The incidence as well as the mortality of breast cancer (BC) is ever increasing with significant variation among different countries. In developing countries, the BC-based mortality rate is moderately higher due to limited treatment regimens and late-stage diagnosis. BC occurs mainly due to somatic, genetic, and epigenetic alterations in a lifetime (i.e., non-hereditary), while only 5%–10% of BC cases are hereditary [majority detected with mutations in two tumor-suppressor genes: breast cancer gene 1 (BRCA1) and breast cancer gene 2 (BRCA2)]. For BC patients, the 5-year survival rate is approximately 99% for localized BC, 86% for regional diseases, and 27% for metastatic BC (stage IV).

BC represents a complex and heterogeneous microenvironment with distinct subtypes, identified through hormonal gene expression profiling. There are four molecular subgroups: luminal A (estrogen receptor, ER⁺; progesterone, PR⁺; HER2⁻), luminal B (estrogen receptor, ER⁺; progesterone, PR^+/−; HER2^+/−), human epidermal growth factor receptor 2 (HER2)-positive (the HER2 protein helps breast cancer cells grow quickly), and triple-negative breast cancer (TNBC) (estrogen receptor, ER⁻; progesterone receptor, PR⁻; HER2⁻) ( Figure 1 ). Presently, TNBCs are further divided into the following transcriptome-based subtypes: basal cell-like type 1 (BL-1), basal cell-like type 2 (BL-2), immune-modulatory (IM), mesenchymal-like (M), mesenchymal stem cell-like (MSL), luminal androgen receptor (LAR), and claudin low (1, 2). Luminal A and luminal B subtypes, due to the presence of hormone receptors, respond to anti-estrogen or anti-progesterone therapies. TNBC, on the other hand, due to the absence of hormone receptors, is difficult to target and is considered as the most aggressive subtype of breast cancer (1, 2).

Breast cancer stem cells (BCSCs) are a rare population of undifferentiated cells that reside within the tumor and exhibit properties similar to normal stem cells and are referred to as tumor-initiating cells (3). They possess the capacity for self-renewal, which refers to a cell’s ability to divide endlessly in an undifferentiated condition (3). They replicate seldom or slowly and have infinite potential for proliferation (3). They can divide asymmetrically to produce daughter cells with the capacity to differentiate (3). Clinically, BCSCs are responsible for treatment resistance and cancer relapse because of their relative resistance to radiation, chemotherapy, and molecular targeted therapy (3). These residual cells after therapy have all the hallmarks of epithelial to mesenchymal transition (EMT) with increased metastasis capacity (3). Also, BCSCs are capable of proliferating in low adherence cell culture conditions in the presence of growth factors like epidermal growth factor and basic fibroblast growth factor to generate floating spheroids called tumorspheres or mammospheres (3). The number of spheroids represents the number of BCSCs, while their size depicts their proliferative capacity (3) ( Figure 2 ).

There are several theories regarding the origin of BCSCs ( Figure 3 ). One theory supports that BCSCs arise from the dedifferentiation of non-stem cells (mammary epithelial cells) (4). Genetic and epigenetic alterations as well as changes within the tumor microenvironment (TME) contribute to the dedifferentiation of non-stem cells to the BCSC phenotype (4). Another theory suggests the presence of multipotent mammary stem cells as well as unipotent luminal and basal progenitor cells within the mammary glands (5). Accumulation of mutation in these progenitor cells may give rise to BCSCs. Growing bodies of evidence suggest that BCSCs can arise from normal stem cells by accumulating mutations (6, 7).

