The Role of Breast Cancer Stem Cells as a Prognostic Marker and a Target to Improve the Efficacy of Breast Cancer Therapy

Breast cancer is the most common form of tumor in women and the leading cause of cancer-related mortality. Even though the major cellular burden in breast cancer is constituted by the so-called bulk tumor cells, another cell subpopulation named cancer stem cells (CSCs) has been identified. The latter have stem features, a self-renewal capacity, and the ability to regenerate the bulk tumor cells. CSCs have been described in several cancer types but breast cancer stem cells (BCSCs) were among the first to be identified and characterized. Therefore, many efforts have been put into the phenotypic characterization of BCSCs and the study of their potential as prognostic indicators and therapeutic targets. Many dysregulated pathways in BCSCs are involved in the epithelial–mesenchymal transition (EMT) and are found up-regulated in circulating tumor cells (CTCs), another important cancer cell subpopulation, that shed into the vasculature and disseminate along the body to give metastases. Conventional therapies fail at eliminating BCSCs because of their quiescent state that gives them therapy resistance. Based on this evidence, preclinical studies and clinical trials have tried to establish novel therapeutic regimens aiming to eradicate BCSCs. Markers useful for BCSC identification could also be possible therapeutic methods against BCSCs. New approaches in drug delivery combined with gene targeting, immunomodulatory, and cell-based therapies could be promising tools for developing effective CSC-targeted drugs against breast cancer.


Introduction
Breast cancer is the highest incidence cancer and the leading cause of cancer-related mortality among women [1]. It cannot be considered a single disease because genetic and genomic variability together with clinicopathological features determine different stages and prognoses [2]. Even though in breast cancer the major cellular burden is constituted by the so-called bulk tumor cells, other cell subpopulations with stem features, self-renewal capacity, and the ability to regenerate the bulk tumor cells can be identified [3]. Because of similarities with the stem progenitors of normal tissues, these cells have been defined cancer stem cells (CSCs). They have been described in several cancers but breast cancer stem cells (BCSCs) were among the first to be identified and characterized [4].
The first evidence of their role in tumorigenesis were inferred from the injection into a xenograft mouse model. CSCs were able to recreate the tumor while other cellular subtypes from the tumor bulk Similarly, the genotypic analysis of CTCs for mutations of estrogen receptor gene ESR1 [43] for the altered HER2 expression [41], for mutations in the PIK3CA gene [44], and for the expression of stemness markers, as with the ALDH-1 gene [45], is gaining ever more attention.
The different cellular populations among CTCs, which recapitulate those present in the primary tumor, are another critical point that should be considered. It is possible to distinguish epithelial CTCs, mesenchymal CTCs, and CTCs with a staminal phenotype, which could be both epithelial-mesenchymal (EM) or mesenchymal-epithelial (ME) [19,46,47]. In this sense, a merely quantitative evaluation of the CTCs is not able to distinguish the prognostic meaning of the different CTC subpopulations.
CTC analysis can give a better insight into CSCs and their dynamic evolution during a cancer's natural history. The identification of stemness and EMT markers on CTCs could possibly help us in order to create an in depth profile of the primary tumor and the metastases and to target more precisely those cell populations, which are mostly responsible of the resistance to therapies, of the disease dissemination, and ultimately of a worse prognosis [4,24,48]. For instance, 41% of patients with aggressive breast cancers (triple negative or HER2 + tumors) presented mesenchymal CTCs, which are not measured by the standard systems [19].
AKT2, PI3Kα, and twist-related protein 1 (TWIST1) are the three main markers that are expressed by the epithelial-mesenchymal (EM) CTCs. These markers were evaluated by several groups in order to identify EM-CTCs [24,49,50]. Patients with nodal involvement and metastatic breast cancer had a higher presence in systemic circulation of CTCs with the EM phenotype [51,52]. An observational study analyzed epithelial CTCs, EM CTCs, and purely mesenchymal CTCs on 56 metastatic breast-cancer patients [53]. Both epithelial CTCs and EM CTCs were significantly associated with a poorer OS, while only EM CTCs were correlated with a reduced PFS.
Another multicentric prospective study used a microfluidic system for the detection of different CTC populations and tried to establish a threshold in order to predict the PFS after one year [54]. The patients with a total number of CTCs equal or superior to 10 and with a proportion of mesenchymal CTCs greater than 10.7% had a worse median PFS compared to the patients who did not meet these criteria.
Nonetheless, the methods used to count the different CTC populations are yet to be standardized and validated, and their prognostic and predictive value is unclear.
Among CTCs, the main markers associated with circulating CSCs are ALDH-1, CD44, and CD24, which are able to identify the cells with a higher metastatic potential [17,55,56]. The breast CSCs can switch easily in between the EM phenotype (which is EpCAM − CD49f + and expresses the CSC markers CD44 + /CD24 − ) and the ME phenotype (which is EpCAM + CD49f + and expresses the CSC marker ALDH-1 + ) [20]. Circulating CSCs with the EM phenotype can be identified with EpCAM-based systems, but it is not possible to distinguish them from the purely epithelial CTCs without further analysis [6]. Matrigel invasion assays with these two circulating CSC subpopulations (EM-CSCs and ME-CSCs) showed that CSCs with an EM phenotype have a greater invasive capacity than ME CSCs [20]. This could be in favor of the theory that hypothesizes that CSCs from the primary tumor undergo EMT on the invasive front and enter the circulation, thus spreading to distant sites. Micro-metastases are quiescent until these CSCs revert to a mesenchymal/epithelial, "self-renewing" phenotype, and originate the new bulk tumor. Nonetheless, the clinical impact of these markers expressed on CTCs remains unclear.
The analysis and the dosage of circulating tumor DNAs and micro-RNAs are other promising options in order to profile CSCs and to evaluate the minimal residual disease [57]. A Taiwanese study dosed the levels of two micro-RNA (miR-9 and miR-221) in 206 patients. MiR-9 and miR-221 have been associated with stemness features, elevated metastatic potential, and EMT activation. In this study, high levels of MiR-9 and miR-221 were independently associated with a poorer OS and DFS after 8 years of follow-up [58]. Nonetheless, the prognostic and predictive values of these markers are still far to be fully understood [59].

