Cancer stem cells in breast and prostate: Fact or fiction?

Abstract

Since the introduction of the cancer stem cell (CSC) hypothesis, accumulating evidence shows that most cancers present stem-like niches. However, therapies aimed at targeting this niche have not been as successful as expected. New evidence regarding CSCs hierarchy, similarities with normal tissue stem cells and cell plasticity might be key in understanding their role in cancer biology and how to efficiently eliminate them. In this Chapter, we discuss what is known in breast and prostate CSCs from their initial discoveries to the current therapeutic efforts in the field. Future challenges towards better CSC identification and isolation strategies will be key to shed light into how CSCs could accurately be targeted in combination to traditional therapies to ultimately prolong patient survival.

1. Origin and evolution of cancer stem cells: Consensus and controversies on single vs. multi-potent progenitors and stem cell hierarchy

All of our tissues are formed thanks to the activity of stem cells. The mother of all cells is, in fact, a stem cell, that has all the requisite information necessary to generate an entire organism. The first time the term stem cell was used in the scientific literature was by Ernst Häckel (in German, ‘Stammzelle”). He employed it to refer to a common unicellular ancestor from which he imagined all multicellular organisms evolved, inspired by Darwin’s stem trees (in German, “Stammbaüme”) that represented the evolution of organisms (Häckel, 1868; Ramalho-Santos & Willenbring, 2007). Häckel later on, once again comparing evolution to embryology, proposed that the fertilized egg also be called a stem cell (Ramalho-Santos & Willenbring, 2007). Around the same time, thanks to the contributions of Ehrlich (1879), the question was raised of whether a common precursor of the various blood cell types existed. In the beginning of the 20th century, several researchers began to use the term stem cell to refer to the common precursor of the blood system (Ramalho-Santos & Willenbring, 2007).

What defines whether a single stem cell is able to regenerate a whole organism or simply regenerate one specific tissue is the stem cell hierarchy. This is what defines whether the stem cell is totipotent, pluripotent, multipotent or simply oligopotent -or even unipotent. Therefore, to be considered “stem” a cell should have these capacities: (1) Be able to self-renew; (2) Be able to regenerate certain tissues/organs/organism; (3) Generate daughter cells that have a lower regeneration capacity or hierarchy (i.e., if a stem cell can regenerate the whole mammary gland, its daughter cells can at the most regenerate certain cell layers of the gland, but are not able to regenerate the whole organ) (Kreso & Dick, 2014).

Cancer stem cells (CSCs) were first identified by John Dick in acute myeloid leukemia in the late 1990s (Bonnet & Dick, 1997; Dick, 1996). Research in CSCs dramatically increased since the beginning of the 21st century (Fig. 1A and B). In fact, publications in both breast and prostate cancer followed a similar trend, starting to spike around 2007–2008 coinciding with a renewed interest and controversies around CSCs and their potential for new therapies (Visvader & Lindeman, 2008). CSCs came back to the spotlight around that year influenced by several discoveries: (a) The discovery of a highly tumorigenic subpopulation of breast cancer cells identified as CD44+/CD24 (−/low) by the Clarke group (Al-Hajj, Wicha, Benito-Hernandez, Morrison, & Clarke, 2003; Liu et al., 2007) further confirmed as resistant to chemotherapy in the beginning of 2008 by Chang and collaborators (Li, Lewis, et al., 2008); (b) By the end of 2017, the discovery of ALDH1 as a marker of normal and malignant stem cells (Ginestier et al., 2007); (c) The identification of embryonic stem cell markers (such as Nanog, Oct4, Sox2 and c-Myc) present in poorly differentiated tumors (Ben-Porath et al., 2008); (d) A new assay to study the capacity of isolated tumor cells to re-generate the original tumor, using new more severely immunocompromised mice (NOD-SCID-IL2Rγ^null or NSG) and novel implantation strategies that led to the discovery that CSCs are not as “rare” within tumors as it was previously suggested (Quintana et al., 2008); and (e) The discovery that differentiated cells could be reprogrammed into pluripotent stem cells in 2006 (Takahashi & Yamanaka, 2006) led several researchers to speculate that the tumor microenvironment could reprogram cancer cells into stem cells.

Fig. 1.

Evolution in the number of published articles in breast and prostate cancer stem cells overtime. (A) Results from the database “Dimensions” (https://www.dimensions.ai/); (B) Results from “Pubmed” (https://www.ncbi.nlm.nih.gov/pubmed/). Insets in both cases show the trends for breast and prostate CSCs with higher detail. In each case, the number of publications was corrected by the total number of publications each year and normalized to the maximum number of publications with the query “Cancer stem cell”.

