New insights into prostate cancer stem cells

Abstract

Prostate cancer (PCa) remains one of the most prevalent malignancies affecting men in the western world. The etiology for PCa development and molecular mechanisms underlying castration-resistant progression are incompletely understood. Emerging evidence from many tumor systems has shown the existence of distinct subpopulations of stem like-cancer cells termed cancer stem cells (CSCs), which may be involved in tumor initiation, progression, metastasis and therapy resistance. Prostate cancer stem cells (PCSCs) have also been identified using different experimental strategies in distinct model systems. In this brief review, we summarize our current knowledge of normal prostate stem/progenitor cells, highlight recent progress on PCSCs, expound on the potential cell-of-origin for PCa and discuss the involvement of PCSCs in PCa progression and castration resistance. Elucidation of the phenotypic and functional properties and molecular regulation of PCSCs will help us better understand PCa biology and may lead to development of novel therapeutics targeting castration-resistant PCa cells.

Introduction

Prostate cancer (PCa) is one of the most commonly diagnosed cancers affecting men in the western world,¹ with an estimated 241,740 new cases and 28,170 cancer-related deaths in the United States in 2012.² PCa is a multifocal disease, and its clinical progression is closely related to aging.¹ PCa develops over a long period of time from normal prostate to prostatic intraepithelial neoplasia (PIN) then to early- and late-stage PCa and finally to metastatic PCa.³ Localized early-stage PCa can be well treated by radical prostatectomy with good prognosis.^4,5 By contrast, advanced PCa is mainly treated with androgen-deprivation therapy (ADT), but most cases will eventually fail ADT, recur and result in a lethal disease termed “castration resistant prostate cancer (CRPC).”^1,4-7 Both primary PCa and CRPC can metastasize to many different organs, including the bone, lung, liver, etc., and once metastasis occurs the disease becomes incurable.⁵ Despite a great deal of progress made in clinical treatments as well as in basic research on PCa in the past decades, the underlying mechanisms of PCa development, progression, recurrence and metastasis are incompletely understood.^1,6

Like many other cancer types, PCa is heterogeneous with respect to both histo-structures and cellular composition. For example, although most localized, early-stage PCa are mainly composed of differentiated glandular structures in which cells are positive for prostate-specific antigen (PSA) and androgen receptor (AR), poorly differentiated or undifferentiated areas can also be detected where cells are largely negative for PSA and AR expression. On the other hand, advanced PCa mainly consist of poorly differentiated or undifferentiated areas, but differentiated regions can also be seen.^8-10 The heterogeneous expression patterns of PSA and AR in PCa cells imply the existence of distinct cell subpopulations, i.e., PSA⁺AR⁺, PSA⁺AR^-, PSA^-AR⁺ and PSA^-AR^-.⁸ A recent study analyzed the Docetaxel-tolerant phenotype in both untreated primary human PCa (HPCa) and metastatic samples, and found that all tumors contained two subpopulation of PCa cells, i.e., cytokeratin (CK) 18⁺/19⁺ and CK18^-/CK19^- cells.¹¹

Tumor cell heterogeneity can be explained by two models: the clonal evolution (stochastic) model and the cancer stem cell (hierarchical) model. The two models may not necessarily be mutually exclusive.^12-14 The clonal evolution model postulates that genetic and epigenetic alterations in some cancer cells confer a growth/survival advantage resulting in a dominant clone, and all tumor cells in the clone have similar tumor-initiating capacities.^12,15,16 In comparison, recent evidence suggests that cancers cells in many human tumors are organized hierarchically, and only a subset of the cancer cells is endowed with tumor-initiating and long-term tumor-propagating activity. These tumor-initiating cells possess many phenotypic and functional properties associated with normal stem cells (SCs) and thus are often termed cancer stem cells (CSCs). CSCs have been proposed to be responsible for tumor initiation, progression, metastasis and treatment resistance.^12-14,17-22 The first direct evidence of CSCs came from seminal work by Dick and his colleagues in the 1990s.^23,24 Since then, CSC have been reported in many solid tumors, including cancers of the breast,^25-27 colon,^28-30 brain,^31-33 prostate,^8,11,34-55 pancreas,⁵⁶ lung^57,58 and liver.^59,60 The revived studies on CSCs have generated novel insights on our understanding of the etiology and mechanisms of cancers and shed new light on developing therapeutics targeting the population of cancer cells that is largely refractory to the current standard-of-care therapies. In this brief review, we summarize our current knowledge of normal prostate stem/progenitor cells, highlight recent progress on prostate CSCs (PCSCs), expound on the cell-of-origin for PCa and PCSCs and discuss the potential involvement of PCSCs in PCa progression, castration resistance and distant metastasis.