Analysis of the Cancer Stem Cells in the Primitive Tumor
Even though it is easier to access and to monitor blood parameters like stem CTCs over time, the analysis of the tumoral specimens still remains of pivotal importance in order to characterize CSCs and try to understand disease prognosis and the possible response to therapies. ALDH-1, CD44, and CD24 were among the first markers that were evaluated regarding primary tumors through immunohistochemistry in order to identify the CSC population, even though the data about their clinical prognostic value are still contradictory. Quantitative immunofluorescence for ALDH-1 + and CD44 + /CD24 − cells was retrospectively evaluated on 639 patients with 12.6 years of follow up. Co-expression of these markers correlated significantly with a worse outcome independently of the tumor size, grade, nodal status, and HER2/neu and receptor status, while the ALDH-1 alone did not significantly predict an outcome [60]. In a study on 144 patients with invasive ductal carcinoma, neither immunohistochemical ALDH-1 + nor CD44 + /CD24 − correlated with a difference in OS [61]. In another study on 121 patients, a positive immunohistochemical staining, both for ALDH and CD44/CD24, was evaluated. A positive staining for ALDH-1 was significantly correlated with a higher rate of metastasis or recurrence [62]. The staining for CD44 + /CD24 − cells was not significantly associated per se with metastasis or recurrence even though a higher proportion of these cells in the tumor showed a significant association with metastatic disease and recurrences.
Other authors highlighted that ALDH-1 is a marker of invasiveness and metastatic potential, while the CD44 + /CD24 − ratio indicates mainly a "self-renewal" capacity, thus these two markers are assigned different functions during the tumor progression. Therefore, they advocated always combining the use of both the markers for the sake of a better understanding of the stem population [63].
Many authors tried to develop a molecular "fingerprint" of the tumor in order to have a deeper comprehension of CSC role in breast cancer evolution and to draw reliable prognostic conclusions. Gwak et al. analyzed the tumoral expression of several transcription factors proper of the embryonic stem cells, including Oct4, Sox2, Nanog, Bmi1, and Klf4 [64]. Expression of Oct4 correlated with ALDH-1 positivity, a high Ki-67 and a high histological grade, and it was an independent prognostic factor for a reduced DFS. These associations were found specifically in the hormone receptor (HR)-positive group and in the HR-positive patients in treatment with tamoxifen. This is possibly another hint in favor of a previously suggested association between Oct-4 and tamoxifen resistance [65].
It was proposed that one should combine different multi-gene prognostic signatures that address 17 specific genes (HTICS), each of them with roles in three different major pathways in the tumor biology: immune response, cell migration, and cell proliferation [66]. These genes were selected among those expressed more by the tumor-initiating cells than by the non-tumor-initiating ones, thus obtaining a prognostic tool that was able to predict metastasis-free survival (MFS) and OS in HER2 + /ERα − cancer patients. Despite the reduction of the signature to a six-gene panel, the authors highlighted that there was still a significant prognostic value, even if reduced.
The importance of the selection of a genetic signature of CSCs and not only of the bulk tumor has been recently demonstrated and clinically validated for a panel of 20 genes in order to add prognostic and predictive value to clinical models. [67]. This 20-gene signature was selected among a set of stem-cell-specific genes overexpressed in mammary stem cells. The genetic signatures derived from bulk tumoral cells frequently overlapped clinicopathological features, thus reducing their prognostic ability in cases like triple-negative breast cancers, which lack expression of hormone receptors and have a high proliferation rate. This panel was clinically validated on a cohort of 2453 breast cancers and it was able to predict the risk of distant metastases in triple-negative (TNBC) and luminal breast cancers, independent of standard clinicopathological parameters.
A better understanding of CSCs both circulating and quiescent in the primitive tumor bulk can lead to a better prognosis prediction, therapy allocation, and ultimately to the development of targeted treatments for these cells, which are often resistant to conventional chemotherapy.

