The Great Immune Escape: Understanding the Divergent Immune Response in Breast Cancer Subtypes

Breast cancer is a highly heterogeneous disease that affects one in eight women in the United States alone (1). Although great progress has been made in the early detection and treatment of breast cancer, it still remains the second largest cause of mortality due to cancer in women after lung cancer (2).

Breast cancers are typically classified according to both molecular and histologic subtypes (3, 4). Molecularly, tumors are divided into hormone receptor–positive (HR⁺), HER2⁺, and triple-negative (TNBC). Histologically, invasive lobular carcinoma (ILC) is the most common special subtype in addition to the no special subtype (NST), also frequently called invasive ductal cancer (IDC). These molecular and histologic classifications confer different prognoses and treatment options. Although breast cancer can perhaps be distinguished as one of the first malignancies to be approached with precision medicine strategies through the use of estrogen receptor (ER)– and HER2-targeted therapies, additional strategies are required to improve long-term outcomes of patients with this disease. Our current classification of breast cancers is largely naïve to the immune state of each tumor type. Nevertheless, the genomic landscape of HR⁺, HER2⁺, and triple-negative cancer cells, along with the tumor immune microenvironment (TIME), often leads to different immune infiltration and functionality. It stands to reason that breast cancer classification will eventually be refined not only to include the current cancer cell–centric molecular and histologic features but also to incorporate the characteristics of the accompanying immune infiltrate at the single-cell and spatially resolved levels in an effort to better understand the ecosystem of the tumor (5, 6).

Immune-targeting therapies have now become a mainstay of treatment for many patients with cancer; despite early success in melanoma and lung cancer, immune-targeting therapy uptake in breast cancer has been relatively slow, as breast cancer has historically been classified as an “immune cold” or “immune inert” cancer type since the 1980s (7). In breast cancer, PD-1– or PD-L1–targeting drugs such as pembrolizumab and atezolizumab have shown some initial promise; however, rates of overall response remain underwhelming when used as a single agent or in combination with traditional cytotoxic chemotherapies (8, 9). Thus, there is an urgent need to expand immune-targeted agents beyond T cells and gain a better understanding of the diversity in the TIME in various breast cancer subtypes, which may yield better drugs, better trial designs, and ultimately improved outcomes for patients (10).

In this review, we aim to compare and contrast the immune response across the different molecular and histologic subtypes of early-stage breast cancer, including HR⁺ (Luminal A and Luminal B), HER2⁺, and TNBC, as well as between HR⁺ IDC and HR⁺ ILC (given the majority of ILC diagnoses are HR⁺; Fig. 1). We will discuss cancer cell–intrinsic factors [including immunogenicity, tumor mutation burden (TMB), PD-L1 expression, and secreted factors], cancer cell–extrinsic factors (including immune infiltration landscape and functionality), and the TIME in totality (including influence of the spatial architecture and stroma). Finally, we will summarize current challenges, barriers, and next-generation opportunities to harness the immune system for the treatment of breast cancer.

Owing to the advancement of bulk and single-cell sequencing technologies, a wealth of genomic and transcriptomic data are now available that underscore the differences between breast cancer subtypes. These underlying differences in the genomic landscape, including TMB, HLA expression, and display of neoantigens, may contribute to divergence in immune infiltration and responses. Further, differences in metabolic reprogramming between the three main subtypes may also affect tumor-infiltrating lymphocyte (TIL) composition and response to immune-targeting agents. Although breast cancers have been historically considered to have a low TMB and immunogenicity compared with other tumor types, emerging evidence suggests resolving the cellular composition and heterogeneity between clonal regions in breast cancer may allow for a more precise understanding. In this section, we aim to shed light on how cancer cell–intrinsic properties may lead to differences in immune interactions.

Efforts to further refine molecular classifications based on intrinsic subtypes have dominated over the past decade. In TNBC, four major subtypes have been identified using RNA sequencing (RNA-seq), which suggests there may be differences in immune response even between subclassifications within TNBC: (i) basal-like immune suppressed (chemotherapy most effective); (ii) immunomodulatory, defined by high TIL presence (immune-checkpoint therapy may be most effective according to the authors of the study); (iii) mesenchymal subtype, defined by high JAK/STAT activation (suggesting STAT3 inhibitors may be most effective); and (iv) luminal androgen receptor (AR), defined by AR expression and ERBB2 activating mutations (suggesting hormone therapy or ERBB2-targeting tyrosine kinase inhibitors may be effective; ref. 11). Similarly, not all HER2⁺ disease has the same genomic profile, and only 50% of clinically HER2⁺ tumors actually fall into the HER2-enriched (HER2-E) intrinsic subtype category (12, 13). HER2-E, which is typically HR-negative, is characterized by the highest levels of ERBB2 mRNA and therefore the highest activation of the HER2 signaling pathway (including FGFR4 and EGFR activation). This leads to higher proliferation rates compared with HER2⁺ tumors that may fall into the Luminal B category and ultimately to different levels of immune interaction (more immune infiltration in HER2-E tumors compared with those that are Luminal B; ref. 14).

Further, it is now appreciated that the traditional receptor-based classification of breast cancer has limitations, and recent data show that classification based on gene modules can better connect tumors with immune status and is associated with prognosis. For example, a recent single-cell sequencing analysis of intrinsic subtype classification showed significant heterogeneity among cancer epithelial cells traditionally belonging to the Luminal, HER2-E, and TNBC subtypes (15). Based on the gene modules, Luminal A cells were enriched in pathways related to estrogen response but also enriched with respect to TNF signaling through NF-κB, hypoxia, and apoptosis, suggesting a role for secretory products interacting with the immune system in the HR⁺ subtype (Fig. 2). Basal-like TNBCs exhibited a strong intrinsic phenotype toward IFN response (15). Interestingly, interrogation of HER2-E tumors showed more enrichment of proliferation pathways rather than specifically immune-related or cytokine secretory products (15). Significantly, very few tumors are homogenous entities—cellular heterogeneity shows an amalgamation of different intrinsic subtypes within the TIME, which may confer different interactions with nontumor cells depending on clone location and may create many local environments within a single TIME (15, 16). In an analysis using The Cancer Genome Atlas (TCGA) data, TNBC and HER2⁺ tumors that had increased immune metagene expression had lower clonal heterogeneity and neoantigen loads, suggesting a strong immunoediting effect in these tumor types that were not seen in ER⁺ tumors (17). This may underlie why stronger immune responses are observed in TNBC and HER2⁺ tumors. Taken together, these data suggest that efforts to better classify tumors beyond receptor status have shed light on subtypes that may be more immune stimulatory or responsive. In the future, clinical trials and studies should provide subanalyses to determine if agents targeted for each subtype validate these preclinical data.

