Molecular genetics of prostate cancer: new prospects for old challenges
- Michael M. Shen1,2,3,6 and
- Cory Abate-Shen3,4,5,7
- 1Department of Medicine, Columbia University Medical Center, New York, New York 10032, USA;
- 2Department of Genetics and Development, Columbia University Medical Center, New York, New York 10032, USA;
- 3Herbert Irving Comprehensive Cancer Center, Columbia University Medical Center, New York, New York 10032, USA;
- 4Department of Urology, Columbia University Medical Center, New York, New York 10032, USA;
- 5Department of Pathology and Cell Biology, Columbia University Medical Center, New York, New York 10032, USA
Abstract
Despite much recent progress, prostate cancer continues to represent a major cause of cancer-related mortality and morbidity in men. Since early studies on the role of the androgen receptor that led to the advent of androgen deprivation therapy in the 1940s, there has long been intensive interest in the basic mechanisms underlying prostate cancer initiation and progression, as well as the potential to target these processes for therapeutic intervention. Here, we present an overview of major themes in prostate cancer research, focusing on current knowledge of principal events in cancer initiation and progression. We discuss recent advances, including new insights into the mechanisms of castration resistance, identification of stem cells and tumor-initiating cells, and development of mouse models for preclinical evaluation of novel therapuetics. Overall, we highlight the tremendous research progress made in recent years, and underscore the challenges that lie ahead.
Keywords
In 2009, there were ∼192,280 new cases of prostate cancer reported and 27,360 related deaths in the United States (American Cancer Society 2009). Although the age-adjusted rate of cancer deaths has decreased steadily in the past 10 years, prostate cancer remains the second leading cause of cancer death in men. Since we last reviewed this topic 10 years ago (Abate-Shen and Shen 2000), there have been numerous advances in basic research on prostate cancer initiation and progression, as well as new clinical advances that have improved patient outcome. Below, we review the principal features of prostate cancer, highlighting key molecular events of initiation and progression and major targets for clinical intervention. When feasible, we cited primary references for the key findings discussed, particularly those published within the past 10 years.
Major clinical challenges in prostate cancer
Prostate cancer has been recognized as a clinical entity since antiquity, when it was first described by the ancient Egyptians, while surgical procedures to remove the prostate were developed >100 years ago (Capasso 2005). However, the availability of a highly accessible blood test for prostate-specific antigen (PSA) has revolutionized the diagnosis of prostate cancer over the past three decades. PSA is a kallikrein-related serine protease that is produced in normal prostate secretions, but is released into the blood as a consequence of disruption of normal prostate architecture (Lilja et al. 2008).
Men that have elevated PSA levels typically undergo biopsy to assess the potential presence of prostate cancer. Following biopsy, histopathological grading of prostate tissue is performed by Gleason scoring, which classifies tumors from 1 to 5 (most to least differentiated) based on their most prevalent architecture, and assigns a combined score that is the sum of the two most common patterns (Mellinger et al. 1967; Epstein 2010). Patients are also diagnosed by the status of their primary tumors, from organ-confined to fully invasive (T1–4), with or without lymph node involvement (N0 or 1), and the presence and degree of distant metastases (M0 and 1a–c) (Ohori et al. 1994). If prostate cancer is diagnosed, conventional treatment regimens include surgical excision of the prostate (radical prostatectomy), or irradiation through external beam therapy or implantation of radioactive “seeds” (brachytherapy). In the case of advanced cancer, these regimens are usually followed or substituted with androgen deprivation therapy, which initially will reduce tumor burden and/or circulating PSA to low or undetectable levels, but ultimately the disease will recur in most cases.
At present, there are several major clinical challenges associated with this conventional paradigm for prostate cancer diagnosis and treatment. Each of these significantly impacts the effective management of prostate cancer, and is the subject of investigations in basic research on prostate tumor biology.
Distinguishing indolent vs. aggressive disease
Recent changes in recommendations that now suggest later and less frequent PSA screenings highlight a major clinical challenge for prostate cancer diagnosis and treatment (Wolf et al. 2010). These new recommendations were proposed because the widespread use of PSA testing has led to a vast increase in the diagnosis of patients with clinically localized low Gleason grade carcinomas that may not require treatment, since their tumors are relatively indolent. In particular, patients with a Gleason pattern of 3 or less almost never relapse after local therapy, and very likely can be managed conservatively with “watchful waiting”; nonetheless, a small fraction of these tumors will progress rapidly and require immediate treatment (Albertsen et al. 2005; Eggener et al. 2007; Lu-Yao et al. 2009).
Consequently, a major clinical challenge is posed by the current inability to readily distinguish indolent from aggressive tumors in prostate cancer patients who present with low Gleason grade tumors upon biopsy (Sartor et al. 2008). The absence of this prognostic information has led to a significant “overtreatment” of patients who would otherwise require only conservative management. Thus, the impact of treatment on prostate cancer survival is small, most likely because overdiagnosis and overtreatment dilutes the benefits of therapy for those who require intervention. This prognostic challenge could be addressed by better understanding of the molecular basis of cancer initiation, which should ultimately lead to the identification of biomarkers that distinguish between indolent and aggressive forms of prostate cancer. At present, however, available panels of molecular biomarkers do not provide greater prognostic significance than Gleason grade determination (True et al. 2006).
Castration-resistant prostate cancer
Circulating androgens are essential for normal prostate development as well as the onset of prostate cancer through their interactions with the androgen receptor (AR). As shown by Huggins and colleagues in the 1940s (Huggins and Hodges 1941), removal of testicular androgens by surgical or chemical castration will lead to regression of prostate tumors. However, androgen depletion is usually associated with the recurrence of prostate cancer, as monitored by rising PSA levels, and this recurrent disease is termed “castration resistant.” (The term “castration resistance” has generally replaced “androgen independence” in usage, as it has become apparent that advanced prostate cancer remains dependent on AR function, as discussed below.) Unfortunately, castration-resistant prostate cancer has been essentially untreatable, with the most effective standard chemotherapeutic regimens resulting in a mean increase in survival time of 2 mo (Petrylak et al. 2004; Tannock et al. 2004). Therefore, a second major clinical challenge that could be significantly impacted by basic research in prostate cancer biology is the elucidation of pathways of castration resistance, which could lead to the identification of new therapeutic approaches.
Bone tropism of prostate cancer metastasis
A third major clinical challenge corresponds to the propensity for advanced prostate cancer to metastasize to bone, which is primarily responsible for its effect on patient morbidity as well as mortality. Thus, unlike other epithelial tumors that occasionally metastasize to bone, metastatic prostate cancer almost invariably metastasizes to bone, and furthermore displays characteristic osteoblastic rather than osteolytic lesions (Logothetis and Lin 2005). Despite the clinical relevance of bone metastasis, the molecular mechanisms that underlie the bone tropism of prostate cancer are not well understood. This gap in knowledge is due in part to difficulties in obtaining metastatic tissue from patients, as well as to difficulties in generating mouse models that display bone metastasis.
At present, relatively little is known about the molecular mechanisms underlying the bone tropism of prostate cancer metastasis. Experimental models for investigation of bone metastases are limited to a small range of xenograft models that typically rely on intracardiac or intratibial injection of highly transformed tumor cells to induce metastases (Corey et al. 2002; Singh and Figg 2005). To date, despite the availability of genetically engineered mouse models that display secondary metastases, there is no autochthonous model that reliably generates bone metastases at an appreciable frequency.
Development of the prostate gland
Anatomy and histology
In men, the prostate gland is a walnut-sized tissue surrounding the urethra at the base of the bladder, and produces important components of the seminal fluid. Although the adult prostate lacks discernible lobular structure, the classic work of McNeal (1969, 1981, 1988) defined the human prostate as having a zonal architecture, corresponding to central, periurethral transition, and peripheral zones, together with an anterior fibromuscular stroma (Timms 2008). Importantly, the outermost peripheral zone occupies the most volume, and harbors the majority of prostate carcinomas. In contrast, benign prostatic hyperplasia (BPH), a common nonmalignant condition found in older men, arises from the transition zone.
Unlike the human prostate, the mouse prostate consists of multiple lobes that have distinct patterns of ductal branching, histological appearance, gene expression, and secretory protein expression (Cunha et al. 1987). These correspond to the ventral, lateral, dorsal, and anterior lobes, with the dorsal and lateral lobes often combined as the dorsolateral lobe for analysis. Although it is sometimes asserted that the mouse dorsolateral lobe is most analogous to the human peripheral zone, particularly with respect to prostate cancer, there is no consensus agreement among pathologists to support this conclusion (Shappell et al. 2004). However, analyses of gene expression profiling data support the idea that the dorsolateral lobe is most similar to the peripheral zone of the human prostate (Berquin et al. 2005).
At the histological level, both the mouse and human prostate contain a pseudostratified epithelium with three differentiated epithelial cell types: luminal, basal, and neuroendocrine (CS Foster et al. 2002; van Leenders and Schalken 2003; Hudson 2004; Shappell et al. 2004; Peehl 2005). The luminal epithelial cells form a continuous layer of polarized columnar cells that produce protein secretions and express characteristic markers such as cytokeratins 8 and 18, as well as high levels of AR. Basal cells are located beneath the luminal epithelium, and express p63 and the high-molecular-weight cytokeratins 5 and 14, but express AR at low or undetectable levels. Finally, neuroendocrine cells are rare cells of unknown function that express endocrine markers such as chromogranin A and synaptophysin, but are AR-negative.
Epithelial–mesenchymal interactions
The prostate is an endodermal tissue that arises during late embryogenesis through ductal budding from the anterior urogenital sinus epithelium. Formation of the prostate is an inductive event that requires reciprocal interactions between the urogenital sinus mesenchyme and epithelium, and is dependent on testicular androgen synthesis. The fundamental parameters of these epithelial–mesenchymal interactions were defined in classical tissue recombination studies by Cunha and colleagues (Cunha et al. 1987; Cunha 2008). These studies demonstrated that an AR-dependent signal from the urogenital mesenchyme is required for prostate formation, while AR is not initially required in the urogenital epithelium for prostate organogenesis, but is subsequently necessary for epithelial differentiation and secretory protein expression. Thus, androgens act indirectly on the urogenital mesenchyme to mediate prostate induction. These findings have been subsequently confirmed by conditional gene targeting of AR in the prostate epithelium (Wu et al. 2007).
More recently, molecular analyses have implicated several developmental signaling pathways in mediating epithelial–mesenchymal interactions during prostate organogenesis, including the Wnt, fibroblast growth factor (FGF), and Hedgehog pathways (Marker et al. 2003; Prins and Putz 2008). For example, ligands and inhibitors for the canonical Wnt/β-catenin as well as noncanonical pathways are expressed in both epithelial and mesenchymal compartments during early prostate organogenesis (Pritchard and Nelson 2008), and abrogation of noncanonical Wnt5a signaling leads to defects in ductal morphogenesis (Huang et al. 2009). In addition, the FGF pathway is clearly required for prostate formation, as null mutants for the mesenchymally expressed Fgf10 mostly lack prostate budding (Donjacour et al. 2003), while conditional deletion of Fgfr2, which encodes the receptor for FGF10, or the downstream signaling component Frs2α in prostate epithelium results in defects in branching morphogenesis (Lin et al. 2007; Zhang et al. 2008). Finally, the Hedgehog signaling pathway is also involved in prostate formation, as the Shh ligand is expressed in urogenital epithelium; the downstream components Smo, Ptc1, and Gli1 are expressed in urogenital mesenchyme (Lamm et al. 2002; Freestone et al. 2003; Berman et al. 2004); and loss of Shh pathway activity results in loss of prostate formation and/or defective ductal branching (Podlasek et al. 1999; Freestone et al. 2003; Berman et al. 2004). However, it remains unclear whether these phenotypes are mediated directly through redundant ligands functioning through the Hedgehog pathway (Doles et al. 2006), or indirectly through a reduction in androgen signaling (Freestone et al. 2003; Berman et al. 2004).
