CD44 is a cell-surface hyaluronan acid (HA) receptor that contains an HA-binding site in its extracellular domain. Notably, HA is the major component of ECM. Hence, the CD44-HA interaction not only contributes to the cell adhesion to ECM components but also to tyrosine phosphorylation of cytoskeletal proteins, activation of RhoA/RhoC, Rac1, and Cdc42, fueling invasion and metastasis (43). Activation of these signaling pathways is essential for actin cytoskeletal remodeling, actin filament assembly, tumor cell migration, and invasion. CD44 is overexpressed in bCSCs and interestingly involves the Src kinase family proteins to initiate BC progression via Twist signaling (44). Moreover, CD44 contributes to chemoresistance since it upregulates the expression of multidrug resistance receptors by activating Nanog (45).

CD133 is another bCSC-specific marker found to be enriched in basal-like, HER2+, luminal, and TNBC subtypes. CD133-high bCSCs have been documented in tumor cell proliferation, vasculogenic mimicry, invasion, metastasis, and drug resistance (6). Croker et al. identified CD44+CD133+ALDHhigh bCSC-like cells as crucial mediators of BC metastasis (46). BRCA1-associated murine breast tumors consist of CD44+/CD24− and CD133+ cells with bCSC-like features, showing a greater intrinsic colony-forming potential that can regenerate breast tumors in NOD/SCID mice (47). Interestingly, CD133-high bCSCs augments endocrine resistance in metastatic BC via the IL-6/Notch signaling (48). Xenograft initiating CD44+CD49fhighCD133/2 high cells display self-renewal in vivo and greater tumorigenicity in ER− breast cancer (49).

Aldehyde dehydrogenase 1 is a member of NAD(P)+-dependent cytosolic isoenzymes, which is critically responsible for the oxidation of retinol to retinoic acid, required for early differentiation of stem cells. ALDH1 is a general marker of both human normal mammary stem cells and malignant mammary stem-like cells, and high ALDH1 activity is an independent predictor of poor clinical outcome and survival of BC patients (50). Interestingly, only a fraction of CD44+CD24−/low bCSC BC cells are ALDH1+ and display the highest tumorigenic potential, when compared with ALDH1− population (50). Moreover, ALDHhighCD44+CD24− and ALDHhighCD44+CD133+ bCSCs, isolated from MDA-MB-231 and MDA-MB-468 cell lines, respectively, demonstrated enhanced growth, adhesion, migration, colony formation, and invasion profile, compared with ALDHlowCD44−/low cells (46). Therefore, inhibition of ALDH activity can effectively reverse doxorubicin/paclitaxel resistance of ALDHhighCD44+ human bCSCs (51). ALDH+ bCSCs with an increased expression of interleukin-1 receptor (IL1R1) are enriched following antiestrogen therapy and held responsible for the treatment failure (52). Hence, targeting the ALDH+IL1R1+ bCSCs is crucial to reverse the drug resistance exerted by antiestrogens.

EpCAM is a glycosylated type 1 glycoprotein, expressed by human epithelial cells, and functions as an oncogenic signal transducer (53). Al-Hajj et al. showed that EpCAM+CD44+CD24−/lowLineage− fraction had a >10-fold higher frequency of tumor-initiating cells compared withto EpCAM−CD44+CD24−/lowLineage− fraction (4). Interestingly, EpCAM overexpressing BC cells can withstand greater radiation stress compared with EpCAMlow cells (54). Hence, EpCAMhigh BC cells retain the ability to form a higher number of mammospheres. Activation of the AKT pathway is also observed in EpCAM overexpressing ZR-75-1 breast cancer cell line, compared with parental cell line. Moreover, EpCAM overexpression also reflects a higher percentage of cells with an E/M hybrid state, encouraging EMT, invasion, and metastasis. EpCAM+ circulating tumor cells isolated from primary human luminal BC patients’ blood contain metastasis-initiating cells, leading to bone, lung, and liver metastasis in mice (55). Since survivin has a crucial role in bCSC chemoresistance, EpCAM aptamer-mediated survivin silencing can sensitize bCSCs to doxorubicin and reverse chemoresistance (56).

The chemokine receptor C-X-C chemokine receptor type 4 (CXCR4) is considered to be a prognostic marker of bCSCs. The metastatic cascade is initiated via a series of sequential steps that include local invasion and intravasation (transendothelial migration) of cancer cells from the primary tumor site into the circulation, followed by extravasation at distant sites and subsequent organ colonization (homing) (57). Cancer cells at the growing front of the tumor undergo EMT, which degrades the underlying basement membrane and ECM before intravasation. The CXCR4 receptor and its ligand, CXCL12 (SDF-1) play an important role in the dissemination of BC cells from the primary site, transendothelial migration, and eventually trafficking and homing of bCSCs. Chemokines are 8–12-kDa chemoattractant cytokines that contribute to differentiation, cell activation, and trafficking. Notably, the chemoattractant CXCL12 provides directional guidance to CXCR4+ bCSCs toward the secondary metastatic site and initiates metastasis (58–60). Hence, targeting the CXCR4-CXCL12 signaling axis could serve as an alternative approach to restrict bCSC-driven drug resistance. Additionally, CXCR4+ bCSCs show a higher vimentin/E-cadherin ratio, indicating EMT. Interestingly, CXCR4 inhibition can enhance the infiltration of cytotoxic T-cell lymphocytes (CTLs) and improve the responses to immune checkpoint blockers in metastatic BC (61). A quantitative phosphoproteomic study by Yang et al. validated the importance of CXCR4-SDF1 signaling in bCSCs and also identified several important signaling pathways in bCSCs, downstream of CXCR4-SDF1 (62).