As with normal stem cells, hierarchy has also been attributed to CSCs (Dick, 2009; Reya, Morrison, Clarke, & Weissman, 2001). The main distinction to normal stem cells that needs to be clarified here is that with normal stem cells, the pluripotent stem cell that gives rise to the entire tissue could phenotypically be the same as the one that maintains it. However, this is not the case for CSCs. The CSC model proposes that a small subpopulation of cells can regenerate the entire tumor heterogeneity (Shackleton, Quintana, Fearon, & Morrison, 2009). Even though this definition could be applied to both the origin of the tumor and the cells that maintain it, the first cell is almost certainty phenotypically different to the cell that maintains the tumor, since the latter have more likely accumulated a set of mutations.

Hierarchy within breast cancer has been the subject of extensive studies (Al-Hajj et al., 2003; Fridriksdottir et al., 2017; Gupta et al., 2011; Kreso & Dick, 2014; Visvader, 2009). In 2012, research from Mina Bissell and Ole Petersen’s groups using 60 primary breast tumors as well as diverse breast cancer cell lines, challenged the CSC model showing that tumor cells from different hierarchies (basal-like CD271⁺—p75^NTR—and luminal-like MM⁺—milk mucin-) can initiate a tumor in vivo (Kim et al., 2012). Moreover, tumors generated by the luminal-like cells were larger and more invasive than the ones triggered by their basal-like counterpart. This indicates that while there is a differentiation hierarchy within these tumors, there does not seem to be a stem cell hierarchy since all of these cells were able to act as stem cells, capable of both self-renewal and functioning as “tumor-initiating” cells, supporting the clonal evolution model (Campbell & Polyak, 2007). However, it is true that the cells with stem markers (basal-like cancer cells) were able to recapitulate the full heterogeneity of the tumor, which might indicate that even though all of the tumor cells were able to initiate the tumor, only the more undifferentiated cells held the capacity to act as pluripotent progenitors. However, it seems clear that basal-like cells are not a requirement for breast tumor aggressiveness.

One alternative explanation would be that both differentiated and undifferentiated cancer cells can potentially convert into tumor-initiating CSCs in a reversible manner, termed cancer cell plasticity (Batlle & Clevers, 2017; Kreso & Dick, 2014). The role of the microenvironment has been shown to be key in determining tumor cell transitioning between stem and non-stem states, as will be discussed below.

The debate around a stem cell hierarchy of breast cancer stem cells remains a matter of intensive research with the potential to unravel key insights for patient care. A recent relevant study determined a novel differentiation hierarchy of breast cancer cells and found that CSCs and early progenitors correlated with metastasis and poorer outcome in a cohort of 20 breast cancer patients (Bliss et al., 2018).

In prostate, the identification and characterization of CSCs seems to be different between androgen-dependent and castrate-resistant tumors (Taylor, Toivanen, & Risbridger, 2010). In androgen-dependent prostate cancers, the normal basal stem cell marker CD133 was also identified as a marker of CSCs. A CSC population expressing the basal cell markers CD133, CD44 and/or integrin α₂β₁ was able to regenerate tumors that resembled the original tumor when xenografted in animals (Gu, Yuan, Wills, & Kasper, 2007; Li, Zhou, et al., 2008; Miki et al., 2007; Rybak, Bristow, & Kapoor, 2015). However, immortalized human prostate cells expressing integrin α₂β₁, but not CD133/AC133, which is consistent with a more differentiated proliferative progeny, also formed xenografts with malignant characteristics (Taylor, Toivanen, Frydenberg, Pedersen, & Harewood, 2012). Human prostate cancer cells derived from androgen-insensitive DU145 cells and propagated under serum-free conditions as non-adherent spheres express the surface markers CD44, integrin α₂β₁ and CD24, both basal and luminal cytokeratins, and display increased tumorigenicity in vivo (Rybak, He, Kapoor, Cutz, & Tang, 2011). Therefore, it is evident that CSCs from different hierarchies are capable of regenerating tumors. The recent generation of patient-derived prostate cancer organoid cultures that recapitulate the diverse mutational landscape and histology observed in their original prostate cancer tissue samples will likely further aid in the characterization of prostate CSCs, as well as improve the understanding of the molecular determinants of therapeutic and castration resistance (Gao et al., 2014; Rybak et al., 2015).