Natural history of prostate cancer
Latent and clinical cancer
Prostate cancer is generally regarded as multifocal, since primary tumors often contain multiple independent histologic foci of cancer that are often genetically distinct (Aihara et al. 1994; Bostwick et al. 1998; Macintosh et al. 1998; Mehra et al. 2007a; Clark et al. 2008). In contrast, despite the phenotypic heterogeneity of metastatic prostate cancer (Shah et al. 2004), molecular and cytogenetic analyses show that multiple metastases in the same patient are clonally related, indicating that advanced prostate cancer is monoclonal (Mehra et al. 2008; Liu et al. 2009). These findings suggest that metastatic prostate cancer may arise from the selective advantage of individual clones during cancer progression; however, this process of clonal evolution may also represent the consequence of therapeutic interventions such as androgen deprivation, which may differentially target cells of varying malignant potential.
The heterogeneity of prostate cancer is potentially relevant for understanding the distinction between latent and clinical disease, and the strong correlation between prostate cancer progression and aging (Fig. 1). Although prostate cancer is a disease of older men, studies of prostate specimens from healthy men in their 20s to 40s show the frequent presence of histologic foci of prostate cancer (Yatani et al. 1989; Sakr et al. 1994; Shiraishi et al. 1994), suggesting that cancer initiation has already taken place at a relatively early age. Combined with the evidence that prostate cancer is multifocal, it appears that the prostate gland can be the site of multiple neoplastic transformation events, many of which give rise only to latent prostate cancer that does not progress to clinically detectable disease. It is conceivable that clinical prostate cancer initiates from a different pathogenic program than latent prostate cancer. Alternatively, most latent prostate cancer foci may not undergo critical activating events that lead to clinical disease, or may remain under active suppression sufficient to maintain these foci in a subclinical state. As discussed above, the advent of PSA screening has led to a vast increase in the diagnoses of prostate cancer, many of which presumably represent latent or indolent forms of the disease that at present are difficult to distinguish from cancers that will become more aggressive; this highlights the critical need for improved molecular markers and/or other approaches to augment the histological assessment of prostate cancer for more effective diagnosis and management.
Progression pathway for human prostate cancer. Stages of progression are shown, together with molecular processes and genes/pathways that are likely to be significant at each stage. Adapted from Abate-Shen and Shen (2000).
Prostatic intraepithelial neoplasia (PIN) and prostate cancer
It is widely accepted that PIN represents a precursor for prostate cancer, although this relationship has not been demonstrated conclusively (Bostwick 1989; DeMarzo et al. 2003). PIN is generally characterized at the histological level by the appearance of luminal epithelial hyperplasia, reduction in basal cells, enlargement of nuclei and nucleoli, cytoplasmic hyperchromasia, and nuclear atypia; in addition, high-grade PIN lesions generally display marked elevation of cellular proliferation markers (Bostwick 1989; Shappell et al. 2004). In contrast with prostate cancer, however, basal cells are reduced in number in PIN, but are not absent.
Although human prostate cancer displays significant phenotypic heterogeneity, >95% of prostate cancers are classified pathologically as adenocarcinoma, which has a strikingly luminal phenotype (Fig. 2). In biopsy specimens, prostate adenocarcinoma diagnosis can be confirmed by the absence of immunostaining using p63 and cytokeratin 5/14 antibodies, both of which detect basal cells (Humphrey 2007; Grisanzio and Signoretti 2008). In addition, a diagnosis of prostate cancer is supported by elevated immunostaining for α-methylacyl-CoA racemase (AMACR), a luminal marker that is overexpressed in carcinoma (Luo et al. 2002; Jiang et al. 2005; Humphrey 2007). Similarly, prostate cancer arising in many mouse models displays a relatively luminal phenotype (Kim et al. 2002d; X Ma et al. 2005). However, the overt histological appearance of prostate carcinoma in most genetically engineered mouse models often differs from that of typical human prostate cancer (Fig. 2).
Histopathology of human and mouse prostate cancer. (A–D) Hematoxylin-eosin-stained sections of human prostate. (A) Benign normal tissue, with representative basal (bas) and luminal (lum) cells indicated. (B) PIN; arrows indicate regions of hyperplastic epithelium. (C) Well-differentiated adenocarcinoma. (D) Poorly differentiated adenocarcinoma. (E–H) Hematoxylin-eosin-stained sections of anterior prostate from genetically engineered mouse models. (E) Normal tissue, with characteristic papillary tufting (arrowheads). (F) High-grade PIN. (G) Prostate carcinoma. (H) Prostate carcinoma with an invasive phenotype. We thank Dr. Robert Cardiff and Alexander Borowsky (School of Medicine, University of California at Davis) for providing images of human prostate specimens.
Subtypes of prostate cancer
A notable difference between prostate cancer and other epithelial tumors, such as breast cancer, is the lack of distinguishable histopathological subtypes that differ in their prognosis or treatment response. The vast majority of prostate cancers correspond to acinar adenocarcinomas that express AR, while other categories of prostate cancer—such as ductal adenocarcinoma, mucinous carcinoma, and signet ring carcinoma—are extremely rare (Grignon 2004). Perhaps the most significant histological variant is neuroendocrine prostate cancer, which is generally classified as either small cell carcinoma or a carcinoid tumor, and represents <2% of prostate cancer cases (Grignon 2004). However, focal regions of neuroendocrine differentiation are more commonly observed in prostate adenocarcinoma, particularly following recurrence after prostatectomy and androgen deprivation therapy (Yuan et al. 2007; Komiya et al. 2009), and expression of the neuroendocrine marker chromogranin A is associated with the development of castration-resistant tumors and shortened time to disease recurrence (Kokubo et al. 2005; Berruti et al. 2007). This prevalence of neuroendocrine differentiation after recurrence may be due to the lack of AR expression by neuroendocrine cells, which are inherently castration-resistant.
While evidence of major subtypes of prostate cancer is lacking at the histopathological level, recent genomic analyses have provided increasing evidence for molecularly defined subtypes (Tomlins et al. 2008b; Palanisamy et al. 2010; Taylor et al. 2010). In particular, expression profiling analyses of prostate cancer specimens have not strictly defined molecular signatures associated with distinct cancer subtypes that specifically correlate with disease outcome (Singh et al. 2002; Lapointe et al. 2004; Tomlins et al. 2007b). However, oncogenomic pathway analyses that integrate analyses of gene expression, copy number alterations, and exon resequencing may provide a unified approach for distinguishing prostate cancer subtypes and stratifying patient outcome (Taylor et al. 2010). Furthermore, the existence of molecular subtypes has been supported by analyses of chromosomal rearrangements associated with prostate cancer; for example, prostate cancers containing the TMPRSS2-ERG translocation may be distinct from those that up-regulate SPINK1, which encodes a secreted trypsin inhibitor (Mehra et al. 2007a; Tomlins et al. 2008b). Ongoing studies will undoubtedly assess whether these subtypes correlate with disease outcome or treatment response.
Metastasis
As noted above, although common sites of secondary metastasis for prostate cancer are lung, liver, and pleura, if prostate cancer metastasizes, it invariably goes to bone, where it forms characteristic osteoblastic lesions (Bubendorf et al. 2000; Logothetis and Lin 2005). Given the clinical importance of metastasis for patient outcome, the ability of prostate tumor cells to disseminate into the bone marrow and peripheral blood has been investigated in detail. Notably, a recent study showed that bona fide circulating tumor cells could be detected in the bone marrow of a significant proportion of patients with localized disease, suggesting that disseminated tumor cells have not attained full metastatic capability (Holcomb et al. 2008). Consistent with this interpretation, circulating tumor cells from patients with metastatic disease show multiple chromosomal rearrangements typical of advanced prostate cancer, consistent with genomic instability acquired during cancer progression (Holcomb et al. 2008; Attard et al. 2009c; Leversha et al. 2009). However, the relationship of disseminated tumor cells to the formation of metastases remains unresolved, and the molecular factors that promote metastases of prostate cancer to bone are poorly defined.
Mouse models of prostate cancer
Xenograft models
Traditionally, in vivo studies of prostate cancer have extensively used xenograft models of human prostate cancer, using cell lines or prostate tumors implanted into immunodeficient mice, either orthotopically into the prostate or transplanted onto the flank. For instance, xenografts derived from LNCaP cells have been used to generate genetically-related lines that vary in their androgen responsivity and metastatic potential (Thalmann et al. 2000). Similarly, several xenografts have been developed by transplantation of human prostate tumors, including the LuCaP and LAPC series, which display a spectrum of prostate cancer phenotypes (Ellis et al. 1996; Craft et al. 1999b). Several of these human tissue xenografts have also given rise to prostate cancer cell lines, such as the VCaP line, which was derived from a bone metastasis (Korenchuk et al. 2001).
Analyses of xenografts have yielded a vast amount of information about molecular mechanisms of prostate cancer, and have been useful for chemotherapeutic approaches. However, xenograft models are limited by a heterologous microenviroment (since human cells/tissues are grafted in mice), an inability to analyze stromal components (unless orthotopic grafting is employed), the lack of endogenous immune response (since the host mice are immunodeficient), and the lack of diversity of available established cell lines. In particular, the limited number of available prostate cancer cell lines is likely related to inherent difficulties in culturing luminal epithelial cells (Peehl 2005) (existing prostate cell lines have been reviewed in detail) (Sobel and Sadar 2005a,b). Of particular concern is that existing cell lines may have uncertain origins, as has been demonstrated for at least one “prostate” cell line that was actually derived from a bladder carcinoma line (van Bokhoven et al. 2001). Additionally, cell lines may have anomalous molecular properties (e.g., loss of AR expression and lack of TMPRSS2-ERG fusions) when compared with most human prostate tumors (Sobel and Sadar 2005a,b), which may limit their applicability. Nonetheless, xenograft systems remain popular for studies of chemotherapeutics, primarily due to their ease of use. Moreover, since they are of human origin, xenografts may be more likely to recapitulate molecular events involved in human prostate tumorigenesis than other experimental models.
Tissue reconstitution models
The tissue recombination and renal grafting methods originally developed to study epithelial–mesenchymal interactions during prostate organogenesis can be extended for investigation of prostate tumorigenesis in vivo. In particular, immortalized human and mouse prostate epithelial cell lines are available that can reconstitute benign prostate tissue with relatively normal histology following recombination with rodent embryonic urogenital mesenchyme and grafting into immunodeficient recipients (Hayward et al. 2001; Gao et al. 2004a; Jiang et al. 2010). Using such cell lines, gene expression can be directly manipulated in culture by overexpression or knockdown methods, followed by analysis of potential tumor phenotypes in vivo. Similarly, stromal components can also be investigated in tissue recombinants using immortalized urogenital mesenchyme cell lines to facilitate genetic manipulation (Shaw et al. 2006), or using carcinoma-associated fibroblasts (Olumi et al. 1999). Furthermore, this overall approach can been greatly extended by efficient lentiviral infection of dissociated prostate epithelial cells, followed by tissue recombination, allowing for overexpression and knockdown approaches for analysis of gene function (Xin et al. 2003, 2006; Zong et al. 2009). Finally, grafting methods can be used to evaluate the relative contribution of epithelial and stromal prostatic components for cancer progression (Kim et al. 2002a; Jeong et al. 2008), as well as to study cancer phenotypes in the prostate glands of mouse strains that display embryonic lethality, even prior to prostate formation (Wang et al. 2000). Thus, tissue reconstitution methods represent powerful approaches to studying cancer mechanisms in both mouse and human prostates.