Normal Prostate Stem/Progenitor Cells

Both human and mouse prostatic glands contain three distinct, differentiated epithelial cells: basal, luminal and neuroendocrine cells.^1,61 Basal cells are localized in the basal layer above the basement membrane and express markers such as CK5, CK14, CD44, CD133, p63, Bcl-2 and low levels of AR. Luminal cells sit above the basal layer, secrete prostatic proteins and express markers such as CK8, CK18, CK19, PSA, prostatic acid phosphatase (PAP), CD57 and high levels of AR. Neuroendocrine cells are rare, and are positive for synaptophaysin and chromogranin A.^1,61 Newly emerged evidence reveals that epigenetic modifications may be involved in prostate tissue differentiation. For instance, the levels of 5-hydroxymethylcytosine seem to be more abundant in normal human prostate luminal cells, compared with the low levels in basal cells.⁶² This finding suggests the potential use of certain epigenetic modification(s) as a marker for prostate luminal cell differentiation.

SCs are defined as cells that have both self-renewal and multi-lineage differentiation abilities.²⁰ Among the various ways to functionally characterize SCs, transplantation and lineage tracing are the most popular approaches.^13,63 Transplantation assays use florescence-activated cell sorting (FACS) to isolate candidate SCs via cell surface markers, followed by functional characterization of SC properties in a xenograft, syngeneic or allogeneic model to see whether these cells have the ability to reconstitute tissues/organs from which they are derived. Transplantation assays tend to over-estimate the frequency of SCs. It is also unclear if the transplantation-based assay accurately mimics the natural regeneration process of the SCs in the orthotopic site. Finally, this method induces trauma in the recipient mice and leads to a wound-healing response, which may compound data interpretation. By comparison, lineage tracing is utilized to characterize SC development and properties in vivo by using a stem cell-specific Cre reporter line in the tissue of interest. Limitations of this approach include choice of promoters, specificity and sensitivity of promoters, incomplete tracing of cell lineages due to variability of promoter strength and the uncertainty of its relevance to humans. In addition to transplantation and lineage-tracing approaches, the label-retaining strategy is used to enrich SCs based on their slow-cycling and quiescent traits.⁶⁴ Moreover, side population (SP) and ALDEFLUOR assays have been widely employed to enrich SCs by virtue of the heightened detoxification capacities in the SCs.^65,66

Normal adult murine prostate regresses after castration with more than 90% of luminal cells and a small population of basal cells undergoing apoptosis. However, the prostate returns to normal size when androgen is restored,^63,67 and such regression-regeneration manipulations can be repeated for multiple cycles, providing indirect early evidence for the presence of prostate SCs (PSCs) that may resist castration. Many recent studies, by employing the above-mentioned approaches, have provided more direct evidence for murine PSCs. In tissue recombination assays, a PSC-containing cell population is isolated by cell surface markers and combined with embryonic urogenital mesenchyme (UGM). The cell mix is then transplanted to the kidney capsule of recipient mice to test if the candidate PSCs can form new prostatic ducts.⁶⁸ For instance, Sca-1⁺ prostatic basal cells are quiescent and positive for integrin α6 (CD49f) and Bcl-2 and possess a higher capacity to regenerate prostatic gland-like structures compared with the corresponding Sca-1^- cells.^69,70 Prospectively purified Lin^-Sca-1⁺CD49f⁺ basal cells (LSCs) can establish spheres and colonies in vitro as well as regenerate prostatic ducts in renal grafts,⁷¹ and murine PSCs could be further enriched by sorting for Trop2^hi LSC cells.⁷² A recent study shows that single Lin^-Sca-1⁺CD133⁺CD44⁺CD117⁺ prostatic cells are capable of reconstituting a prostate in the kidney capsule of recipient mice, and CD117 is preferentially expressed in murine prostatic basal cells and exclusively in human prostate basal cells.⁷³ Early studies in Wilson’s lab demonstrated that slow-cycling (label-retaining) PSCs were localized mostly in the basal layer of the proximal prostate tubules, and that the relative quiescence of the murine PSCs was maintained by TGFβ produced from the stroma.⁷⁴ In support, a recent study in animal models shows that the dormancy of adult murine PSCs is likely to be mediated by a positive reciprocal regulatory loop formed from Notch and TGFβ signaling pathways.⁷⁵