Although intrinsic differences exist between breast cancer subtypes and can contribute to differences in immune interactions, more traditional markers of TMB or HLA expression may also contribute to differing TIMEs. A large analysis of 3,969 patients with breast cancer identified subtype differences in TMB, with median TMB in TNBC higher than in HER2⁺ and HR⁺, with HR⁺ being the lowest (18). Across all included samples, the median TMB was 2.63 mutations per megabase (mut/Mb). When grouping TMB categorically, Luminal A and Luminal B tumors have significantly lower proportions of tumors classified as high TMB (which was defined by the authors as greater than 1.63 mut/Mb, the mean TMB rate of the BRCA cohort in TCGA) than HER2⁺ and TNBC tumors (19). Although cutoffs may be helpful for the purposes of research studies, in reality, TMB should be considered as a continuous variable with immune responses varying accordingly across the TMB spectrum as well. Across these studies, it becomes apparent that compared with other cancers, breast cancers as a whole harbor a much lower level of hypermutation, with only around 5% of all tumors classified as having TMB >10 mut/Mb (18). Breast cancers that do have a high TMB were more often enriched with favorable immune infiltrates, including the presence of CD8⁺ T cells, CD4⁺ T cells, activated natural killer (NK) cells, and γδ T cells (18). Consistent with an antigen-stimulated immune response, patients with a high TMB and high immune infiltrate survived longer than those with weak or poor immune infiltrates. Although important, TMB may not tell the whole story: For example, one study demonstrated that 36% of basal-like tumors were classified as having high TMB, yet only 24% of these high-TMB tumors had strong immune infiltrates. This suggests that only a relatively small fraction of these tumors may be particularly immunogenic (19).

Cancer cell HLA class I expression may also drive differences in antigen presentation and subsequent T-cell interactions. In an IHC analysis of tumors from patients in the GeparTrio trial, HR⁺/HER2⁻ tumors showed significantly lower HLA class I expression compared with other subtypes. Although lower in the HR⁺ subtype, HLA downregulation is likely involved in immune evasion in other subtypes as well (20). Evidence from the AURORA US Network investigating HLA-A dysregulation in TNBC found genetic (deletion) and epigenetic (hypermethylation) alterations in HLA-A leading to a general decrease in HLA-A gene expression and an associated lower number of MHC-I–associated neoantigens, independent of TMB (21). This was also associated with lower immune cell infiltrates using DNA methylation–based assessments of leukocyte infiltration. In TCGA, primary tumors that had HLA-A methylation had associated decreased expression of multiple adaptive immunity signatures, possibly indicating evidence of immune escape (21). Speculatively, the authors state that immune-checkpoint inhibitors (ICI) may have little effect in these HLA-A–low, TNBC tumors, which may explain modest response rates in clinical trials for TNBC. Although it is difficult to say that ER⁺ tumors are MHC-I “low,” it is likely that TNBCs are just comparatively higher because of more IFN signaling, which is negatively associated with ER signaling and ESR1 expression (22).

An additional layer of cancer cell–intrinsic immune differences may be due to underlying expression of stimulatory or inhibitory ligands, the most characterized of which is PD-L1 (encoded by the CD274 gene). Although PD-L1 may be expressed on many different cells in the TIME (23), we focus here on PD-L1 expression on tumor cells given our focus on cancer cell–intrinsic properties in this section. Appreciable differences are observed between the three main molecular breast cancer subtypes (24): TNBCs seem to harbor the highest levels of both PD-L1⁺ tumor and immune cells, with concomitant decreases in PD-L1⁺ tumor cells seen in HER2⁺ and ER⁺ disease when analyzed by IHC and gene expression (25, 26). Luminal A tumors, characterized by high ER positivity and low Ki-67 indices, have the lowest proportion of PD-L1⁺ tumor and immune cells in an analysis of multiplex IHC (mIHC; refs. 27, 28). In multivariate analysis, PD-L1 positivity was associated with positive lymph node metastases, higher histologic grades, ER negativity, and TNBC (26). As noted above, it could be that PD-L1 expression is highly influenced by strong IFN signaling in TNBCs (a phenomenon that is observed less so in HER2⁺ and ER⁺ tumors). No relationship was observed between PD-L1 positivity and HER2 status. At the gene expression level, recent studies have shown that TNBCs harbor higher levels of CD274 compared with non-TNBCs (29). The prognostic significance of PD-L1 positivity on breast tumor cells remains unresolved.

An additional layer of complexity is that subpopulations of breast cancer cells, such as cancer stem cells, may harbor up to 3-fold higher levels of PD-L1 than their more differentiated counterparts when measured by quantitative immunofluorescence (30). This additional layer of immune evasion may help tumors sustain the tumorigenic process. As TNBCs seem to harbor higher amounts of cancer stem cells, this subgroup of cells may help to drive immune evasion, and combinatorial targeting of newly identified pathways, such as Notch3 through mTOR or the β-catenin pathway, in addition to anti–PD-L1 agents may better target this group (31, 32).

As breast epithelial cells progress from noncancerous to cancerous, changes in cellular metabolism occur to meet new demands for energy and nutrients. As such, intrinsic factors, such as genetic and epigenetic alterations, may contribute to the unique metabolic programs of the breast cancer subtypes (33). Recent efforts have focused on identifying metabolic pathway–driven heterogeneity within the subtypes. Metabolic dysregulation has been most pronounced in TNBC, which has shown to be most distinct metabolically from non-TNBC and normal samples in a pathway enrichment analysis using gene expression data from tumors included in TCGA (34). Further scrutiny of TNBCs, again using pathway-based analysis, revealed three distinct metabolic pathway–based subtypes (MPS), including one with upregulated lipid metabolism, one with upregulated carbohydrate and nucleotide metabolism, and one with mixed pathway activation and dysregulation (34). Interestingly, the MPS group of TNBCs with upregulated glycolytic and nucleotide metabolism had the worst prognosis out of the three identified groups and high rates of homologous recombination deficiency (34). Recent data suggest that highly glycolytic breast tumors, such as TNBCs, create a local lactic acidosis in the TIME, which is a known driver of immune evasion (35). It is then not surprising that these tumors also had higher grades and were most responsive to anti–PD-1 therapy in vitro in patient-derived organoids, suggesting that therapy targeting lactate dehydrogenase (which would decrease the local lactic acidosis) in combination with anti–PD-1 agents may be an optimal combination for further testing (36). Additionally, ectonucleotidases, specifically CD73 and CD39, have received increased attention. These ectonucleotidases rapidly convert extracellular ATP into immunosuppressive adenosine (37, 38). Given that TNBCs seem to harbor the most immunosuppressive TIME, inhibiting ectonucleotidases to enhance the immunotherapy effect (reprogramming the TIME to be less immunosuppressive) is actively being tested in vitro and in clinical trials (39). On the other hand, metabolomic efforts to identify distinct dysregulation in ER⁺ disease have been more elusive. However, amino acid pathway metabolism may show differences across the subtypes, with the lowest levels of amino acid metabolism in the Luminal A subtype compared with all others (40, 41). Estrogen and ER signaling itself may also alter cellular metabolism, and the potential between-pathway cross-talk is an active area of investigation. Further research is needed to determine how intrinsic metabolic reprogramming can lead to promising new combination therapies with currently used immune-targeting agents.