Vasculogenic mimicry (VM) is a recently defined pattern of tumor microvascularization that refers to the ability of cancer cells to organize themselves into vascular-like structures to procure nutrients and oxygen independently of normal blood vessels (76, 77). Unlike the concept of angiogenesis or vasculogenesis where the endothelial cells participate in blood vessel formation, VM particularly depends on the participation of highly aggressive tumor cells, having the endothelial phenotype, to form vessel-like structures ( Figure 3A ). VM has been reported in different types of solid aggressive tumors including BC (78–80). CD133+ breast CSCs reflect VM, with a higher expression of vascular endothelial-cadherin (VE-cadherin), along with an upregulated expression of matrix metalloproteinase, MMP-2, and MMP-9 in TNBC (81). Notably, both MMP-2 and MMP-9 are critical players in cellular plasticity and VM formation. According to Sun et al., it is the bCSCs that line the VM channels in breast tumor tissues from TNBC patients (82). Additionally, bCSCs produce more VM-related molecules like CD133 and ALDH1, to synergize VM formation (82, 83). However, cells participating in VM formation lack the classical endothelial marker CD31, and thus, administration of angiogenesis inhibitors does not affect VM formation. In this context, a phytocompound, thymoquinone (TQ), has been reported to exhibit an inhibitory effect on VM and promotes mesenchymal–epithelial transition (MET) in bCSCs derived from MDA-MB-231, in a dose-dependent manner. Moreover, CD44+CD24− bCSCs when incubated with TQ can interfere with rhodamine-123 efflux and decrease stemness. This observation indirectly denotes that thymoquinone relieves the drug-resistance properties of bCSCs (84). Mechanistically, TQ suppresses the PI3K and Wnt3a signaling, leading to the reduction of the p-Akt/Akt ratio, and has the potential to reduce the number of bCSCs.

Ferroptosis is an iron-dependent mechanism of regulated cell death, which is characterized by the intracellular accumulation of lipid-based ROS, ultimately resulting in the loss of membrane integrity ( Figure 3B ) (85). Notably, lipid-ROS is detoxified in a GPX4-catalyzed enzymatic reaction, which uses glutathione as a reducing agent. Hence, ferroptosis can be triggered either by inhibiting GPX4 enzymatic activity or depleting glutathione. Type I ferroptosis-inducing compounds, including sulfasalazine and erastin block the amino acid transporter required for cysteine import to synthesize glutathione. Type II drugs, such as RSL3, interfere with GPX4 peroxidase activity. Mechanistically, the execution of ferroptosis requires a high concentration of intracellular iron. Ferritin, the intracellular iron-storing protein, can release iron to initiate ferroptosis. The released iron can yield lipid-ROS in an autoamplifying manner. Ferroptosis can be inhibited by the presence of iron chelators and activated by transferrin and its receptor (86). Hence, sensitizing tumor cells to ferroptosis appears as a possible therapeutic approach for BC treatment. Notably, drug-tolerant BC cells show a dependency on the GPX4 activity, thus inhibition of GPX4 can potentially overcome BC drug resistance (87). Taylor et al. reported an array of ferroptosis-inducing small molecules that can selectively kill bCSCs with the mesenchymal phenotype in vitro (88). TNBC cells are highly susceptible to cysteine starvation, leading to ferroptosis and necroptosis, via the GCN2-eIF2α-ATF4-CHAC1 pathway (89). Since cysteine serves as the substrate for glutathione synthesis to prevent ferroptosis, depleting the pool of cysteine can sensitize BC cells to ferroptosis (89). Another synthetic derivative from salinomycin, ironomycin (AM5), can trigger cell death in bCSCs, both in vitro and in vivo, by sequestering iron in lysosomes, which further indicates that iron homeostasis plays a crucial role in bCSC survival (90). A novel nanoparticle, ferritin-bound erastin, and rapamycin (NFER) has shown robust ferroptosis-inducing properties by interfering GPX4 in 4T1 orthotopic BC mouse model (91).

Autophagy is an evolutionarily conserved self-degradation process that recycles intracellular nutrients, growth factors, and energy to sustain survival and cellular activities during stress like hypoxia, nutrient deprivation, and ischemia (92). Autophagy provides bCSCs with metabolic flexibility that becomes a prerequisite for their survival in oxygen- or nutrient-poor TME (93). Autophagy contributes to bCSC dormancy, stemness, maintenance, and drug resistance (94–96). Chaperone-mediated autophagy (CMA) and macroautophagy are two different modes of the autophagy process, documented in mammalian cells. Autophagy can elevate bCSC number, and thus develop drug resistance to conventional chemotherapies (97). Furthermore, autophagy-related genes (ATGs) such as ATG4A, ATG5, ATG12, LC3-B, and Beclin1 are expressed in dormant bCSCs, promoting bCSC survival and sustaining bCSCs over the progression of BC (96). Expression of Beclin1 is noted to be higher in mammospheres derived from BC cell lines, MCF-7 and BT474, compared with the adherent cultures (98). Moreover, the expression of lysosome-associated membrane protein type 2A (LAMP2A), involved in the CMA pathway, is augmented in the course of BC metastasis (99). Autophagy of cancer-associated fibroblasts (CAFs) also contributes to TNBC proliferation and progression. Notably, autophagy-relevant Beclin1 and LC3-II/I protein conversion levels are higher in CAFs, compared with the normal fibroblasts in TNBC (100). Since autophagy serves as one of the factors in malignant growth, inhibition of autophagy can suppress tumor growth. Pharmacological targeting of the autophagic flux with salinomycin can reduce bCSC-driven drug resistance, interfere with their stemness, and also compromise the bCSC tumorigenic potential (101).