Breast Cancer Stem Cell-Targeting: New Strategies in Drug Development for Therapy Resistance
Conventional drugs targeting the tumor bulk are ineffective at eradicating CSCs [7]. In particular, it has been reported that anti-mitotic agents, such as taxanes (paclitaxel and docetaxel), cannot target quiescent CSCs inside the tumor bulk [23], leading to the reconstitution of the initial tumor cell population, increasing the adhesiveness of CTCs and the disease progression [68,69]. Based on this evidence, some studies have focused on directly targeting CSC subpopulation with promising results from preclinical experiments and clinical trials [24][25][26]. Novel therapeutic strategies for BCSC-targeting are based on the combined use of new and conventional drugs [27]. In particular, the main targets for BCSCs are Notch, Hedgehog, Wnt, PI3K/Akt/mTOR, IL-8, HER2, and TGF-β signaling, which are implicated in BCSC self-renewal and survival [27]. Moreover, the application of nano and biotechnologies, combined with gene targeting, represents a promising strategy for the development of effective BCSC-targeted drugs. An overview of the main strategies based on BCSC-targeting is shown in Figure 1.

Breast Cancer Stem Cell-Targeting: New Strategies in Drug Development for Therapy Resistance
Conventional drugs targeting the tumor bulk are ineffective at eradicating CSCs [7]. In particular, it has been reported that anti-mitotic agents, such as taxanes (paclitaxel and docetaxel), cannot target quiescent CSCs inside the tumor bulk [23], leading to the reconstitution of the initial tumor cell population, increasing the adhesiveness of CTCs and the disease progression [68,69]. Based on this evidence, some studies have focused on directly targeting CSC subpopulation with promising results from preclinical experiments and clinical trials [24][25][26]. Novel therapeutic strategies for BCSC-targeting are based on the combined use of new and conventional drugs [27]. In particular, the main targets for BCSCs are Notch, Hedgehog, Wnt, PI3K/Akt/mTOR, IL-8, HER2, and TGF-β signaling, which are implicated in BCSC self-renewal and survival [27]. Moreover, the application of nano and biotechnologies, combined with gene targeting, represents a promising strategy for the development of effective BCSC-targeted drugs. An overview of the main strategies based on BCSCtargeting is shown in Figure 1.

Notch Signaling
The most clinically developed approach is the inhibition of the Notch signaling by γ-secretase inhibitors (GSIs). Notch receptors are cleaved by γ-secretase, which determines the release of the Notch intracellular domain (NCID), and subsequently, Notch signaling activation. Then, NCID is translocated to the nucleus where it induces gene transcription by interacting with other co-factors [70,71]. Notch signaling is highly active in BCSCs and it associates with tumor invasiveness [72]. The use of γ-secretase inhibitors has been proven effective in blocking the Notch pathway [72] and BCSC capability to form mammospheres in vitro [73]. Different phase I/II clinical trials using the γ-secretase inhibitor MK-0752 (Merck) in combination with docetaxel are ongoing for the treatment of metastatic breast cancer [73]. Enrolled patients' biopsies demonstrated a significant reduction in BCSC number, sustaining the advantages of Notch pathway inhibitors for BCSC-targeted therapy. In addition, other GSIs are in use to treat metastatic breast cancer such as RO4929097 (also combined with paclitaxel and carboplatin), PF-03084014, LY3039478, and CB-103 (Table 2).