Using lineage tracing with different promoters, several studies have generated evidence that the prostate stem/progenitor cells with regenerative activity may also be localized in the luminal layer. For example, Wang et al. used an Nkx3.1-Cre mouse line to identify a small subset of luminal cells that survive castration (thus called CARNs for castration-resistant Nkx3.1-expressing cells), can self-renew in vivo and regenerate a prostate in renal grafts.⁷⁶ Choi et al. showed that there exist lineage-restricted unipotent prostate progenitor cells in both basal and luminal layers.⁷⁷ Liu et al. conducted lineage tracing in their PSA-CreER(T2) mouse model to show that luminal cells can survive, proliferate and regenerate during cycles of regression and regeneration.⁷⁸ Finally, Ousset et al. demonstrated that there are multipotent stem cells in the basal layer that can differentiate into all epithelial prostate cells (basal, luminal and neuroendocrine cells) as well as unipotent basal and luminal progenitors, which together contribute to the prostate postnatal development.⁷⁹

The PSCs in humans are mainly inferred from correlative and in vitro experiments and all seem to have a basal origin. These include cells that express high levels of integrin α2β1,⁸⁰ CD133,⁸¹ ABCG2³⁷ and Trop2.⁷² Using marker-independent methods, Garraway et al. have shown that a small population of human prostate cells is able to generate prostaspheres bearing a basal phenotype (CD44⁺CD49f⁺CK5⁺p63⁺Trop2⁺CK8^-AR^-PSA^-), and these sphere-forming cells are capable of inducing prostatic gland structures in vivo.⁸²

PCSCs

A CSC is defined as a cell within a tumor that is endowed with unlimited self-renewal ability and can also regenerate non/low tumorigenic progeny.^13,14,20,21 The “gold” standard to characterize CSC activities is to examine whether a candidate CSC-containing cell population has the ability to initiate serially transplantable tumors in immunodeficient mice that histopathologically resemble the parental tumor.¹⁷ Most CSC studies have employed similar experimental strategies that are used to enrich normal SCs,^13,14,18,21 including transplantation assays using FACS-sorted cells via cell surface markers,^23-25 SP³⁴ and ALDEFLUOR assays⁸³ as well as lineage-tracing studies.^84-86

Several populations of PCSCs have been reported.^8,11,34-55 For example, the CD44⁺α2β1⁺CD133⁺ HPCa cells are highly clonogenic and seem to have certain CSC properties.³⁸ In long-term cultured PCa cell lines (e.g., Du145) and xenograft models (i.e., LAPC9, LAPC4), the CD44⁺ PCa cells preferentially express “stemness” genes, are largely quiescent and more tumorigenic and metastatic than the isogenic CD44^- PCa cells.³⁵ Follow-up studies reveal that the CD44⁺α2β1⁺ PCa cell population in the LAPC9 model is further enriched in PCSCs.³⁶ Other PCSCs have been uncovered via different cell surface markers such as CD133,^43,87 CXCR4^43,54 and TRA-1-60/CD151/CD166.⁴⁹