CSCs often express a higher level of ATP-binding cassette (ABC) transporters that facilitate them to survive chemotherapy, aiding in the survival of drug-resistant CSCs (102). ABC transporters can efficiently expel chemo drugs like anthracycline or taxanes out of the cells and can eventually lead to the acquisition of drug resistance phenotype in bCSCs ( Figure 3C ). This group of proteins is located on the cell membrane and thus can allow transmembrane transportation of different toxic molecules. Notably, CSCs exhibit a higher expression of ABC transporters, such as ABCB1 (MDR1), ABCG2 (BCRP1), ABCC11 (MRP8), and ABCB, which are positively correlated with CSC-mediated drug resistance (103, 104). The activity of drug efflux proteins can be monitored through the transport of fluorescent dyes like rhodamine and Hoechst 33342 (105). Based on this property, the CSCs can be isolated from non-CSCs by fluorescent-activated cell sorter (FACS). This fraction of cells termed as “side population” (SP), is identified as CSCs since these cells fall along the side of the cellular distribution on the FACS profile. Downregulating ABCG2 with wedelolactone-encapsulated PLGA nanoparticles increases the chemosensitivity of bCSCs (106). The Sox2-ABCG2-TWIST1 axis contributes to chemoresistance and stemness in TNBC, indicating the importance of ABCG2 as a potential bCSC-specific target in TNBC patients (107). Moreover, simultaneous blocking of ABCG2 and antiapoptotic gene BCL2 with SiRNA in bCSCs leads to better chemotherapeutic response to doxorubicin (108). Dofequidar, an ABC transporter inhibitor, increases the chemosensitivity of bCSCs in advanced or recurrent BC patients, when administered in combination with chemo drugs like doxorubicin, fluorouracil, and cyclophosphamide (109). ABC transporters not only participate in establishing the drug resistance via increased efflux of chemo drugs but also contribute to EMT (110). ABCB1 is another group of ABC transporter implicated in the chemoresistant nature of CSCs and induction of EMT (111, 112). Therefore, the combined application of chemotherapeutic drugs and ABC inhibitors should be employed to kill the bCSCs (111).

Although cancer cells show a reduced DNA damage repair (DDR) mechanism and reflect many mutations and genomic instability, CSC-like cells exhibit a highly dynamic DDR system that protects the DNA effectively (113). Both chemotherapy drugs and radiotherapy can induce DNA damage. Mechanistically, radiotherapy contributes to DNA damage through the production of water-derived free radicals and ROS that avidly interacts with DNA, protein, and lipids. The generation of ROS, in turn, switches on the DDR pathway. However, in contrast to non-CSCs, CSCs have a low ROS burden and augmented DNA repair system (refer to Figure 3D ). An increased level of ROS scavengers in CSCs maintains low levels of ROS that protect the CSCs from ROS-mediated DNA damage and apoptosis. The ROS scavenger, N-acetylcysteine, can restore both CSC and EMT phenotypes (114). A novel compound, salinomycin, can effectively target the CSC niche and kill CD44highCD24−/low bCSCs as it upregulates the ROS levels (115). The radioresistance property of CSC-like cells is linked with cell-cycle kinetics, which is reflected by a significant increase in the doubling time and Chk1/Chk2 basal activation level (116). The elongated cell-cycle window, therefore, offers more time to repair the genetic defects in CSCs. When the DNA defects are corrected, CSCs again enter into the cell cycle from the quiescent state and escape apoptosis. Therefore, targeting DDR could reverse therapeutic resistance. Frizzled 5 (FZD5), a member of the FZD family, contributes to DDR, G1/S transition, proliferation, stemness, and chemoresistance in TNBC (117). FZD5 knockdown suppresses the expression of CD133, ALDH, EpCAM, and Oct-4, thus potentially overcoming chemoresistance and recurrence in TNBC. Likewise, CCR5 directs the DDR mechanism and bCSC expansion (118). CCR5 antagonists, vicriviroc and maraviroc, can substantially increase cell death caused by DNA-damaging chemo drugs. MYC and MCL1 also cooperatively function in bCSC maintenance in TNBC patients via increasing ROS production and HIF-1α expression (119).

CSCs are a crucial driver in immune evasion, metastasis, and drug resistance. Substantial evidence suggests a reciprocal interaction between CSCs and immune cells, and that CSCs adopt different strategies to circumvent immune attacks mediated by different immune cell types within TME. This, in turn, contributes to CSC expansion and mediate protumorigenic immune function, leading to CSC-specific avoidance of immune detection and destruction of immune cells. Tumor-associated macrophages (TAMs), tumor-associated neutrophils (TANs), myeloid-derived suppressor cells (MDSCs), DCs, tumor-infiltrating lymphocytes (TILs), B cells, natural killer (NK) cells, and T regulatory (Treg) cells, and proinflammatory cytokines secreted by these cells, are therefore crucial for maintaining an immune-resistant phenotype. Treatment of bCSCs with conditioned medium from TAMs results in an upregulated expression of Oct3/4, Sox2, Nanog, and ALDH1 activity (120). MDSCs also lead to the enrichment of bCSCs via IL-6/STAT3 and NO/Notch signaling, leading to the suppression of T-cell activation (121). The T-cell inhibitory molecule, PD-L1 is overexpressed on the bCSC cell surface, compared with their differentiated counterparts, and is dependent on PI3K/AKT and Notch signaling pathway (122). Interestingly, PD-L1 expression is upregulated in response to EMT induction and facilitates the immune escape of bCSCs (123, 124). Moreover, bCSCs are not only resistant to chemotherapy but also immunotherapy (125). Therefore, restoring the T-cell activity by manipulating immune checkpoint molecules, targeting either PD-1/PD-L1 (nivolumab/pembrolizumab) or CTLA-4 (ipilimumab) can be an effective strategy ( Figure 3E ). Also, immunosuppressive cytokines secreted by breast CSCs (IL-4, IL-6, IL-8, IL-10, and IL-13) result in therapeutic resistance, increased EMT, metastasis, and recruitment of immunosuppressive immune cell types like Treg and MDSCs (126). A high circulating level of IL-6 contributes to disease recurrence, tamoxifen resistance in luminal BC, and trastuzumab resistance in HER-2 enriched BC. Targeting IL-6 receptor with monoclonal antibody tocilizumab, hence, suppresses metastatic potential of bCSCs and enhances the cytotoxicity of cisplatin against TNBC (127).