Taken together, the cancer cell–intrinsic factors in HR⁺ tumors, HER2⁺ tumors, and TNBCs are one component in the varying interactions between tumor and nontumor cells in the TIME (Fig. 2A). TNBCs harbor higher TMB rates, HLA downregulation (either through genetic or epigenetic mechanisms) likely indicating immune escape, higher PD-L1 expression, and higher energy demands. Evidence already shows a role for immunoediting in the TIME for these tumors, and finding the right populations of patients for immune-targeted therapy is important given the heterogeneity of the disease. The picture is less clear for HER2⁺ and ER⁺ tumors. Although ER⁺ tumors are thought to be the “coldest” out of the three, there is clearly a subset of these tumors that do have strong immune interactions even beyond expression of PD-L1, and though not harboring as high of a value of TMB, PD-L1 expression, or metabolic demand as TNBCs, HER2⁺ tumors lie somewhere between TNBCs and ER⁺ tumors. Although T cell–targeted therapy may not be the most effective immune-targeted therapy for HER2⁺ or ER⁺ tumors, assessment of immune cell infiltration and functionality may yield additional insights (42).

Breast cancer was historically thought to be immunologically silent, but a growing body of literature over the last decade demonstrates a role for immune cells in the development, progression, and therapy responsiveness of the disease (43–46). Through multiple retrospective analyses using a wide array of methods (including but not limited to IHC, flow cytometry, single-cell RNA-seq, and imputation from bulk RNA-seq), it is now well established that all subtypes of breast cancer show varying degrees of the presence of both innate and adaptive immune cells. These cells have been shown to be involved in patient responses to not just immunotherapy but also chemotherapy in both the adjuvant and neoadjuvant settings across all subtypes of breast cancer (47). However, the functional impact of these cells, the quality of this immune response, and prognostic value remain matters of great debate and curiosity. Additionally, during the inflammatory reaction to the tumor, nearly all immune cell types can be recruited into the TIME, and across immune cell types, it generally holds true that infiltration and the number of cells follow a similar pattern: TNBCs >HER2⁺ tumors >ER⁺ Luminal B >ER⁺ Luminal A. Specifically, ER⁺ Luminal A tumors generally have poor infiltration but much better prognoses due to endocrine therapies. However, any tumor features that oscillate a tumor toward a more aggressive “basal-like” state will typically elicit a stronger immune reaction.

Further contributions to the TIME include the surrounding stromal cells, including cancer-associated fibroblasts (CAF), which can be active cytokine and chemokine secreters that can influence TIL infiltration and the balance of the local inflammatory reaction (48, 49). CAFs are typically believed to orchestrate tumor-promoting inflammation and modulate the TIME toward immunosuppression, functions that are mediated through intricate reciprocal signaling interactions with cancer cells, matrix components, and infiltrating immune cells. Tumors hijack the inherent repair mechanisms in fibroblasts, as their natural state is to sense tissue damage and promote repair through inflammatory responses. CAFs may also recruit myeloid and regulatory T cells (Treg) and promote M2 macrophage polarization and infiltration (48).

In the following subsections, we discuss a key component of the TIME: TILs. Although TILs is a broad term, and their importance and association with survival have led to contradictory results in various studies, recent efforts to reproducibly standardize TIL terminology and evaluation have been spearheaded by an international group of oncologists and pathologists (50). The studies cited in the section below regarding the prognostic or predictive value of TILs were largely performed using these criteria, which include technical considerations, how to perform calculations, what cell types are to be included, and inclusion/exclusion considerations. In some cases, TILs were assessed in the tumors themselves; in most cases, we report on stromal TILs, which is a distinct metric from those in the tumor. It should also be cautioned that, to date, no phase III studies have identified TIL scoring to be associated with benefits for patients taking immunotherapies. Although TILs themselves are likely quite important in the TIME, quantitative TIL scoring may have little immunobiological association, and further insight into TIL functional status and cellular composition may be more biologically meaningful.

Of all the breast cancer subtypes, TNBC is the most extensively studied tumor type with respect to the role of the immune system. This is partially due to the observation that the majority of TNBCs are highly immunogenic and harbor significantly higher stromal TILs, which are consistently shown to be associated with good prognosis (51–54). Recent analyses indicate that an association between stromal TILs and response to neoadjuvant chemotherapy is strongest in patients with TNBC (55). In this analysis, stromal TILs were quantified using hematoxylin and eosin (H&E) slides both as a continuous measurement and by predefined cutoffs (low 0%–10%, intermediate 11%–59%, and high 60%–100%; ref. 55). Patients with TNBC had higher stromal TIL infiltration overall, and 10% increases in TILs were associated with both disease-free survival (DFS) and overall survival (OS; ref. 55).

At the cellular level, TNBCs have significantly higher stromal and intratumoral CD8⁺ and CD4⁺ T-cell infiltration compared with HR⁺ tumors. CD8⁺ T cells and tissue-resident memory CD8⁺ T cells (TRM) are independently correlated with better prognosis (53, 56–59). Although CD4⁺ T-cell quantities were not prognostic, they have been shown to correlate with CD68⁺ macrophages as well as B cells within the TIME. The TNBC TIME also shows a significant presence of tumor-infiltrating B cells (TIL-B) by IHC and B cell–associated metagene signatures from single-cell RNA-seq and from bulk RNA-seq (60), which are positively associated with both improved OS and DFS (referring to general prognosis, as this was analyzed retrospectively for patients included in the BIG 02–98 trial; refs. 61, 62). B cells together with CD4⁺ follicular helper T cells have in fact been shown across multiple TNBC genetically engineered mouse models (GEMM) to mediate response to immune-checkpoint blockade therapy (63). The presence of peritumoral tertiary lymphoid structures (TLS), which are ectopic lymph node–like aggregates of CD4⁺ T cells and B cells that can be identified by multiplexed immunofluorescence, with clonal expansion of B cells and associated chemokine CXCL13 expression, demonstrates the potential for ongoing humoral immune responses at the tumor site, stroma (including those mediated by CAFs), and sentinel lymph nodes in human disease and mouse models of TNBC (61, 64–67). This might be particularly interesting in the context of potentiating the antitumor immune response beyond the current T cell–targeting approaches. Recent reports that utilize single-cell and spatial mapping technologies, such as cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq), largely corroborate findings from deconvolution of bulk sequencing, IHC, and immunofluorescence. By single-cell sequencing and spatial mapping, enrichment of TILs and CD8⁺ T cells in TNBC was observed, and these T cells exhibited substantially higher dysfunction scores in patients with TNBC (15).