Hedgehog Signaling
Hedgehog signaling is implicated in the maintenance of CSC stemness and in the regulation of self-renewal, survival, angiogenesis, EMT, and cell invasion [74]. Different inhibitors of the Hedgehog pathway, such as vismodegib, have been investigated, also in clinical trials, for their anti-CSC activity. In particular, an ongoing phase I clinical trial is evaluating the effects of vismodegib and RO4929097 on BCSC differentiation markers ( Table 2). Sims-Mourtada et al. co-treated breast cancer cells with vismodegib and docetaxel and found a decrease in the BCSC number and mammosphere formation that were increased instead by docetaxel alone [75]. In addition, they reported a Hedgehog signaling-dependent induction of multi-drug resistance 1 (MDR1) and ATP-binding cassette super-family G member 2 (ABCG2) in BCSCs. The Hedgehog inhibitor vismodegib (GDC-0449) has also been proven to counteract tumor growth in tamoxifen-resistant breast cancer xenografts [76].
LGK-974 has proven effective in different in vitro and in vivo cancer models [79]. Moreover, two clinical trials are using Foxy-5, a Wnt5a mimicking peptide, as an anti-metastatic cancer drug. In a study by Hallett et al., the Wnt signaling inhibitor PKF118-310 has been reported to be effective in reducing tumor growth and BCSC number in HER2-overexpressing breast cancer xenografts [80]. ROR1 is a type I orphan receptor expressed exclusively on different tumor cell types [81]. ROR1 is mainly involved in EMT and metastasis and its silencing in breast cancer cells counteracted these processes in vivo [81].

PI3K/Akt/mTOR Pathway
The PI3K/AKT/mTOR pathway regulates BCSC functions [52]. The inhibition of the PI3K/Akt/mTOR pathway by everolimus (RAD001) has been reported to counteract BCSC proliferation in primary breast cancer cells, especially in combination with docetaxel [114]. Combined treatment of everolimus and an aromatase inhibitor increased the progression-free survival in advanced hormone receptor-positive breast cancer patients [93] (Table 2).

Ephrin/Ephrin Receptor Pathway
Ephrin receptors, belonging to the largest family of receptor tyrosine kinases (RTKs), have been reported to influence BCSC activity [115] and several molecules targeting this pathway are being tested in clinical trials, especially the tyrosine kinase inhibitors. Ephrin A4 has been reported to be a potential therapeutic target for BCSCs [116,117]. An antibody-drug conjugate targeting Ephrin A4, named PF-06647263, consisting of a humanized monoclonal antibody anti-Ephrin A4 conjugated to the DNA-damaging agent calicheamicin, has been developed and tested to induce tumor regression in TNBC xenografts in vivo [117]. Moreover, PF-06647263 is currently being evaluated in a phase I clinical trial in metastatic TNBC patients ( Table 2).

Chemokine Ligand 8-Chemokine Receptor Type 1/2 (CXCL8-CXCR1/2) Axis
Many studies have focused attention on the role of IL-8 (CXCL8) in the biology of BCSCs. CXCL8 is a chemokine whose biological effects are mediated by two G-protein-coupled receptors: CXCR1 and CXCR2 [118]. CXCL8 has been reported to play multiple roles in cancer, such as increasing proliferation, angiogenesis, and metastases [119], as well as in mammosphere formation in HER2-positive breast cancer [120]. The adding of exogenous CXCL8 has proven to be effective in promoting the proliferation of CSCs in vitro; this growth was prevented by the presence of CXCR1/2 antagonists, such as reparixin [121] or a monoclonal antibody anti-CXCR1 (but not anti-CXCR2) [122]. These results have also been confirmed in breast cancer cell lines and in breast cancer patient-derived xenografts, in which the combination of docetaxel and reparixin was more effective in reducing tumor growth than either treatment alone, with a significant decrease in CSC number through apoptosis by activating Fas/FasL signaling [122] (Table 2). In addition, the combined treatment with reparixin and paclitaxel in the human TNBC cell line MDA-MB231 showed a synergistic effect, as proven for mammosphere activity and cell cycle arrest, likely mediated by the inhibition of focal adhesion kinase (FAK)/AKT and cyclin B1 signaling [123]. Neutralizing anti-CXCR1 and anti-CXCL8 monoclonal antibodies induced the same results [123]. SCH563705, another CXCR1/2 inhibitor, has been reported to counteract the effects of exogenous CXCL8 on BCSC mammosphere activity [120]. Currently, two clinical trials (phase I and II) are evaluating the efficacy of reparixin on BCSC survival (the first in combination with paclitaxel) ( Table 2). In the ongoing phase Ib clinical study [124], patients with HER-2 negative metastatic breast cancer who received reparixin and paclitaxel showed no cytotoxic effects and are in long-term remission [124]. However, it was not possible to collect serial biopsies of tumor tissue at baseline and during treatment. No significant differences in CTC number, ALDH expression, and EMT transcription factors were observed, likely due to the small sample size and high baseline heterogeneity.
A pilot trial (NCT01861054) of single agent reparixin that is evaluating the efficacy of reparixin to eliminate CSC in primary operable breast cancer [97] found the same limitations. A randomized, placebo-controlled clinical trial is using paclitaxel with and without reparixin in a front-line treatment of metastatic TNBC with good tolerability [124].