Like normal tissue stem cells, EMT and the reverse process MET are critical to CSC features. Substantial evidence exists that correlates EMT plasticity to the emergence of dedifferentiated cells with CSC phenotype, ultimately driving metastasis and drug resistance (128, 129). EMT inducers such as TGF-β and receptor tyrosine kinase (RTK) ligands modulate gene expression patterns through complex signaling networks (112). This results in the upregulation of transcriptional repressors like Snail, Slug, Zeb1/2, Twist, and E47. This group of proteins, then, interacts with the promoter sequence of adherens junction protein, E-cadherin, recruits histone deacetylases (HDACs), and induces its chromatin condensation, leading to transcriptional repression of E-cadherin. EMT involves the dissolution of cell–cell adherens junction barriers, loss of apico-basolateral polarity of epithelial cells, along with increased expression of mesenchymal markers such as fibronectin and vimentin ( Figure 3F ) (128). This, in turn, aids in gaining motile characteristics of cancer cells post-EMT. Mani et al. indicated that when the EMT program is induced in immortalized human mammary epithelial cells through ectopic expression of Twist, Snail, or TGF-β treatment, the cells exhibited mesenchymal appearances, developed many stem-like properties, and had the potential to form mammary tumors in mice (128). Importantly, the metastatic cancer cells with mesenchymal-like features generated post-EMT, exhibited CD44high/CD24low signature, and formed mammospheres whereas CD44low/CD24high cells could not. Al-Hajj et al. reported that disseminated BC cells found in pleural effusions are enriched in CD44highCD24−/low bCSCs (4). Interestingly, the EMT-associated emergence of bCSCs is induced by CD8+T cells that stimulate dedifferentiation of BC cells into bCSCs (130). An increased expression of stroma cell-related genes, attributed to the EMT program, could be linked to drug resistance in BC (131). Hence, blocking the EMT program can eventually interfere with CSC maintenance and innate or acquired drug resistance (70). Therapeutic intervention of micro-RNAs can provide an additional strategy to disrupt the EMT-CSC deadly axis. Application of HDAC inhibitors and “differentiation-inducing” agents are also believed to fetch clinical benefits to BC patients.

Paclitaxel, a first-line therapeutic agent for the treatment of metastatic BC, is a microtubule-stabilizing agent. It interferes with microtubule dynamic instability at nanomolar concentrations, thus leading to G2/M mitotic arrest and apoptosis in BC cells ( Figure 4A ). Paclitaxel resistance seems to be one of the primary obstacles that lead to chemotherapeutic failure in BC. HER2/β-catenin pathway mediates paclitaxel resistance in BC cells, and hence suppression of the HER2/β-catenin signaling can overcome paclitaxel resistance (132). Penfluridol (PFL) treatments that suppress both HER2/β-catenin pathways significantly inhibit the survival of paclitaxel-resistant BC cells. Notably, paclitaxel resistance increases both CD44+CD24− bCSC content and sphere-forming ability in the paclitaxel-resistant SUM159 metastatic TNBC cell line (133). According to this report, dasatinib, an Src family kinase inhibitor, induces epithelial differentiation of mesenchymal TNBC cells and sensitizes TNBC cells to paclitaxel therapy through targeting bCSCs. An increased association of ALDH1 expression has been noted in paclitaxel-resistant BC patients (134). TNBC cell lines, such as MDA-MB-231, SUM-149, and SUM-159 show an enhanced activity of hypoxia-inducible factors (HIFs) and their target gene products, with chronic exposure to paclitaxel therapy. Furthermore, chemotherapy-induced HIF activation results in bCSC enrichment through IL-6 and IL-8 signaling and enhanced expression of MDR-1 (135). Hence, combinatorial therapies including HIF inhibitors along with paclitaxel chemotherapy are being tested in clinical trials to explore its efficacy in BC patients.

Platinum-based chemotherapeutic drugs, such as cisplatin, oxaliplatin, carboplatin, nedaplatin, and lobaplatin are frequently the choice of drugs for treating advanced BC cases, including TNBC (136, 137). Mechanistically, platinum-based drugs interact with guanine and adenine nucleotides, forming platinum-DNA nonfunctional adducts, disrupting DNA double-helical structure, and eventually inhibiting cell division ( Figure 4B ) (138). However, drug resistance associated with platinum therapy and the numerous side effects that it causes have been a long-standing concern for BC patients. According to Sledge et al., only 47% of the BC patients with metastatic BC are partially sensitive to platinum therapy (139). Several lines of evidence indicate a crucial contribution of bCSCs in developing and maintenance of platinum resistance. In this context, a novel drug disulfiram (DSF) can reverse cisplatin resistance in different BC cell lines through inhibiting ALDH enzymatic activity and interfering with the expression of Oct4, Sox2, and Nanog in bCSCs. IL-6 secreted by breast tumor-derived mesenchymal stem-like cells (MSCs), augments cisplatin resistance via STAT3 signaling (140). Although neutralizing IL-6 can partially interfere with the IL6-STAT3 axis and reverse the cisplatin resistance, the specific role of bCSCs remains unclear. Notably, there is the active involvement of PI3K/AKT/NF-ĸB signaling in enrichment as well as maintenance of breast CSCs. Cisplatin is known to stimulate transcriptional upregulation of PI3KCA, thereby triggering PI3K/AKT signaling in platinum-resistant cells. However, a recent report emphasizes that mechanistically cisplatin leads to CSC enrichment in platinum-resistant cells through the NF-ĸB-TNF-α-PI3KCA loop (141).

Anthracyclines, antibiotics extracted from Streptomyces bacteria, administered as broad-spectrum chemotherapeutic drugs in BC patients, are topoisomerase II inhibitors ( Figure 4C ) (142). Since, topoisomerases modify DNA topology by breaking or rejoining DNA double strands, inhibiting its catalytic activity stabilizes DNA breaks, eventually causing cell death. This class of drugs includes doxorubicin, daunorubicin, and epirubicin. Resistance to anthracyclines has been linked with multiple factors, such as the acquisition of MDR due to overexpression of drug efflux pumps and permeability glycoprotein-1, alteration of topoisomerase II activity, CSC enrichment, altered DNA repair, and metabolic reprogramming (143). A 4-day exposure to doxorubicin and paclitaxel, followed by a 2-day recovery, leads to significant enrichment of CD44highCD24−/low bCSCs (144). In this context, cardamonin, a small molecule, significantly prevents bCSC enrichment, when administered along with chemotherapeutic drugs, via downregulation of IL-6, IL-8, NF-ĸB, and STAT3 signaling (144). Knockdown of Annexin A3 also influences the drug sensitivity of bCSCs to doxorubicin via an upregulation of drug uptake, inhibits metastasis, and exhibits a change in heterogeneity and plasticity in bCSCs (145).

