

This was an open-label, single-arm, single-center, phase Ib/II trial of pembrolizumab, leuprolide, and exemestane in premenopausal/perimenopausal patients with ER-positive/HER2-negative metastatic BC. The protocol received approval from the institutional research ethics committee (201610003MIFA) and was registered on clinicaltrials.gov (NCT02990845).

The primary eligibility criteria were ER positivity (≥1%) and/or PR positivity (≥1%), HER2 negativity, resistance to front-line hormone therapy (defined as the development of locally advanced or metastatic disease within 1 year of adjuvant hormonal therapy, failure of at least two lines of prior hormonal therapy for locally advanced BC or MBC, or progression within 6 months during first-line hormone therapy for locally advanced or metastatic disease), and an Eastern Cooperative Oncology Group performance-status score of 0 or 1.

Because of changes in the reimbursement policy in Taiwan, the trial was stopped prematurely. In October and December 2019, ribociclib and palbociclib were covered by the Taiwanese National Health Insurance only for postmenopausal ER+/HER2− MBC patients, respectively. After the reimbursement of CDK4/6 inhibitors for postmenopausal patients, a significant number of premenopausal ER+/HER2− MBC patients opted for oophorectomy for surgical menopause in Taiwan to access CDK4/6 inhibitors. Due to this external factor affecting patient recruitment, study enrollment significantly slowed down, leading to the early closure of the study.

Procedures

In the 28-day treatment cycle, pembrolizumab at 150 mg was administered through intravenous injection on days 1 and 15, exemestane at 25 mg was administered orally daily, and leuprolide at 3.75 mg was administered through subcutaneous injection on day 1. Treatment continued until disease progression, unacceptable toxicity, death, or discontinuation for any other reason. At screening, tumor assessment was conducted using CT or MRI every 8 weeks in the first eight cycles and then every 12 weeks after eight cycles of treatment until disease progression, withdrawal of consent, loss to follow-up, or death. Data were acquired and interpreted by the investigator. Adverse events were monitored throughout the study and were graded according to the Common Terminology Criteria for Adverse Events version 4.03.

Endpoints and sample size calculation

The primary endpoint was the investigator-evaluated progression-free survival (PFS) rate at 8 months. The tumor response was evaluated according to RECIST 1.1. Based on the efficacy from previous clinical trials for pretreated ER+/HER2- MBC, if more than 50% of the patients remain progression-free at 8 months, we will consider the treatment as a potentially effective strategy. This study arbitrarily selected 25 as the sample size, and provided the estimated PFS rate and calculated the corresponding binomial 80% confidence interval from the number of patients without disease progression on the 8th month evaluation in the protocol (Table S7). The use of hormonal therapy, leuprolide plus exemestane, was not expected to increase the toxicity of immunotherapy. In case of unexpected high toxicity rate resulted from the pre-defined dose combination of treatment drugs occur, a3+3 dose de-escalation design was applied for dose selection of the treatment of following recruited patients. As a phase Ib/II study, the power of the study was not sufficient to draw final conclusion, but it will provide pivotal evidence to determine if it worths further investigation in future phase III study for combining hormone therapy and immunotherapy in ER+/HER2- MBC. Secondary endpoints included adverse events, ORR based on RECIST 1.1 and irRECIST, clinical benefit rate (CBR), duration of overall response, median PFS, and median overall survival. As a pilot study, the power of the study was not sufficient to draw final conclusion, but it will provide preliminary evidence to determine if it worths further investigation in future phase III study for combining hormone therapy and immunotherapy in ER+/HER2- MBC.

Tumor sample collection

Fresh tumor samples were obtained at screening before treatment, at cycle 3 before treatment, and at the study’s conclusion. If acquiring a fresh tumor biopsy before treatment proved impractical, archival tumor tissue obtained before treatment was procured.

PD-L1 and stromal TILs analysis

Immunohistochemistry (IHC) was performed using a Ventana autoimmunostainer (Ventana BenchMark). PD-L1 staining was performed using clone 22C3 antibody (Dako, Santa Clara, CA, USA) in accordance with its manufacturer’s instructions. PD-L1 expression was quantified as combined positive score, defined as the number of PD-L1 staining cells (tumor cells, lymphocytes, macrophages) divided by the total number of viable tumor cells, multiplied by 100. Stromal TILs (sTILs) were evaluated from hematoxylin-eosinstained slides following the guidelines of the International TILs Working Group. sTIL abundance was measured as a percentage of Ic in the stromal tissue immediately associated with the tumor. The number of sTILs was analyzed as a continuous variable. PD-L1 and stromal TILs were both read aby breast pathologist.

RNA sequencing and data analysis

RNA quantification was conducted using NanoDrop One (Thermo Scientific) and Bioanalyzer 2100 with RNA Nano Chip (Agilent Technologies). Library preparation and sequencing were performed following Illumina’s protocol. All libraries were constructed using the Illumina TruSeq Stranded mRNA kit with the 150 bp paired-end sequencing mode on Novaseq 6000. The sequencing-by-synthesis technology was applied for sequencing, which was conducted using the NovaSeq 6000 S4 Reagent Kit v1.5 (300 cycles). Raw sequences, which were expected to generate 20M (million reads) per sample, were obtained from Illumina Pipeline software.