EGFR/HER2 and TGF-β Signaling
Several molecules targeting EGFR/HER2 signaling (downstream of IL-8) have been intensively investigated in breast cancer. In particular, Lapatinib (a HER2 inhibitor) has proven effective in counteracting tumor growth of HER2-positive and negative breast cancers, showing moderate toxicity and a decrease in brain metastases [125]. Currently, some clinical trials are evaluating the efficacy of Lapatinib on BCSCs [25] (Table 2). Trastuzumab (Herceptin), an inhibitor of HER2, has been proven to target HER2-expressing BCSCs and inhibit the tumor growth of patient-derived xenografts from HER2-negative breast cancer [126].
Given the critical role of TGF-β signaling in EMT and CSC activity, its inhibition has proven to be a promising strategy against drug resistance in chemotherapy [92]. The blocking of TGF-β signaling by a TGF-β type I receptor kinase inhibitor, EW-7197, suppressed paclitaxel-induced EMT and CSC mammosphere formation, reducing the number of lung metastases and increasing survival time in vivo [127]. Moreover, the cross-talk between the TGF-β pathway and Notch signaling in BCSCs has been demonstrated [128,129]. Currently, many different inhibitors of TGF-β pathway are being tested. One of the main is the TGF-β type I RTK inhibitor galunisertib/LY2157299. It has been reported that LY2157299 inhibited CSC expansion induced by paclitaxel alone in TNBC cell lines and in mouse xenografts [92]. Moreover, LY2157299 prevented tumor regrowth after paclitaxel treatment [92]. The evidence of a cardiovascular toxicity of LY2157299 has emerged in some preclinical studies [130,131], but now with the appropriate administration protocols, this issue has been overcome [132]. Some clinical trials are studying the efficacy of LY2157299 but none of these trials is explicitly referring to its anti-CSC activity (Table 2).
Disulfiram, used for chronic alcoholism treatment, is a dithiocarbamate that acts as an inhibitor of ALDH activity [133]. It has been reported that disulfiram inhibited TGF-β-induced EMT and CSC markers in breast cancer [134].
In addition, antibodies targeting clusterin, a stress-activated and apoptosis-associated molecular chaperone also overexpressed in breast cancer [135], have been reported to inhibit TGF-β-induced EMT and to reduce lung metastasis in breast cancer models [136,137]. In particular, a humanized anti-clusterin mAb (AB-16B5) has been tested in patients with advanced solid tumors that showed an inhibition of EMT markers in the tumor biopsies after treatment [91] (Table 2).

Angiogenic Signaling Pathway
Angiogenesis is critically required for cancer development and its inhibition by blocking vascular endothelial growth factor (VEGF) with the monoclonal antibody bevacizumab has been studied in the treatment of different types of cancers [138]. However, these studies reported conflicting opinions on its efficacy. In fact, the antiangiogenic therapy has reported to promote BCSC proliferation driven by hypoxia, limiting the efficacy of antiangiogenic drugs [139]. These findings suggest that antiangiogenic drugs should be combined with CSC-targeted therapies to improve cancer patient outcome. BCSC activity can be induced by hypoxia through a hypoxia-inducible factor 1α (HIF-1α) mediated pathway [140,141]. The blocking of HIF-1α by specific inhibitors, such as ganetespib (a second-generation HSP90 inhibitor), has been reported to be effective in counteracting chemoresistance induced by a paclitaxel or gemcitabine treatment, as demonstrated by in vitro and in vivo studies [140,141]. A phase I clinical trial evaluated the efficacy of a ganetespib treatment in trastuzumab-resistant HER2-positive metastatic breast cancer patients, reporting a good tolerability and slowing down of the progression [96] (Table 2). Another hypoxia-related strategy consists of inhibiting HIF-1αand HIF-2α-dependent expression of AlkB homolog 5 (ALKBH5) leading to the downregulation of NANOG, a pluripotency key gene in CSCs [142].

