Bone Microenvironment Signaling of Cancer Stem Cells as a Therapeutic Target in Metastatic Prostate Cancer

Abstract

Prostate cancer (PCa) is one of the most prevalent cancers and the second leading cause of cancer death among U.S. males. When diagnosed in an early disease stage, primary tumors of PCa may be treated with surgical resection or radiation, sometimes combined with androgen deprivation therapy, with favorable outcomes. Unfortunately, the treatment efficacy of each approach decreases significantly in later stages of PCa that involve metastasis to soft tissues and bone. Metastatic PCa is a heterogeneous disease containing host cells, mature cancer cells, and subpopulation of cancer stem cells (CSC). CSCs are highly tumorigenic due to their self-renewing and differentiating potential, clinically resulting in recurrence and resistance to standard therapies. Therefore, there is a large unmet clinical need to develop therapies, which target CSC activity. In this review, we summarize the main signaling pathways that are implicated in the current preclinical and clinical studies of recurrent metastatic PCa within the bone microenvironment targeting CSCs and discuss the trajectory of therapeutics moving forward.

Introduction

Prostate cancer (PCa) is one of the most commonly diagnosed cancers in American males (Denmeade and Isaacs, 2002). With the advent of prostate specific antigen (PSA) testing and increased screening, a significant increase in the diagnosis of early stage PCa has occurred (Grönberg, 2003). In organ confined, low grade PCa, surgical resection or radiation therapy have been highly successful in treating the disease with long term cancer free survival (Bedard and Chow, 2013). However, a larger or higher grade primary tumor increases the chance of recurrence, which is due to the radio-resistant cancer stem cell (CSC) subpopulations located within the tumor (Li et al., 2013). The radio-resistance is often associated with increased CSC self-renewal activity, increased DNA repair capacity, and enhanced reactive oxygen species (ROS) defenses (Ogawa et al., 2013).

There are several mechanisms whereby castrate resistant cells may develop including the reduction of androgen receptor (AR) expression, which occurs in the vast majority of the tumors (Tsujimura et al., 2002). These cells are capable of both self-renewal and possess the ability to differentiate to AR+ PCa cells (Collins et al., 2005; Moltzahn and Thalmann, 2013). In fact, androgen targeted therapies appear to enhance the PCa stem cells or progenitor cell populations, which ultimately contribute to castration resistance (Germann et al., 2012). While the mechanisms for development of mCRPC remain unclear, several intrinsic factors and signaling pathways within the tumors facilitate CSC expansion with an epithelial-mesenchymal transition (EMT) phenotype (Ojo et al., 2015). As a result, the link between mCRPC and CSC activity is critical to understand and target.

One of the major pathways utilized by resistant CSC populations is the engagement or activation of a non‐proliferative state (quiescence, dormancy, or G0 phase), which is highly conserved without multiple inputs and disrupt quiescence (Takeishi and Nakayama, 2016). As such, CSCs are able to bypass therapies that use the traditional anti-proliferative approaches, suggesting that new approaches are needed to identify and ultimately target CSCs directly (Jaworska et al., 2015).

The identification of the source and phenotype of CSCs in PCa has been investigated but remains in controversial. PCa CSCs can be distinguished from non-CSCs based upon the expression of CD44, CD133, and α2β1 integrin (Collins et al., 2005; Sharpe et al., 2013). Qin et al. proposed an integrative definition of cells, which highly expressed CD44, ALDH, and α2β1, and expressed little or no PSA (Qin et al., 2012). Another group proposed that mitotic quiescence rather than surface phenotype was more effective at defining CSCs (Patrawala et al., 2007; Wang et al., 2015a). Intracellularly, the expression of stemness-related transcriptional factors has also been considered to identify and potentially serve as therapeutic targets for CSCs. These include high expression of NANOG, MYC, OCT4, SOX2, and KLF4, which have shown to be expressed by CSCs in significant levels compared to non-CSCs of PCa cells (Jeter et al., 2011).

Many patients with metastatic PCa present with bone lesions (Logothetis and Lin, 2005). The source of pain and reduced quality of life are often associated with skeletal related event (SRE), including pathological fractures, spinal cord compression, and increasing pain (Coleman, 2016). Although SREs are partially managed using anti-osteoclast agents (Zoledronate and Denosumab), it is important to target the source of the bone dysfunctions by preventing or treating PCa (Maier et al., 2016; Zhang et al., 2017c). Signaling in the bone microenvironment is known to play an important role in initiation and maintenance of PCa metastasis and fate of CSCs (Casimiro et al., 2009; Holen, 2016).

Here we explore several pathways, which contribute to a CSC phenotype in the bone marrow: Notch, Wnt, Hedgehog, fibroblast growth factor (FGF), transforming growth factor β (TGFβ), stromal derived factor 1 (SDF-1; CXCL12), and growth arrest protein 6 (GAS6) (Figure 1). Further we summarize current preclinical and clinical studies, which target these pathways in relation to metastatic PCa and/or CSCs that contribute to PCa recurrence (Table 1), and discuss novel routes forward for the future therapeutics.

Figure 1. Core signaling pathways of cancer stem cells are associated with the recurrence of metastatic PCa in the bone microenvironment.