As a countermeasure for this immune activation, patients with TNBC have been shown to harbor an increased number of Tregs, myeloid-derived suppressor cells (MDSC), and tumor-associated macrophages (TAM) in the TIME as well as in systemic circulation (68–73). Neutrophils, typically assessed by a trained breast pathologist from H&E-stained tumor sections, which can play both pro- and antitumor roles, have been observed to be enriched in TNBC tumors compared with HR⁺ and HER2⁺ tumors and are found to be associated with disease progression (74). Mechanistically, this effect is attributed to the support of metastatic niche formation by neutrophils and other myeloid cells and their associated cytokines and chemokines like CCL9 and Prokineticin-2, as observed in a murine model (4T1) and in vitro cell line models (e.g., MDA-MB-231) of TNBC (75–77). Paradoxically, despite being immunosuppressive and protumorigenic, Tregs when considered as a single variable are associated with better outcomes in TNBC. This surprising correlation may be explained by the finding that elevated Tregs are also associated with a high CD8⁺ T-cell:Treg ratio, suggesting that the cytotoxic, antitumor CD8⁺ T-cell effect is still predominant over the regulatory effect of the Tregs (69, 78, 79).

Complementing and contributing to the immune milieu and function is the local CAF population. The cross-talk between immune cells and CAFs has also been most studied in the context of TNBC (80). In addition to their roles in driving protumorigenic inflammation, CAFs may affect MHC-II presentation and immune regulation along with extracellular matrix remodeling (48). Four primary CAF subsets have recently been identified in each of the subtypes of breast cancer using multiple methods, including multicolor flow cytometry, IHC verification, and RNA-seq (81). These four CAF populations were distinguishable using six markers: CD29, FAP, αSMA, PDGFRβ, FSP1, and CAV1. Interestingly, in TNBC, both the CAF-S1 (characterized by high expression of all markers except for CAV1) and CAF-S4 (characterized by the highest levels of CD29 expression and αSMA) subsets were identified. If the TNBC TIME was enriched for CAF-S1 fibroblasts, an inflammatory myofibroblast subset, the balance between Tregs and CD8⁺ T cells was tipped in favor of increased Tregs, leading to a highly immunosuppressive locale. Recruitment of Tregs and differentiation of CD4⁺ T cells to Tregs was in part mediated by the chemokine CXCL12. On the other hand, some TNBC TIMEs were enriched for the CAF-S4 subset, which was associated with more CD8⁺ T cells.

Overall, in TNBC, there is a high degree of infiltration of T cells (including CD8⁺, CD4⁺, and Tregs) compared with other subtypes. Despite this, only a small proportion of TNBCs respond favorably to current T cell–targeted immunotherapies. A sizable proportion of TNBCs seems to harbor an immunosuppressive TIME characterized by dysfunctional CD8⁺ T cells, Tregs, and CAF-S1 fibroblasts, suggesting potential synergistic opportunities to boost anti–PD-1/PD-L1 agents (82).

The HER2⁺ TIME is more immunogenic than HR⁺ TIME, and, similarly, ER⁻/HER2⁺ tumors show greater infiltration of TILs compared with ER⁺/HER2⁺ when evaluated immunohistochemically (51, 83, 84). There is a growing body of evidence supporting the idea of intrinsic immunogenicity of HER2⁺ tumors and a role for adaptive immune responses mediated by T and B lymphocytes in these tumors (24). However, compared with TNBCs, HER2⁺ tumors seem to be less likely to exhibit high TIL levels (24).

In the HER2⁺ TIME, the HER2 protein itself serves as a unique target that has been shown to be recognized by T cells and B cells and is amenable to immunotherapeutic intervention (85). Increased CD8⁺ T cells, including CD8⁺ TRMs, and higher immune cell expression of IFNγ have been correlated with better prognosis, improved pathologic complete responses (pCR) with therapy, and significantly increased OS in HER2⁺ disease after analysis of stromal TILs in the GeparQuattro and GeparQuinto trials (51, 86–91). Similarly, Tbet⁺ (a hallmark transcription factor that controls the expression of IFNγ) type 1 CD4⁺ T cells and HER2-specific CD4⁺ T cells in the peripheral circulation of patients with HER2⁺ disease were also indicative of better outcomes (92). An abundance of TIL-Bs in the HER2⁺ TIME was demonstrated by IHC and using in silico B cell–gene signature analysis, both associating with better DFS (60, 62, 93). From the gene signature analysis, more B-cell receptor chain variations were found in HER2⁺ tumors, indicative of somatic hypermutation and suggesting a role for a productive antibody response in antitumor immunity (62). Although not significantly associated with outcome, an increased number of TLSs have also been observed in the HER2⁺ TIME via IHC (83). HER2⁺ tumors respond well to neoadjuvant chemotherapy (NAC), and there is evidence for cooperative antitumor immunity in patients treated with NAC. In that setting, patients with HER2⁺ breast cancer who achieved pCR harbored increased NK cells, B cells, and Th17 CD4⁺ T-cell frequency, as assessed by multispectral flow cytometry, as well as polyfunctional CD8⁺ T cells capable of producing IFNγ upon stimulation in vitro (88, 94, 95). Additionally, iNOS⁺ CD68⁺ M1-like macrophages and increased CD8⁺ effectors as a combined factor were significantly associated with improved survival, suggesting the importance of a multiparametric approach to understanding the impact of various immune cells (96).

HER2-targeted therapies also rely on interactions with the immune system to enhance efficacy. Interestingly, through an immune gene signature analysis on a set of HER2⁺ samples treated with the HER2-targeted monoclonal antibody trastuzumab, it was demonstrated that therapeutic benefit was correlated with immune gene enrichment, suggesting that HER2-directed antibody benefit is at least in part immune cell–mediated (87, 97–99). One such mechanism of anti-HER2 therapy is now recognized to be antibody-dependent cellular cytotoxicity (ADCC). In tumors with HER2 overexpression, the trastuzumab or pertuzumab therapeutic antibodies bind to HER2 receptors expressed on the surface of cancer cells. This binding facilitates NK-cell recruitment via FcγR ligation, bringing NK cells into proximity to cancer cells, thereby allowing for a prolonged cytotoxic interaction (98). Beyond ADCC, a novel approach that combines the tumoricidal effects of chemotherapy with the benefit of targeted therapy has resulted in the development of HER2-directed antibody–drug conjugates (ADC) like T-DM1 (ado-trastuzumab emtansine), which contains trastuzumab conjugated with emtansine, a microtubule inhibitor. After binding the HER2 receptor, T-DM1 is internalized and degraded inside the cell, which in turn releases DM1. Preclinical studies have shown that T-DM1 enhances T-cell infiltration and turns the TIME into a T cell–inflamed phenotype (100). Recent investigations have focused on the role of trastuzumab deruxtecan (T-DXd), which has recently been shown to be superior to T-DM1 in the metastatic setting (although associated with increased risk of interstitial lung disease and pneumonitis; ref. 101).