Marker-independent approaches have also been employed to enrich PCSCs. For instance, SP cells are much more tumorigenic (> 100-fold) than non-SP cells in the LAPC9 model.³⁴ ALDH⁺ PCa cells possess high tumorigenic and metastatic potential.^47,48 A recent study has shown that HPCa samples contain cells capable of establishing serially passagable prostaspheres, which bear the phenotype of CD49b⁺CD49f⁺CD44⁺DeltaNp63⁺Nestin⁺CD133⁺.⁴⁶ PCa cell holoclones harbor long-term self-renewing CSCs in contrast to cells in either meroclones or paraclones.⁴² Using a lentiviral reporter system in long-term PCa cell lines (e.g., LNCaP), xenograft models (e.g., LAPC9, LAPC4) and primary HPCa samples, we have recently shown that the less differentiated PCa cell population expressing little differentiation marker PSA (i.e., PSA^-/lo) harbors long-term self-renewing PCSCs that resist castration.⁸ Compared with the isogenic PSA⁺ PCa cells, the PSA^-/lo cells possess multiple SC features. First, PSA^-/lo PCa cells are more clonal and clonogenic (sphere-forming). Second, PSA^-/lo cells are more quiescent and slow cycling. Third, PSA^-/lo cells overexpress dozens of SC-related genes (e.g., Nanog, CD44, SPP1, etc.). Fourth, PSA^-/lo cells are more resistant to stress (e.g., paclitaxel) and castration. Fifth, PSA^-/lo cells regenerate serially transplantable tumors that histologically resemble parental tumors. Finally, a small population of PSA^-/lo cells can undergo asymmetric cell division to self-renew and generate PSA⁺ cells. Importantly, the PSA^-/lo PCa cell population is still heterogeneous with true PCSCs that possess high castration-resistant tumor-regenerating activity, representing an even smaller subpopulation. In support of this conjecture, we could further fractionate a subpopulation of ALDH⁺CD44⁺α2β1⁺ cells from the PSA^-/lo cell population, and show that these cells could regenerate serially transplantable tumors in fully castrated mice.⁸ Another recent study found a small subset of CK18^-/CK19^- cells in docetaxel-resistant cell lines (Du145-DR and 22Rv1-DR) and primary as well as metastatic PCa patient samples that express low levels of differentiation marker (HLA).¹¹ Using a lentiviral reporter system, this group has shown that CK18^-/CK19^- cells in both Du145-DR and 22Rv1-DR models can survive docetaxel exposure and acquire chemoresistance via upregulation of Notch and Hedgehog signaling, whereas targeting these two pathways can abolish the chemoresistance both in vitro and in vivo.¹¹ Of note, a small subset of HLA^- cells from both xenografts and primary patient samples are much more tumorigenic than the bulk HLA⁺ cells,¹¹ suggesting that HLA^- PCa cells may be enriched in PCSCs that are chemoresistant. Taken together, these studies^8,11 imply that less differentiated PCa cells (e.g., PSA^-/lo, HLA⁺), compared with the bulk differentiated PCa cells, may be highly enriched in PCSCs that drive tumor development and mediate resistance to both chemotherapy and castration.

Recently accumulated evidence has implicated several molecules and signaling pathways in regulating PCSC activity. For example, Nanog, a homeodomain transcription factor that plays a vital role in regulating ES cell pluripotency, seems to be also important for PCSC activity and tumorigenicity. Thus, its expression is enriched in CD44⁺ PCa cells, and its knockdown inhibits tumor regeneration, whereas its inducible overexpression confers castration-resistant PCa development.^40,41 An unbiased microRNA (miRNA) library screening led to the identification of a tumor-suppressive microRNA, miR-34a, that was depleted in the CD44⁺ PCa cells.⁵⁰ Subsequent studies reveal that miR-34a functions as a powerful negative regulator of CD44⁺, tumorigenic and metastatic PCSCs by directly targeting CD44.⁵⁰ The above discussions illustrate a positive (i.e., Nanog) and negative (i.e., miR-34a) regulator of PCSCs. “Less differentiated” CK18^-/CK19^- (HLA^-) PCa cells are highly resistant to docetaxel treatment as a result of activation of both Notch and Hedgehog pathways.¹¹ Consequently, targeting these two signaling pathways abrogates docetaxel resistance in CK19^- cells via inhibition of AKT and Bcl-2.¹¹ Other work has generated evidence that PCSCs may also be regulated by the PTEN/PI3K/AKT⁴⁵ and NFκB pathways.⁴⁹ Potentially, these reported regulators of PCSCs may all become therapeutic targets (for positive regulators) or tools (for negative regulators such as miR-34a) for the treatment of PCa, especially in the context of CRPC.