ADAM: a disintegrin and metalloproteinase; TACE: tumor necrosis factor-alpha converting enzyme; NICD: intracellular domain of the notch protein; HAT: acetyltransferase; SKIP: skeletal muscle and kidney enriched inositol phosphatase; CSL: CBF-1, suppressor of hairless, lag-2; MAML: mastermind-like; LRP: lipoprotein receptor-related protein; FZD: Frizzled; DSH: dishevelled; GSK3: glycogen synthase kinase 3; APC: adenomatous polyposis coli; TCF/LEF: T cell factor/lymphocyte enhancer binding factor; PTCH1: patched 1; SMO: smoothened; SUFU: suppressor of fused homolog; Gli1/2: glioma-associated oncogene 1/2; FGF: fibroblast growth factor; FGFR: fibroblast growth factor receptor; FOXO: forkhead box; TKs: tyrosine kinases; PLCγ: phospholipase C gamma; PKC: protein kinase C; MAPKK: mitogen-activated protein kinase kinase; MAPK: mitogen-activated protein kinase; PI3K: phosphoinositide 3-kinase; TGFβ: transforming growth factor beta; GTP: guanosine triphosphate; NFKB: nuclear factor kappa-light-chain-enhancer of activated B cells; GAS6: growth arrest specific 6; mTOR: mechanistic target of rapamycin.

Table 1.

The pre-clinical and clinical studies targeting PCa cells associated with signaling pathways in the bone microenvironment.

Pathway	Drug/agent	Target PCa	Drug development stage	Description	Citation
Notch	PF-03084014	mCRPC, CSC	Pre-clinical	γ-secretase inhibitor	Cui et al., 2015
	GSI	mCRPC, CSC	Pre-clinical	γ-secretase inhibitor	Stoyanova et al., 2016
	DBZ	Metastatic	Pre-clinical	Notch inhibitor	Qui et al., 2018
	RO4929097	mCRPC	Clinical (phase II) NCT01200810	γ-secretase/Notch inhibitor	NCI, 2010–2017
Wnt	LGK974	Metastatic	Pre-clinical	Wnt inhibitor blocks porcupine-a membrane-bound O-acyltransferase	Ma et al., 2016
	ICG00I	Metastatic, CSC	Pre-clinical	β-catenin inhibitor	Zhang et al., 2018
	Ipafricept (OMP-54F28)	Advanced, CSC	Clinical (Phase I) NCT01608867	frizzled 8 receptor fused IgG1 Fc	Jimeno el al., 2017
	Foxy-5	Metastatic	Clinical (Phase I) NCT02020291 NCT02655952	Formylated 6 amino acid peptide targets Wnt-5a	WntResearch AB, 2013–2016; WntResearch AB, 2016–2017
Hedghog	GANT61	Metastatic	Pre-clinical	Gli inhibitor	Lauth et al., 2007; Rimkus et al, 2016
	Sonidegib (erismodegib, LDE225)	Metastatic, CSC	Pre-clinical	Hh inhibitor	Nanta et al., 2013
	Sonidegib	Advanced	Clinical (Phase I) NCT021111870	Hh inhibitor	Ross et al., 2017
	Vismodegib	mCRPC	Clinical NCT02115828	Hh inhibitor	Maughan et al., 2016
	Itraconozale	Advanced	Clinical (Phase II) NCT018787331	Antifungal agent	Lee et al., 2019
FGF	AZ8010	Metastatic	Pre-clinical	FGFR inhibitor	Feng et al., 2012
	PD173074/CH5183284	Metastatic	Pre-clinical	FGFR antagonist targeting FGFRI	Bluemn et al., 2017
	Dovitinib (TK12S8)	Metastatic and osteoblasts	Pre-clinical	FGFR inhibitor	Wan et al., 2014;Yadav et al., 2017
	Dovitinib	mCRPC	Clinical (Phase II) NCT01741116	FGFR inhibitor	Choi et al., 2018
	Nintedanib (BIBF1120)	Metastatic	Pre-clinical	FGFR inhibitor	da Silva et al., 2017, 2018
	Nintedanib	mCRPC	Clinical (Phase I, II) NCT02856425 NCT00706628	FGFR inhibitor	Bousquet el al., 2011; Droz el al., 2014; Gustave Roussy, Paris 2016–2018; Boehringer Ingelheim, 2008–2016
TGFβ	Tranilast	Metastatic	Pre-clinical	Anti-allergic agent targets TGFβl	Izumi et al., 2009
	IN-1130	Metastatic	Pre-clinical	TGFβRI ALK-5 kinase inhibitor	Lee et al., 2008
	SB-431542	Metastatic	Pre-clinical	TGFβRI ALK-5 kinase inhibitor	Miles et al., 2012
	LY2109761	Metastatic	Pre-clinical	TGFβRI & RII kinase inhibitor	Wan et al., 2012
	Galunisertib (LY2157299)	Metastatic	Pre-clinical	TGFβRI kinase inhibitor	Paller et al., 2018
	Galunisertib	mCRPC	Clinical (Phase II) NCT02452008	TGFβRI kinase inhibitor	Johns Hopkins, 2016–2021
CXCL12	CTCE-0908 (CTE9908) or Plerixafor (AMD3100)	Metastatic	Pre-clinical	CXCR4 antagonist	Gravina et al., 2015
	Plerixafor	Metastatic. CSC	Pre-clinical	CXCR4 antagonist; DTC mobiliser by blocking the CXCR4	Shiozawa et al., 2011; Dubrovska et al., 2012; Gravina et al., 2015; Wang et al., 2015; Conley-LaComb et al., 2016; Jung et al., 2018
	Plerixafor	Metastatic	Clinical NCT02478125	CXCR4 antagonist; DTC mobiliser by blocking the CXCR4	Johns Hopkins. 2015–2017
GAS6	Bemcentinib (R428, BOB-324)	Metastatic	Pre-clinical	Axl inhibitor blocks autophosphorylation of Axl	Bansal el al., 2015
	Amuvatinib (MP470)	Metastatic	Pre-clinical	Axl inhibitor blocks ATP-binding cassette Bl (ABCB1)	Lin et al., 2017
	UNC1062	Metastatic, CSC	Pre-clinical	Mer inhibitor blocks autophosphorylation of Mer	Shiozawa et al., 2016; Jung et al., 2016
	Cabozantinib (XL-184, Exelixis)	mCRPC	Pre-clinical	Tyrosine kinase inhibitor (c-Met, VEGFRs, Alk, Ron, and Axl)	Graham et al., 2014; Dai et al., 2014; Stern and Alvares, 2014; Haider et al., 2015; Leibowitz-Amit et al., 2016
	Cabozantinib	mCRPC	Clinical (Phase II, III) NCT01605227 NCT01522443	Tyrosine kinase inhibitor (c-Met, VEGFRs, Alk, Ron, and Axl)	Exelixis, 2012–2018; Smith et al., 2013; Smith et al., 2014; Smith et al., 2016
	Foretinib (GSK1363089, XL880)	Advanced	Clinical (Phase II) NCT00726323	Tyrosine kimse inhibitor (c-Met, VEGFRs, Alk, Ron, and Axl)	GlaxoSmithKline, 2008–2017; Choueiri et al., 2013