BCSCs are characterized based on the expression of several markers derived from breast cancer cell lines, transgenic mouse models, and patient-derived tumors. Among these, the most commonly used markers are CD44⁺/CD24^−/low and alcohol dehydrogenase 1 (ALDH1⁺). ALDH belongs to the family of NAD(P)⁺-dependent enzymes which are involved in the detoxification of a wide variety of aldehydes to their corresponding carboxylic acids. It mainly functions in converting vitamin A (retinol) to retinoic acid (8–10). It maintains the characteristics of cancer stem cells including drug resistance, thereby contributing to disease relapse (11). CD44⁺/CD24^−/low and ALDH1⁺ represent two distinct subpopulations of BCSCs that are different from one another (9). These two states are highly dynamic and interchangeable (9). Their number varies among different subtypes of BC. Among all subtypes, luminal-A BC has the lowest proportion of BCSCs which contributes to its best prognosis (9). Luminal B, however, has higher proportions than luminal A but lesser than TNBC or HER2⁺ breast cancers ( Figure 4 ). The HER2⁺ BC is characterized by the presence of ALDH1⁺ epithelial BCSCs leading to its poor prognosis. TNBC is the most heterogeneous subtype and is characterized by the presence of the highest population of BCSCs (12–14). Claudin-low TNBC has a high proportion of mesenchymal BCSCs (CD44⁺/CD24^−/low), whereas the basal-like TNBC has higher proportions of ALDH1⁺ epithelial BCSCs and also certain amounts of mesenchymal BCSCs (12).

In this review, we have tried to demonstrate the underlying mechanism of the crosstalk between BCSC with various immune cells of relevance, the major determinant factor of tumor fate. Focus has also been given to their impact on the plasticity of certain immune cells and the stemness of BC cells to keep immune surveillance in check so that tumor proliferation and progression can be encouraged without immune hassle. Finally, some BCSC-based therapeutic strategies have been discussed that might help in regulating BC severity.

Cancer stem cells (CSCs) exhibit the capacity of self-renewal, initiating and promoting primary tumor growth, and can drive metastases at distal sites. In 2003, BCSCs were first isolated using cell surface markers for CD44⁺/CD24^−/low. There exist two distinct developmental states for BCSCs, and due to cellular plasticity, these two states are reversibly interchangeable (9). One BCSC state is mesenchymal-like (CD44⁺/CD24⁻) with low proliferative activity and located mainly at the tumor-invasive edge (9, 15, 16). The other BCSC population is epithelial-like (ALDH1 expressing) which is highly proliferative and mainly found at the center of the tumor (9, 15, 16). Cytokine signaling such as inhibitor of DNA binding 1 (Id1) reversibly regulates the transition between two BCSC states (9). BCSCs which are CD44⁺/CD24⁻ and ALDH⁺ have robust tumor-initiating capacity and are involved with the incidence of chemoresistance and breast cancer relapse. When small numbers of isolated BCSCs were injected into immune-compromised mice, they exhibited a high capacity to develop tumors compared to when the system was injected only with a high number of cancer cells (17). Tumors generated by BCSCs maintained the cell type heterogeneity of the primary tumor.

Research studies of the last two decades have replaced the traditional view of breast cancer as a homogeneous system of rapidly proliferating neoplastic cells and uncovered the true face of the disease and established that BC is composed of cancer cells along with a variety of immune and other host cells that altogether create the breast TME. The crosstalk between all these cellular and structural components in the BC TME is cardinal for the growth and progression of BC. Here, we are specifically focusing on the interactions between BCSCs along with various immune cells in order to understand the consequences of such interactions and their role in the determination of breast cancer fate.

The interaction between cancer cells and immune cells is highly complex and can be both beneficial and detrimental for tumor cells. The final outcome of this interaction depends on various environmental stimuli within the TME. “Escape of immune surveillance” is one of the major hallmarks of tumor cells, and they use various techniques like secretion of immune-suppressive molecules like IL-6 and TNF-α, while others edit the host immune system by targeting the regulatory T cells. Defective antigen presentation by targeting the MHC-I pathway is another strategy used by tumor cells to escape the host immune system (78). CSCs also are endowed with properties that help tumors in immune evasion (83). CSCs may secrete immunosuppressive factors and also recruit non-cancerous immune-suppressive cells (83). CSCs may downregulate MHC molecules and tumor-associated antigens, thereby inhibiting/escaping from the immune response of T cells (83).