Genetically engineered models
The use of genetically engineered transgenic and knockout mice to produce autochthonous models of prostate cancer has represented a major avenue for prostate cancer investigations (Table 1). Most first-generation prostate cancer models used transgenes that overexpress potent viral oncogenes, resulting in highly aggressive disease that can often lead to metastatic cancer (Winter et al. 2003; Kasper 2005). Among these models are the well-studied TRAMP (transgenic adenocarcinoma of the prostate) mouse, which carries a minimal probasin promoter driving both SV40 large T and small t antigen (Greenberg et al. 1995), and the LADY models that use a larger probasin promoter and express large T antigen only (Masumori et al. 2001). However, transgenic mice that overexpress SV40 large T antigen typically have short latency, and develop cancer with features of neuroendocrine differentiation (Kaplan-Lefko et al. 2003; Shappell et al. 2004). Nonetheless, these first-generation models have provided numerous important insights into prostate cancer mechanisms.
Representative genetically engineered mouse models of prostate cancera
A second generation of prostate cancer models has used loss-of-function mutations in candidate genes implicated in the genesis of human prostate cancer (Table 1). Several popular models have employed null mutations in genes of interest, including Nkx3.1 and Pten (phosphatase and tensin homolog deleted from chromosome 10) (discussed below); for example, Nkx3.1; Pten double mutants show accelerated formation of high-grade PIN and invasive cancer (Kim et al. 2002d; Abate-Shen et al. 2003). Other commonly used models have employed conditional deletion mediated by the Pb-Cre4 transgene, which uses a modified probasin promoter (ARR2PB) to drive Cre expression in the prostate epithelium (Wu et al. 2001), although a potential concern is that this Cre allele also drives recombination in the stroma (X Wang and MM Shen, unpubl.). The Pb-Cre4 driver has been used by many laboratories for the conditional deletion of Pten as well as other genes of interest (Wang et al. 2003; Z Chen et al. 2005; Bruxvoort et al. 2007). Another popular Cre driver is the Nkx3.1-Cre knock-in allele, which expresses Cre recombinase specifically in the prostate epithelium, but also in several other tissues during embryogenesis (Stanfel et al. 2006; Lin et al. 2007; Thomsen et al. 2008; Zhang et al. 2008).
Despite much recent progress, current genetically engineered models have several important limitations. First, constitutive conditional gene deletion systems generally result in deletion from early stages of prostate organogenesis, and cannot be initiated in the adult, or in a stochastic manner, as is the case with somatic mutations in human cancer. Thus, the development of an inducible gene targeting system that can be induced in adult prostate epithelium would allow the investigation of gene function in situations where normal prostate organogenesis would not proceed in the absence of the gene of interest. Recent publications have described the generation of tamoxifen-inducible Cre drivers that are likely to be suitable for such approaches (Luchman et al. 2008; Ratnacaram et al. 2008; Birbach et al. 2009; Z Wang et al. 2009). Second, the use of a reversible transgene expression system, such as those driven by a tetracycline-regulated promoter, would allow the modeling of targeted therapeutic interventions on cancer growth, as well as potential acquisition of drug resistance and treatment failure. Such tetracycline-regulated models have been used successfully to investigate oncogene addiction in melanoma and mammary cancer models (Chin et al. 1999; Moody et al. 2002), but have not yet been employed in the prostate. Third, existing Cre drivers to investigate stromal function in prostate carcinogenesis are limited and nonspecific (Jackson et al. 2008), with the best available driver corresponding to the FSP1-Cre transgene (Bhowmick et al. 2004). Fourth, all current prostate cancer models use androgen-dependent promoters to drive the cancer phenotype, either directly or indirectly, and thus are poorly suited to investigate the effects of modulating androgen levels, since androgen deprivation will simultaneously down-regulate transgene expression. Finally, at present, there is no autochthonous model that reliably displays bone metastasis, which represents a major limitation in the study of advanced prostate cancer.
Nonetheless, despite their limitations, analyses of genetically engineered mouse models of prostate cancer have significantly advanced our understanding of the molecular pathways of prostate cancer initiation, progression, and castration resistance. Additionally, investigations of genetically engineered mice have led to the identification of biomarkers that can predict disease recurrence, and have provided valuable preclinical resources for investigations of novel therapies and analyses of chemopreventive agents. Examples of each of these applications are described below.
Processes that promote prostate carcinogenesis
The single most significant risk factor for prostate cancer is advanced age. While men who are younger than 40 have a one in 10,000 chance of developing prostate cancer, this risk increases to one in seven by the age of 60 (American Cancer Society 2009). However, prostate cancer is not simply a by-product of aging, since the incidence varies considerably among different populations. More likely, the relationship of prostate cancer to advanced age likely reflects the interplay of environmental, physiological, and molecular influences with normal consequences of aging that presumably exacerbate the effects of these influences. Moreover, while the precise molecular consequences of aging as they pertain to prostate cancer have not been elucidated, various studies have described gene expression changes associated with aging, particularly in the prostatic stroma, including genes involved in inflammation, oxidative stress, and cellular senescence (Begley et al. 2005; Bavik et al. 2006; Bethel et al. 2009).
Below, we discuss some of the major processes that have been implicated in prostate carcinogenesis (Fig. 1). Perhaps not surprisingly, these are interrelated and involve key regulatory molecules that have been associated with cancer initiation and progression, which will be discussed in the next section.
Inflammation
Various lines of epidemiological, pathological, and molecular evidence have supported the idea that chronic inflammation is causally linked to prostate carcinogenesis (Haverkamp et al. 2008; Klein and Silverman 2008; Bardia et al. 2009). For example, expression of certain chemokines is a predictor of biochemical disease recurrence in human prostate cancer (Blum et al. 2008). Moreover, administration of the potent heterocyclic amine PhIP (2-amino-1-methyl-6-phenyl-imidazo[4,5-b]pyridine), results in chronic inflammation and promotes prostatic hyperplasia and PIN in rodents (Borowsky et al. 2006; Elkahwaji et al. 2007; Nakai et al. 2007; Elkahwaji et al. 2009; Khalili et al. 2010). One of the most well-studied genes involved in protection against carcinogens such as PhIP is GSTP1, which encodes a member of the glutathione S-transferase family that is responsible for detoxification of reactive species, and is epigenetically silenced in a majority of prostate cancers by DNA methylation (Nakayama et al. 2004). Most notably, regions of focal atrophic prostate epithelium can often be identified in aging men, frequently in association with an inflammatory response. Such regions usually display increased epithelial proliferation, and have been termed “proliferative inflammatory atrophy” (PIA) (De Marzo et al. 1999). Regions of PIA are also often located in proximity with PIN and adenocarcinoma, and thus PIA has also been proposed to represent a precursor lesion for prostate cancer (De Marzo et al. 1999, 2003).
These observations have raised interest in the potential causes of prostate inflammation, which may include hormonal perturbations such as altered androgen and estrogen levels, or infection by bacterial or viral agents, physical trauma, or dietary factors (De Marzo et al. 2007b). Indeed, the susceptibility of the prostate gland to infection is known from the incidence of chronic bacterial prostatitis, and a potential role for bacterial infection in prostate carcinogenesis has been suggested by the identification of multiple bacterial species in most prostatectomy samples examined (Sfanos et al. 2008). Interestingly, a recent study has shown that induced bacterial prostatitis in mice can result in histological changes reminiscent of human PIA and down-regulation of expression of the Nkx3.1 homeoprotein (Khalili et al. 2010), while the levels of NKX3.1 expression in human prostate cancer cell lines have been shown to be regulated by inflammatory cytokines (Markowski et al. 2008).
Support for a role for viral infection as contributor to chronic inflammation of the prostate has been provided by studies of rare hereditary prostate cancer (HPC) that have identified several loci associated with increased familial prostate cancer risk (Schaid 2004). Of these loci, HPC1 has been shown to correspond to RNASEL, which encodes 2′-5′-oligoadenylate (2-5A)-dependent RNase L, an endoribonuclease for ssRNA that is a component of the interferon response to viral infection (Carpten et al. 2002). Further studies have shown that the R462Q variant of RNASEL confers decreased protein activity and is associated with increased risk of sporadic prostate cancer (Casey et al. 2002; Xiang et al. 2003), although the relevance of this allele as well as other RNASEL variants for sporadic prostate cancer has been disputed (Wiklund et al. 2004; Li and Tai 2006). Notably, a significant percentage of prostate tumors from patients carrying the variant RNASEL allele have been reported to contain a novel gammaretrovirus, termed xenotropic murine leukemia virus-related virus (XMRV) (Urisman et al. 2006; Dong et al. 2007), which is expressed in ∼20% of prostate cancer samples, but is not correlated with the presence of the RNASEL R462Q variant (Schlaberg et al. 2009). At present, there is considerable interest in the possibility that XMRV infection may play a role in prostate cancer initiation through stimulation of an inflammatory response, but further studies are required to replicate these initial findings and evaluate a possible causal link (Silverman et al. 2010).
Oxidative stress and DNA damage
Several lines of evidence have suggested that one of the major aging-associated influences on prostate carcinogenesis is oxidative stress and its cumulative impact on DNA damage (DeWeese et al. 2001; Khandrika et al. 2009; Minelli et al. 2009). Oxidative stress results from the imbalance of reactive oxygen species (ROS) and detoxifying enzymes that control cellular levels of ROS, which leads to cumulative damage to lipids, proteins, and DNA. The prostate appears to be exceptionally vulnerable to oxidative stress, perhaps as a consequence of inflammation, hormonal deregulation, diet, and/or epigenetic modifications such as silencing of GSTP1. Evidence linking oxidative stress and prostate cancer initiation include correlative studies showing that major antioxidant enzymes are reduced in human PIN and prostate cancer, together with a coincidental increase in the oxidized DNA adduct 8-oxy-7,8,dihydro-2′-deoxyguanosine (8-oxy-dG) (Bostwick et al. 2000). Furthermore, APE/Ref1, a multifunctional enzyme involved in redox control of key enzymes and base excision repair, is up-regulated in prostate cancer, while polymorphisms in the APE gene are associated with increased prostate cancer risk (Kelley et al. 2001; L Chen et al. 2006). Similarly, perturbations in oxidative stress response pathways have been observed in genetically engineered mouse models of prostate cancer coincident with cancer progression (Ouyang et al. 2005; Frohlich et al. 2008). Interestingly, loss of function of the Nkx3.1 homeobox gene in the mouse prostate leads to deregulated expression of oxidative damage response genes and increased levels of 8-oxy-dG, correlated with the onset of PIN (Ouyang et al. 2005), while gain of function of the NKX3.1 homeobox gene has also been shown to protect against DNA damage in prostate cancer cell lines (Bowen and Gelmann 2010). Since NKX3.1 is frequently down-regulated in early stages of prostate cancer, its inactivation may contribute to the observed vulnerability of the prostate to oxidative stress as well as to DNA damage associated with cancer initiation.
Telomere shortening
Another event that has been implicated in prostate cancer initiation is the shortening of telomeres, which is generally associated with DNA damage and may lead to chromosomal instability (Meeker et al. 2002, 2004; Vukovic et al. 2003). Telomere length has been correlated with disease outcome (Fordyce et al. 2005; Joshua et al. 2007), while prostate carcinomas as well as many high-grade PINs display increased telomerase activity, which is not observed in benign prostate tissue (Sommerfeld et al. 1996; Koeneman et al. 1998). These observations suggest that telomere length is actively modulated during prostate cancer progression, but the mechanistic relationships between telomere shortening and cancer initiation, or to induction of cellular senescence (see below), are presently unclear. Nonetheless, various strategies to regulate telomere length are being investigated as potential therapeutic agents (Asai et al. 2003; Chen et al. 2003).