ER-α-positive BC cases constitute around 70%–75% of overall BC incidence. Although, tamoxifen (TAM) has been the fundamental mode of endocrine therapy for the treatment of ER+ BC patients for the last three decades, acquired TAM resistance is frequently held accountable for the disease relapse (155). TAM competitively inhibits the interaction of estrogen ligand with ERs ( Figure 4E ). Most importantly, an increased proportion of bCSCs in advanced BC patients have a potential contribution to TAM resistance and breast tumorigenesis (156). Poorly differentiated breast tumors contain a higher percentage of CSC-like cells than well-differentiated breast tumors (157). TAM-resistant BC cells retain stem-like properties (158). Notably, TAM-resistant MCF-7 cells showed increased proliferation rate, enhanced mammosphere formation ability, increased mRNA expression of OCT-4, SOX-2, and CD133, and increased EMT signature, compared with wild-type MCF-7 cells (158). In a parallel study, Wang et al. indicated that TAM-resistant MCF-7 cells contain a higher proportion of CD44+CD24−/low bCSCs, exhibit lesser sensitivity to Adriamycin compared with wild-type MCF-7 cells, and express SOX-2 as a biomarker for TAM resistance (159). Serine phosphorylation, particularly at Serine 118, has been documented for activating the N-terminal transcriptional function of ER-α. SOX-2 can reprogram the non-genomic estrogen signaling and augment bCSC proportion through phosphorylation of ER-α at serine 118, making it hypersensitive to circulating estrogen (160). Phosphorylation, ubiquitination, and other posttranslational modifications play an important role in activating ER and its coregulators and can influence the sensitivity to different endocrine therapies (161). Therefore, inhibition of SOX-2 could restore the sensitivity of BC cells to TAM (162). Furthermore, ER splicing variants, including estrogen-related receptors (ERRs), and the recently identified estrogen receptor α-variant (ER-α36) are involved in TAM resistance and estrogen hypersensitivity (163). However, the contribution of ER-β in bCSC-mediated TAM resistance is still under investigation. Upregulation of different growth factors including HER2, epidermal growth factor receptor (EGFR), and insulin-like growth factor 1 receptor (IGF1R) has been documented in BC endocrine resistance, although direct evidence has been found in support of PI3K-mediated TAM resistance (164). Hence, targeting PI3K and IGF1R is considered a major therapeutic target to reverse TAM resistance in bCSCs. Different signaling pathways, such as Wnt and Notch, induce TAM resistance, promoting bCSC activity in TAM-resistant MCF-7 cells, while inhibition of these pathways could overcome TAM resistance (165, 166). A positive correlation has been noted between activation of Hedgehog (Hh) signaling and reduction in disease-free or recurrence-free survival in BC patients, which can even result in TAM resistance (167). The intervention of Hedgehog (Hh) signaling, thus, can potentially interfere with bCSC proliferation, migration, and invasion and reverse TAM resistance. Therefore, inhibition of Notch, Hedgehog (Hh), and Wnt/β-catenin signaling pathways should serve as another strategy to overcome TAM resistance in bCSCs.

Fulvestrant is a selective estrogen receptor degrader (SERD) administered in both first and subsequent lines of treatment in ER-α+ metastatic BC patients (149, 168). Fulvestrant competitively inhibits estrogen to occupy the ER, eventually promotes degradation of the receptor, and thus interferes with estrogen signaling in breast tumor tissues (169). Unfortunately, there has not been extensive research done on fulvestrant resistance in bCSCs as well as the molecular mechanisms responsible for the resistance. Dysregulation of both G protein-coupled estrogen receptor-1 (GPER) and CDK6 are associated with fulvestrant resistance in BC (170). Notably, GPER-induced signaling is essential for the survival of bCSCs (171). Very recently, Kaminska et al. reported that cyclin E2 overexpression has been recognized as a biomarker for persistent fulvestrant-resistant metastatic BC and reduced disease-free survival (172). However, AI-resistant BC cells, having a higher proportion of bCSC-like cells and increased stemness, are inhibited by fulvestrant (173). Several signaling pathways, such as MEK/ERK, NF-ĸB, EGFR, PI3K/AKT, and β-catenin have been implicated so far to fulvestrant resistance in BC. MiRNA-221/222 confers estrogen-independent growth and fulvestrant resistance in BC through multiple signaling networks. Strikingly, miR-221/222 contributes to acquired fulvestrant resistance through activation of the β-catenin pathway, and miR-221/222 has recently been documented in CD44+CD24−/low bCSCs (174).

AIs constitute the first-line therapeutic approach for the treatment of ER+ BC in postmenopausal women (175, 176). AIs deplete the circulating level of estrogen in the human body by interfering with estrogen biosynthesis through blocking aromatase activity (177). Hence, in the presence of AIs, estrogen production is inhibited, which slows down tumor progression in ER+ BC settings. When treating early-stage ER+ BC, AIs are frequently the choice of hormonal therapy over TAM due to the fewer side effects it causes. However, acquired AI resistance may develop in over 20% of early-stage BC patients and found to be inevitable in metastatic BC patients (178). Acquired AI resistance involves a switch from dependence on ER signaling to growth-factor-mediated signaling, such as HER2 signaling (179). Both cancer-cell intrinsic (enhanced activity of FGFR, ERBB2, and IGF1R and the downstream signaling of PI3K-AKT-mTOR and MAPK pathways) and extrinsic mechanisms (interaction of TME with other cell types) cumulatively coordinate the development and maintenance of AI resistance (180). AIs are classified into 2 subtypes—steroidal (type I) and nonsteroidal (type II). The three different AIs anastrozole (nonsteroidal), letrozole (nonsteroidal), and exemestane (steroidal) are being used in adjuvant therapy as the first line of treatment modality for both early and metastatic BC in postmenopausal women. BCSCs in ER-α+ settings reflect an activated PI3K signaling, which confers endocrine resistance including AI resistance (181). Notably, different PI3K inhibitors such as alpelisib, buparlisib, and taselisib (https://clinicaltrials.gov/ct2/show; ClinicalTrials.gov Identifier: NCT02437318, NCT01610284, and NCT02340221) are being administered as novel therapeutic drugs in phase III clinical trials for the treatment of breast cancer AI resistance ( Table 1 ). The expression of HIF-1α has also been recognized as a biomarker and therapeutic target that promotes AI resistance (179).