The two types of BCSCs—ALDH1⁺ and CD44⁺/CD24⁻—have different mechanisms to escape the immune response. ALDH1⁺ CSCs downregulate the TAP gene and CD80, thereby targeting the antigen presentation pathway (84), while the CD44⁺/CD24⁻ group upregulates the expression of CXCR4 which is involved in EMT and the induction of the CSC phenotype in BC (83, 85, 86). BCSCs are also capable of inhibiting various cytolytic granules like perforin and granzyme B. Several studies support that CSCs from estrogen receptor-α-positive cell line inhibit granzyme B by upregulating the expression of protease inhibitor 9 which targets granzyme B (87, 88). CD44⁺/CD24⁻ BCSCs express higher levels of TGF-β (a potent inducer of EMT) than the non-CSC population (84, 89, 90). Moreover, Sca-1 is found to be involved in TGF-β signaling (88, 90). In a human BC model, Sca-1 homologs Ly6K and Ly6E play important roles in TGF-β pathway activation and, thus, immune evasion (90, 91). Another pleiotropic cytokine, IL-6, secreted from TAM, induces and maintains the CSC phenotype via STAT3 signaling (92–94). Hypoxia induces the CSC phenotype in BC cells by activating the NOTCH1 signaling pathway (94, 95), while CCL20 induces the BCSC phenotype via the NF-κB pathway (90, 93, 94, 96). PD-L1, an important checkpoint molecule, is positively correlated with BCSC expression (97, 98). Thus, BCSCs employ a variety of mechanisms to bypass antitumor immunity. Therefore, targeting these CSCs to restore the host immunity will be an effective therapeutic strategy.

The transcription factors OCT4, SOX2, NANOG, BMI1, and KLF4 are associated with the stemness and aggressiveness of BC. OCT4 and SOX2 are associated with high histologic grade tumors, whereas NANOG, BMI1, and KLF4 are common in low histologic grade tumors and hormone receptor-positive BCs (99). The expression of these transcription factors varies among different molecular subtypes of BC. The expression of OCT4 and SOX2 is the lowest in the luminal A subtype. However, luminal A and luminal B subtypes have a higher expression of NANOG, BMI1, and KLF4 than the HER2 and TNBC subtypes (99). Stemness and EMT are two correlated events. OCT4 is an important inducer of EMT (100). In lung cancer, it upregulates the expression of mesenchymal markers like vimentin and N-cadherin and promotes the degradation of the β-catenin/E-cadherin complex, thereby facilitating the invasion of the CSCs (100, 101). However, in breast tumors, the result is quite contrasting. In a study conducted by Hu et al., it was shown that silencing OCT4 induced EMT and promoted the invasion of the human breast cancer cell line MCF7 (101). Like OCT4, SOX2 is also associated with tumor aggressiveness, whereas NANOG is associated with lymph node metastasis of breast tumors (99–103).

In recent years, the development of molecular subtype-specific therapies has significantly increased the overall survival rate (ORR) for BC patients, particularly with metastatic spread. These therapies principally encompass third-generation hormonal therapies against hormone-susceptive BC, HER2-targeting agents against HER2-overexpressing BC, and PARP inhibitors to counter TNBC (104). Despite these advancements, many patients still experience relapse. Accumulating pieces of evidence suggest that a very small malignant subpopulation with stem cell properties residing within BC contributes to these phenomena. On account of their relative resistance to conventional therapies and tumorigenic potency, these BCSCs persist against genotoxic hits and contribute to treatment resistance and induce relapse. That is why framing treatment strategies against BCSCs is urgently needed.