New Nano and Biotechnologies Applied to BCSC-Targeting Therapies
Nanotechnology could offer potential solutions for the specific targeting of BCSCs. The aim of nanoparticle technology is to promote the delivery of chemotherapeutic drugs to the tumor site using engineered drug-loaded nanoparticles targeting BCSCs. Currently, an ongoing phase I clinical trial is using lyso-thermosensitive liposomal doxorubicin (LTLD, ThermoDox; Celsion Corporation, NJ, USA) to achieve higher local drug concentrations in metastatic HER2-negative breast cancer patients (NCT03749850) [143]. In addition, nanoparticle albumin-bound (nab)-paclitaxel, in association with atezolizumab, has been proven effective in metastatic TNBC patients (NCT02425891) [144]. Moreover, nanoparticles are commonly used as RNA/DNA carriers in order to prevent degradation and ensure the delivery to the tumor site (see Section 3.2.3 on gene-targeted therapies). Several preclinical studies gave promising results using nanoparticles. In particular, salinomycin-loaded PEGylated polymeric micelles have proven effective in targeting BCSCs in vivo more than salinomycin alone [145]. Other strategies are based on the identification, by specific antibodies or ligands loaded onto nanoparticles, of particular receptors overexpressed on BCSCs [145]. For example, an anti-CD44 antibody conjugated to gold nanorod has been used to target and photo-ablate CD44 + subpopulations from three-dimensional MCF-7 mammospheres [146]. Promising results come from the use of salinomycin-loaded nanoparticles coated with hyaluronan (HA), a ligand of CD44. The treatment with these nanoparticles has proven to increase the cellular uptake and specifically target BCSCs [147]. Recently, Han et al. reported the efficacy of HA-conjugated liposomes loaded with gemcitabine in killing BCSCs with a lower systemic toxicity compared with the drug alone in experimental models [148]. Different strategies for HA targeting are under consideration; in particular, the use of small HA oligosaccharides competing with an endogenous HA polymer [149] and the use of antibodies blocking the HA-binding site of CD44 [150] have proven efficacious.

Immunomodulatory Therapy
The genetic engineering of autologous T lymphocytes or dendritic cells (DCs) and cancer vaccines (anti-CSCs or associated-individual proteins) represent promising strategies to elicit a specific antitumor immune response against cancer [151].
The high levels of lymphocytic infiltration is significantly associated with a more-favorable prognosis in patients with early stage TNBC and HER2-positive breast cancer [152]. This infiltration indicates a host antitumor immune response, fundamental for the survival outcome. Recent trials have shown that the targeting of the PD-1 and PD-L1 axis was also clinically effective in metastatic TNBC [152]. In particular, the transduction of T-cell with siRNA against PD-1 ligands or a PD-1/CD28 fusion receptor represents a novel immunotherapy application in breast cancer [153,154].
Moreover, T-cells can be genetically modified to express a novel T cell receptor (TCR) or a chimeric antigen receptor (CAR) that specifically recognize a tumor-associated antigen, inducing the cytolysis of the target cell [155]. However, some limitations are due to the downregulation of HLA class I (antigen presentation) and the required compatibility between patient's HA haplotype and the TCR [155]. CAR-T cells are engineered T-cells able to target a specific tumor protein expressing chimeric receptors (CARs) that combine both antigen-binding sites and T-cell activating functions by intracellular signaling motifs [155]. This system is independent from HLA and can recognize many targets other than peptides [156].
DCs, involved in the antigen processing and presentation, are currently exploited for their potential antitumor activity. In particular, some studies reported the genetic engineering of tumor cells and their fusion with DCs [157,158]. In addition, DCs can be loaded with tumor antigens or peptides and transfected with tumor-derived RNA or DNA [157,158]. In breast cancer, DCs expressing breast cancer antigens and transfected with siRNA against indoleamine 2,3-dioxygenase have been proven effective in reducing tumor growth and increasing survival in mouse models [159,160].
Therapeutic vaccines are able to stimulate a reactive and competent immune response against tumor antigens. Vaccines can be derived from whole tumor cell lysates, proteins, peptides, DNA, or DCs [151]. Immune cells are able to affect CSCs in vitro and are promising candidates for new strategies in breast cancer immunotherapy [151]. Currently, whole-cell vaccines showed inconsistency regarding the clinical efficacy in cancer patients [161]. Several vaccinations against anti-CSC individual proteins have been investigated in breast cancer (Table 2). In a preclinical study, vaccination against sodium-independent cystine-glutamate antiporter (xCT), the functional subunit of the cysteine/glutamate antiporter system xc-, has been proven effective in the inhibition of mammosphere formation, xenograft growth, and metastasis in EGFR-positive breast cancer cells that overexpressed xCT [162].