2. Are cancer stem cells related to the normal stem cells within the niche?

CSCs are different from normal stem cells as much as cancer cells are distinctive from normal cells. According to the CSC model, normal stem and progenitor cells are considered to be the most likely targets of transformation; however, no normal cells in particular are identified as such by the clonal evolution model (Campbell & Polyak, 2007). Evidence supporting the relationship between normal stem cells and cancer stem cells include the notion that normal stem cells within epithelia are “long-lived” cells compared to differentiated cells, making it more likely that these cells will acquire the multiple mutations needed to become cancer (Miller, Lavker, & Sun, 2005). Normal stem cells have enormous proliferative potential, but tend to be slow-cycling within epithelia and can divide symmetrically to self-renew or produce more differentiated daughter cells. Being long-term residents of the epithelium, stem cells are uniquely susceptible to the accumulation of oncogenic changes. These traits could also explain their proliferative capacity, and the phenotypic heterogeneity observed in tumors. These cells also share many overlapping traits, including induction of angiogenesis, resistance to apoptosis and drugs, and cell migration (Allinen et al., 2004; Bapat, 2007; Bissell & Labarge, 2005; Campbell & Polyak, 2007; Hu et al., 2005; Miller et al., 2005; Spradling, Drummond-Barbosa, & Kai, 2001; Wicha, Liu, & Dontu, 2006).

In the breast, normal stem cells give rise to the progenitors of the luminal and basal myoepithelial compartment, although controversies continue regarding the existence of a common bipotent progenitor between the luminal and myoepithelial compartments and whether stem cells that maintain both compartments are localized in each one of those (Bach et al., 2017; Visvader & Clevers, 2016; Visvader & Lindeman, 2006, 2011; Visvader & Stingl, 2014; Wang, Christin, Oktay, & Guo, 2017).

Evidence in favor of normal stem cells being the target of initial transformation include the fact that most breast cancers arise from the segment of terminal ductules where normal stem cells are believed to be located (Villadsen, 2005; Villadsen et al., 2007). In addition, a marker of normal breast cells, CD44, as mentioned above, is used to identify breast cancer stem cells (Al-Hajj et al., 2003; Shipitsin et al., 2007). Also, cancer CD44⁺ cells may give rise to a hierarchy of tumor cells resembling those in the normal mammary gland, since cancer CD24⁺ cells are similar to normal CD24⁺ cells and have genetic alterations not present in cancer CD44⁺ cells (Shipitsin et al., 2007). Interestingly, most of the surface markers used to identify CSCs in different types of tumors are derived from known normal embryonic or adult stem cell surface markers, suggesting that CSCs might originate from normal stem cells (Kim & Ryu, 2017; Tirino et al., 2013).

An alternative explanation for the similarities between CSCs in breast tumors and the normal stem cells within the tissue is that, as discussed above, plasticity of cancer cells could convert any cells within the niche into a CSC through a combination of genetic alterations and microenvironmental cues (Batlle & Clevers, 2017; Cabrera, Hollingsworth, & Hurt, 2015; Cazet et al., 2018; Garg, 2017; Meacham & Morrison, 2013). This supports the possibility that differentiated tumor cells can reversibly convert to CSCs and re-generate a tumor or conquer a metastatic niche. In this model, there is no unidirectional stem cell hierarchy within the tumor.

In prostate, there is substantial controversy related to the normal stem cells within the tissue, which has led to multiple proposed differentiation hierarchies (Taylor et al., 2010). Linear vs. bidirectional models have been proposed to explain the generation of this tissue. The linear model of differentiation was established in 1989 where stem cells within the basal compartment either self-renew to generate more basal cells or give rise to a multipotent progenitor that can then differentiate to luminal or neuroendocrine cells in a linear manner (Isaacs & Coffey, 1989). The bidirectional model proposes that a common stem cell gives rise to lineage specific progenitors that, in turn, will generate the luminal, basal and neuroendocrine compartments (Wang, Hayward, Cao, Thayer, & Cunha, 2001). There is a third model, called the independent lineage model, established upon the discovery that there are also stem cells within the luminal compartment. Therefore, each cell type (basal, luminal or neuroendocrine) will derive from their own stem progenitor cells. It is also possible that basal and/or luminal stem cells can be multipotent and also generate the opposing lineage. Overall, it is likely that cells with stem properties occur in diverse cell types and that the characteristics that define a stem cell can be turned on or off depending on their response to extrinsic or intrinsic factors.

Without a clear definition of stem cells in normal prostate (and considering there may be more than one cell type), it is difficult to determine whether the cancer cell of origin in prostate cancer is a stem cell, multipotent progenitor/transient amplifying cell, or a more differentiated progeny cell. There is evidence indicating that tumor initiating cells can include both basal (cytokeratin—CK—CK5⁺/CK8⁻) and luminal cells (CK5⁻/CK8⁺) (Taylor et al., 2010). Intermediate cells express CKs of both basal and luminal cells (CK5⁺/CK8⁺). However, a diagnostic identifier of human prostate cancer is the loss of basal cells. Therefore, prostate cancer can potentially arise from oncogenic transformation of CK5⁺ CK8⁻ basal cells resulting in rapid differentiation into a luminal phenotype, or alternatively from stem or multipotent progenitor cells within the CK5⁺ CK8⁺ intermediate or CK5⁻ CK8⁺ luminal populations. Accumulating evidence shows that prostate cancer initiation can occur from luminal (and potentially intermediate) cells (Ellwood-Yen et al., 2003; Iwata et al., 2010; Ma et al., 2005). Alternatively, a body of work by the Witte Laboratory provides evidence suggesting that basal cells can initiate preneoplastic and cancerous lesions (Lawson et al., 2010; Wang et al., 2006). Therefore, similar to breast cancer, there may be more than one cancer cell type capable of generating human prostate cancer, which may correlate with tumor phenotype, but this possibility requires formal experimental documentation.