Bone Marrow Signaling Pathways Targeting Cancer Stem Cells

The bone marrow is comprised of diverse marrow cells including all of the hematopoietic lineage cells, hematopoietic stem cells (HSCs), mesenchymal stem cell (MSCs), bone marrow stromal cells (BMSCs), osteoblasts, endothelial cells, osteoclasts, and immune cells and extracellular matrix (ECM), which releases many different growth factors, cytokines, interleukins, adhesion molecules, as well as ECM proteins. BMSCs express many growth factors and adhesion molecules such as Hedgehog (Hh), FGFs, and Jagged 1. Endothelial cells produce vascular endothelial growth factor (VEGF) and CXCL12. Osteoblasts also produce a multitude of growth factors and cytokines such as Wnt, FGFs, TGFβ, CXCL12, GAS6, bone morphogenic proteins (BMP), osteopontin (OPN), interleukins, and receptor activator of nuclear factor kappa-B ligand (RANKL). Osteoblasts also express adhesion molecules (e.g., Jagged1, Annexin2) and integrins, which bind to different ECM proteins, such as collagen, laminin, vitronectin, and fibronectin. Many of factors from osteoblasts or calcified bone matrix activate osteoclast formation, which release the growth factors and proteolytic enzymes (Decker et al., 2017; Pedersen et al., 2012) (Figure 2).

Figure 2. Bone marrow signaling interacts with metastatic PCa CSCs.

BMSC: bone marrow stromal cell; HSC: hematopoietic stem cell; MSC: mesenchymal stem cell; ECM: extracellular matrix; VEGF: vascular endothelial growth factor; CXCL12: stromal derived factor 1 (SDF-1); GAS6: growth arrest protein 6; TGFβ: transforming growth factor beta; FGFs: fibroblast growth factors; BMPs: bone morphogenic proteins; IL: Interleukin, OPN: osteopontin; RANKL: receptor activator of nuclear factor kappa-B ligand.

Uniquely, PCa cells can convert to a CSC phenotype defined by dual expression of CD133 and CD44 (CD133+/CD44+) under the influence of the marrow (Shiozawa et al., 2016). These observations suggest that the metastatic bone marrow niche for tumor cells may contribute or even direct the establishment of PCa CSCs (Shiozawa et al., 2016). The fate of CSCs relies on marrow microenvironment and the signals from marrow control CSC activities including the protection against the environmental stress. Thus, development agents that target these signaling pathways will be the critical steps to eradicate CSCs in metastatic diseases.

Notch Pathway

Notch is a transmembrane receptor, which plays a critical role in embryogenesis, stem cell self-renewal, and differentiation in many tissues. The canonical Notch signaling pathway is activated through binding of Notch ligands (Jagged1/2 and Delta-like 1/3/4) to Notch receptors (1/2/3/4). The canonical Notch signaling pathway plays a critical role in stem cell activity (Gu et al., 2012). The non-canonical Notch signaling is thought to act in a ligand- or transcription-independent manner (Andersen et al., 2012).