The majority of the aforementioned studies on human PCSCs employed xenograft transplantation models with the knowledge that the residual immune response in recipient mice may confound data interpretation due to species incompatibility. Therefore, studies on mouse PCSCs may circumvent this problem. The Sca-1⁺, but not Sca-1^-, murine prostate cells, upon lentiviral-mediated overexpression of AKT1, can initiate PIN,⁷⁰ a precursor lesion to PCa, suggesting that Sca-1⁺ cells with high AKT expression/activity could function as the cells-of-origin for PCa development. In the Pten-null mouse PCa model, a small population of PCa cells manifested as Lin^-Sca-1⁺CD49f^hi (i.e., LSC) possess PCSC properties, as these cells have high sphere-forming and tumorigenic potential compared with other isogenic subpopulations.⁸⁸ As with the human PCSCs, potential regulators of mouse PCSCs have also been reported. For example, protospheres from Pten/p53-null mice harbor tumorigenic stem/progenitor cells with heightened levels of AKT and AR signaling pathways.⁸⁹ Bmi-1, a Polycomb group transcriptional repressor, helps to maintain the self-renewal capacities of adult mouse p63⁺ PSCs⁹⁰ and, of note, inhibition of Bmi-1 appears to slow down PCa initiation in Pten-null mice.⁹⁰ This study⁹⁰ also suggests that PCSC properties may be regulated at the chromatin level. Recently, the same group reported that the intracellular domain of Trop2, an EpCAM family member of cell surface receptors, is important for regulating the self-renewal capacity of normal PSCs and, when overexpressed, is sufficient to initiate PIN in vivo,⁹¹ implying that Trop2 is not just a phenotypic marker but may be functionally indispensible for the activity of normal PSCs and PCSCs.

Cell-of-Origin for PCa and PCSCs

In the past decade, the topic of ”cell-of-origin” for PCa has attracted enormous attention in the field. As of now, which subpopulation of the cells represents the real PCa cell-of-origin is still under debate. At the histopathalogical level, > 95% of untreated primary PCa present as acinus carcinoma (i.e., adenocarcinoma) in which luminal-like cells expressing AR and PSA predominate and basal-like cells are rare.¹ Hence it has been presumed that normal prostate luminal cells may serve as an oncogenic transformation target for PCa. There are several pieces of evidence to support this postulation. For instance, Shen’s group deleted Pten in the CARNs of castrated host (Nkx3.1^CreERT2/+;Pten^+/+), and found that such a deletion leads to high-grade PIN and carcinoma after regression-regeneration cycles,⁷⁶ suggesting that the CARNs may serve as the cell-of-origin for PCa. Choi et al. used the lineage-tracing strategy in both basal-specific (K14-CreER^Tg/Tg; mTmG^Tg/Tg) and luminal-specific (K8-CreERT2^Tg/Tg; mTmG^Tg/Tg) inducible Cre systems, and observed that prostate luminal cells are more susceptible to direct malignant transformation upon Pten deletion, whereas basal cells appear to need to first differentiate into the transformation-competent luminal cells before oncogenic transformation can take place.⁷⁷ Interestingly, transgenic overexpression of certain gene(s) (e.g., PKCε) in murine prostatic luminal cellular compartment could also lead to PIN.⁹² Taken together, these animal model studies suggest that murine prostate luminal cells can function as the cells of origin for PCa.