3. Similarities and differences between cancer stem cells and undifferentiated aggressive cancer cells: Dormancy and beyond

During tissue and organ development, epithelial cells undergo a physiological process called epithelial-to-mesenchymal transition (EMT) through which they lose cell polarity, cell-cell contacts and cell-matrix adhesions, acquiring a more motile mesenchymal-like phenotype. Cancer cells that reside within an epithelium can acquire deregulated EMT programs that convert them into migratory cells and this can, eventually, contribute to metastasis (Larue & Bellacosa, 2005; Nistico, Bissell, & Radisky, 2012). It is believed that CSCs share molecular pathways with EMT that regulate metastatic cell dissemination and the establishment of a new tumor niche (Agliano, Calvo, & Box, 2017; Wang, Jiang, Liang, & Tang, 2015). These pathways include: activation of TGFβ, Wnt, Notch, Hedgehog and tyrosine kinase receptors (e.g., EGFR, MET among others). It is also likely that even if not occurring in the same cell, EMT mediators might contribute to modify the microenvironment and, in turn, promote CSC phenotype, further conferring resistance to different therapies and increased ability of tumor regrowth (Nistico et al., 2012).

Mitotic arrest in G0-G1—quiescence—for long periods of time is one operational definition of “dormancy”. Adult stem cells in different tissues spend certain periods of dormancy, awaiting the right signals to differentiate and repopulate the tissue (Essers & Trumpp, 2010; Fuchs, 2009). On the basis of this premise, stem cells are often identified by their propensity to retain DNA labels much longer than their rapidly proliferating off spring.

A fraction of cells within each tumor is believed to undergo certain periods of dormancy (Talukdar et al., 2019). This dormant state is reversible and tumor cells can recover the ability to proliferate once closed vessels reopen or new vasculatures reach the hypoxic areas. Chaffer and Weinberg (2011) reported that CSCs have enhanced tumor-initiating potential, but with temporary growth arrest within a tumor. A fraction of tumor cells can migrate outside the primary tumor, and the majority of disseminated tumor cells retain a state of dormancy and these are believed to be responsible for cancer metastasis and relapse (Sosa, Bragado, & Aguirre-Ghiso, 2014). In addition, some findings indicated that disseminated tumor cells could present a CSC phenotype (Gao, Zhang, Tang, & Liang, 2017; Goss & Chambers, 2010).

Dormancy may also be a mechanism for the resistance of CSCs to antiproliferative chemotherapy (Talukdar et al., 2019). Moreover, if indeed CSCs appear in a dormant state, this could explain local recurrence or distant metastasis after long lag periods (Clevers, 2011).

In breast cancer in particular, several similarities between CSCs and metastatic cancer cells have been proposed. Evidence has been provided that induction of EMT in human or mouse mammary epithelial result in the generation of cells with stem cell properties, such as self-renewal and resistance to toxins (Mani et al., 2008; Nistico et al., 2012). Also, Balic et al. showed that bone marrow metastases are enriched in CD44^high/CD24^low cells, markers associated with CSCs (Balic et al., 2006). Along these lines, tumorigenic CD44^high/CD24^/low cells have been found in metastatic pleural fluid in patients with breast cancers (Al-Hajj et al., 2003).

Also, breast tumor epithelial cells that undergo EMT express markers of CSCs (CD44^high/CD24^low) and display tumor-initiating potential, such as mammosphere formation, and drug resistance (Gupta et al., 2009). Furthermore, Chaffer et al. showed a ZEB-1-dependent conversion of epithelial non-CSCs to mesenchymal CSC-like state in basal breast cancer (Chaffer et al., 2013).

In order to sustain growth of the metastatic lesions, a proportion of cancer cells would need to retain CSC features to maintain the proliferative reservoir and, potentially, to recapitulate the heterogeneity of the primary tumor. Cell plasticity complicates these interactions, as cancer cells may modify their transcriptional pattern depending on the physiological conditions. In support of this viewpoint, as discussed above, it has been shown that luminal-like breast tumor cells can trigger metastasis without acquiring stem-like markers (Kim et al., 2012). This could indicate either that it is not a prerequisite condition for disseminated tumor cells to present with CSC phenotypes in order to colonize a new niche or that, due to tumor cell plasticity, a fraction of the disseminated cells can reversibly convert into CSCs and initiate the new tumor.