BCSCs being a very small subset within a malignant bulk require proper identification for therapeutic targeting. The development of biomarkers to identify BCSCs has fast-tracked such as identification and characterization. The BCSC marker CD44 has been indicated as a biomarker for diagnostic, therapeutic, and prognostic purposes ( Table 1 ). A significant decrease in sphere formation and tumor growth in severe combined immunodeficiency (SCID) mice was observed after the knockdown of CD44 (110). Various CD44 targeting avenues include CD44-specific antibodies, competitive proteins saturating CD44-binding sites, chemotherapy agents, and CD44-siRNA coupled to CD44-binding partners such as hyaluronic acid (108). The biomarker CD133 (prominin-1) is linked with high tumorigenicity and spheroid-forming ability (109). The therapeutic approaches to target CD133 govern antibody-based targeting and CAR T-cell therapy (clinical trial no. NCT02541370) (6, 111) ( Table 1 ). In BC, epithelial cellular adhesion molecule (EpCAM) expression positively correlates to CSC-like phenotypes which further promote bone metastasis (112). Targeting EpCAM⁺ BCSCs ( Table 1 ) with multiple antibody formats including EpCAM toxin-conjugated antibodies like oportuzumab monatox [scFv antibody and Pseudomonas exotoxin A (ETA)], citatuzumab bogatox (Fab fragment with bouganin toxin), and immuno-conjugate antibody tucotuzumab (monoclonal antibody with IL-2) have yielded positive response (113).

The second approach to counter BCSCs is to target signaling networks that govern the CSC phenotype and functionality. In BC, the NOTCH, WNT/β-catenin, and Hedgehog (Hh) pathways ( Table 1 ) regulate the self-renewal of BCSCs. Aberrant activation of the NOTCH signaling pathway promoting chemo-/radioresistance in BSCSs is prevented by gamma-secretase inhibitors (GSI). GSI administered alone or in combination with chemotherapeutic agents like docetaxel (GSI MK-0752), paclitaxel, and carboplatin has proven its efficacy and currently undergoing clinical trial (114).

The WNT/β-catenin pathway regulates not only BCSC maintenance but also cellular plasticity through EMT modulation (115). Aberrant activation of WNT has been neutralized by WNT inhibitors, which are now entering phase I clinical trials (116, 117). Hh signaling proteins like SMO and GLI1 selectively upregulate BCSCs, particularly in TNBC through paracrine secretion. SMO inhibitors like vismodegib and sonidegib are passing through clinical trials for basal cell carcinoma. Additionally, clinical trials are now focusing on the combination therapy of SMO inhibitors and chemotherapeutic drugs in TNBC. GLI1 inhibitors like GANT58 and GANT61 are also under preclinical drug development (118). Several studies have also reported that HER2-signaling-regulated BCSC activities, including self-renewal and radioresistance, have been counteracted through HER2-specific antibodies like trastuzumab, which is going through a phase III clinical trial (119, 120).

Another outlook in BCSC targeting is neutralizing their overexpressed drug efflux pumps like ABC transporters. Because of their strong drug efflux properties, they serve as a strong protective mechanism against xenobiotic altercations, particularly during conventional chemotherapies. Utilizing ABC inhibitors like pheophorbide (ABCG2 inhibitor) and camptothecin (ABCC1 inhibitor), BCSCs are sensitized (121, 122).

Another perspective that is employed in BCSC targeting is immunotherapy. Immunotherapy targets BCSCs by modulating immune cells such as cytokine-induced killer (CIK) cells, NK cells, γδ T cells, and CD8⁺ T cells and through DC-based vaccines (123, 124). Adoptive transfer of reinfused TILs, e.g., CAR T (chimeric antigen receptor T) cells or CIK cells, has shown therapeutic efficacy. Adoptive transfer of CD8⁺ T cells and γδ cells has elicited antigen-specific killing by CD8⁺ T cells in BCSCs (71). Very often, immunotherapies are used in combination with chemotherapy, viz., atezolizumab (a PD-L1 inhibitor) coupled with albumin-bound paclitaxel to treat metastatic TNBC (125, 126). BCSCs also exhibit selective upregulation of various cancer testis (CT) antigens like MAGEA12, PA17, PLU-1, and TEX15 (127). Since these antigens are normally expressed only in germline cells, which remain devoid of MHC molecules, immune targeting via CT-specific T cells appears to be promising. So far, MAGE-A and NY-ESO-1-based vaccines are currently under clinical trials (127).