Senescence
Cellular senescence corresponds to a form of cell cycle arrest in which cells remain fully viable, but are nonproliferative despite exposure to mitogenic signals (Courtois-Cox et al. 2008; d'Adda di Fagagna 2008; Evan and d'Adda di Fagagna 2009). Much recent work has identified cellular senescence as a potent mechanism of tumor suppression that prevents manifestation of the malignant phenotype after oncogenic insults. In particular, activated oncogenes are believed to induce senescence through a variety of molecular mechanisms, including replicative stress or formation of ROS, or as a response to DNA damage. Thus, oncogene-induced senescence may play a central role in preventing the progression of preneoplastic lesions to the fully malignant state.
In the prostate, cellular senescence has been shown to occur during aging-related prostate enlargement, and has been implicated as a tumor suppressor mechanism for prostate carcinogenesis. Thus, SA-β-Gal, a commonly used biomarker of senescence, is frequently detected in BPH in the human prostate (Choi et al. 2000; Castro et al. 2003). Moreover, other markers of senescence, including p14arf and p16ink4a, are increased with aging and particularly in nonmalignant cancers, suggesting these may represent markers that distinguish indolent from more aggressive forms of the disease (Zhang et al. 2006). In addition to senescence-related changes observed in epithelial cells, senescent primary prostatic fibroblasts display gene expression signatures associated with oxidative damage and DNA damage, which may in turn influence the invasive behavior of epithelial cells (Bavik et al. 2006). Notably, gene expression changes affecting oxidative damage and DNA damage responses are also observed in prostatic stroma from aged, tumor-prone rats (Bethel et al. 2009), as well as in reactive stroma from human prostate tumors (Dakhova et al. 2009).
Studies in genetically engineered mice have provided mechanistic insights into the role of senescence for prostate tumorigenesis. In particular, complete conditional inactivation of the Pten tumor suppressor gene results in PIN lesions that display a senescence phenotype that can be overcome by inactivation of p53 (Z Chen et al. 2005), but is enhanced in combination with inactivation of the Skp2 E3-ubiquitin ligase (Lin et al. 2010). Based on these findings in mouse models, one possible interpretation for the temporal difference between the occurrence of latent prostate cancer and the appearance of clinical prostate cancer is that cellular senescence may be involved in suppressing progression to aggressive disease, while additional oncogenic events may be required to bypass the senescence mechanism to promote disease progression.
Genomic alterations
Extensive genomic analyses of prostate cancer have identified copy number alterations and chromosomal rearrangements associated with prostate carcinogenesis. In particular, a number of important somatic alterations have been identified by comparative genomic hybridization (CGH) as gains or losses of chromosomal regions, including gains at 8q and losses at 3p, 8p, 10q, 13q, and 17p (Dong 2001; Lapointe et al. 2007; Taylor et al. 2010). Importantly, several of these genetic alterations have also been identified in PIN as well as PIA lesions, which has supported the precursor relationship of these lesions to prostate cancer and has emphasized their relevance for promoting cancer progression. Finally, several key regulatory genes have been mapped to within these chromosomal regions undergoing copy number alterations, including NKX3.1 at 8p21, PTEN at 10q23, and MYC at 8q24. In contrast, however, targeted resequencing studies have suggested that somatic point mutations may be relatively infrequent in prostate cancer, with tumor suppressor genes such as TP53 undergoing alterations of copy number instead (Taylor et al. 2010).
Genetic factors
Extensive efforts have been made to identify genetic susceptibility loci for prostate cancer, both through analyses of hereditary factors associated with familial risk of early-onset disease, and more recently through genome-wide association studies. In particular, prostate cancer susceptibility loci associated with HPC have been mapped to 1q24-25 (HPC1), 17p11 (HPC2), and Xq27-28 (HPCX) (Xu et al. 1998, 2001a,b), which correspond to RNASEL (HPC1), which was discussed above, and ELAC (HPC2), a gene of uncertain function. Additionally, genome-wide association studies have identified numerous single-nucleotide polymorphisms (SNPs) that are associated with cancer risk (Thomas et al. 2008; Eeles et al. 2009; Gudmundsson et al. 2009; Kader et al. 2009). In particular, a major locus identified in these genome-wide association studies is identified by multiple sequence polymorphisms at 8q24, proximal to MYC, and is discussed below. Disappointingly, however, many of the other loci identified in genome-wide association studies have not been easily replicated in other population-based studies, including analyses of groups with high-risk for prostate cancer, such as African-Americans (Hooker et al. 2010). Finally, studies of individual genetic loci have identified rare sequence polymorphisms associated with increased cancer risk, as has been shown for the C154T polymorphism at the NKX3.1 locus (Gelmann et al. 2002).
Epigenetic alterations
Epigenetic perturbations are also believed to represent important contributing factors in prostate carcinogenesis, and may provide useful biomarkers for disease progression (Li et al. 2005; Nelson et al. 2007, 2009). For example, DNA methylation has been implicated in silencing genes involved in signal transduction, hormonal response, cell cycle control, and oxidative damage response, such as GSTP1. Furthermore, prostate tumors display global changes in chromatin modification coincident with cancer progression (Kondo et al. 2008; Ke et al. 2009) that presumably result in significant perturbations in the gene expression program of tumor cells. One key modification associated with prostate carcinogenesis is trimethylation of lysine residue 27 of histone H3 (H3K27-me3), which is mediated by the histone methyltransferase enzyme Ezh2, a key oncogenic driver of advanced disease and metastasis (Varambally et al. 2002). Since the H3K27-me3 mark is associated with transcriptional repression, increased levels in prostate cancer are associated with repression of tumor suppressor genes such as DAB2IP, a member of the Ras GTPase family (H Chen et al. 2005). Global changes in histone modifications are also associated with cellular senescence, through the development of senescence-associated foci (SAHF), which include epigenetic marks of chromatin silencing (Funayama and Ishikawa 2007). In the future, global analyses of histone modifications by next-generation sequencing approaches may provide broad insights on the cumulative influences of these modifications for prostate carcinogenesis.
Molecular mechanisms of prostate cancer initiation and progression
Below, we discuss several molecular events that are believed to occur in a large percentage of prostate carcinomas, focusing on their relationships to key processes discussed in the preceding section (Fig. 1). Although each event has been associated with a possible role in cancer initiation or progression, it is unknown whether there is a temporal sequence associated with these events, or whether there is a causal relationship between them.
NKX3.1 down-regulation
Down-regulation of the NKX3.1 homeobox gene represents a frequent and critical event in prostate cancer initiation, and is likely to involve multiple mechanisms (Abate-Shen et al. 2008). NKX3.1 is localized within a 150-Mb minimal deleted region of chromosome 8p21.2 that displays loss-of-heterozygosity (LOH) in up to 85% of high-grade PIN lesions and adenocarcinomas (Emmert-Buck et al. 1995; Vocke et al. 1996; Haggman et al. 1997; Swalwell et al. 2002; Bethel et al. 2006). However, although LOH of 8p21 progressively increases in frequency with cancer grade, the remaining allele of NKX3.1 remains unmutated (Vocke et al. 1996; Voeller et al. 1997; Ornstein et al. 2001; Bethel et al. 2006). In addition, whether or not 8p21 LOH has occurred, there is substantial evidence that NKX3.1 undergoes epigenetic down-regulation, perhaps through promoter methylation (Asatiani et al. 2005). Although earlier studies had suggested that NKX3.1 expression is completely lost in advanced cancers (Bowen et al. 2000), recent analyses using a highly sensitive antibody indicate that low levels of NKX3.1 expression can be demonstrated in nearly all prostate cancers and metastases examined (Gurel et al. 2010). Thus, there appears to be a selection for reduction, but not loss, of NKX3.1 expression throughout prostate cancer progression.
These findings are highly suggestive, since Nkx3.1 has been shown to be a critical regulator of prostate epithelial differentiation and stem cell function in mouse models. During development, Nkx3.1 is expressed in all epithelial cells of the nascent prostate buds from the urogenital sinus, and represents the earliest known marker for the prostate epithelium (Bhatia-Gaur et al. 1999). In the absence of Nkx3.1, there is a significant decrease in prostatic ductal branching, as well as in production of secretory proteins (Bhatia-Gaur et al. 1999; Schneider et al. 2000; Tanaka et al. 2000). Notably, young adult Nkx3.1 heterozygous and homozygous mutants frequently display prostate epithelial hyperplasia and dysplasia, and often develop intraductal neoplasia (PIN) by 1 year of age (Bhatia-Gaur et al. 1999; Schneider et al. 2000; Tanaka et al. 2000; Abdulkadir et al. 2002; Kim et al. 2002a). These findings are consistent with the tumor suppressor activity of NKX3.1 in cell culture and xenograft assays (Kim et al. 2002a; Lei et al. 2006). Finally, recent work has shown that Nkx3.1 expression in the androgen-deprived prostate marks a rare population of prostate epithelial stem cells that is a cell of origin for prostate cancer in mouse models (Z Wang et al. 2009).
Analyses of Nkx3.1 function in human tumor cells and genetically engineered mice have provided insights into its potential roles in cancer initiation. In particular, Nkx3.1 inactivation in mice results in a defective response to oxidative damage, while its expression in human prostate cancer cell lines protects against DNA damage and is regulated by inflammation (Ouyang et al. 2005; Markowski et al. 2008; Bowen and Gelmann 2010). A causal role for Nkx3.1 in these processes has been suggested by analyses of genes that are dysregulated following perturbation of Nkx3.1 expression in mouse models or human cell lines (Magee et al. 2003; Ouyang et al. 2005; Muhlbradt et al. 2009; Song et al. 2009). These and other findings have led to a model in which NKX3.1 represents a haploinsufficient tumor suppressor gene that acts as a “gatekeeper” gene for prostate cancer initiation (Kim et al. 2002a; Gelmann 2003; Magee et al. 2003).
Myc up-regulation
It has long been known that the 8q24 chromosomal region encompassing the MYC oncogene is somatically amplified in a subset of advanced prostate tumors (Jenkins et al. 1997; Sato et al. 1999). However, recent studies have suggested a role for MYC overexpression in cancer initiation, as nuclear MYC protein is up-regulated in many PIN lesions and the majority of carcinomas in the absence of gene amplification (Gurel et al. 2008). These findings may be consistent with the identification of a major susceptibility locus at 8q24 in several large-scale genome-wide association studies of prostate cancer as well as other epithelial cancers (Amundadottir et al. 2006; Freedman et al. 2006; Gudmundsson et al. 2007, 2009; Haiman et al. 2007; Yeager et al. 2007, 2009; Al Olama et al. 2009). Multiple SNPs associated with prostate cancer risk alleles are clustered within three independent regions of a gene-poor genomic locus spanning ∼1.2 Mb between FAM84B and MYC, with MYC located ∼250 kb away from the closest SNP marker. Detailed analyses have not yet revealed any correlation between risk alleles and MYC RNA expression levels in prostate tumor samples, or the presence of any non-protein-coding genes such as microRNAs (miRNAs) (Pomerantz et al. 2009). Nonetheless, long-range regulatory elements for MYC have been identified recently in this region, raising the possibility that the risk alleles may alter the regulation of MYC expression (Jia et al. 2009; Sotelo et al. 2010). Interestingly, another recent study has found that the X-linked gene FOXP3 encodes a winged helix transcription factor that represses MYC expression (although apparently not through distant enhancer-binding sites), and itself is mutated in prostate cancer (L Wang et al. 2009).