An additional indirect evidence of dormancy in breast CSCs relates to the gene signature of normal quiescent mammary gland stem cells present within cultured mammospheres, which correlate with CSC behavior when applied to breast cancers (Pece et al., 2010). In another study, it was found that the human breast cancer cell line MDA-MB-231 can stably survive by entering into a dormant state after several cycles of hypoxia, that selects for the CSC population (Carcereri de Prati et al., 2017). Hypoxia was also shown to induce the expression of CSC-like phenotype in dormant MCF-7 cells, promoting EMT and escape from dormancy to metastatic outgrowth (Weidenfeld et al., 2016). A role for hypoxia-inducible factors in promoting breast cancer stem cell specificities and maintenance in response to hypoxia or cytotoxic chemotherapy has also been documented (Xiang & Semenza, 2019). In prostate, CSCs are believed to be involved in metastasis and therapeutic relapse and recapitulation of the primary tumor heterogeneity in the secondary site, although their link with disseminated tumor cells is unclear (Bjerkvig, Tysnes, Aboody, Najbauer, & Terzis, 2005; Lathia, 2013; Reya et al., 2001; Schilling et al., 2012; van der Toom, Verdone, & Pienta, 2016). Several markers associated with tumor progression and therapeutic resistance can be found in disseminated tumor cells in a patient’s bone metastases. Primary tumor expression of CXCR4, EpCAM and EZH2—all associated with a CSC phenotype-correlated with increased distant metastasis and local recurrence during patient follow-up (Conley-LaComb et al., 2012; Harris & Kerr, 2017; Massoner et al., 2014; Matsika et al., 2015; Mochizuki et al., 2004) indicating that CSCs may drive metastasis.

Among the different cell types present in prostate cancers, a patient’s prostate cancer typically contains differentiated cancer cells expressing high levels of PSA (i.e., PSA+), as well as stem-like cells that express little or no PSA (i.e., PSA−/low). The tumor-regenerating ability of PSA+ cells gradually declines, suggesting that PSA−/low cells possess long-term tumor-propagating capacity, consistent with CSCs. The PSA−/low cells appear to be rare in early-stage tumors, but more abundant in high-grade and locally advanced tumors, and some tumors may even lack PSA expression (Liu et al., 2015; Qin et al., 2012; Shah et al., 2004). Low PSA expression correlates with poorer clinical outcomes, including metastasis, recurrence, and reduced patient survival.

4. How does the microenvironment determine cancer stem cell fate?

The tumor microenvironment is essential in providing a fertile ground for CSC selection and development (Aguirre-Ghiso, Bragado, & Sosa, 2013; Carcereri de Prati et al., 2017). Different signals that originate from the tumor niche regulate CSC self-renewal, survival, and ability to invade tissues and promote metastasis (Clarke et al., 2006; Jaworska, Krol, & Szliszka, 2015; Williams, Motiani, Giridhar, & Kasper, 2013; Yu et al., 2012; Zenzmaier, Untergasser, & Berger, 2008). There are several stem cell and self-renewal associated pathways that are activated in CSCs upon interaction with the surrounding tumor microenvironment that correlate with aggressive phenotypes, such as Wnt, Notch-1, PI3K, Hedgehog, CXCL12 and induction of pluripotency-associated transcription factors, such as Oct-3/4, Nanog and Sox-2 (Albini et al., 2015; Clevers, 2011; Liu et al., 2006; Mimeault & Batra, 2013; Philip, Ito, Moreno-Sanchez, & Ralph, 2013; Schwitalla et al., 2013). There are several cells and/or signals from the environment that trigger these responses. Cancer-associated fibroblasts (CAFs) or adipocytes secrete a variety of cytokines related to tumor progression, such as Platelet-Derived Growth Factor, Vascular Endothelial Growth Factor (VEGF), Hepatocyte Growth Factor (HGF), platelet-derived growth factor (PDGF) and a variety of ECM proteins that enhance tumor growth (Polanska & Orimo, 2013); also, CAFs exert a metabolic reprogramming of cancer cells by inducing a reverse Warburg phenotype. Acidity and hypoxia are often characteristics of the tumor microenvironment. Hypoxia-inducible factors, HIF1a and HIF2a, are also sensitive to cellular pH conditions (Parks, Mazure, Counillon, & Pouyssegur, 2013). Both hypoxia and changes in pH can regulate stem cell behavior by altering their metabolic status and promoting metabolic reconfiguration of cancer cells toward glycolysis, induction of the EMT phenotype (including C-X-C-chemokine receptor 4 (CXCR4), Snail and Twist gene expression) and expression of stem-prone transcription factors.