Targeting BCSC markers like ALDH1, CD44, CD133, and CD24 with specific agents and antibodies combined with chemotherapy can achieve a good therapeutic effect (6). Positive results were obtained by combining anti-CD44 antibody and gemcitabine with magnetic nanoparticles (MNPs) (105). Similarly, low-dose decitabine and doxorubicin along with nanoparticles yielded better results in ALDH^high CSCs (105). Normal stem cells with many different signaling pathways can be turned into CSCs. Signaling pathways such as TGF-β, NOTCH, JAK-STAT, WNT, Hh, and NF-κB regulate the survivability of CSCs. Ligand antibodies and GSI are found to inhibit the NOTCH signaling pathways. MK-0752 along with docetaxel is found suitable for the treatment of BC. Targeting the DLL4 protein (an angiogenic regulator) can also block the NOTCH pathway. CSC stemness was also regulated by smoothened (SMO) protein through Hh signaling. Sonidegib (an antagonist of SMO) with docetaxel is mainly used to treat a TNBC patient. WNT induces EMT and cellular plasticity leading to the generation of CSCs. Hence, WNT inhibitors restrained the growth of BCSCs, and their effect was prominent on both in-vitro and xenograft models. Targeting the HER2 signaling pathways is very successful in BC evasion. Lapatinib and trastuzumab and HER2 targeted antibodies increased patient survival with advanced stage ER⁺ cancer, reducing the CSC population. Lapatinib reduced the CD44⁺/CD24^low population and the self-renewing ability, and this was found through mammosphere formation (106). Treating antibodies against the IL-8 receptor CXCR1 and repertaxin (CXCR1/CXCR2 inhibitor) targets BCSCs in xenograft models, thus reducing the EMT and tumor growth.

Immunotherapy is also a potential strategy to counteract tumor growth and recurrence. It was reported that chimeric antigen receptor therapy T (CAR T) is a successful therapy for leukemia, so these treatments can also be examined in BCSC therapy. Reports suggest that DC vaccines elicit a better immune response by increasing the stimulation of interferon-γ expression. Blocking T-cell regulation by targeting PDL, PDL-1, and CTLA-4 showed a positive response in BCSC treatment. The FDA has approved a combined therapy of atezolizumab (a PD-L1 inhibitor) along with albumin-bound paclitaxel to treat metastatic TNBC (107, 128–130). Consequent in-vitro upregulation of PD-L1 in BCSCs has been reported, which however needs to be validated by further studies (107, 129, 130).

The heterogeneity of the cellular population and their function and in some cases the plasticity of the phenotypic behavior in TME plays a key role in determining the evolution and fate of BC. The complex network of interactions between BCSCs and various immune cells impacts heavily on tumor initiation as well as the metastatic behavior of BC. These modulated TME components signal for the acquisition and/or promotion of BCSC traits. This interaction also encourages EMT traits in the local and distant sites, thereby spreading the cancer. Understanding the molecular mechanism of such interaction will definitely help in preparing tumor-eradicating therapy not only by reducing tumor but also by eliminating future possibilities of recurrence and metastasis. Pointing out of these overlooked areas will help control BC in better non-conventional ways.

SBa, AG, and KG conceived and planned the article. AG, JS, PC, MC, KG, and SBa wrote the manuscript. NG prepared the graphical illustration. ABh, AS, SBe, SD, JD, TD, RB, and ABo edited the manuscript. All authors contributed to the article and approved the submitted version.