At the functional level, transgenic mice overexpressing human MYC display rapid formation of PIN followed by progression to invasive adenocarcinoma with rare metastases (Ellwood-Yen et al. 2003), while forced expression of MYC is sufficient to immortalize nontumorigenenic human prostate epithelial cells (Gil et al. 2005). Interestingly, bioinformatic analyses identified an expression signature characterized by down-regulation of Nkx3.1 and up-regulation of Pim1, which is known to collaborate with Myc in lymphomas (Ellwood-Yen et al. 2003). Consistent with these data, lentiviral coexpression of human MYC with mouse Pim1 in tissue recombinants results in cooperative formation of carcinomas with neuroendocrine differentiation (Wang et al. 2010).
TMPRSS2-ERG translocations
Important recent studies have identified chromosomal rearrangements that activate members of the ETS family of transcription factors (ERG, ETV1, and ETV4) in the majority of prostate carcinomas (Tomlins et al. 2005, 2007a; Iljin et al. 2006; Mehra et al. 2007b; Mosquera et al. 2007; Hu et al. 2008; Rouzier et al. 2008; Saramaki et al. 2008). The most common of these rearrangements creates a TMPRSS2-ERG fusion gene, resulting in expression of N-terminally truncated ERG protein under the control of the androgen-responsive promoter of TMPRSS2 (Tomlins et al. 2005; Iljin et al. 2006; Perner et al. 2006; J Wang et al. 2006; Clark et al. 2007). As TMPRSS2 and ERG are located ∼3 Mb apart on chromosome 21q, this rearrangement occurs through either an interstitial deletion, which is more common, or an unbalanced interchromosomal translocation (Iljin et al. 2006; Perner et al. 2006). The frequency of these TMPRSS2-ERG fusions is ∼15% in high-grade PIN lesions, and ∼50% in localized prostate cancer (Clark et al. 2008; Mosquera et al. 2008; Albadine et al. 2009), suggesting that this rearrangement either occurs after cancer initiation, or alternatively corresponds to an early event that predisposes to clinical progression. Interestingly, formation of these chromosomal rearrangements may be an indirect consequence of AR function, as studies in androgen-responsive LNCaP cells have shown that AR binding induces chromosomal proximity between the TMPRSS2 and ERG loci that can lead to formation of TMPRSS2-ERG fusions following DNA damage (Lin et al. 2009; Mani et al. 2009). In addition, androgen signaling can recruit topoisomerase II to AR-binding sites, leading to induction of double-stranded breaks even in the absence of genotoxic stress (Haffner et al. 2010).
Despite the prevalence of these genomic rearrangements, the functional significance of the TMPRSS2-ERG fusion and other ETS rearrangements in prostate cancer is still not fully resolved. Recent whole-genome chromatin immunoprecipitation analyses have shown that ERG can bind to AR downsteam target genes and disrupts AR signaling in prostate cancer cells through epigenetic silencing, consistent with a role in inhibiting prostate epithelial differentiation (Yu et al. 2010). Furthermore, analyses of ETS gene activation in cell culture assays as well as transgenic mice have suggested that ETS activation promotes EMT and tumor-invasive properties (Tomlins et al. 2007a, 2008a; Klezovitch et al. 2008; J Wang et al. 2008), although the effects are relatively moderate. In transgenic mice, expression of truncated human ERG transgenes results in a minimal or weak PIN phenotype (Tomlins et al. 2007a, 2008a; Klezovitch et al. 2008). However, expression of truncated ERG synergizes with loss of Pten to result in high-grade PIN and carcinoma in mice (Carver et al. 2009; King et al. 2009). In addition, recent findings suggest that TMPRSS2-ERG-positive tumors are also associated with the deletion of a small genomic region on 3p14, suggestive of another cooperative interaction in tumorigenesis (Taylor et al. 2010). Taken together, these findings suggest that ETS rearrangements are selected primarily for their ability to disrupt differentiation programs and/or to promote prostate cancer progression through cooperative interactions with other transforming events.
PTEN
PTEN was originally identified as a tumor suppressor that is frequently mutated or deleted in many cancers, including prostate (Salmena et al. 2008). The relevance of PTEN loss for prostate cancer was initially inferred from its location on chromosomal region 10q23, which frequently undergoes allelic loss in prostate cancer, as well as by its reduction or loss of expression in prostate tumors (Wang et al. 1998; Whang et al. 1998; McMenamin et al. 1999; Dong et al. 2007). Earlier studies had generated conflicting data regarding whether both alleles of PTEN are deleted in prostate cancer, or, if one allele is deleted, whether the remaining allele is mutated, or if the expression of PTEN protein is reduced, inactivated, or altered in subcellular localization. To resolve these issues, recent studies have investigated PTEN copy number, mutational status, and/or protein expression in primary or castration-resistant tumors using multiple experimental approaches (Verhagen et al. 2006; Schmitz et al. 2007; Sircar et al. 2009; Taylor et al. 2010). In combination with the consensus of previous reports, these studies support the conclusion that PTEN undergoes copy number loss as an early event in prostate carcinogenesis, and is correlated with progression to aggressive, castration-resistant disease. Interestingly, these studies have also suggested that low levels of PTEN activity may be retained in prostate cancer—an observation that parallels the haploinsufficiency of NKX3.1 and the p27 cell cycle regulator (Gao et al. 2004a; Abate-Shen et al. 2008), and which may reflect the relative indolence of prostate tumors.
Analyses of Pten deletion in genetically engineered mouse models have uncovered its cooperativity with inactivation of other key genes that are deregulated in prostate tumorigenesis, and have also provided insights into new therapeutic options for the treatment of prostate cancer. Germline loss of Pten in heterozygous mutants or conditional deletion in the prostate epithelium results in PIN and/or adenocarcinoma (Di Cristofano et al. 1998a; Podsypanina et al. 1999; Trotman et al. 2003; Wang et al. 2003). Inactivation of Pten has been shown to cooperate with loss of function of the Nkx3.1 homeobox gene, up-regulation of the c-Myc proto-oncogene, or the TMPRSS-ERG fusion (Kim et al. 2002c, 2009; Carver et al. 2009; King et al. 2009). Additionally, investigations of Pten loss—together with perturbations of cell cycle regulators such as p27, p18ink4c, and p14arf (Di Cristofano et al. 2001; Bai et al. 2006; Z Chen et al. 2009), or components of key signaling pathways such as Rheb, TSC2, and Rictor (L Ma et al. 2005; Nardella et al. 2008; Guertin et al. 2009)—have further emphasized the significance of haploinsufficiency in prostate cancer. Interestingly, the requirement of the mTORC2 complex as well as the p110β isoform of PI3K for tumor formation following Pten loss suggests that these signaling components may provide additional and/or alternative targets for therapeutic intervention (Jia et al. 2008; Guertin et al. 2009). Moreover, the observation that complete inactivation of Pten in mouse prostate tumors leads to cellular senescence (Z Chen et al. 2005) has led to the idea that novel therapeutic approaches might promote senescence for selective targeting of prostate tumor cells through knockdown of Pten function (Alimonti et al. 2010) or targeting of Skp2 (Lin et al. 2010). Notably, PTEN reduction or loss in prostate cancer predisposes to the emergence of castration-resistant prostate cancer (Mulholland et al. 2006; Shen and Abate-Shen 2007). In particular, perturbation of PTEN expression in human prostate cancer cell lines or targeted deletion of Pten in mouse prostate cancers is sufficient for the development of castration resistance (Lin et al. 2004; Bertram et al. 2006; Gao et al. 2006b; Wu et al. 2006). While this may reflect the ability of PTEN to interact directly with AR, the mechanistic details by which PTEN loss promotes castration resistance remain to be resolved.
Signaling pathways—Akt/mTOR and MAPK signaling
As noted above, considerable evidence indicates that Pten loss of function results in up-regulation of the Akt/mTOR signaling pathway in prostate cancer, primarily through activation of Akt1 (Thomas et al. 2004; ML Chen et al. 2006; Mulholland et al. 2006; Shen and Abate-Shen 2007). Up-regulation of this pathway in prostate cancer can also take place through activating mutations of Akt1 (Boormans et al. 2008), or through activation of the p110β isoform of PI3K (Hill et al. 2010; Lee et al. 2010). The functional consequences of Akt/mTOR pathway activation are particularly relevant for castration-resistant prostate cancer, as has been shown in genetically engineered mouse models, in gain-of-function studies with orthotopic grafting or tissue recombination models, as well as in human cell lines (Majumder et al. 2003; Uzgare and Isaacs 2004; Gao et al. 2006a; Xin et al. 2006). Activation of Akt occurs primarily at the cell membrane, and is consequently sensitive to levels of cholesterol in prostate cancer cells (Zhuang et al. 2005; Adam et al. 2007); however, Akt has additional functions in the nucleus that are dependent on the levels of PML (Trotman et al. 2006). The consequences of Akt activation are mediated in part by activation of NF-κB signaling via stimulation of IKK (Dan et al. 2008). Conversely, functional studies in mouse models and correlative studies in human prostate cancer have implicated deregulated NF-κB signaling in mediating androgen responsivity, metastasis, and disease outcome (Fradet et al. 2004; Ismail et al. 2004; Lessard et al. 2006; Luo et al. 2007; Zhang et al. 2009).
In addition to Akt/mTOR signaling, Erk (p42/44) MAPK signaling is also frequently activated in prostate cancer, particularly in advanced disease, and is often coordinately deregulated together with Akt signaling (Abreu-Martin et al. 1999; Gioeli et al. 1999; Paweletz et al. 2001; Malik et al. 2002; Thomas et al. 2004; Kinkade et al. 2008). Simultaneous activation of these signaling pathways promotes tumor progression and castration resistance in prostate cancer cell lines and mouse models (Uzgare and Isaacs 2004; Gao et al. 2006a), while combinatorial inhibition of these pathways inhibits castration-resistant prostate cancer in genetically engineered mice (Kinkade et al. 2008). In contrast with Akt/mTOR signaling, the upstream events that lead to activation of Erk MAPK signaling are less well defined, but are thought to be linked to aberrant growth factor signaling (Gioeli 2005). Although mutations of RAS or RAF are rarely found in human prostate cancer, the pathway is frequently perturbed in advanced prostate cancers (Taylor et al. 2010). Notably, expression of activated forms of either Raf or Ras in the mouse prostrate epithelium results in MAPK activation and promotes cancer formation (Jeong et al. 2008; Pearson et al. 2009). Interestingly, a small percentage of aggressive prostate tumors contains a translocation of B-RAF or C-RAF that results in activation (Palanisamy et al. 2010), suggesting that perturbations of Ras or Raf signaling may occur in prostate cancer through mechanisms other than activating mutations.
Oncogenic tyrosine kinases
The deregulated expression of oncogenic tyrosine kinases has been studied extensively in many cancers, since these can represent targets for therapeutic intervention (Gschwind et al. 2004). In prostate cancer, aberrant tyrosine kinase signaling, particularly through Her2/Neu or SRC tyrosine kinases, has been implicated in aggressive disease, progression to metastasis, and castration resistance, and, consequently, has been implicated as a key therapeutic target in patients with advanced disease (Mellinghoff et al. 2004; Fizazi 2007). In particular, stimulation of AR signaling leads to activation of SRC in prostate cancer cells, which can lead to phosphorylation of AR, castration resistance, and cellular proliferation and invasiveness (Migliaccio et al. 2000; Agoulnik et al. 2005; Kraus et al. 2006). However, most functional analyses of SRC and other oncogenic tyrosine kinases have been limited to studies of prostate cancer cell lines in culture or in xenografts, and further insights will require analyses of in vivo models and correlative studies of clinical specimens.