Cell-Based Therapy
Different and conflicting data from the literature indicate that MSCs can promote tumor growth and progression through their ability to home in on the tumor microenvironment [163]. It is thanks to this capacity that preclinical studies have suggested MSCs as anti-cancer drug delivery carriers [164][165][166]. In particular, it has been reported that MSCs can uptake and subsequently slowly release paclitaxel through exosomes, inhibiting the proliferation of different cancer cells [164][165][166][167][168][169]. As reported by Scioli et al., adipose-derived stem cells (ASCs) can uptake and release paclitaxel inhibiting CG5 breast cancer survival and proliferation, with no effects on ASC viability and cell cycle [170]. Moreover, it has been demonstrated that gold nanorod embedded hollow periodic mesoporous organosilica nanospheres (GNR@HPMOs) possess high paclitaxel-loading capability, excellent photothermal transfer ability upon near-infrared (NIR) light irradiation, and are well-retained by MSCs after internalization without affecting their viability and tumor-homing capability [171]. Some experiments revealed that GNR@HPMOs-paclitaxel loaded MSCs showed synergistic chemo-photothermal killing effects on breast cancer cells in vitro and in vivo [171]. Genetically engineered ASCs overexpressing TNFα were able to induce apoptosis, via caspase 3/7 activation, in human breast cancer cells and melanoma xenografts [172]. However, there are several limitations, principally regarding safety, that currently prevent the application of MSC-based strategies in clinical trials.

Gene-Targeted Therapies
Currently, therapeutic small interfering RNAs (siRNAs) and microRNAs (miRNAs) represent potential tools for specific gene targeting. Many preclinical studies have been carried out to explore their efficacy in cancer therapy [173]. Commonly, nanoparticles are used as carriers in order to prevent RNA degradation and ensure the delivery to the tumor site [174]. Several siRNA-based therapeutics are in use in cancer patients, instead of miRNA applications, which are still in the preclinical stage. In 2008, the first targeted delivery of siRNA was accomplished in humans. A phase I trial using the siRNA CALAA-01 showed an inhibition of tumor growth by targeting ribonucleotide reductase in patients with advanced solid tumors ( Table 2). The nanoparticles, which contained the siRNA, was made of a cyclodextrin-based polymer, a human transferrin protein (TF)-targeting ligand to engage TF receptors on the surface of the cancer cells and a hydrophilic polymer (polyethylene glycol, PEG) used to promote nanoparticle stability in biological fluids. Instrumental analysis confirmed the presence of nanoparticles in the tumor site with low toxicity [109,110].
Atu027 is a siRNA directed against serine/threonine-protein kinase N3 (PKN3), an angiogenic regulator expressed in the vascular endothelium [111]. The siRNA, formulated as liposomal particles (AtuPLEX18), is currently used in an ongoing phase I study in patients with advanced refractory solid tumors, including breast cancer, and has been proven effective as an anti-tumor agent [111].
Only one phase I/II clinical trial has been performed using a liposomal miR-34a mimic, MRX34, which showed antitumor activity in patients with advanced solid tumors, including breast cancer [112]. miR-34 is a tumor suppressor and it has been found to be lost or repressed in cancer patients [175]. It has been reported that miRNAs are involved in the post-transcriptional regulation of breast-cancer-related genes [176]. Antagomirs are small synthetic oligonucleotides involved in the silencing of endogenous miRNAs. The treatment with the anti-miR21 antagomir of MDA-MB-231 cells resulted in the reversion of the EMT and CSC phenotypes [177]. In addition, anti-miR10b antagomirs prevented metastasis formation in a mouse mammary tumor model [178].
Small hairpin RNA lentivirus (shRNA) particles for CD44 knockdown in BCSCs have been proven to induce differentiation into non-CSCs with a lower aggressiveness [179]. It has been reported that autophagy, the lysosomal degradation of cellular components, is involved in the survival and maintenance of BCSCs [180]. The knockdown of autophagy specific genes increased the expression of CD24 and the epithelial-like CD44 + /CD24 + phenotype [180]. However, a study by Kumar et al. reported that the induction of early stage autophagy triggered apoptosis in CD44 + /CD24 − BCSCs and the inhibition of autophagosome formation prevented this phenomenon [181].
The use of shRNA for ganglioside GD3 synthase reduced the CSC population and CSC-associated markers in breast cancer cell lines and completely hindered tumor formation in vivo [182]. OncoGenex Technologies Inc. and Isis Pharmaceuticals Inc. developed OGX-011, a clusterin-inhibiting antisense oligonucleotide, a potential sensitizer of solid tumors that are resistant to conventional cancer therapeutics. A phase II clinical trial of OGX-011 in combination with chemotherapeutic drugs is underway for breast cancer patients [113].