Although CAF characterization is less well described in HER2⁺ tumors, CAF-S4 fibroblasts seem to be the most predominant type of CAF in the HER2⁺ TIME (81). Unlike their CAF-S1 counterparts, CAF-S4 fibroblasts were less able to attract FOXP3⁺ T cells and enhance their functioning to inhibit the proliferation of effector T cells. Although deeper subtyping has been studied for the CAF-S1 cells in the context of immunotherapy resistance, further investigation into the role of CAF-S4s is certainly warranted, as recent evidence suggests that stromal cells may be a better predictor of HER2-directed therapy efficacy than looking at just the tumor epithelial component (102).

HR⁺ breast cancer encompasses two distinct molecular subtypes (Luminal A and Luminal B). It is hypothesized that TIL levels and function are quite different in HR⁺ tumors. For example, in a recent analysis of stromal TILs, increased TILs in HR⁺ tumors were not linked to DFS and were actually associated with decreased OS (55). These data suggest a different immunobiology for HR⁺ tumors with a different cellular composition (103).

Although limited, the existing data suggest that the immune infiltrate in the primary HR⁺ TIME is dominated by CD8⁺ and CD4⁺ T cells and TAMs (104–108).

Paradoxically, despite having antitumor effector functions, higher TILs have been shown to be associated with poor prognosis and worse breast cancer–free survival in the ER⁺ setting (55, 105, 109, 110). In a single-cell RNA-seq analysis of the tumor and immune ecosystem in breast cancers, more than half of ER⁻ tumors but only 12% of ER⁺ tumors had more than 10% of T cells that express PD-1. When comparing Luminal A and Luminal B tumors, a higher proportion of Luminal B tumors had more than 10% of T cells expressing PD-1 (16). Beyond T cells, in HR⁺ tumors, the immune cell type most associated with poor clinical outcomes was the increased presence of TAMs. Interestingly, when comparing the most immune-rich HR⁺ tumors against immune-rich TNBCs, M2-like macrophages were particularly increased in HR⁺ tumors with a corresponding enrichment of TGFβ-related genes (111). However, recent evidence indicates that heterogeneity exists with respect to different macrophage populations within luminal breast cancer, and a recent study indicates that an enrichment of FOLR2⁺ macrophages is associated with higher levels of CD8⁺ T-cell infiltration and better OS (112). Recent developments indicate ER⁻ tumors are enriched for PD-L1⁺ TAMs compared with ER⁺ tumors. Using the TCGA database, a different study found that M2-like, protumor macrophage polarization markers were significantly upregulated in both Luminal A and Luminal B tumors compared with basal-like tumors (113). If validated further, it might offer insight into the cancer cell–intrinsic suppressive ability of luminal tumors and the observed lack of benefit from T-cell infiltration as a result of the presence of more immunosuppressive M2-like macrophages.

As in TNBC and HER2⁺ tumors, CAFs are believed to play an important role in tumor immune–stromal interactions in ER⁺ tumors as well. Several studies have shown different proportions and characteristics of the CAFs in ER⁺ tumors (48, 81, 114). In ER⁺ tumors, the predominant CAF is the CAF-S2, which is a less activated and less inflammation-promoting myofibroblastic CAF compared with the CAF-S2 and CAF-S4 subsets (81). Spatial pathologic evidence also suggests that Luminal A tumors harbor the highest content of fibroblasts compared with other breast cancer subtypes, and using a novel spatial model for better characterizing the interactions between TILs and fibroblasts, Nederlof and colleagues identified heterogeneity across an array of ER⁺/HER2⁻ tumors (115). Interestingly, poorer survival was observed for patients with high CAFs but low TILs, and a better survival rate was noted for patients with TIMEs characterized by spatial mixing of CAFs and TILs (115).

In addition to ER, other hormone receptors, such as the progesterone receptor (PR) and the AR, may have immunomodulatory properties in breast cancer (116). Several studies have suggested that the progesterone (P4)–PR signaling axis has immunosuppressive and immune-evasion properties that are mediated through the dampening of IFN–STAT1 sig­naling (117). Like estradiol (E2), P4 also seems to suppress the immune system, which is beneficial during pregnancy and fetal development. However, this effect can be pathologic and can contribute to increased tumorigenesis (118). This suggests that high levels of P4 in the TIME may lead to fewer TILs and decreased immune gene expression signatures, which could thus be a potential target to boost immunotherapy efficacy. In addition to PR, a significant number of breast tumors express AR independent of ER, PR, and HER2. However, the role of AR in ER⁺ breast cancer is currently controversial, and these findings have prompted clinical trials for both AR agonists and AR antagonists. Recent findings shed light on this question, suggesting AR is a tumor suppressor in ER⁺ breast cancer, including endocrine-resistant tumors, thereby supporting a role for AR agonism (119). Additional work is warranted to investigate AR interactions with the immune system in this setting, but new work in the setting of castration-resistant prostrate cancer showed an association between AR downregulation in intratumoral T cells and PD-L1 responsiveness (120, 121). In patients who were resistant to anti–PD-L1 therapy, evidence supported a potential role for AR-mediated suppression of antitumor genes such as TNF, granzyme B, and IFNγ. These findings may necessitate a look at the role of ER expressed in TILs.

An additional layer of complexity in breast cancers is the presence of hormones, such as E2, and hormone receptor signaling. Studies have also shown that immune cells, especially macrophages and MDSCs, express ERα and are affected by estradiol in the HR⁺ TIME (122–124). E2 is historically known to be an anti-inflammatory molecule, which may in part explain poor immune infiltration, poor functioning, and protumor immune cells in the TIME (125). Increased estradiol in the breast TIME has been shown to enhance the polarization of monocytes into M2-like macrophages, differentiation of CD4⁺ T cells into Tregs, and increased infiltration and suppression by Tregs and MDSCs, making the TIME highly immunosuppressive for effective antitumor immunity (126–129). This also suggests that hormone receptor signaling and the presence of hormones in the TIME not only shape tumor cell phenotypes but also have strong effects on immune infiltrates. Given that ER⁺ breast cancer is an age-related disease and the peak incidence of ER⁺ tumors occurs around the age of 70, further research is also warranted to characterize how chronic age-related inflammation may impact TIL proportion and functioning (130).