In bone, BMSCs and osteoblasts express Jagged1, the ligand for Notch, and enhanced Jagged1 expression promotes Notch-expressing PCa cell bone metastasis (Delury et al., 2016). Several studies indicate that the Notch signaling plays a significant role in the development of an EMT phenotype and associated with CSC activities. Further, high levels of Notch1 expression are associated with an EMT phenotype frequently found in metastatic bone lesions compared with primary PCa tissues (Sethi et al., 2010). Down-regulation of Notch1 by microRNA-200 is partially responsible for the inhibition of sphere-forming ability and PCa cells (Kong et al., 2010). Importantly, when the ligand binds to its receptor, a disintegrin and metalloproteinases (ADAMs)/tumor necrosis factor-alpha converting enzyme (TACE) and gamma secretase cleave the Notch receptor, releasing the Notch intracellular domain (NICD) from the plasma membrane and facilitating its translocation into the nucleus (Takebe et al., 2015). A recent study also demonstrates that expression of NICD1 is highly elevated in high-risk PCa and castration-resistant PCa (Stoyanova et al., 2016). Moreover, the tumors driven by the combination of NICD1 with multiple (Ras/Raf/MAPK, myrAKT, or Myc) pathways have an increased capacity for self-renewal with enhanced metastatic potential (Stoyanova et al., 2016). Overexpression of Notch1 and NICD in PCa cells enhanced migration and invasion, and increased sphere formation of PCa cells (Zhang et al., 2017b). Expression of Notch2 also induces the stem-like properties and enhances docetaxel-resistance in PCa cells (Qiu et al., 2018).

In pre-clinical studies, a combination of a γ-secretase inhibitor (PF-03084014) and docetaxel reduced both docetaxel-sensitive and docetaxel-resistant CRPC tumor growth in soft tissues and bone better than either agent alone. Targeting this pathway with a γ-secretase inhibitor was associated with reduced survival, increased apoptosis, reduced microvessel density, reduced EMT, and reduced cancer stem-like cells (Cui et al., 2015). Further work on γ-secretase inhibitor (GSI) in vitro and murine model demonstrated inhibition of Notch1 signaling and γ-secretase, resulting in the delay of PCa cell growth and inhibition of colony formation (Stoyanova et al., 2016). Use of a γ-secretase inhibitor, Dibenzazepine (DBZ) with Docetaxel reduced tumor burden and prolonged survival following injection of PCa cells or patient-derived primary prostate carcinoma-1 (PPC1) cells in s.c. tumor xenograft models (Qiu et al., 2018). RO4929097, an inhibitor of γ-secretase in Notch signaling pathway, was tested in phase II clinical trial (clinicaltrials.gov: NCT01200810). This study evaluated efficacy of anti-androgen Bicalutamide with RO4929097 in patients with PCa recurrence after surgery or radiation; however, it was terminated early due to lack of sufficient drug supply.

Wnt Pathway

Wnt plays an important role in embryonic development and adult tissue homeostasis. Wnt signaling is mediated through at least three major pathways; canonical Wnt pathway (regulation of gene transcription), non-canonical planar cell polarity pathway (regulation of cytoskeleton), and non-canonical Wnt/calcium pathway (regulation of calcium). The canonical Wnt signaling pathway is known to play a critical role in self-renewal and maintenance of stem cells (Han et al., 2013).

In bone, Wnts promote the differentiation of pre-osteoblasts into mature osteoblasts through the β-catenin dependent canonical pathway (Kobayashi et al., 2008) and suppress bone resorption by altering RANKL/osteoprotegerin (OPG) ratios through the β-catenin dependent canonical pathway (Hall et al., 2006). Wnt signaling has been linked to PCa bone metastasis with induction of osteoblastic activity (Yardy and Brewster, 2005). In fact, β-catenin expression decreases during PCa progression and nuclear β-catenin decreases in PCa bone metastasis, which results in masking of osteogenic Wnts, thereby, switching the phenotype to osteolysis. However, during metastases progress, the expression of the Wnt antagonist Dickkopf-1 (DKK1) is decreased, which unmasks Wnt’s osteoblastic activities. Therefore, DKK1 is believed to serve as an important player switching from osteolytic to osteoblastic activities at the metastatic site during PCa bone metastasis (Hall et al., 2008; Hall et al., 2006). Critically, the Wnt/β-catenin signaling pathway is known to be highly activated in CSCs, and may play an important role in prostate stem cell self-renewal (Bisson and Prowse, 2009). Specifically, activation of Wnt3 signaling stimulates a significant induction of nuclear β-catenin, keratin 18, CD133, and CD44 expression in PCa cells, which regulates the sphere size and enhanced self-renewal capacity (Bisson and Prowse, 2009). Interestingly, a series of studies have shown that manipulation of the Wnt pathway is capable of regulating the CSC phenotype. For example, microRNA-320 suppresses the stem cell-like phenotype of PCa cells by downregulating Wnt/β-catenin signaling pathway (Hsieh et al., 2012). Additionally, AR79, an inhibitor of GSK-3β (a negative regulator of Wnt signaling), increased ALDH+/CD133+ stem-like population from tumor cells, which effectively promoted tumor growth in xenograft models (Jiang et al., 2013). Furthermore, high levels of PHD finger protein 21B (PHF21B) expression are associated with poor recurrence-free survival in PCa patients and PHF21B overexpression activated Wnt/β-catenin signaling pathway, promoting stem-like properties in PCa cells (Li et al., 2017). Recently, Wnt/β-catenin has been shown to enhance human telomerase reverse transcriptase (hTERT) activity in PCa cells and thereby regulate the self-renewal and differentiation activity of PCa cells (Zhang et al., 2017a). A novel Wnt co-activator, increases the canonical Wnt activity, which maintains the prostate CSC subpopulations in PCa (Pai et al., 2019). Together, all of these data support that Wnt signaling may be involved with PCa tumorigenesis with an induction of stem cell phenotype.