Amplification of ERBB2 (HER2) is associated with clinically aggressive breast tumors, shorter disease/recurrence-free survival, and poor overall survival (183). Trastuzumab is a recombinant humanized IgG1 monoclonal antibody that interacts with extracellular domain IV of ERBB2, inhibiting dimerization between ERBB2 and other EGFR family members ( Figure 4F ) (184). Although HER2+ BC responds quite well to trastuzumab (Herceptin™) therapy plus chemotherapy in the early stages of the disease, acquired resistance, however, to trastuzumab after 1–2 years of treatment is a frequent event following metastasis (185). Factors like HER2 degradation, overexpression of other RTKs, mutation of PI3KCA (PI3K catalytic subunit p110α), and loss of Phosphatase and Tensin Homolog Deleted on Chromosome 10 (PTEN) tumor-suppressive function have been linked with trastuzumab resistance (186). Continued application of trastuzumab in HER2+ cells with loss of PTEN encourages EMT and transforms HER2+ BC to TNBC (187). Strikingly, these transformed cells frequently exhibit mesenchymal features along with mesenchymal-specific gene expression profile, although the parental HER2+ cells show epithelial morphology with epithelial-specific gene signature. Since bCSCs exhibit chemoresistance to small-molecule targeted therapy, exploring the mechanism of trastuzumab resistance must have clinical implications. BCSCs confer drug resistance by activation of different prosurvival pathways, such as PI3K/AKT, NFĸB, and JAK/STAT pathways (188). Thus, CD44+CD24−/low bCSC phenotype serves as a prognostic factor for clinical outcome and predictive factor for poor trastuzumab response in patients with HER2+ BC. Importantly, PI3K/AKT/mTOR activation has been implicated in both de novo and acquired trastuzumab resistance (189). Since PTEN loss and mutation of PI3KCA lead to aberrant downstream activation of PI3K/AKT/mTOR pathway, which in turn sustains bCSC population, both the factors correlate with trastuzumab resistance (190). Therefore, combining PI3K/AKT/mTOR inhibitors along with HER2 targeting drugs to overcome trastuzumab resistance provides an active area of research. Pan-class I PI3K inhibitors, such as buparlisib and pilaralisib when administered with trastuzumab (189), lapatinib (191), or trastuzumab and paclitaxel (192), are proven to be safer and successful in HER2+ advanced stage BC patients. IL-6-mediated bCSC expansion is another independent mechanism resulting in trastuzumab resistance (36). Moreover, STAT3 activation also stimulates breast cancer stem-like properties resulting in HER2 overexpression and trastuzumab resistance (193). Hence, targeting JAK/STAT3 pathway or administering IL-6 receptor-targeted antibody should overcome the trastuzumab resistance by reducing the bCSC burden. Additionally, CD47 blockade with trastuzumab also eliminates HER2+ BC cells, overcoming trastuzumab tolerance (194).

Lapatinib is a reversible and selective receptor tyrosine kinase inhibitor, targeting both epidermal growth factor receptor (EGFR) and HER2 (195). In contrast to trastuzumab, lapatinib blocks kinases’ active ATP-binding site, thus interfering with receptor phosphorylation ( Figure 4G ). However, despite the initial response in HER2-overexpressing BC, acquired resistance to lapatinib turns out to be a frequent event in the course of treatment. Liu et al. have isolated and characterized several lapatinib-resistant HER2/ER+ BC clones from lapatinib-sensitive BT474 cells through chronic exposure to lapatinib. This group has identified that activation of AXL is associated with lapatinib resistance in these resistant BT474 clones (196). Evidence indicates a close association of breast CSCs in exerting lapatinib resistance. In this context, a recent study suggests that miR-205-5p is highly expressed in bCSCs. miR-205-5p represses ERBB/HER receptors in bCSCs, leading to resistance to targeted therapy (197). Silencing miR-205-5p in bCSCs, followed by lapatinib treatment, significantly reduces BC proliferation, resensitizing BC cells toward EGFR/anti-HER2 treatments. Furthermore, knockdown of miR-205-5p by locked oligonucleotides significantly reduces EMT and metastatic potential exerted by bCSCs (198). TGF-β-SMAD3 signaling also contributes to trastuzumab and lapatinib resistance, maintaining CSC phenotype in HER2+ settings (199). CD24 supports the expression of HER2 along with activation of PI3K/AKT signaling, resulting in lapatinib resistance (200). Hence, small-molecule inhibitors of SMAD3, or targeting CD24, can attenuate lapatinib resistance and increase the sensitivity of HER2+ BC cells to lapatinib.

Deregulated Notch signaling in bCSCs represents poor clinical outcomes in drug-resistant BC. Evolutionarily conserved Notch signaling is linked with cell differentiation and cell fate decisions. Notch signaling pathways include 4 receptors (Notch1–4) and 5 ligands such as delta-like ligand (DLL)1, DLL3, DLL4, JAG1, and JAG2 (230, 231). Interaction with the Notch ligand leads to the release of its intracellular domain (NICD), which then translocates to the nucleus and impacts gene expression in association with different transcription factors. Studies have established links between bCSCs, aberrant Notch signaling, and radio-/endocrine-/chemoresistance. A significantly higher expression of activated Notch1 is noted in the culture media of bCSCs, postradiation (66). A substantial induction of the JAG1 ligand is also evident on the surface of nonadherent bCSCs after fractionated radiation (232). Both JAG1 and Notch pathway contributes to chemoresistance in BC metastasizing to bone (233). Moreover, suppression of Notch1 signaling enhances antitumor efficacy of chemotherapy agents via reduction of bCSCs in TNBC (234). Notably, Notch ligand DLL1+ quiescent bCSCs drive chemoresistance via NFĸB pathway in BC (235), and disease progression in ER+ BC is dependent on DLL1-mediated Notch1 signaling in bCSCs (236). Notch1 ligands, JAG1, and JAG2 are also overexpressed in endocrine-resistant luminal BC, resulting in an increased bCSC activity (237). Therefore, the blockade of Notch signaling is of clinical importance to eradicate resistant bCSCs and offer long-term disease-free survival.