There is also compelling evidence suggesting a role of the endothelial compartment in supporting CSC phenotypes, activating Notch signaling among other pathways through the secretion of VEGF and HGF ( Jeon et al., 2014; Keith & Simon, 2007). Immune cells at the tumor site, such as NK or tumor-associated macrophages (TAMs) usually show altered phenotypes compared to those in healthy tissues. NK cells are able to modulate metastatic dissemination in different tumor types, including breast and prostate cancer (Liu et al., 2013; Wang et al., 2014). TAMs have been shown to contribute to tumor growth and metastasis by producing antiinflammatory cytokines that promote tumor cell escape from immune surveillance (Allen & Louise Jones, 2011; Solinas, Germano, Mantovani, & Allavena, 2009; Ugel, De Sanctis, Mandruzzato, & Bronte, 2015).

Additionally, cell metabolism regulates cell proliferation and differentiation in CSCs, which have different metabolic states compared to their differentiated progeny (Carey, Finley, Cross, Allis, & Thompson, 2015; Vlashi et al., 2014). Further, CSCs establish a reinforcement mechanism by secreting cytokines that help recruit and activate specific cell types that modulate their differentiation states (Plaks, Kong, & Werb, 2015). At the site of metastasis, it has been shown that the microenvironment within the bone marrow plays a crucial role in the growth of disseminated tumor “stem” cells (Pedersen, Shiozawa, Pienta, & Taichman, 2012; Rahim et al., 2014; Shiozawa et al., 2011). These cells take over the bone marrow hematopoietic stem cell niche, competing directly for the niche.

In breast cancers in particular, CAFs contribute to CSC proliferation, invasion and metastasis by producing a variety of factors including IGF, PDGF, Wnt, Notch ligand, Hedgehog ligands, and matrix metalloproteases (MMPs) (Fig. 2). Together with these factors, ECM proteins secreted primarily by activated fibroblasts are known to contribute to tumor progression (Dufraine, Funahashi, & Kitajewski, 2008; Kenny & Bissell, 2003; Pontiggia et al., 2012; Sampayo et al., 2018). In this context, increased stromal collagen content positively correlates with stemness and also enhances CSC properties of breast cancer cells in culture (Pang et al., 2016). A recent study demonstrated that stromal cues from Hedgehog-activated CAFs, including FGF and collagen-rich ECM, are able to induce and maintain a stem-like phenotype in triple negative breast cancer cells in vivo (Cazet et al., 2018).

Fig. 2.

Key players in the interaction between CSCs and their microenvironment in breast and prostate cancers.

Mesenchymal stem cells (MSCs) have been shown to be recruited from the bone marrow or from the normal breast stroma (Korkaya, Liu, & Wicha, 2011; Liu et al., 2011). Within the tumor, MSCs interact with breast CSCs through cytokine loops that include IL-6 and CXCL7, to stimulate CSC self-renewal. Furthermore, MSCs can cause elevated miR-199a expression in breast cancer cells, suppressing FOXP2 expression and thereby imparting CSC properties to tumor cells (Cuiffo et al., 2014). The immune system seems to exert both inhibitory and stimulatory effects on breast tumors, and the balance between these effects dramatically influences tumor growth. Inflammatory cytokines IL-6 and IL-8 produced by immune cells promote CSC self-renewal (Sansone et al., 2007). IL-8 receptor, CXCR1, has been found to be highly expressed on breast CSCs (Ginestier et al., 2010). It has been reported that breast CSCs fail to express inhibitory NK ligands, which is consistent with metastatic spread (Wang et al., 2014).

In prostate cancer, CAFs contribute to enhance the growth potential of CSCs by increasing spheroid formation and cancer cell proliferation through different paracrine signals (Adisetiyo et al., 2014; Liao, Adisetiyo, Liang, & Roy-Burman, 2010) (Fig. 2). Moreover, co-injection of prostate cancer CSCs and CAFs into immune-compromised mice increase the number of neoplastic lesions as compared to normal fibroblasts, further supporting the significance of CAFs in regulating CSCs phenotypes (Liao et al., 2010). Cancer cell subpopulations can interact with other normal prostate cells present in the tumor environment and take advantage of these cells for tumor growth. An example of such cooperative interactions was described between two clonal subpopulations of the PC-3 prostate cancer cell line (Mateo et al., 2014). They found that the invasiveness of a CSC-enriched subpopulation is enhanced by a non-CSC subpopulation, resulting in an increase in tumorigenic and metastatic potential of CSCs. An ECM protein, SPARC (also known as osteonectin), was found to be the main factor mediating the cooperation between CSCs and non-CSCs (Mateo et al., 2014).