In preclinical models, the Wnt inhibitors 1) porcupine- a membrane-bound O-acyltransferase (LGK974) (Ma et al., 2016) and 2) β-catenin inhibitor (ICG001) (Zhang et al., 2018) have been shown to decrease CSC function and tumor growth. In fact, the inhibition of Wnt/β-catenin sensitizes PCa cells to Enzalutamide (Zhang et al., 2018). At a clinical level, Ipafricept (OMP-54F28), a recombinant fusion protein of frizzled 8 receptor fused to a human IgG1 Fc fragment that binds Wnt ligands, was evaluated in a phase I trial for treatment of solid tumors including PCa (clinicaltrials.gov: NCT01608867). Ipafricept was well tolerated by patients with solid tumors (Jimeno et al., 2017). The Foxy-5, a formylated 6 amino acid peptide fragment, is effective as a Wnt-5a inhibitor and as such was developed as an anti-metastatic agent. It was tested in two phase I clinical trials (clinicaltrials.gov: NCT02020291 and NCT02655952) to establish the recommended dose for a clinical phase II study, but in trial the results were inconclusive and no recommended phase II trial initiated.

Hedgehog Pathway

Hh is a conserved regulator for stem cell activities during animal development. Hh signaling is mediated through the transmembrane proteins Patched1 (PTCH1)-associated Smoothened (SMO), which activates glioma-associated oncogene (Gli) transcription factor leading to the activities related to proliferation and pluripotency (Hadjimichael et al., 2015). As such the Hh signaling pathway is essential for the maintenance of CSCs in human cancers (Chang et al., 2011; Han et al., 2013).

In bone, Hh is produced by bone marrow-derived mesenchymal stromal cells which supports cell survival in hematological disorders (Zou et al., 2015). Further, Hh signaling-activated Gli2 transcription induces BMP2 expression, which enhances osteoblast differentiation (Kesper et al., 2010; Zhao et al., 2006). Gli1 upregulates osteopontin (OPN), a factor linked to malignant behavior of cancer cells (Das et al., 2009). In advanced PCa lesions, overexpression of PTCH1 and SMO with P63+ basal/stem cells differentiate into AR+ and AR− progeny of metastatic tumor cells, which significantly increased the metastatic potential (Chang et al., 2011). Hh signaling is upregulated in prostate CSCs (Hurt et al., 2008; Klarmann et al., 2009).

In preclinical studies, the cell-permeable hexahydropyrimidine Gli inhibitor GANT61 strongly reduced PTCH1 expression, which decreased PCa growth and proliferation (Lauth et al., 2007; Rimkus et al., 2016). Sonidegib (Erismodegib, LDE225), an Hh inhibitor, that antagonizes SMO, downregulates pluripotency-maintaining factors NANOG, OCT4, SOX2, c-MYC and thereby inhibits CSC activity and ultimately tumor growth in vitro and a xenograft mouse model (Nanta et al., 2013). Critically, was tested in phase I trial (clinicaltrials.gov: NCT021111870) for men with localized high-risk PCa undergoing radical prostatectomy. Hh activity was detectable at baseline in men with localized high-risk PCa (as measured by Gli1 expression). Sonidegib penetrated into prostatic tissue and induced a >60-fold suppression of the Hh signaling (Ross et al., 2017). Vismodegib, an Hh inhibitor, has been evaluated in men with mCRPC (clinicaltrials.gov: NCT02115828). Vismodegib suppressed Hh signaling as evaluated by Gli1 mRNA expression in mCRPC tissues from the majority of patients. Despite this pharmaco-dynamic response, unfortunately this was little evidence of clinical activity (Maughan et al., 2016). Itraconozale is an antifungal agent tested in patients with biochemically relapsed PCa in a phase II trial (clinicaltrials.gov: NCT018787331). Results indicate that Itraconazole modulate serum PSA levels without affecting circulating serum testosterone levels (Lee et al., 2019).

FGF Pathway

FGF is involved in the regulation of many developmental processes including morphogenesis, differentiation, cell proliferation, and migration. FGFs are comprised of a gene family consisting of 22 structurally related proteins that are divided into 7 subfamilies based on sequence homology (Corn et al., 2013). FGF signaling is activated by ligand binding to four cognate high-affinity tyrosine kinases receptors (FGFR1-4). These interactions result in activation of multiple signal transduction pathways including the MAPK (Erk), PLCγ-PKC, and PI3K-AKT pathways (Kwabi-Addo et al., 2004).

In bone, multiple FGFs are produced by BMSCs and immune cells (Ishino et al., 2013; Itkin et al., 2013). FGF signaling in the bone microenvironment promotes castrate-resistant PCa growth (Lescarbeau et al., 2012; Li et al., 2008) and bone biopsies of mCRPC patients demonstrate greater frequency of FGF8 expression than patients with the primary tumors (West et al., 2001). FGF2, FGF8, and FGF9 are produced by PCa cells during bone metastases and are known to participate in osteogenesis and bone remodeling (Koeneman et al., 1999; Valta et al., 2008). A recent study has been reported a phenotypic shift occurs in metastatic PCa with the AR-null and neuroendocrine (NE)-null phenotype through elevated FGF and MAPK signaling pathways (Bluemn et al., 2017). At present more investigation is needed to elucidate the role of FGF signaling in CSC development in bone metastatic PCa.