Taken together, the TIME varies markedly between the three major molecular breast cancer subtypes (Fig. 2B; ref. 131). In TNBC and HER2⁺ tumors, evidence supports the enrichment of TILs, which is prognostic with respect to DFS and OS (45). In HR⁺ tumors, evidence is controversial for the prognostic role of TILs, but it is clear that these tumors harbor lower levels overall of TILs. Additionally, key features of TNBC and HER2⁺ tumors compared with HR⁺ tumors include increased TLSs, increased immunoediting, increased numbers of PD-L1⁺ TAMs, and a higher proportion of CAF-S1 (immunosuppressive; predominantly in some TNBCs) and CAF-S4 (primarily immunostimulatory; observed in both TNBC and HER2⁺ tumors) cells. HR⁺ tumors accumulate higher amounts of M2-like macrophages, which seem to drive a more immunosuppressed TIME, and the spatial mixing of CAFs and TILs seems to be particularly important for these tumors. Despite similarities between TNBC and HER2⁺ tumors compared with HR⁺ tumors, NK-cell engagement and enhancement may be particularly promising for HER2⁺ tumors given the use of trastuzumab and may be a target to further enhance ADCC action. Additional studies of the immunosuppressive nature of macrophages, the role of E2 in the TIME, and the interplay between age-related chronic inflammation and TIL functioning may also facilitate additional therapeutic opportunities. These data should be used to guide new subtype-specific, immune-targeting therapies.

The two main histologic subtypes of breast cancer, IDC and ILC, are also associated with differences in the immune state within their respective TIMEs. Overall, ILCs represent 10% to 15% of all breast cancer diagnoses; the distinguishing hallmark of these tumors is the loss of CDH1, which leads to cancer cell discohesion and the characteristic single-file infiltration of cancer cells within the stroma (132, 133). The vast majority of these tumors are HR⁺ and fall into the Luminal A PAM50 subtype. Distinct from patients with IDC, patients with ILC are more often diagnosed at older ages, typically have larger, multifocal tumors, and may have a unique propensity for late tumor recurrences (Fig. 1). It is hypothesized that the spatial architecture of these tumors and influence of the stroma (surrounding nontumor cells and extracellular matrix) also lead to differences in immune infiltration, functionality, and response. Given that the vast majority of ILCs are considered HR⁺, we will discuss differences between HR⁺ ILC and HR⁺ IDC in this section.

One of the most comprehensive studies to date analyzing immune cell infiltration in HR⁺ breast cancer, accounting for over 750 samples each for ER⁺ ILC and IDC that assessed TILs using both IHC and deconvolution by CIBERSORT for tumors included in TCGA and the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC), showed that ER⁺ ILCs have statistically significantly lower total immune cell infiltrates compared with ER⁺ IDCs (105). Importantly, this study also analyzed the relative distribution of CD3⁺ T cells, CD8⁺ T cells, CD68⁺ macrophages, CD20⁺ B cells, and FOXP3⁺ Tregs within intraepithelial, adjacent stromal, and distant stromal compartments and found that ILC had lower levels in all measures of CD3⁺, CD8⁺, and CD68⁺ cells compared with IDC. Thus, as it stands now, ILCs are believed to harbor fewer immune cells overall than IDCs. Along this line, a recent study from our group suggests that although 86% of IDCs harbor higher levels of total immune cells in the TIME compared with normal breast tissue from mammoplasty, only 52% of ILCs harbor this increase (134).

Despite the low overall infiltration, there are perhaps subsets of ILC that do elicit a strong immune response; it is not known what proportion of ILCs falls into this category. An “immune-related” (IR) subtype within ER⁺ ILC was discovered by in silico genomic and proteomic approaches on three large cohorts of samples: TCGA, METABRIC, and the Rational Therapy of Breast Cancer (RATHER) consortium (135, 136). This IR subtype showed higher reads in markers for CD4⁺ and CD8⁺ T cells along with increased expression of T-cell inhibitory markers including PD-1, PD-L1, and CTLA-4 compared with the other ILC genomically identified subtypes. There is limited evidence suggesting that CD8⁺ T-cell infiltration and PD-L1 positivity of tumor and immune cells are correlated in ER⁺ ILC but not in IDC (137). As a step toward identifying patients who may benefit from immune-targeted agents, a recent large study focusing on 459 ILCs found that stromal TILs assessed by IHC were associated with younger age, larger tumors, lymph node involvement, and HER2 amplification (138). A corroborating study also showed similar findings, with younger age, lymph node involvement, and highly proliferative tumors (such as those with pleomorphic features) demonstrating more TIL involvement (105). Recent studies have also shown that there are subsets of Luminal ILCs that harbor enriched immune cell infiltration and high immune-checkpoint gene expression (139). For these cases, CD274, PDCD1, and CTLA4 in high immune ILCs were comparable to levels seen with basal-like and HER2-E tumors. These enriched immune cases were largely due to increased dendritic cell and NK-cell infiltration, identified by immune gene signatures, and less driven by T-cell differences between ILC and IDC. In total, these findings suggest that the negative association of TILs with ILC may reflect the fact that TILs are found in the more aggressive ILC subtypes but are largely absent in classic ILCs.

In our recent study that utilized fresh treatment-naïve tumors to examine differences between HR⁺ IDC and ILC, M2-like macrophages were found to be a key cell type enriched within the TIME of ILCs (134). This enrichment was seen in both the tumor and surrounding stromal regions. Functional analysis suggested that ILC tumor cells specifically could be driving macrophages toward the M2 phenotype. Cellular neighborhood analysis using data from mIHC revealed the importance of a tissue architecture–wide approach in understanding differences between ILC and IDC, as specific cell type and neighborhood frequencies, specifically the interaction between T cells and macrophages, showed differential association with outcome in the two histologic subtypes.

With the histologic differences between ILC and IDC, ILCs typically have a more prominent stromal (nonepithelial) component than IDCs (140), and studies suggest that the loss of CDH1 drives an entirely different TIME structure with changes to associated infiltrative immune cells (141, 142). This may provide one reason why ILCs exhibit fewer infiltrating immune cells, as some believe the stromal component provides a physical barrier that is more difficult for immune cells to penetrate through (143). ILCs can be additionally categorized into a “reactive” breast cancer subtype, which is defined by a high intratumoral stromal component without many infiltrating immune cells (Fig. 3). This stromal component is associated with good prognosis and high breast cancer–specific survival (144). Interestingly, given the unique interactions with the stroma in ILC, treatment targeting stromal contents such as LOXL1 decreased tumor growth, invasion, and metastasis in preclinical studies (145). Additionally, a recent analysis showed a unique transcriptome for ER⁺ ILC CAFs compared with ER⁺ IDC CAFs (146). The PAPPA was the most enriched gene in the ILC stroma compared with the tumor epithelium, and higher PAPPA was associated with increased IGF1 bioavailability. Targeting this stromal or CAF compartment of ILCs, with recent evidence pointing to LOXL1 or the IGF pathway, may also be a therapeutic opportunity. Further investigation into the spatial architecture of ILCs and how this leads to fewer immune cells but a better prognosis is certainly warranted.

Although important albeit small subsets of these tumors exhibit higher immune infiltration, these results could be explained by the finding that a very small fraction (about 5%) of primary ER⁺ IDC and ILC tumors displays high (>10 mut/Mb) TMB and potential neoantigen load with ER⁺ IDC >ER⁺ ILC (147, 148). Interestingly, there are new reports suggesting that metastatic HR⁺ disease and metastatic tumors that harbor ER mutations specifically display higher TMB (ILC >IDC) and are better infiltrated than the primary disease (149, 150).