By contrast, studies using tissue recombination/transplantation assays suggest that prostate basal cells are more likely the targets of malignant transformation. Lawson et al. have reported that overexpression of transcription factor ERG1, the fusion partner of TMPRSS2, in murine prostate basal/stem cells results in dysplasia and PIN, whereas the similar phenotype cannot be seen with luminal or stromal cells.⁹³ Moreover, they found that combinatorial overexpression of AKT and AR in murine basal/stem cells, but not luminal cells, leads to poorly differentiated carcinoma.⁹³ Remarkably, when overexpressing ERG, AKT and AR in benign human prostate basal cells (CD49f^hiTrop2^hi) and luminal cells (CD49f^loTrop2^hi), only basal cells are susceptible to malignant transformation and can initiate PCa in immunodeficient mice, regenerating PCa resembling patient tumors histologically.⁹⁴ A recent report has shown that recombination of cancer-associated fibroblasts with integrin α2β1^hi human prostate basal cells (from non-tumorigenic BPH-1 cells) regenerates tumor grafts.⁹⁵ In summation, these results suggest that both human and murine prostate basal cells can serve as cells-of-origin for PCa.

Regardless of the cell-of-origin for PCa, it is critical that the concept must not be confused with PCSCs. In former studies, we generally focus on a subpopulation of normal prostate cells that has the potential to serve as the cellular targets of malignant transformation to become a cancer cell upon specific genetic alteration(s) and in certain experimental models. However, in the latter studies, PCSCs are referred to as the subsets of cancer cells in established tumors that possess certain SC activities. Certainly, it is possible that the cells of origin for PCa may have acquired SC features and have thus become PCSCs.

Studies of CSCs in other cancer types suggest that CSCs may originate from their normal counterparts for normal SCs, and CSCs in some tissues seem to share phenotypic markers.^23,24 Nevertheless, CSCs may also originate from progenitors or differentiated cells.^96,97 Interestingly, recent work from our lab and others has hinted that PCSCs appear to be generally less differentiated, manifested by no or low levels of expression of differentiation makers such as PSA⁸ and CK18/CK19 (HLA).¹¹ Moreover, the abundance of immature PCSCs seems to correlate with tumor aggressiveness,^8,11 consistent with CSCs in other tumors.⁹⁸

PCSCs in CRPC

Both androgen and AR are crucial in the development of normal prostate and PCa.¹ ADT is the mainstay treatment for advanced PCa patients by either surgical and/or chemical castration.¹ However, most PCa patients eventually fail ADT and develop CRPC, which is untreatable. CRPC represents one of the major clinical challenges, and the exact etiology remains elusive. Many possible mechanisms have been put forth to explain the emergence of CRPC, most of which center on AR and AR signaling and include AR amplification, AR mutation, overexpression of ligand-less AR splice isoforms and increased AR-independent and survival pathways.^1,6,7 Studies on PCSCs, however, may help explain some uncertainties and offer fresh insights regarding CRPC development.

A study from Tanaka et al. indicates that the expression levels of N-cadherin are high in CRPC xenografts and primary metastatic HPCa samples.⁹⁹ Ectopic expression of N-cadherin in androgen-dependent PCa (ADPC) cells leads to a castration-resistant phenotype, whereas specific monoclonal antibody targeting of N-cadherin dramatically delays CRPC progression.⁹⁹ Of interest, the relative abundance of N-cadherin⁺ cells gradually increases in the late-stage CRPC xenografts.⁹⁹ These data indicate that N-cadherin⁺ cells and/or the N-cadherin molecule itself plays a vital role in CRPC development. Another recent study shows that stem-like cancer cells exist in the BM18 xenograft model, which express both SC-related markers (e.g., Nanog) and luminal markers (e.g., Nkx3.1), but not basal or neuroendocrine markers.⁵³ Moreover, cancer cells expressing a luminal phenotype can regenerate new tumors after androgen replacement,⁵³ suggesting that SC-like cancer cells with the luminal progenitor phenotypes might be candidate cells of origin for CRPC. A cell surface marker, CD166, is found to be highly expressed in human CRPC samples, and, importantly, LSC^hiCD166^hi cells in the Pten-null model have much enhanced sphere-formation abilities compared with the other subpopulations,⁵⁵ suggesting that the CD166^hi cell population may be enriched in castration-resistant PCSCs. Work from our lab indicates that PSA^-/lo PCa cells are more clonal and clonogenic than the corresponding bulk PSA⁺ PCa cells in androgen-deficient conditions in vitro and are more tumorigenic than PSA⁺ cells in fully castrated mice.⁸ These observations suggest that the PSA^-/lo PCa cell population likely harbors castration-resistant variants that pre-exist in the tumors. Remarkably, castration-resistant PCa cells can be further enriched in the PSA^-/lo cell population by fractionating for ALDH⁺CD44⁺α2β1⁺ cells, which are significantly more tumorigenic and can regenerate serially transplantable tumors in fully castrated hosts,⁸ raising the possibility that such cells may be future therapeutic targets for CRPC.