Nevertheless, pre-clinical and clinical studies of FGF signaling pathway have been demonstrated in PCa. AZ8010, a selective inhibitor of FGFR1 and 4, inhibited invasion and tumor growth in vitro and in vivo tumor xenograft models (Feng et al., 2012). Combination of inhibitors of FGFR1 (PD173074 or CH-5183284) and/or MAPK (U0126) represses the growth of double-negative (AR-null and NE-null) metastatic PCa in vitro and in vivo tumor xenograft models (Bluemn et al., 2017). Dovitinib (TKI258), the FGFR antagonist targeting FGFR1-3, inhibits FGFR signaling in osteoblasts rather than in cancer cells, thereby improving bone quality and blocking cancer-bone interaction (Wan et al., 2014). Although Dovitinib shows efficacy in mouse-model of PCa bone metastasis, this Dovitinib induces NE phenotype from AR+ and AR-PCa cells, perhaps as a way of developing resistance (Yadav et al., 2017). A phase II study of Dovitinib in patients with mCRPC (clinicaltrials.gov: NCT01741116) demonstrated a modest antitumor effect (Choi et al., 2018). Dovitinib has manageable toxicities in men with mCRPC (Choi et al., 2018). A recent report shows that Nintedanib (BIBF1120), the tyrosine-kinase inhibitor of VEGF, FGF1-3, and PDGF signaling pathways, inhibits PCa growth by modulating both cell cycle and angiogenesis. Moreover, Nintedanib delays tumor progression in transgenic adenocarcinoma of the mouse prostate (TRAMP) (da Silva et al., 2017) as well as migration and invasion of PCa in nude mice (da Silva et al., 2018). Further, two studies show the efficacy and safety of two doses of Nintedanib in phase I and phase II trials. First study demonstrated in a phase I trial that Nintedanib (200mg twice daily) combined with Docetaxel (75mg/m2, every 3 weeks) was well tolerated to the patients, with promising results in chemo-naive hormone-refractory PCa (Bousquet et al., 2011). Second study demonstrated in a phase II trial that Nintedanib (250mg) showed only modest activity with manageable adverse events in patients with mCRPC with post-Docetaxel treatment (Droz et al., 2014). More recently, Nintedanib was tested in phase I and II clinical trials for the patients with hormone-resistant PCa (clinicaltrials.gov: NCT02856425, NCT00706628), although results have not been posted.

TGF Pathway

TGFβ is a secreted protein that performs many cellular functions including proliferation, differentiation, and migration (Massague, 2012). TGFβ has three isoforms, which bind to several different homo- or heterodimeric TGFβ receptors (Doré et al., 1998). TGFβ signaling has dual functions in terms of PCa; in the SMAD-dependent pathway TGFβ serves as a tumor suppressor in early stage PCa development and later TGFβ switches to non-SMAD pathways and serve as a tumor promoter in advanced PCa (Lebrun, 2012).

In bone, TGFβ is primarily produced by active osteoblasts and bone stromal cells and is a potent paracrine stimulator of PCa growth (Festuccia et al., 2000; Horner et al., 1998; Meng et al., 2018). In clinical studies, TGFβ plasma levels are markedly elevated in men with metastatic PCa to regional lymph nodes and bone and high preoperative TGFβ plasma levels correlate with disease progression after prostatectomy (Shariat et al., 2001). TGFβ signaling may play a significant role in establishing a CSC phenotype. Activation of TGFβ signaling induces significant enrichment of CD44+/CD133+/CD24− populations by downregulating poly r(C) binding protein (PCBP)-1 (Chen et al., 2015). Further, downregulation of TGFβ2 is associated with escape from dormancy in PCa xenograft models (Ruppender et al., 2015). Another member of the TGF superfamily BMP7 has been shown to play a critical role in dormancy and recurrence of prostate CSCs in bone via BMPR2 mediated p38 MAPK signaling (Kobayashi et al., 2011).

Pre-clinical studies have demonstrated that TGFβ signaling plays a significant role in PCa bone metastases. For example, Tranilast (N-3, 4-dimethoxycinnamoyl-anthranilic acid) is a therapeutic agent used in the treatment of allergic diseases. Tranilast suppresses TGFβ1 secretion from bone-derived stromal cells thereby inhibiting hormone refractory PCa cell growth (Izumi et al., 2009). Alternatively, IN-1130, a novel TGFβRI kinase (activin receptor-like kinase 5, ALK-5) inhibitor, decreases the PCa growth in mice in part by activating an host immune response (Lee et al., 2008), and SB-431542, another TGFRI ALK-5 inhibitor, also suppresses growth and motility of PCa cells (Miles et al., 2012). Inhibition of TGFβ signaling by a selective TGFβRI kinase inhibitor, LY2109761 was shown to control PCa growth in bone, while increasing the mass of normal bone (Wan et al., 2012). More recently, the combination of the TGFβR1 inhibitor Galunisertib (LY2157299) with Enzalutamide, significantly suppressed PCa growth by increasing apoptosis (Paller et al., 2018). At the clinical level, currently combination of Galunisertib and Enzalutamide in a phase II clinical trial (clinicaltrials.gov: NCT02452008) is ongoing for targeting mCRPC patients.