Overall, ILCs have a distinct interaction with the immune system given the enrichment of stromal content (Fig. 3). This may impart less immune cell infiltration compared with IDC but confers a better prognosis for HR⁺ ILC. Immune-targeting agents may be better geared toward the advanced setting for patients with ILC. We continue to advocate for additional and lobular-specific results from immunotherapy clinical trials to drive data for this group of patients (151).

Because primary breast tumors are routinely resected and breast cancer mortality is nearly exclusively a consequence of metastatic disease, an additional point of consideration should be the distinct organs in which metastases occur. First, it is worth noting that intrinsic subtype switching may occur during metastatic spread, and this switching can affect tumor–immune cell interaction in the host organ. Basal-like TNBCs tend to be the most stable and remain basal-like in the metastatic site, whereas luminal and HER2-E tumors seem to exhibit higher rates of switching. A recent analysis from the AURORA US Network, which used supervised learning of gene expression data transformed into a set of 740 signatures, showed that signatures of stromal cells, endothelial cells, and many adaptive immune cells are lower in metastatic sites than in the paired primary tumors (21). A subtype-specific approach revealed that whereas basal-like/TNBCs had far less enrichment of T cells, B cells, NK cells, and antigen presentation in metastases, ER⁺ tumors showed little change with respect to these signatures in metastases. Further analysis by the site of metastasis showed that immune features were much lower in liver and brain metastases, but little was changed between primary tumors and associated lung metastases. This may have a profound impact on those who receive immunotherapy in the metastatic setting, and initial clinical observations from other primary tumor types confirm these findings of immunosuppression in the liver (152).

Given the heterogeneity of the tumor and the associated TIME in ER⁺ tumors, HER2⁺ tumors, and TNBC, it is not surprising that the immune response itself is also quite diverse with some unique areas of interest in each molecular and histologic subtype. Increasing insight from preclinical and clinical research focusing on each subtype has confirmed that breast cancer is in fact a group of different tumor types that require distinctive approaches for optimal treatment. Given the wealth of clinical trial data over the past 5 years, only modest benefit has been seen from T cell–targeted agents such as pembrolizumab and atezolizumab. Even clearer is the fact that benefit from anti–PD-1 or anti–PD-L1 monotherapy is minimal; thus, approaches that harness other aspects of the immune response and in tumor subtype–specific settings will likely be necessary. In the subsequent section, we provide a primer of some emerging approaches that provide opportunities for future immune-targeted therapies guided by the aforementioned observations.

Although monotherapy with ICI has been limited, many trials are now sprouting that leverage a series of combination therapies to boost ICI efficacy. At the top of the list are conventional cytotoxic agents and radiotherapy. Immunostimulation with chemotherapeutics is thought to release immune-enhancing molecules from dying cancer cells (153). In this case, cytotoxic agents have been shown, mainly in the setting of lung tumors, to convert “cold” TIMEs that are poorly infiltrated into a “hot” TIME with an abundance of dendritic cells and CD8⁺ TILs (154, 155). Interestingly, in the recent TONIC trial, which tested a number of immune induction strategies, doxorubicin administration rendered the metastatic TIMEs more sensitized to anti–PD-1 therapy. This study further showed that following doxorubicin or cisplatin induction, there was enrichment of genes and gene signatures related to inflammation, JAK–STAT, and TNFα pathways (156). Although immune induction or stimulation with cytotoxic agents may be beneficial for its systemic effects, local immunostimulation may be achieved with radiotherapy. Preclinical data from a poorly immunogenic mouse model of TNBC showed that radiotherapy induced new tumor mutations, and these new mutations rendered the tumor more immunogenic and recognizable to CD4⁺ and CD8⁺ T cells, allowing for more effective tumor control (157). Critically, stimulation of the immune system may also be mediated by DNA damage and local inflammation, which appear to upregulate the local IFN responses (158). Overall, whether utilized in the pretreatment of tumors with cytotoxic agents or radiotherapy or at the same time as ICI, chemoimmunotherapy or radiation-containing ICI regimens hold potential for further exploration.

In addition to rescuing dysfunctional T cells using ICI, adoptive T-cell therapies are another avenue, with some promising results in patients with breast cancer having mainly been tested in the advanced setting. In a recent study, neoantigen-specific TILs were cultured and used for treatment in combination with pembrolizumab (159). Interestingly, there was heterogeneity in tumor response by subtype, with a patient with ER⁺ cancer having the longest duration of response. Limitations of this approach include its time- and resource-intensive nature as well as the limited number of patients who actually qualify for the therapy after screening for TILs recognizing immunogenic somatic mutations. For subtypes like HER2⁺ and TNBC with slightly higher mutational burden, generating a tumor-specific (c-Met or mesothelin) chimeric antigen receptor T cell has opened the door to a newer area of immunotherapy for patients with advanced disease (43, 160). A case report describing complete and durable (ongoing) regression of chemorefractory metastatic HR⁺ breast cancer upon infusion of expanded autologous T cells (specific for somatic mutations in four genes) in conjunction with IL2 and ICI represents an exciting and novel frontier for the application of immunotherapeutic treatment for advanced HR⁺ tumors (161).

A real challenge to the efficacy of effector cells is the protumor, immunosuppressive microenvironment and immune evasion within breast cancer. Many breast TIMEs (ex vivo and in vivo in mouse models) have already been shown to be rich in protumor molecules that modulate immune suppression like IL4, IL6, IL8, IL10, IL17, TGFβ, and IDO (162–166). These are generally higher in TNBC compared with HER2⁺ and HR⁺ breast cancer in different stages across primary and metastatic disease (167). As such, dissecting these underlying factors of resistance is an area of intense ongoing research. Some examples of potential strategies targeting resistance that are being tested include anti-TGFβRI and TGFβRII for tumor-secreted TGFβ, anti-CSFR1 for TAMs, and histone deacetylase inhibitors to decrease MDSC burden, which could specifically be geared toward the HR⁺ TIME (103, 168). Interestingly, in a recent phase I trial testing fresolimumab, a TGFβ-blocking antibody, patients receiving higher doses of the drug experienced longer median OS than the lower dose group (169). Both groups experienced favorable systemic immune responses.

Targeting TAMs and myeloid cells may be another strategy to dampen the protumor TIME. As key protumor cells, TAMs may function primarily by promoting extracellular matrix remodeling, supporting angiogenesis, sustaining inflammation, and by acting as a hub for regulating antitumor adaptive immune responses (170). Emerging strategies to target TAMs include the use of anti-CD47 therapies and CSF1R inhibitors. Anti-CD47 checkpoint immunotherapy inhibits the CD47 surface protein and messages macrophages to proceed with phagocytosis. Typically, tumor cells coat themselves with the CD47 molecules, which act as a “don't eat me” signal to macrophages. Although piloted in HER2⁺ disease (171), additional research is warranted in the ER⁺ setting given the enrichment of TAMs. CSF1R inhibitors, such as pexidartinib, target monocyte homing to tumors prior to differentiation into TAMs (172, 173). Early responses in phase Ib clinical trials warrant further testing.