Developmental signaling pathways
Molecular analyses of prostate development are likely to be informative for prostate carcinogenesis, as recent studies have shown that prostate tumors express a wide range of genes normally expressed during embryonic/neonatal organogenesis, suggesting that cancer progression reactivates embryonic developmental programs of gene expression (Schaeffer et al. 2008; Pritchard et al. 2009). In particular, elevated canonical Wnt signaling may play a role in the emergence of castration resistance (G Wang et al. 2008), while prostate cancer in mice can result from inactivation of Apc or overexpression of a constitutively active β-catenin together with activated K-ras (Bruxvoort et al. 2007; Pearson et al. 2009; Yu et al. 2009). In contrast, however, evidence from human tumors suggests that nuclear localization of β-catenin is inversely correlated with tumor progression (Horvath et al. 2005; Whitaker et al. 2008), suggesting that canonical Wnt signaling may not play a significant role in prostate cancer progression. With respect to the Hedgehog pathway, although there is considerable evidence that activation of Hedgehog signaling plays a significant role in prostate cancer progression, it remains unclear as to whether this occurs through an autocrine mechanism in epithelial cells (Karhadkar et al. 2004; Sanchez et al. 2004), or, alternatively, through paracrine signaling involving stromal components (Yauch et al. 2008; Shaw et al. 2009). Finally, paracrine FGF signaling has also been implicated in prostate cancer in mouse models, through either epithelial activation of FGFR1 or stromal overexpression of FGF10 (Acevedo et al. 2007; Memarzadeh et al. 2007). This up-regulation of FGF signaling may provide a mechanism for the activation of Erk MAPK pathway activity observed in prostate cancer progression.
Ezh2
The Polycomb group gene EZH2 encodes a histone lysine methyltransferase that is frequently up-regulated in advanced prostate cancer, in some cases through gene amplification (Varambally et al. 2002; Saramaki et al. 2006), and is associated with aggressive tumors (Bachmann et al. 2006). EZH2 expression is negatively regulated by miR-101, and miR-101 expression decreases during cancer progression, concomitant with somatic loss of one or both miR-101 alleles (Zhao et al. 2007). Among the targets of EZH2 is NKX3.1, which is repressed via expression of ERG and is dependent on H3K27 trimethylation (Kunderfranco et al. 2010). Other EZH2 target genes in prostate cancer are associated specifically with metastasis, including E-cadherin (Cao et al. 2008) and DAB2IP (H Chen et al. 2005), which promotes prostate cancer metastasis through activation of Ras and NF-κB pathways (Min et al. 2010). However, Ezh2 has also been shown to function in the cytoplasm to control actin polymerization in prostate and nonprostate cells (Su et al. 2005; Bryant et al. 2008), and therefore analyses of target genes may not fully explain EZH2 function in cancer progression.
miRNAs
miRNAs regulate normal processes of growth and development as well as pathogenic processes associated with cancer, and analyses of their expression patterns have been effective for stratifying human cancers (Lu et al. 2005; Volinia et al. 2006). Expression profiling studies of human prostate tumors and xenografts have suggested that the expression patterns of miRNAs may distinguish indolent from aggressive tumors (Porkka et al. 2007; Ambs et al. 2008; Ozen et al. 2008; Coppola et al. 2009; DeVere White et al. 2009), and have implicated specific miRNAs in castration-resistant prostate cancer (Shi et al. 2007; Sun et al. 2009). Consistent with these findings, key enzymatic components of miRNA synthesis and processing such as Dicer are up-regulated during prostate tumor progression (Chiosea et al. 2006; Ambs et al. 2008; Poliseno et al. 2010a), while functional analyses of mice with conditional deletion of Dicer support a role for miRNAs in prostate epithelial proliferation (Zhang et al. 2010). Furthermore, miRNAs have specific roles in regulation of critical target genes, as the cluster miR-106b-25 negatively regulates PTEN expression (Poliseno et al. 2010a), while genomic loss of miR-101 leads to up-regulation of EZH2 in prostate cancer progression (Varambally et al. 2008). In addition, functional analyses of the miR-15a–miR-16-1 cluster in regulating expression of CCND1, WNT3A, and BCL2 in prostate cancer have provided an example of potential therapeutic benefit in restoring expression of miRNAs (Bonci et al. 2008), while detection of miRNAs in human plasma has been proposed as a platform for blood-based detection of human cancer (Mitchell et al. 2008). Notably, a recent study has shown that the expressed pseudogene PTENP1 can regulate the expression of PTEN in prostate cancer by competing for miRNA binding (Poliseno et al. 2010b), demonstrating a new mechanism for regulating gene expression in human tumors.
AR function and castration resistance
AR and the emergence of castration resistance
AR is a nuclear hormone receptor whose signaling plays a key role in both normal prostate development and prostate cancer. The most abundant androgen is testosterone, which is synthesized by the testis and converted into the more active metabolite dihydrotestosterone in prostate tissue through the activity of 5a-reductase. In addition, the adrenal gland synthesizes minor androgen species, including androstenedione and dehydroepiandrosterone (DHEA), which can be converted into testosterone. In human patients, androgen deprivation therapy is usually performed by administration of gonadotropin-releasing hormone analogs and/or surgical castration (orchiectomy), often in combination with anti-androgens such as flutamide or bicalutamide.
Following androgen deprivation, the androgen dependence of prostate tissue is manifested by rapid cellular apoptosis and involution to the regressed state (Fig. 3). In culture, however, androgen-dependent prostate cell lines cease proliferation when androgens are removed, but do not undergo apoptosis (Watson et al. 2005; Gao et al. 2006a). In contrast, in tissue reconstitution experiments, the same prostate cell lines will display apoptosis following androgen deprivation, indicating that the apoptotic response is induced by stromal tissue (Gao et al. 2006a). These findings are consistent with earlier tissue reconstitution experiments that analyzed recombination of AR-null mutant epithelium with wild-type stroma (Kurita et al. 2001). Thus, androgen dependence of prostate epithelium in vivo requires paracrine activity of stromal AR, similar to the requirement for mesenchymal AR in epithelial–mesenchymal interactions during early prostate organogenesis (Shen and Abate-Shen 2007). Consistent with this conclusion, conditional deletion of AR in both epithelium and stroma of TRAMP mice resulted in smaller tumors with decreased proliferation relative to those formed after epithelial-specific AR deletion (Niu et al. 2008b).
Role of AR in castration-resistant prostate cancer. (A) AR maintains homeostasis of both epithelial and stromal tissues in the normal prostate. (B) Following androgen ablation, stromal cells produce paracrine proapoptotic signals that act on neighboring epithelial cells, promoting regression of normal prostate. (C–F) Castration resistance can occur through a variety of molecular mechanisms, including AR amplification (C); gain-of-function mutation of AR mutation (D); ligand-independent AR activation by up-regulation of other signaling pathways, such as the Akt/mTOR and Erk MAPK pathways (E); or endogenous biosynthesis of androgens by tumor cells (F). Adapted from Shen and Abate-Shen (2007); © 2007 American Association for Cancer Research.
In normal prostate epithelium, AR suppresses cellular proliferation, as probasin-Cre-mediated conditional deletion of AR leads to increased proliferation accompanied by decreased expression of differentiation markers (Wu et al. 2007). In prostate cancer, however, AR suppresses proliferation of basal cells, supports survival of luminal cells, and promotes metastasis, as shown by analyses of AR conditional deletion in the context of the TRAMP model (Niu et al. 2008a). This complex loss-of-function phenotype contrasts with more straightforward gain-of-function studies, as transgenic mice overexpressing wild-type AR under the control of the probasin promoter develop PIN (Stanbrough et al. 2001), while overexpression of an AR missense mutation results in prostate cancer (Han et al. 2005). Overall, it appears that AR is likely to play different cell type-specific roles in both normal and cancer cells, which are modulated by interactions with other key regulators of prostate epithelial fate. For example, Nkx3.1 negatively regulates AR transcription and signaling activity (Lei et al. 2006), while genomic analyses of AR enhancer-binding sites reveal likely interactions with Nkx3.1 and FoxA1, another key transcriptional regulator of prostate epithelial differentiation (Gao et al. 2005; He et al. 2010).
Retention of AR signaling in castration resistance
Even when prostate cancer progresses to castration resistance, AR activation and signaling remains sustained through a variety of mechanisms (Fig. 3; Taplin and Balk 2004; Attard et al. 2009a; Bonkhoff and Berges 2010). Notably, castration-resistant tumors express AR as well as AR target genes such as PSA, indicating that pathway activity is intact (Gregory et al. 1998). These findings have been most strongly supported by key experiments showing that xenografts that have been selected for castration resistance primarily differ from their parental androgen-dependent lines with respect to levels of AR expression (Chen et al. 2004). Thus, androgen signaling switches from a paracrine mechanism involving the stroma in androgen-dependent cells to an autocrine mechanism in castration resistance (Gao et al. 2001).
Several molecular mechanisms have been described for the ability of AR to retain signaling activity in castration-resistant prostate cancer. These mechanisms include the amplification of AR gene copy number in approximately one-third of castration-resistant carcinomas (Visakorpi et al. 1995; Koivisto et al. 1997; Linja et al. 2001). Another 10%–30% of tumors have gain-of-function mutations of AR that may confer increased protein stability, greater sensitivity to androgens, novel responses to other steroid hormones, ligand-independent activity, or increased recruitment of AR coactivator proteins (Taplin et al. 1995, 2003; Zhao et al. 2000; Robzyk et al. 2007; Brooke et al. 2008; Steinkamp et al. 2009). In addition, recent studies have shown that expression of alternative splice isoforms encoding constitutively active AR variants also occurs in castration-resistant cancer (Dehm et al. 2008; Guo et al. 2009; Hu et al. 2009). Finally, an unusual mechanism for increased AR signaling activity is the endogenous expression of androgen synthetic enzymes by tumor tissue, which can lead to de novo androgen synthesis or conversion of weaker adrenal androgens into testosterone and dihydrotestosterone (Titus et al. 2005; Stanbrough et al. 2006; Locke et al. 2008; Montgomery et al. 2008).
Ligand-independent activation of AR activity can also take place through activation of growth factor signaling pathways. Notably, up-regulation of the PI3K pathway through Pten deletion appears to be particularly effective, as PIN lesions in Nkx3.1; Pten double-mutant mice display castration resistance prior to carcinoma formation (Gao et al. 2006b). Furthermore, analysis of androgen-dependent cell lines in tissue reconstitution assays has shown that castration resistance can be induced by activation of the PI3K pathway, and is synergistically enhanced by up-regulation of MAPK signaling, but remains dependent on AR function (Gao et al. 2006a; Jiao et al. 2007). At the molecular level, growth factor signaling can up-regulate AR transcriptional activity through increased tyrosine phosphorylation, or perhaps elevated ubiquitination of AR (Guo et al. 2006; Xu et al. 2009).
Finally, castration resistance can be enhanced through an increased inflammatory response. For example, production of interleukin-1β by infiltrating macrophages can lead to derepression of the AR corepressor complex in prostate tumor cells, thereby converting AR antagonists into agonists (Zhu et al. 2006). In addition, production of inflammatory cytokines by B lymphocytes can lead to nuclear translocation of IKKα and castration resistance in mouse prostate tumor cells and allografts (Luo et al. 2007). Moreover, analyses of TRAMP mice and cell lines have shown that nuclear IKKα can enhance prostate cancer metastasis through down-regulation of Maspin (Luo et al. 2007). Consequently, the emergence of castration resistance and metastasis may be coordinately linked at the molecular level through interactions with the tumor microenvironment.