CXCL12 Pathway

CXCL12 is a chemokine, which is expressed in both embryonic and adult tissues including brain, heart, lung, liver, kidney, spleen, and bone marrow. CXCL12 plays a key role in tissue homeostasis, survival, trafficking of hematopoietic and immune cells, and cancer progression and metastases (Lataillade et al., 2004). CXCL12 binds primarily to CXC receptor 4 (CXCR4; CD184). CXCR4 is expressed in a variety of cell types including lymphocytes, hematopoietic stem cells, endothelial cells, epithelial cells, and cancer cells (Rossi and Zlotnik, 2000).

In bone, CXCL12 is highly expressed by pre-osteoblasts and endothelial cells and is down-regulated upon maturation and matrix synthesis (Jung et al., 2015). Importantly, this signaling pathway plays critical roles in the CSC activities that are responsible for the establishment and development of PCa bone metastases (Cojoc et al., 2013). Activation of CXCR4 by CXCL12 in the CD44+/CD133+ prostate CSCs induce self-renewal and tumorigenesis of PCa cells through activation of PI3K/AKT signaling and downregulation of the FOXO transcriptional factor, which ultimately impacts therapy resistance (Cojoc et al., 2013; Dubrovska et al., 2010; Dubrovska et al., 2009; Trautmann et al., 2014). High levels of CXCR4, CD44, and integrin α2/β1 expressing quiescent cells are significantly more tumorigenic than fast growing cells, indicating that CXCR4 is a CXC12 signaling mediator, which facilitates the regulation of CSC fate in bone metastases (Wang et al., 2015b). Interestingly, an intracellular splice variant of CXCL12, CXCL12γ induces the expression of a CSC (CD44+/CD133+) and neuroendocrine (high levels of CD44/CD133 expression) phenotypes of PCa cells through CXCR4-mediated PKCα and NFκB signaling. This activity results in the development of aggressive metastatic resistance of PCa (Jung et al., 2018).

Several pre-clinical studies of CXCR4 inhibitors have been shown to be significantly more effective at eradicating metastatic tumors when used in combination with conventional chemo-or radiation therapies that target PCa and/or PCa CSCs. For example, CXCR4 antagonist, CTCE-9908 (CTE9908) or Plerixafor (AMD3100) blocks CXCL12 binding to CXCR4 resulted in a significant improvement in the overall survival of mice and reduction of intra-osseous PCa tumor growth. In conjunction, decreased osteolysis, serum mouse tartrate-resistant acid phosphatase (mTRAP), and type I collagen fragments were observed (Gravina et al., 2015). Furthermore, AMD3100 alters the homing of CXCR4 expressing-quiescent tumor cells to the bone and reduces binding of tumor cells to the bone surface, although no significant reductions in the total number of tumor cells were detected in marrow (Wang et al., 2015b). AMD3100 also decreases the initial establishment of bone tumors without affecting the expansion of existing bone tumors in tumor xenograft models in a metastatic setting (Conley-LaComb et al., 2016). Not surprisingly, given the role in homing, AMD3100 is able to mobilize disseminated tumor cells (DTCs) out of the bone marrow niche in tumor xenograft models (Shiozawa et al., 2011). Critically, AMD3100 is able to target the induction of prostate stem-like population in PCa cells in vitro and in vivo tumor xenograft models (Dubrovska et al., 2012). AMD3100 inhibits the sphere formation in PCa cells and sensitizes PCa cells to Docetaxel chemotherapy (Jung et al., 2018). However, AMD3100 had minimal impact altering CXCL12γ-mediated prostatosphere formation, given the primary function of CXCL12γ is to signal intracellularly (Jung et al., 2018). Co-treatment of AMD3100 with granulocyte colony-stimulating factor (G-CSF) increases the mobilization of hematopoietic stem cells (HSCs), which is important as a source of HSCs for transplantation. AMD3100 is currently being used for the patients with lymphoma and multiple myeloma (Cashen et al., 2008; DiPersio et al., 2009). Recently, a clinical trial was conducted to evaluate whether CXCR4 inhibition in PCa patients by AMD3100 was able to mobilize DTCs out of the bone marrow (clinicaltrials.gov: NCT02478125), although results have not been posted.

GAS6 Pathway

GAS6 is a vitamin K dependent grow factor, which is expressed in many human tissues and regulates proliferation, survival, and migration (Goruppi et al., 1999). GAS6 binds to its receptors, TRYO3, AXL, and MER (TAM), which are known to play important roles in regulating DTC dormancy and CSCs in a number of settings (Taichman et al., 2013).