PI3K is a family of lipid kinases that phosphorylate phosphatidylinositol (PI) at the intracellular membrane and plasma membranes. An increased PI3K/AKT/mTOR signaling in bCSCs has been documented over the years, contributing to survival, proliferation, metastasis, and drug resistance in BC cells (238, 239). Mutations, specifically in its catalytic domain, p110α, are the most frequent genetic events, affecting around one-third of BC patients. Alterations in the PI3K/AKT/mTOR pathway in bCSCs result in the TAM resistance in ER+ BC (240, 241). The interaction between PI3K and Wnt/β-catenin pathway is responsible for stemness and self-renewal abilities of bCSCs (242). Therefore, small-molecule inhibitors targeting the key players, PI3K, AKT, and mTOR can reverse the drug resistance and self-renewal abilities of breast cancer stem-like cells. Pan-PI3K inhibitors, such as buparlisib and pictilisib (inhibiting p110α/β/γ/δ); PI3K isoform-specific inhibitors such as alpelisib and taselisib (inhibiting p110α and p110α/γ/δ, respectively); AKT inhibitors such as ipatasertib, capivasertib (AZD5363), and vevorisertib (MK-2206); PI3K/AKT dual inhibitor gedatolisib (PF-05212384); and mTOR inhibitors such as everolimus, vistusertib, and sapanisertib are currently available for the treatment of BC (243). B591, a novel PI3K inhibitor, has shown promising results in targeting breast CSCs in the mouse xenograft model, affecting both its self-renewal potential and EMT (244). However, despite substantial preclinical evidence, the innate and acquired resistance has limited the application of this group of inhibitors in BC.

Wnt/β-catenin signaling contributes to self-renewal, migration, and invasion of bCSCs, leading to systematic dissemination in BC. A significantly higher level of Wnt/β-catenin signaling is noted in bCSCs compared with the bulk of the tumor (245). Hence, Wnt/β-catenin signaling serves as a novel target for restricting BC progression. A highly potent small-molecule inhibitor CWP232228 can preferentially inhibit bCSC proliferation via antagonizing the binding of β-catenin to T-cell factor (TCF) in the nucleus (246). Another natural product, gomisin M2, downregulates Wnt/β-catenin signaling and inhibits bCSC proliferation, mammosphere formation, and self-renewal (247). Studies indicate multiple interaction points or crosstalk between Notch and Wnt/β-catenin signaling pathways, and thus it is essential to focus on Notch-Wnt synergies in BC progression (248). In a normal mammary setting, in response to Notch ligand DLL1, macrophages express Wnt ligands (Wnt3, Wnt10A, and Wnt16), important for mammary stem cell numbers and activity (249). Thus, the proteins exerting regulatory effects on both these pathways should serve as a novel therapeutic target and targeted in BC. GSK3β is one such protein that regulates β-catenin stability as well as phosphorylates Notch ICD (250).

The Hh signaling is another novel target in BC since it is frequently upregulated in bCSCs and contributes to CSC self-renewal and stemness maintenance. The cancer-associated fibroblasts (CAFs) within TME support the maintenance of CSC function in breast tumors via their regulation of both Wnt/β-catenin and Hh signaling. Briefly, CAFs promote BC progression through proliferation, invasion, matrix remodeling (via matrix production and crosslinking, matrix stiffness, force-mediated matrix remodeling), macrophage, and endothelial cell crosstalk (via secretion of VEGF, exosomes, HGF production), chemoresistance, and immunosuppression (251–253). Notably, BC shows divergent CAF phenotypes, including FAP-positive (fibroblast-activating protein α1) CAFs driving immunosuppression and resistance to PD-L1 therapy (254). According to Friedman et al., two distinct subpopulations of CAFs (S100A4+ and PDPN+) exist in human breast tumors, where their ratio decides the clinical outcomes across subtypes and is highly correlated with BRCA mutations in TNBC (255). The interaction between the breast cancer cell and fibroblasts also induces the CAF phenotype through activation of Notch signaling (256). Hence, understanding the full repertoire of CAFs and the dynamic changes as breast tumors evolve can improve the precision of treatment and reverse drug resistance. BCSCs secrete the Hedgehog ligand, SHH, which controls CAFs through activation of Hh signaling (257). The CAFs, in turn, secrete some factors that result in the expansion and self-renewal of bCSCs. Therapeutic targeting of CAFs using the inhibitor molecule of Smoothened, the main effector molecule of the Hh pathway, sensitizes TNBC xenograft models to docetaxel (258). Tetraspanin 8 (TSPAN8), a membrane glycoprotein, enhances BC stemness by activating SHH signaling (259). Activation of Hh signaling results in salinomycin resistance in tumor spheres, generated from the MCF-7 cell line (260). However, the inhibition of the Hh pathway by cyclopamine can sensitize the MCF-7 cells to paclitaxel. Therefore, exploring the detailed mechanisms of Hh-driven bCSC signaling can help in the designing of novel drug candidates to reverse BC drug resistance.