In addition, it has been observed that hypoxia induces, through HIF, the reprogramming of prostate cancer cells increasing the expression of stemness markers like CD44, Oct-3/4, Nanog, and drug resistance-associated molecules or anti-apoptotic proteins (Mathieu et al., 2011; Ravenna et al., 2014). Other cells within the prostate tumor microenvironment such as TAMs, endothelial cells and MSCs also support prostate cancer progression (Luo et al., 2014).

5. Biomarkers consensus: Are we there yet?

Hematopoietic stem cell markers have provided the basis for identifying and isolating CSCs in solid tumors including breast and prostate cancers (Kasper, 2008; Taylor & Risbridger, 2008). In breast, evaluation of combinatorial expression of surface markers has been proven to yield a better prognostic value for identifying breast CSCs. Al-Hajj et al. showed that less than 100 mammary CSC cells with the markers CD44+, CD24−/low, Lin− were able to generate tumors in NOD/SCID mice, whereas CD44−, CD24+ breast cancer cells, even when injected at 100-fold higher cell densities, were not able to generate tumors (Al-Hajj et al., 2003; Patrawala et al., 2005). Breast CSCs expressing CD133+ showed similar behavior, with only a few cells being able to recapitulate the original tumor (Wright et al., 2008). The CD44+, CD24−/low, Lin− subpopulation can be further subdivided based on EpCAM expression. The EpCAM+ , CD44+, CD24−/low, Lin− breast CSCs, but not EpCAM−, CD44⁺ , CD24−/low, Lin− cells, were capable of generating breast tumors and reconstitute the heterogeneity of the initial tumor in NOD/SCID mice (Al-Hajj et al., 2003). Both CD133+ and CD44+, CD24−/low, Lin− populations displayed markedly enhanced self-renewal capacity and shared the expression of sternness genes (OCT4, NOTCH1, ALDH1, FGFR1, SOX1) (Ginestier et al., 2007; Klonisch et al., 2008; Wright et al., 2008). However, not all are associated with aggressive metastatic growth (Lin, Zhong, Guan, Zhang, & Sun, 2012). The other markers that have been used to identify breast CSCs are CD44⁺ CD49f^hi CD133/2^hi phenotype (Meyer et al., 2010), CD49f (Cariati et al., 2008) and CD61 (Bozorgi, Khazaei, & Khazaei, 2015; Vaillant et al., 2008). Desgrosellier et al. demonstrated that CD49f and CD61 were found to be associated with tumor initiation properties through an in vivo study of breast cancer in mice (Desgrosellier et al., 2014; Lo et al., 2012; Sin & Lim, 2017). The breast CSCs markers described above are summarized in Table 1.

Table 1.

Current markers for identification of CSCs in breast and prostate cancers.

CSCs type	Marker
Breast CSCs	EpCAM+ CD44+ CD24−/low Lin−
	CD133+
	OCT4high
	NOTCH1high
	ALDH1high
	FGFR1high
	SOX1high
	CD49f+ (Integrin α6) CD61+
	CD44+ CD49f+ CD133/2+
Prostate CSCs	CD44+
	Sca-1+
	CD133+
	CK5/14
	ALDHhigh
	ABCG2+
	p63high
	Oct-3/4high
	β-cateninhigh
	CD44+ Integrin α2β1high CD133+

In prostate, CD44 is a marker for normal prostatic epithelium stem cell and prostate CSCs (Tang et al., 2007; Yu et al., 2012). CD44⁺ prostate cancer cells are enriched in tumorigenic and metastatic progenitor cells and are more proliferative, clonogenic, tumorigenic, and metastatic than the isogenic CD44− cells (Patrawala et al., 2006). CD44⁺/Androgen receptor-(AR-) tumor progenitor cell population expressed stem-cell-related genes such as OCT3/4 and β-catenin and 1% of the CD44+ prostate cancer cells presented asymmetric cell division (Patrawala et al., 2006). CD44⁺/CD24⁻ cells formed prostaspheres in vitro and a small number of cells was sufficient to initiate tumors in vivo (Hurt, Kawasaki, Klarmann, Thomas, & Farrar, 2008). Stem cell antigen (Sca-1)⁺ cells have increased proliferative capacity, with a subpopulation of Sca-1⁺ cells expressing Bcl-2 and integrin α6 (Xin, Lawson, & Witte, 2005). CD133⁺ prostate normal and cancer cells exhibit stem cell features such as prostasphere formation and they develope prostatic-like acini in immunocompromised mice (Pellacani, Oldridge, Collins, & Maitland, 2013; Richardson et al., 2004). In addition, CD133⁺ cells also co-express CK14 or hTERT and exhibit more developed ducts compared to CD133⁻ cells (Mundy, 2002). The ATP-binding cassette (ABC) membrane transporter, ABCG2, expressed in a subpopulation of prostate cancer cells, enables the efflux of Hoechst 33342 dye suggesting that these cells might present multidrug resistance, which is a characteristic of CSCs (Patrawala et al., 2005). Aldehyde dehydrogenase (ALDH) is an enzyme involved in intracellular retinoic acid production. In prostate CSCs, high expression of ALDH1A1 was found to be positively correlated with Gleason score and pathologic stage and inversely correlated with overall survival, which suggests it may be a potential prostate CSC marker (Li, Cozzi, & Russell, 2010). The basal cell markers CK5/14 and p63 were found in a small fraction of cells in primary prostate carcinoma tissues and in the majority of prostate cancer metastases, suggesting that these might be markers of cells exhibiting prostate CSC features (Janssen et al., 2002; Klonisch et al., 2008).