Other Therapeutic Approaches
Salinomycin, an ionophore antibiotic, has been proven effective in eliminating BCSCs in different breast cancer histotypes, likely by autophagy [183], increasing metastasis-free survival and overall survival, as well as inhibiting mammosphere formation and EMT in vitro [147,184]. Combined treatments with salinomycin, conventional drugs (i.e., doxorubicin or paclitaxel), anti-HER2 targeted therapies (monoclonal antibody trastuzumab and lapatinib), and histone deacetylase inhibitors synergistically counteracted tumor growth [185,186]. In particular, the histone deacetylase inhibitor abexinostat has been proven to promote CSC differentiation in breast cancer cell lines with low X-inactive specific transcript expression [187].
Other strategies involve the ALDH activity in combination with conventional therapies to improve breast cancer patients' outcomes. Croker et al. reported that the inhibition of ALDH activity, by all-trans retinoic acid (ATRA) or diethylaminobenzaldehyde (the specific ALDH inhibitor), counteracted the resistance to chemotherapy (doxorubicin/paclitaxel) and radiotherapy in TNBC cells [188]. In particular, the treatment with ATRA, an inducer of cell differentiation, has been proven to be effective in inhibiting BCSCs [189], but currently, its clinical application (in combination with paclitaxel) has not been successful [190] as an inhibitor. The inhibition of the DNA repair enzyme poly adenosine diphosphate ADP ribose polymerase by Olaparib has proven effective in counteracting CSC activity in breast cancer cells by ERK signaling [191]. It has been reported that the combination with the common chemotherapeutic drug irinotecan induced a decrease in the number of CSCs [191]. Consequently, Olaparib has been proposed as a candidate for the treatment of non-BRCA-related breast cancer [191]. The expression of ATP-binding cassette (ABC) transporters is higher in stem cells compared with normal cells, suggesting a potential role in drug resistance [192,193]. The combined use of dofequidar, an ABC transporter inhibitor, with other chemotherapeutic agents, such as cyclophosphamide, doxorubicin and fluorouracil, demonstrated promising results in patients with advanced or recurrent breast cancer [194]. Dofequidar increased the sensitivity of CSC-like side population cells from different cancer cell lines to anticancer drugs [195]. As described above, conventional chemotherapeutic drugs are not able to target CSCs because of their quiescent state. Therefore, a possible therapeutic strategy is to force CSCs to re-enter the cell cycle, as reported by Gasca et al. [196]. In particular, fbxw7 (F-box protein), a subunit of the stem cell factor SCF-type ubiquitin ligase complex, seems to maintain cell quiescence by reducing the expression of the c-Myc transcriptional factor responsible for the control of the cell cycle and proliferation [197]. Gasc et al. silenced fbxw7 in paclitaxel-resistant TNBC (MDA-MB-468), resensitizing cells to the chemotherapeutic drug item [196]. Moreover, radioresistant BCSCs expressed high levels of ataxia telangiectasia mutated (ATM) (DNA damage surveillance/repair system), and the treatment with an ATM inhibitor has been proven efficacious in re-sensitizing BCSCs to radiation [198].

Conclusions
Important evidence supports the pivotal role of tumor-initiating or cancer stem cells in anticancer drug resistance and recurrence. Many efforts have been made in the isolation and characterization of breast cancer stem cells (BCSCs), as well as in the identification of possible markers to specifically target this cell population. Several targets have been proposed for the development of BCSC-directed therapies; however, a combination approach directed toward multiple and different pharmacological targets is the most promising. Unfortunately, the outcome of the applied therapeutic methodologies is inconsistent because of the difficulty, by the current markers, to identify a single small population of cells with a high plasticity. The heterogeneity of the BCSC population sustains the therapeutic strategy based on the combination of multiple targets. In addition, the research of new circulating markers for monitoring the effect of anti-BCSC agents is constantly at work. Nano-and biotechnologies associated with gene-targeted strategies represent a promising approach in the development of efficacious drugs targeting CSCs and are able to improve breast cancer therapies.