Overall, these findings suggest that AR target genes and regulatory networks should be similar in androgen-dependent and castration-resistant prostate cancer. This conclusion has been supported by expression profiling of tumors with and without neoadjuvant androgen ablation prior to radical prostatectomy, which showed that castration-resistant tumors displayed up-regulation of AR, androgen synthetic enzymes, and known AR target genes (Holzbeierlein et al. 2004). However, recent genomic chromatin immunoprecipitation studies have shown that AR activity in castration-resistant prostate cancer is not identical to that displayed by AR in androgen-dependent cells. In particular, there is a significant alteration of genomic AR-binding targets and associated epigenetic chromatin marks in castration-resistant prostate cancer cell lines, resulting in up-regulation of M-phase-associated cell cycle genes (Q Wang et al. 2009). These findings suggest that AR-interacting proteins and/or histone-modifying enzymes may play a significant role in mediating castration resistance.
At present, it is unclear when castration resistance normally arises within prostate tumors. The conventional “adaptation” model proposes that castration-resistant cells arise through genetic/epigenetic conversion of previously androgen-dependent cells during conditions of androgen deprivation, while the alternative “clonal selection” model suggests that emergence of castration resistance reflects the proliferation of a previously quiescent population of rare castration-resistant cells within an otherwise androgen-dependent tumor (Isaacs and Coffey 1981). Although the former model represents the prevailing view, evidence for the latter model has been provided in studies of the onset of castration resistance in TRAMP mice (Gingrich et al. 1997), as well as through limiting dilution and fluctuation analyses of an androgen-dependent xenograft (Craft et al. 1999a). In addition, analysis of localized human prostate tumors suggests that rare AR mutations can be detected prior to androgen deprivation therapy (Gaddipati et al. 1994; Tilley et al. 1996; Bergerat and Ceraline 2009). Furthermore, the finding that castration-resistant cells such as CARNs (castration-resistant Nkx3.1-expressing cells) represent a cell of origin for prostate cancer also favors a clonal selection model (X Wang et al. 2009) in which the rare castration-resistant population might also correspond to putative cancer stem cells. Thus, while some mechanisms of castration resistance may represent an adaptive response to androgen deprivation therapy, in many cases, increased AR activity may be selected prior to treatment during prostate cancer progression.
Prostate stem cells and tumor-initiating cells (TICs)
Localization of adult stem cells
A tissue stem cell can be defined as a progenitor that is multipotent, being capable of giving rise to distinct cell types of the tissue of interest, and also able to self-renew by maintaining the stem cell phenotype in progeny following cell division (Rossi et al. 2008). In the case of the adult prostate, the existence of epithelial stem cells is implied by the ability of the adult prostate to undergo repeated cycles of extensive regression in response to androgen deprivation, followed by full regeneration following androgen restoration. Consequently, the prostate epithelium should contain a long-term resident pool of stem cells that are castration-resistant (Isaacs 1985). Notably, the majority of luminal cells are androgen-dependent and undergo apoptosis following castration, while most basal and neuroendocrine cells survive and are castration-resistant (English et al. 1987; Evans and Chandler 1987).
Most studies of prostate epithelial stem cells have relied on flow cytometry to purify subsets of epithelial cells based on cell surface marker expression, and have explored their progenitor potential in cell culture or transplantation assays (Lawson and Witte 2007; Kasper 2008). In particular, subpopulations of prostate basal cells isolated using cell surface markers can display bipotentiality and self-renewal in cell culture as well as tissue reconstitution assays (Richardson et al. 2004; Burger et al. 2005; Xin et al. 2005; Lawson et al. 2007; Goldstein et al. 2008). For example, isolation of a Lin−Sca-1+CD49f+ population results in a 60-fold enrichment for stem cells (Lawson et al. 2007). Further enrichment can be obtained using the Trop2 marker, which also allows isolation of a stem cell-enriched Lin−CD49f+Trop2+ fraction from human prostate epithelium (Goldstein et al. 2008). In independent studies, α2β1 integrinhiCD133+ basal cells also correspond to an enriched stem cell fraction in human prostate epithelium (Richardson et al. 2004). Finally, single Lin−Sca-1+CD133+CD44+CD117+ cells, which are predominantly basal in mice and are exclusively basal in humans, have been reported to reconstitute prostatic ducts in renal grafts (Leong et al. 2008).
In contrast, other data have supported a luminal localization of prostate epithelial stem cells, primarily with respect to prostate regeneration. In particular, analyses of grafted tissue from p63-null mice have demonstrated the formation and serial regression/regeneration of prostate tissue in the absence of basal cells (Kurita et al. 2004). Furthermore, recent studies have identified a rare luminal population of CARNs in the regressed prostate epithelium that displays stem cell properties during prostate regeneration (X Wang et al. 2009). In particular, in vivo genetic lineage marking showed that CARNs display bipotentiality and long-term self-renewal during prostate regeneration, and are also capable of reconstituting prostatic ducts following single-cell transplantation.
At present, it is difficult to ascertain the potential overlap as well as lineage relationships of the various candidate stem cells that have been identified, in part due to the distinct methodologies and assays employed (Fig. 4). In addition, individual cell surface markers may lack specificity for stem/progenitor cells, as has been suggested for CD133 (Shmelkov et al. 2008). It is also notable that existing cell culture assays are performed under conditions that select strongly against the growth of luminal cells (Peehl 2005), resulting in a significant bias toward outgrowth of basal cells and basal cell differentiation in assays such as prostasphere formation. Thus, in the absence of comprehensive in vivo approaches to investigate stem cell properties, the present data suggest that there may be multiple independent stem cell populations within the adult prostate epithelium.
Lineage hierarchy in the prostate epithelium and the cell of origin for prostate cancer. Two possible lineage relationships for the adult prostate epithelium are shown, together with the potential roles of Lin−Sca-1+CD49f+ cells (LSCs) and CARNs. Different cell types of origin in the lineage hierarchy might then generate distinct tumor subtypes following oncogenic transformation (red arrows). (A) In this model, LSCs correspond to stem cells, and CARNs correspond to a luminal progenitor that acquires stem cell properties in the context of prostate regeneration (green arrows), thus corresponding to a facultative stem cell. (B) An alternative model is that LSCs and CARNs correspond to independent stem cells that maintain basal and luminal populations, respectively. Adapted from X Wang et al. (2009); © 2009 Nature.
Cell of origin
The tissue localization of prostate epithelial stem cells is highly relevant for investigating the putative cell type(s) of origin for prostate cancer (Lawson and Witte 2007; Kasper 2008; Maitland and Collins 2008). A cell of origin can be defined as a normal tissue cell that can be oncogenically transformed to give rise to a cancer; thus, the cell of origin refers to a cell or cell type that is found in normal untransformed tissue. In principle, cancer could result from transformation of a rare stem cell, and/or could result from transformation of a more restricted cell type (such as a transit/amplifying cell) and its “dedifferentiation” to acquire self-renewal properties characteristic of stem cells (Fig. 4). Indeed, differences in the cell of origin in the stem cell lineage hierarchy have been proposed to represent the basis for distinct tumor subtypes for breast cancer (Visvader 2009).
Given the luminal phenotype of human prostate cancer, the cell of origin should correspond to either a luminal cell, or a basal progenitor that can rapidly differentiate into luminal progeny following oncogenic transformation. A basal cell of origin has been suggested by analyses of Pb-Cre4; Ptenflox/flox mice, which display an expansion of basal cells as well as intermediate cells coexpressing basal and luminal markers in tumors (S Wang et al. 2006). More recently, a comparison of basal and luminal epithelial populations isolated by flow cytometry from the mouse prostate has shown that basal populations are readily transformed by lentiviral expression of ERG and AR in tissue reconstitution experiments, whereas luminal cells are not transformed (Lawson et al. 2010). Importantly, analogous reconstitution assays using normal epithelial cells isolated from the human prostate have shown that transformed basal cells can generate prostate adenocarcinomas with luminal phenotypes (Goldstein et al. 2010).
In contrast, studies of PSA-Cre; Ptenflox/flox mice have suggested a rare luminal Clu+Tacstd2+Sca-1+ population as corresponding to the cell of origin in this model (Korsten et al. 2009). Consistent with these findings, detailed phenotypic analysis of Probasin-Myc and Nkx3.1-Myc transgenic mouse lines also suggests that PIN and prostate cancer originates from luminal cells (Iwata et al. 2010). Notably, CARNs correspond to luminal cells of origin for prostate cancer in mouse models, as evidenced by targeted deletion of Pten resulting in high-grade PIN and invasive carcinoma following androgen repletion and prostate regeneration (X Wang et al. 2009). Additional evidence is suggested by detailed histopathological analysis of MYC expression in high-grade PIN samples, which still retain basal cells, which shows that MYC up-regulation is associated exclusively with luminal cells, and is not detected in their basal neighbors (Gurel et al. 2008); similar findings have also been reported with respect to telomere shortening (Meeker et al. 2002). Also in favor of a luminal cell of origin is the recent finding that AR mediates formation of the TMPRSS2-ERG fusion in human prostate cancer cells (Lin et al. 2009; Mani et al. 2009; Haffner et al. 2010), suggesting that initiating events take place in AR-expressing luminal cells. Thus, based on the available evidence, prostate cancer can indeed arise from distinct cell types of origin, but it remains unclear whether different cells of origin are used in human prostate cancer initiation, or whether they might result in differing molecular subtypes.
Identification of TICs
The cancer stem cell model proposes that cell populations within a tumor have a hierarchical organization, in which a stem cell-like population gives rise to more differentiated derivatives that lack tumor-initiating and/or long-term self-renewal capability (Reya et al. 2001; Pardal et al. 2003; Wicha et al. 2006; Visvader and Lindeman 2008; Marotta and Polyak 2009; Rosen and Jordan 2009). This model has strong translational and clinical relevance, since it would likely have several implications for prostate cancer treatment. First, the identification of appropriate markers would allow the correlation of prostate cancer stem cell status in tumors with histopathology and clinical outcomes, and might also serve as accurate surrogates for the efficacy of cancer treatments. Second, targeted therapeutics for cancer stem cells might be superior to conventional therapies, which usually target cellular proliferation in the bulk tumor, while cancer stem cells may be relatively resistant due to a lower proliferative rate. Finally, the assessment of cancer stem cell numbers and molecular properties among circulating tumor cells might have prognostic value for the risk of metastatic disease, since the ability of circulating tumor cells to generate secondary metastases presumably requires self-renewing cancer stem cells.
The cancer stem cell model is consistent with the observed phenotypic heterogeneity found in many tumors, including prostate adenocarcinoma. In contrast, a stochastic or clonal evolution model of tumor development suggests that the phenotypic heterogeneity of tumors is due to variations in the genetic or epigenetic composition of tumor subpopulations, but that these subpopulations are not hierarchically organized and have similar tumor-initiating ability under appropriate circumstances (Adams and Strasser 2008; Shackleton et al. 2009). In many experimental contexts, cancer stem cells are identified in assays for TICs, using xenotransplantation to isolate cancer cells that can form a tumor after grafting, most rigorously after transplantation of a single cell. However, recent work has questioned the interpretation of such studies, since technical improvements in xenotransplantation can yield significant increases in efficiency, with up to 25% of melanoma cells displaying tumor-initiating properties (Quintana et al. 2008). These and other studies continue to engender doubt as to the existence of cancer stem cells in many solid tumors (Hill 2006; Shackleton et al. 2009).