Combining multiple markers has also improved the identification and isolation of prostate CSCs. CD44⁺ Integrinα2β1^high CD133⁺ rare prostate cell populations possessed a significant capacity for self-renewal in vitro and could regenerate the mixed populations present in the tumor (Collins, Berry, Hyde, Stower, & Maitland, 2005; Moltzahn & Thalmann, 2013; Yun, Lo, & Hsieh, 2016). It was also found that ALDH^high CD44⁺ cells show a higher proliferative, clonogenic, and metastatic capacity in vitro and higher tumorigenicity capacity in vivo than ALDH^low CD44− cells. However, ALDH^low CD44− cells were able to develop tumors, although they had longer latency periods (Yu et al., 2011, 2012). The variability of the different marker combinations suggests that CSC may be more than a distinct subpopulation and might be explained by the fact that CSC exhibit phenotypic plasticity which leads to dynamic phenotypes.

The prostate CSCs markers described above are summarized in Table 1.

6. Impact of the cancer stem cell hypothesis in cancer treatment and future challenges

The cancer stem-cell hypothesis, stating that tumors originate in stem or progenitor cells as a result of dysregulation of the normally tightly regulated process of self-renewal, has major implications for cancer research as well as important clinical implications for prevention and therapy (Kakarala & Wicha, 2008).

The first step in cancer prevention is the reduction of risks. In this sense, the CSC niche could be directly targeted through certain habits that can be promoted to inhibit CSC growth or that should be eliminated because they stimulate CSC proliferation. In this sense vitamin D has been shown to both promote differentiation of hematopoietic stem cells (therefore reducing the stem cell pool (Kakarala & Wicha, 2008; Nagler, Riklis, Kletter, Tatarsky, & Fabian, 1986)) and also has been recently found to directly inhibit the expression of stemness markers in glioma cells (Hu et al., 2019). On the contrary, cigarette smoke has been recently shown to promote stemness in renal cancer stem cells through activation of the Sonic Hedgehog pathway (Qian et al., 2018).

The second step in cancer prevention is early detection. Markers of cancer stem cells or their secreted factors could be used to detect early events of tumor initiation.

In terms of cancer treatment, accumulating efforts are tending to directly target the CSC compartment, either by eliminating it or promoting its differentiation, in order to prevent metastasis and prolong survival. One of the restrictions in traditional therapies is drug resistance, which is due to characteristic properties of CSCs such as high expression of drug-efflux pumps, low rate of division and high efficiency for DNA repairing (Dragu, Necula, Bleotu, Diaconu, & Chivu-Economescu, 2015). Therefore, targeting CSCs might be critical in treating cancer and preventing tumor relapse. Examples of these therapies are Notch signaling inhibitors, such as BMS-906024, which showed promising results for patients with relapsed T-cell acute lymphoblastic leukemia (Venkatesh et al., 2018), inhibition of efflux-pumps in which several generations of ABC blockers have been developed and now small-molecules are being tested to this end (Dragu et al., 2015) or targeting the interaction with the tumor microenvironment such as using CXCR4 agonists to inhibit the CXCL12-CXCR4 axis like CTCE-9908 that proved to be effective in reducing both tumor growth and metastasis in xenograft mouse models of inflammatory breast cancer (Singh et al., 2010) and in decreasing the tumor invasivity and angiogenesis in prostate cancer (Wong, Kandagatla, Korz, & Chinni, 2014).

Among the future challenges that are going to be key for CSC research in order to establish these cells as solid indicators for diagnosis and targets for therapies, the identification of better markers to unmistakably recognize this niche is going to be critical. And this identification should be done by carefully isolating tumor cell populations and testing their capacity to initiate a tumor in different types of immunocompromised mice. Research in this direction will be benefited by current state-of-the-art high-throughput sorting and screening tools.

Future research in the field should also consider increasing the specificity and efficiency of CSCs targeting, reducing toxicity of normal stem cells and also developing new strategies for delivery. These new therapies, combined with traditional therapeutic strategies, may contribute to increase their efficacy for aggressive cancers, therefore preventing tumor relapse and ultimately enhancing patient survival.