Challenges and Perspective

The newly revived CSC concept has attracted much attention, and CSCs have been studied in various tumor systems, including PCa.^12-14,17-22 Elucidating the phenotypic and functional properties as well as molecular regulators of PCSCs should help us to better understand the etiology of and mechanisms responsible for PCa development. As in other tumors, many seemingly divergent PCSC populations have been reported, and one of the main challenges is to delineate the interrelationship between various phenotypically different PCSCs.

Most PCSCs so far reported have been based on studies in long-term cultured cell lines, xenograft models or mouse PCa models.¹³ For example, we have identified distinct subpopulations of PCa cells, mostly in several well-characterized xenograft models, that are enriched in PCSC activity including CD44⁺,³⁵ CD44⁺α2β1⁺³⁶ and PSA^-/lo cells.⁸ Little is known concerning whether human primary PCa also harbors tumorigenic SC-like cancer cells, and whether different patient tumors may have distinct PCSCs. The paucity in our knowledge in this regard is related to one of the major hurdles well-known in the field, i.e., it is extremely difficult to reconstitute HPCa development in an immunodeficient host by using single HPCa cells,¹⁰⁰ explaining why cell lines and xenograft models are typically used for PCSC studies. Future efforts should be directed to developing reliable tumor reconstitution protocols, so that this critical challenge can be directly tackled.

Tissue recombination assays have shown that both normal murine⁹³ and human⁹⁴ prostate basal cells appear to be the primary cells-of-origin for PCa. In contrast, lineage-tracing studies demonstrate that murine prostate luminal cells seem to be more competent in functioning as the cells-of-origin for PCa.^76,77 It is presently unclear why the two experimental approaches give rise to different results and conclusions. Suffice it to say that we should not exclude the possibility that human prostate luminal cells may also be susceptible to oncogenic transformation, given that similar lineage-tracing strategy is not feasible in humans. Furthermore, development of a specific culture medium favoring human prostate luminal cells is needed to conduct side-by-side comparisons in assays based on tissue recombination and transplantations.

CRPC represents one of the most challenging stages for PCa treatment. Although castration-resistant PCSCs have not yet been uncovered in primary human PCa samples, recent studies have provided hints that some PCSC subpopulations may express low levels of AR and intrinsically be resistant to castration.^8,11,55,99 It is conceivable that the expansion of such cells may promote CRPC development. As a result, these castration-resistant PCa cells may represent potential cellular targets for novel drug development. Future studies with improved tumor-reconstitution protocols and model systems are required to identify castration-resistant PCSCs in patient tumors at different clinical stages, and the success of which will facilitate future therapeutic development and benefit PCa patients.

Glossary

Abbreviations:

ADT

androgen-deprivation therapy

ADPC

androgen-dependent prostate cancer

AIPC

androgen-independent prostate cancer

AR

androgen receptor

CARNs

castration-resistant Nkx3.1-expressing cells

CK

cytokeratin

CRPC

castration-resistant prostate cancer

CSCs

cancer stem cells

FACS

florescence-activated cell sorting

HPCa

(primary) human prostate cancer

LSCs

Lin^-Sca-1⁺CD49f⁺ cells

miRNA

microRNA

PCSCs

prostate cancer stem cells

PIN

prostate intraepithelial neoplasia

PSA

prostate specific antigen

PSCs

prostate stem cells

SCs

stem cells

SP

side population

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.