Raw read quality was assessed using FastQC (v0.11.9). Adapter sequences, low-quality bases, and reads were trimmed using cutadapt (v3.5). The clean reads were aligned to the human reference genome GRCh38 using STAR (v2.7.8a) with a 2-pass mapping approach. Gene-level read counts were generated by STAR based on the annotation of Gencode (v35). Protein-coding genes and long non-coding RNAs expressed in at least 50% of the samples were retained for further analysis. The trimmed mean of M-values (TMM) normalization method was applied to account for compositional differences between samples and reduce potential batch effects. Differential expression analysis was performed using limma R package (v3.60.2). The design formula incorporated variables such as age, PAM50 subtype, stage, and treatment to adjust for factors influencing gene expression. Genes with p-value <0.05 and absolute value of log2fold change (log2FC) ≥ 1 were defined as differentially expressed genes (DEGs). Pre-ranked gene set enrichment analysis (GSEA) was performed using ClusterProfilers R package (v4.12.0) on gene-sets from MSigDB (v7.4). In the pre-ranked GSEA, genes were ranked by a score calculated as sign(log2FC) ∗ -log10(p-value), where p-value and log2FC were obtained from the limma analysis. Immune cell composition analysis was estimated by calculating the geometric means of expression value of transcriptomic markers as defined by.⁴⁵

PanCancer IO 360 and breast cancer 360 gene expression panels

Tumor block slides were dispatched for analysis using the nCounter Breast Cancer 360 and PanCancer IO 360 gene expression panels. The procedures for RNA extraction and preparation are detailed in our previous publication.⁴⁶ RNA expression values were normalized, and signatures were calculated using the nSolver analysis software. The procedure for pre-ranked GSEA on the gene sets of interest followed the same methodology as described in the ‘RNA sequencing and data analysis' section.

ACTImmuneTM

The ACTImmuneTM is a multiplexed whole-exome sequencing (>18,000 genes) and human leukocyte antigen (HLA) class I typing to identify somatic mutation and neoantigen prediction. The test uses next-generation "deep" targeted sequencing to analyze whole-exome to characterize mutation burden, the number of predicted neoantigens, and to provide molecular determinants of the response to immune therapy.

WES data analysis

Raw reads generated by the sequencer were mapped to the hg19 reference genome using the Ion Torrent Suite (v. 4.2). Coverage depth was calculated using Torrent Coverage Analysis plug-in. Single nucleotide variants (SNVs) and short insertion/deletions (INDELs) were identified using the Torrent Variant Caller plug-in (version 4.2). Variant Effect Predictor (VEP) was used to annotate every variant with database from COSMIC (v.70), dbSNP (release 138) and 1000 Genomes (phase1). Variants with frequency lower than 5% were filtered out. The variants data were further filtered to remove SNPs, germline mutations, indels and synonymous mutations. Only somatic non-synonymous mutations were retained for subsequent neoantigen prediction and for transition-to-transversion ratio calculation.

HLA typing

2 μg of genomic DNA were used for HLA sequencing-based typing using the HLAssure SE SBT kit (Texas BioGene Inc). In brief, genomic DNA was PCR amplified using generic primers that amplify all of known alleles for HLA-A, -B and -C. Amplified DNA was used as the template for sequencing using Sanger sequencing method. Sample sequence is compared with known HLA allele sequences to assign the HLA type.

Neoantigen prediction

All somatic missense mutations and their neighboring nucleotides were converted into epitopes of 9 amino acids in length, each containing the mutated amino acid (i.e., potential neoepitopes). The method follows as described in the NASeek algorithm (N Engl J Med 2014:371:2189–2199). Potential neoepitopes with the same amino acid sequence as any human protein were discarded. The patient-specific human leukocyte antigens (HLAs) were typed by using Sanger sequencing. The binding affinity of all potential neoepitopes and all patient-specific HLA alleles were predicted using NetMHC 4.0 or NetMHCpan 3.0 (Protein Science 2003:12:1007–1017; Genome Medicine 2016:8:33), which employs a neural network algorithm to predict the binding affinities of peptides to HLA Class I. The neoepitopes ranked in the top first percentile of the binding affinities for a patient-specific HLA allele were considered as candidate neoantigens (N Engl J Med 2014:371:2189–2199). In addition, an alternative prediction tool recommended by International Epitope Database 3.0 (IEDB). The neoantigens predicted were utilized to calculate tumor mutation burden.

Quantification and statistical analysis

Statistical significance was determined using GraphPad Prism 10.0.3 (Graphpad Software, Inc.) and the R 4.3.1 software (R Foundation for Statistical Computing, Vienna, Austria). In statistical testing, a two-sided p value ≤0.05 was considered statistically significant, and 0.05 < p ≤ 0.10 was regarded as a marginal or borderline significance. The distributional properties of continuous variables were expressed by mean ± standard deviation, median, and interquartile range (IQR), categorical variables were presented by frequency and percentage (%), and survival curves of survival outcomes were estimated using the Kaplan-Meier method. The sample statistics presented were mean ± standard deviation for continuous variables and frequency (percentage, %) for categorical variables. The listed p-values of statistical tests were calculated using the Wilcoxon rank-sum test for continuous variables and the Fisher’s exact test for categorical variables.

Additional resources

ClinicalTrials.gov: NCT02990845.

Published: December 26, 2024