Flow cytometry approaches to purifying subsets of epithelial cells based on cell surface marker expression have been combined with xenograft assays to identify putative TICs isolated from mouse prostate cancer models as well as human prostate cancer specimens (Lawson and Witte 2007; Kasper 2008). In the case of mouse prostate cancer, Lin−Sca-1+CD49f+ cells from PbCre4; Ptenflox/flox mice have been shown to have tumor-initiating properties in renal graft and sphere-forming assays, suggesting marker conservation between normal stem cells and cancer stem cells (Mulholland et al. 2009). In human prostate cancer, CD44 has been used as a marker to enrich for TICs from established xenografts (Patrawala et al. 2006), while further enrichment of TICs was obtained in a subsequent study by sorting for α2β1 integrinhiCD44+ cells (Patrawala et al. 2007). Finally, enrichment of CD133+α2β1 integrinhiCD44+ cells from primary prostate tumor biopsies resulted in identification of cells with increased invasiveness and clonogenicity in culture (Collins et al. 2005), while molecular analyses of CD133+α2β1 integrinhi cells revealed a potential cancer stem cell signature that is enriched for components of the JAK–STAT, Wnt, and focal adhesion pathways (Birnie et al. 2008). To date, however, the successful use of cell surface markers to isolate cell populations from primary human prostate cancers with tumor-initiating capabilities in grafting assays has not yet been reported.
Despite these promising findings, it remains unclear whether normal stem cells and cancer stem cells should display conserved marker expression, or whether the markers used display specificity for putative cancer stem cells. Second, the candidate TICs isolated to date display prevalent basal cell differentiation in vivo and in vitro, which is unexpected, since the primary tumors from which these cells were derived presumably lack basal cells. Finally, the putative TICs lack expression of AR, which is surprising given the strong selection for AR activity throughout prostate cancer progression, and the known mechanisms for castration resistance (Sharifi et al. 2006). These concerns suggest that authentic prostate cancer stem cells have not yet been definitively identified.
Translational applications
In recent years, principal areas of translational research on prostate cancer have focused on (1) understanding the dietary/lifestyle/environmental factors that influence prostate carcinogenesis, and identifying strategies to delay its onset or progression; (2) identifying biomarkers that distinguish indolent versus aggressive forms of the disease, and the application of such biomarkers for patient stratification; and (3) developing new therapeutic approaches for the treatment of castration-resistant prostate cancer, as well as for prevention of bone metastases. For instance, one example of a novel therapeutic approach that may be promising is the use of immunotherapy, as exemplified by the recent FDA approval of a therapeutic vaccine (Provenge) for advanced prostate cancer patients (Harzstark and Small 2007; Morse and Whelan 2010). Below, we briefly highlight major directions for translational research, focusing on how they can benefit from basic research, recent technological advances, and/or the application of robust preclinical models for in vivo analyses.
Dietary and lifestyle factors in cancer prevention
Epidemiologic investigations support the idea that dietary/lifestyle factors are major contributors of population differences in the occurrence of clinical prostate cancer (Kolonel et al. 2000, Kolonel 2001). In particular, dietary/lifestyle differences may account for the considerable difference in incidence of clinical prostate cancer between Asian and American populations, reflecting a shift in the rate of cancer detection by ∼10 years; notably, this discrepancy in cancer rate disappears when Asians immigrate to Western countries (Hanenszel and Kurihari 1968; Dunn 1975). However, the molecular/mechanistic bases for these differences have not been fully explained.
Considerable data support the hypothesis that dietary/lifestyle factors affect prostate cancer incidence by promoting chronic inflammation and/or oxidative stress, ultimately leading to DNA damage, epigenetic modifications, or other perturbations associated with cancer initiation (De Marzo et al. 2007a; Nelson 2007). This model has consequently emphasized the role of antioxidants and anti-inflamatory agents in protection against prostate cancer (DeWeese et al. 2001). Some prevention trials testing this model have been successful, including one showing that consumption of large quantities of tomato, which contain the potent antioxidant lycopene, results in reduced prostate cancer incidence (Chen et al. 2001). However, other trials have not shown that supplementation with antioxidants will reduce prostate cancer risk (Kirsh et al. 2006). A particular disappointment was the outcome of the SELECT trial, which found no benefit to supplementation with selenium and vitamin E (Lippman et al. 2009).
Additional studies have addressed the potential efficacy of antioxidants, anti-inflammatory agents, and/or other dietary factors by using epidemiological findings to investigate preclinical mouse models. For example, based on an extensive body of literature indicating that dietary restriction is anti-tumorigenic, analyses of dietary restriction or low-fat diets on cancer progression in genetically engineered mice has revealed the PI3K–Akt signaling pathway as a molecular target for these dietary interventions (Berquin et al. 2007; Kobayashi et al. 2008; Kalaany and Sabatini 2009). Another promising agent is vitamin D, which has been suggested by ample epidemiological evidence to protect against tumorigenesis, but has displayed variable efficacy in clinical trials (Deeb et al. 2007). Notably, analyses in genetically engineered mice that have shown that the timing of vitamin D administration is critical, as its beneficial effects are only realized early in cancer progression, as it promotes expression of the vitamin D receptor in prostate epithelial cells (Banach-Petrosky et al. 2006). These examples highlight the importance of integrating epidemiological analyses with systematic evaluation of mechanisms in preclinical models for effective design and implementation of dietary interventions for cancer prevention.
Biomarker discovery
PSA testing has revolutionized the diagnosis of prostate cancer, since it is now possible to detect most prostate tumors at early stages, unlike other cancers that lack a straightforward method for early detection. However, the early detection of prostate cancer needs to be augmented by improved biomarkers that can stratify patients in conjunction with Gleason grading. The search for effective biomarkers has included gene expression profiling, miRNA expression profiling, serum proteomics, and metabolomics. The latter represents a promising new approach that may allow for the development of noninvasive urine tests for cancer metabolites to detect prostate and other cancers (Sreekumar et al. 2009). More generally, the investigation of potential urine biomarkers has led to the identification of PCA3 (prostate cancer antigen 3), a promising marker for predicting disease outcome (Ploussard and de la Taille 2010).
However, to date, few if any biomarkers are now being used that can predict disease outcome more effectively than Gleason score alone. In principle, suitable combinations of markers may be successful in cumulatively predicting outcome, as enabled by new technologies such as molecular systems pathology (Cordon-Cardo et al. 2007). Alternatively, system biology approaches that identify master regulatory genes of disease progression may enable the effective stratification of patients, as has been applied for other cancer types (Carro et al. 2009). Finally, comprehensive oncogenomic approaches that integrate gene expression and copy number analyses may identify new biomarkers for predicting disease outcome (Taylor et al. 2010).
Manipulating AR signaling for prevention and treatment
The essential role of AR signaling for the development of prostate cancer provided the rationale for a large-scale prevention trial that evaluated the 5α-reductase inhibitor finasteride for prevention of prostate cancer (Higgins and Thompson 2004). The results of this trial were encouraging, since they showed a 24% reduction in prostate cancer incidence, which has led to the recommendation of finasteride administration for men in high-risk categories. As a cautionary note, however, a subset of patients in this trial appeared to develop more aggressive disease (Lucia et al. 2007), which may reflect a selection for men predisposed to limiting levels of androgens, as has been suggested by studies of limiting androgen levels for cancer progression in genetically engineered mice (Banach-Petrosky et al. 2007).
AR has also been a primary target for treatment of patients with advanced disease. Based on the central role of AR in castration resistance, novel AR pathway inhibitors could potentially provide important therapeutics for advanced prostate cancer (Attar et al. 2009; Y Chen et al. 2009; Knudsen and Scher 2009). In this regard, a second-generation AR antagonist, MDV3100, which completely lacks agonist activity and binds AR with greater affinity than bicalutamide, has provided new insights into castration resistance, and has given promising results in mouse models and in a human phase 1–2 trial (Tran et al. 2009; Scher et al. 2010). Other agents that target the N-terminal transcriptional regulatory region of AR are now being evaluated in cell lines and mouse models (Andersen et al. 2010). Another promising AR pathway antagonist is abiraterone acetate, which inhibits the activity of CYP17, an enzyme required for two steps in androgen biosynthesis, and has shown promising results in initial clinical trials (Attard et al. 2009b; Y Chen et al. 2009).
Targeting signaling pathways in treatment of advanced disease
For reasons that are poorly understood, the therapeutic benefits of standard chemotherapy regimens are limited in patients with advanced prostate cancer, although improvements have been made in the past several years (Calabro and Sternberg 2007; Petrylak 2007). Therefore, recent approaches have been aimed at targeting signaling pathways activated in advanced prostate cancer, including the Akt/mTOR and MAPK signaling pathways. The evaluation of Rapamycin and related compounds (Rapalogs) that target mTOR signaling in preclinical trials in genetically engineered mutant mice and in human clinical trials suggest that these may not be effective as single agents (Sawyers 2003; Garcia-Echeverria and Sellers 2008; Morgan et al. 2009). However, combination therapy using Akt/mTOR inhibitors in conjunction with first-line chemotherapy or agents that target other key signaling pathways such as the Erk MAPK pathway may be highly effective, as has been suggested by preclinical studies in which combination therapy effectively blocks castration-resistant prostate cancer in mice (Kinkade et al. 2008). Thus, the development of combination therapy for treatment of advanced prostate cancer will likely benefit from evaluation in robust preclinical models.
Perspectives and conclusions
Considering the tremendous progress made in the past 10 years, we envision continuing advances over the next decade in areas of research that will facilitate effective strategies for the prevention, diagnosis, and treatment of prostate cancer. Among the challenges for future studies will be to integrate epidemiological studies with molecular investigations and clinical analyses to gain fundamental insights into how environmental, dietary, and lifestyle influences contribute to the development of prostate cancer, and to identify the molecular factors that are altered by these influences and how they can be modified by appropriate dietary or chemical interventions. Of paramount importance will be the effective diagnosis of men that have prostate cancer, and their stratification into high-risk and low-risk groups for treatment management. Thus, biomarker discovery will likely represent a considerable emphasis for future research, perhaps focused on identification of master regulator genes that can provide accurate readouts of signaling pathways associated with disease progression. Moreover, considering that prostate cancer is fairly indolent, the development of treatment approaches that delay its onset or progression is likely to have a significant impact on outcome. Finally, more effective strategies will be necessary for preventing the transition to lethal forms of prostate cancer, which will require a deeper understanding of the mechanisms underlying castration-resistant prostate cancer and the bone tropism of prostate cancer metastasis. Thus, while our knowledge of the molecular genetics of prostate cancer has greatly expanded in the past decade, much work remains to be done to enhance the overall rate of prostate cancer survival.
Acknowledgments
We apologize to our colleagues for our inability to discuss and cite numerous important studies, due to length constraints. We thank Robert Cardiff and Alexander Borowsky for assistance with Figure 2, and Arul Chinnaiyan, Angelo De Marzo, Edward Gelmann, and Charles Sawyers for helpful discussions. This work was supported by grants to M.M.S. and C.A.-S. from the NIH, DOD Prostate Cancer Research Program, and NCI Mouse Models of Human Cancer Consortium.
Footnotes
-
↵6 Correspondence.
E-MAIL mshen{at}columbia.edu; FAX (212) 851-4572.
-
↵7 E-MAIL cabateshen{at}columbia.edu; FAX (212) 851-4572.
-
Article is online at http://www.genesdev.org/cgi/doi/10.1101/gad.1965810.
- Copyright © 2010 by Cold Spring Harbor Laboratory Press
References
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