Beyond these strategies, further consideration of various immune-checkpoint markers is also warranted. LAG3, for example, has been shown to be enriched in basal and HER2⁺ tumors and leads to poor survival (174). Given the recent FDA approval of a fixed dose combination of relatlimab (anti-LAG3) and nivolumab (anti–PD-1) for nonresectable or metastatic melanoma (175), further study is warranted for its use in breast cancer.

Interestingly, across all subtypes, despite the higher tumor burden in the metastatic setting, immune infiltration is significantly lower than in matched primary tissues with the exception of macrophages, which tend to increase in metastatic disease (166, 176). One study reported that the magnitude of reduction in infiltration across T and B cells from primary to metastatic disease is more pronounced in TNBC compared with HR⁺ and HER2⁺ disease, a finding that needs to be validated further through cross-platform analyses (177). These key observations, although limited at this point, might have a significant impact on the design of all immunotherapy trials in the advanced setting using only ICI that target just T cells but not protumor macrophages, which might be more abundant in these advanced stages of the disease. In order to significantly impact the outcome for patients with advanced disease, it would be prudent to include a combination of immunotherapeutic agents that not only support immune activation or T-cell stimulation but also block the protumor suppressor cells that form a bigger barrier to an effective immune response in the metastatic setting.

Given the heterogeneity in the CAF subsets observed in breast cancer subtypes, targeting a specific subset (such as the immunosuppressive CAF-S1 group in TNBC) may be challenging. Prevailing hypotheses suggest targeting specific signaling molecules or pathways may be a more viable option (48). For example, most of the immunosuppressive CAFs in TNBC upregulate IL6/IL6R signaling, so that therapeutics, such as tocilizumab, an anti-IL6R antibody, may help to dampen the immunosuppressive TIME. ICIs have also shown limited benefit in tumors that have a prominent stromal component, which provides a physical barrier for immune cells. Targeting the extracellular matrix may also help to remodel this barrier and allow more immune infiltration (48). In total, targeting CAFs may be a tool in a multipronged approach to revert an immunosuppressive TIME.

Recent investigation into the metabolic subpopulations in breast cancer subtypes has yielded some insight into potential targets to boost antitumor responses. For example, in TNBC the MPS subtyping showed that particular tumors may be more sensitive to lactate dehydrogenase therapy in combination with anti–PD-1 agents (34). Another interesting avenue for consideration is the anti-CD73 agents, which target the CD73 ectonucleotidase, preventing downstream adenosine pro­duction into the TIME (178). Adenosine is thought to be protumorigenic in the TIME, as it promotes the generation of Tregs and MDSCs and impairs the functioning of T cells and NK cells (179). Recent studies in EGFR-mutated non–small cell lung cancer have shown that although neither anti–PD-L1 nor anti-CD73 therapy alone significantly inhibited a xenograft mouse tumor model, combination therapy not only prevented tumor growth but also increased the number of CD8⁺ TILs while enhancing secreted IFNγ and TNFα from these cells. With a number of ongoing clinical trials in breast cancer, this could be a promising link to inhibit the metabolic TIME (180, 181).

In summary, within the different breast cancer subsets, there is growing evidence that HER2⁺ and TNBC subsets will benefit most from targeting the effector arm of the immune system, whereas HR⁺ TIME will require immunotherapeutic interventions that focus on antigen presentation and immune cell activation of existing T cells within the TIME. Additionally, all three TIMEs tend to be immunosuppressive to varying degrees and will benefit substantially from therapies that target Tregs (TNBC >HER2⁺ >HR⁺), MDSCs (TNBC >HER2⁺ >HR⁺), and TAMs (HR⁺ >TNBC/HER2⁺).

It has become clear that novel strategies to better manage breast cancer are essential. Immunotherapy is one such powerful weapon in our arsenal against cancer, as demonstrated through its revolutionary impact on outcomes for a number of solid tumors. For the success of immunotherapeutic modulation, it is imperative that future studies are designed with an eye on the unique features for each of the subtype TIMEs. It is becoming increasingly clear that immunotherapy as monotherapy might not be optimal to treat any subtype of breast cancer. Multipronged, innovative approaches for targeting tumor and immune cells together might be the path forward for breast cancer. For the further rational design of immunotherapy with maximal benefit, we must thoroughly characterize the TIME of the different molecular and histologic breast cancer subtypes to understand and address a series of outstanding questions (Fig. 4).

In conclusion, to aid the interrogation of the aspects of the TIME stated above, recently developed, sophisticated approaches for sequencing, imaging, and in vivo syngeneic animal studies that are currently underutilized in subtype-specific breast cancer research must be leveraged. These include mIHC, spatial transcriptomics, multiparametric flow cytometry, and single-cell RNA-seq. Lessons from such studies when translated to the clinic can yield exciting new opportunities to treat patients with breast cancer. It is conceivable that immunotherapy when combined with other targeted agents specific for each breast cancer subtype will develop into an indispensable therapeutic modality for enhancing patient care across all subtypes of breast cancer.

N.M. Carleton reports grants from the NCI during the conduct of the study. P.C. Lucas reports other support from Amgen and personal fees from Schrodinger, Inc. outside the submitted work, as well as a patent for small molecules and their use as MALT1 inhibitors pending. T.C. Bruno reports other support from Walking Fish Therapeutics outside the submitted work. D.A.A. Vignali reports grants, personal fees, and other support from Novasenta and Bristol Myers Squibb, personal fees and other support from Tizona and Werewolf, personal fees from Bicara, FStar, Apeximmune, T7/Imreg Bio, G1 Therapeutics, Inzen Therapeutics, and Incyte, and other support from Trishula and Oncorus outside the submitted work, as well as patent for LAG3 pending, issued, licensed, and with royalties paid from Bristol Myers Squibb. No disclosures were reported by the other authors.

We thank the members of the Vignali Lab (Vignali-lab.com; @Vignali_Lab) and the Lee/Oesterreich Lab for their helpful suggestions and conversations related to this review. We apologize to the researchers whose original work could not be cited due to space restrictions. The authors thank the following funding sources: the NIH/NCI under awards T32CA082084 and F30CA264963 (to N.M. Carleton), R35CA263850 and P01AI108545 (to D.A.A. Vignali), R01CA252378 (to S. Oesterreich and A.V. Lee), and P30CA047904. A.V. Lee and S. Oesterreich are Komen Scholars and Hillman Foundation Fellows and acknowledge additional funding for research related to the topic of this review from the Breast Cancer Research Foundation, the Magee-Women's Research Institute and Foundation, the Shear Family Foundation, and METAvivor.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.