DCs loaded with CSC lysate or mRNA, administered as vaccines, are capable of eliciting cancer-specific immune responses (279). Notably, DCs are the professional antigen-presenting cells (APCs) that process the antigenic material and present peptide antigens to T cells and activate them. Interestingly, DCs in BC patients exhibit decreased antigen uptake, reduced antigen processing, reduced expression of costimulators, weak migration profile, along decreased IL-12 production (280, 281). Patients with advanced breast and ovarian cancers can be successfully vaccinated by DC loaded with HER2/Neu- or MUC1-derived antigenic peptides (282). Phase-I clinical trial with metastatic BC showed that fusion of breast tumor cells with DCs resulted in immunological and clinical antitumor responses (283). DC pulsed with breast tumor lysate has proved to be a standard method for a source of BC antigen, capable of eliciting anticancer immune responses (284). BCSC-RNA-pulsed DC vaccine also effectively kills breast tumor cells through activation of CD4+ T_h lymphocytes and CD8+ cytotoxic T cells. This study highlights the efficacy of bCSC-RNA for priming DC cells in evoking immune response against drug-resistant CSC populations. However, DC-based vaccines present a few drawbacks; they are both cost-effective and time-consuming for patient-specific treatment. Factors like antigenic peptide or CSC-RNA loading on DC, route of administration, and doses are yet to be standardized to resolve these technical limitations.

Adoptive T-cell therapy is a personalized mode of immunotherapy to target CSCs. Here, tumor immune lymphocytes (TILs) with intrinsic antitumor activity are isolated from cancer-bearing patients. Following isolation, TILs are cultured in the presence of IL-2 so that they can recognize tumor-associated antigens on cancer cells, and eventually release cytotoxic cytokines, perforin, and granzymes (285). The recent approach focuses on designing CAR-T cells against the CSC surface antigens in different cancer models to achieve complete regression of tumor (286). CARs generally constitute the extracellular binding domain, a single-chain variable fragment (scFv) specific for a tumor antigen, an extracellular spacer domain, a transmembrane domain, followed by an intracellular signaling domain ( Figure 5D ). EGFR-specific CAR-T cells have shown promising results in high EGFR-expressing TNBC cell lines and patient-derived xenograft mouse models (287). Another recent report highlights promising results for HER2-specific second-generation CAR-T therapy for the treatment of breast-to-brain metastasis (288). Other bCSC markers targeted by CAR-T therapy include c-Met, CD133, CD166, CD47, EpCAM, and LGR5 (55, 289, 290). Despite the remarkable clinical success of CAR-T therapy in hematologic cancers, its application is limited in solid tumors. Due to the lack of chemokine expression required for the infiltration of CAR-T cells into the tumor tissues and dense fibrotic matrix in solid tumors, the ability of CAR to get recruited at the tumor site and infiltrate is considerably affected (291). Frequently, CAR-T cells fail to penetrate the tumor tissues through the vascular endothelium (292). Therefore, instead of systemic administration, regional administration of CAR-T cells in solid cancers will be more effective. Altogether, the major limitation of this approach includes burdensome and expensive preparation to isolate the patient-derived T cells and the major side effects resulting from cytokine release syndrome.

Oncolytic viruses (OVs), a novel class of DNA/RNA-attenuated viruses that selectively infect, replicate inside the tumor cells, and eventually kill them either through modulating the TME or via antitumor response ( Figure 5C ) (293). These naturally occurring or genetically engineered viruses have the potential to convert an “immunologically cold” TME into an “immunologically hot” one by increasing the net influx of TILs, consisting of CD4+ and CD8+ T cells, B cells, and NK cells (293, 294). Activated CD8+ cytotoxic T cells and NK cells are associated with a good prognosis, whereas the presence of Foxp3+ Treg cells within the breast TME is associated with a poor prognosis, due to their role in immunosuppression. Immunologically “cold” tumors exhibit a low mutational burden, poor MHC presentation of tumor antigen, poor migration of TILs, and also have reduced expression of PD-L1 on the surface of tumor cells, thus making the response to ICIs inadequate (295). Interestingly, OVs induce a strong antiviral tumor immune response through the production of cytokines like type-1 interferon that in turn promotes PD-L1 expression on tumor cells and also cytokines, such as CCL3 and CCL4, attracting PD-1+ or CTLA-4+ immune cells within TME (296). Eriksson et al. indicated that OVs, Ad5/3-Delta24 and Ad5.pk7-Delta 24, can selectively kill CD44+ CD24−/low bCSC population. Oncolytic herpes simplex virus, oHSV G47Δ, effectively kills bCSCs both in vitro and in vivo, derived from SK-BR-3 and primary human BC cells (297). A randomized phase II study by Bernstein et al. reported that the combination of oncolytic reovirus (pelareorep) with paclitaxel significantly increased survival of metastatic BC patients (298). Combining pelareoprep with paclitaxel, along with anti-PD-L1 antibody, avelumab (NCT04215146) is presently undergoing phase II clinical study in BC patients (299).

Immune checkpoint ligands such as PD-L1 and PD-L2 are highly expressed on CSCs. The immune cells, on the other hand, express the receptor for these ligands, PD-1. Now, the interaction between PD-1 and PD-L1/PD-L2 interferes with T-cell proliferation and activity, leading to tumor immune suppression, thus serving as a strategy to immune escaping of CSCs (300). Therefore, immune checkpoint blockade of PD-L1/PD-L2 is emerging as a novel therapeutic approach, whereby these CSC-specific ligands are engaged by ICIs, thus making it possible to target CSCs for programmed cell death. Notably, multiple clinical trials on ICIs, targeting CTLA-4, PD-1, and PD-L1 are in progress that are either administered as a single agent or in combination with trastuzumab or with chemotherapeutic drugs, in HER2-enriched and TNBC settings, respectively. In March 2019, FDA has approved the clinical application of anti-PD-L1 antibody, atezolizumab, in combination with nab-paclitaxel, to be administered as the first-line therapy to metastatic or locally advanced PD-L1+ TNBC patients (301). A recent phase Ib clinical study by Nanda et al. explored the antitumor efficacy and safety profile of PD-1 inhibitor molecule pembrolizumab in advanced TNBC patients (302). Furthermore, certain drugs that stimulate PD-L1 degradation can be administered as a combination therapy with ICIs to significantly enhance the efficacy of cancer immunotherapy (303).