In bone, GAS6 is expressed by osteoblasts in marrow and is known to regulate bone metastasis and induction of DTC dormancy (Cackowski et al., 2017; Shiozawa et al., 2010; Taichman et al., 2013). The studies from our group have recently linked GAS6 to a CSC phenotype through MER receptor signaling (Jung et al., 2016; Shiozawa et al., 2016). In the first of these studies, we showed that DTCs recovered from murine marrow were significantly enriched for a CSC phenotype. Critically, the conversion of DTCs to CSCs (CD133+/CD44+) was regulated by niche-derived GAS6 through the MER and mTOR signaling pathways in vitro and in vivo tumor xenograft model. GAS6 triggered mTOR signaling in PCa followed by increases of both mTORC1 and mTORC2, and these were diminished by the mTORC1 inhibitor Rapamycin and the dual mTORC1/2 inhibitor pp242. Inhibition of MER signaling by a MER inhibitor decreased mTORC2 signaling activated by GAS6. Thus, the study demonstrated that the marrow niche plays a significant role in maintaining tumor-initiating PCa and suggests a functional relationship between CSCs and dormancy (Shiozawa et al., 2016). In a second of these studies, we demonstrated that elevated endogenous GAS6 expression can be found preferentially in putative CSCs (CD133+/CD44+) compared to non-CSCs (CD133-/CD44-) isolated from the coculture of PCa cells and osteoblasts in vitro. Similar observations were made in DTCs isolated from the bone marrow at 24 hours post injection in tumor xenograft models. The elevated endogenous GAS6 expression activates the phosphorylation of MER receptor signaling and subsequent induction of a CSC phenotype. The study supports the concept that endogenous GAS6 and MER receptor signaling contributes to the establishment of PCa CSCs in the bone marrow microenvironment (Jung et al., 2016).

Inhibitors of GAS6 and TAM signaling pathways are being tested in PCa tumors in combination with chemotherapeutic drugs in pre-clinical studies (Corno et al., 2016; Varkaris et al., 2011). For example, Bemcentinib known as R428 or BGB-324 is a novel small molecule relatively specific AXL inhibitor that blocks auto-phosphorylation of AXL. The activation of this agent blocks cancer progression, invasion, metastasis, and drug resistance in models of metastatic PCa. R428 sensitizes AXL expressing PCa cells (DU145) to Metformin treatment (Bansal et al., 2015). Thus, a combination of Docetaxel chemotherapy with R428 is significantly more effective to suppress proliferation, migration, invasion, and tumor growth, and resistance in PCa cells (PC3 and DU145) in vitro and in vivo tumor xenograft model. Further, Amuvatinib (MP470), a different AXL inhibitor, significantly blocks expression of the ATP-binding cassette B1 (ABCB1) in the resistant PCa cells (Lin et al., 2017). UNC1062, a MER inhibitor, prevents sphere-forming ability of PCa cells, while it did not affect cell viability (Jung et al., 2016; Shiozawa et al., 2016). Cabozantinib (XL-184, Exelixis) is an orally bioavailable inhibitor of pro-oncogenic tyrosine kinases including c-MET, vascular endothelial growth factor receptors (VEGFRs), ALK, RON, and AXL. Not unexpectedly given the broad range of targets, Cabozantinib has potent effects on both tumor and tumor-induced bone matrix remodeling in the intratibial VCaP xenograft models (Graham et al., 2014). Interestingly Cabozantinib inhibited tumor growth in tumor xenograft models in a metastatic setting and produced changes in the bone microenvironment, including a biphasic effect on an increase of osteoblast activity and inhibition of osteoclast production, resulting in bone remodeling (Dai et al., 2014). Cabozantinib has been shown to decrease RANKL expression in osteoblastic cells and produces inhibition of osteoclastogenesis and parathyroid hormone-related protein (PTHrP)-stimulated bone resorption (Stern and Alvares, 2014). Short-term treatment with Cabozantinib to animals without tumor induced rapid modification of the bone microenvironment including significant elongation of the epiphyseal growth plate and hypertrophic chondrocyte zone with increased osteoblast numbers and reduced osteoclasts as well as altering other marrow elements including megakaryocytes (Haider et al., 2015). In the plasma of Cabozantinib-treated mCRPC patients, significant increase levels of VEGFA, FLT3L, c-MET, AXL, GAS6A, bone-specific alkaline phosphatase, interleukin-8, hypoxia markers CA9 and Clusterin and decrease levels of VEGFR2, TRAP5b, Angiopoietin-2, TIMP-2 and TIE-2 were observed (Leibowitz-Amit et al., 2016).

At the clinical level, in Phase II and III clinical studies (clinicaltrials.gov: NCT01605227, NCT01522443), Cabozantinib showed clinical activity in men with mCRPC, including improvements in bone scans, pain, measurable soft tissue disease, circulating tumor cells, and decreases in biomarkers for bone turnover (Smith et al., 2013; Smith et al., 2016; Smith et al., 2014). Despite these encouraging results and early stage clinical activity, Cabozantinib failed to increase survival of heavily pre-treated mCRPC patients in a phase III trial (Smith et al., 2016). These observations highlight the need to further understand regulation of CSCs and castration resistance in order to best bring effective treatments to patients. For example, based on laboratory studies, Cabozantinib may be best used effectively with other agents. This was recently studied in a phase I trial in combination with second generation anti-androgen Abiraterone (Choudhury et al., 2018). Foretinib (GSK1363089 or XL880) is a multi-targeted tyrosine kinase inhibitor whose targets include c-MET, RON, AXL, VEGFR2, and PDGFR. Foretinib has similar pharmacologic effects to Cabozantinib. Foretinib has been tested in several solid tumors in phase II studies including the few patients with PCa (clinicaltrials.gov: NCT00726323). Foretinib showed modest antitumor activity with manageable toxicity profiles in patients (Choueiri et al., 2013).

Conclusion and Future Direction

A poor prognosis and reduced quality of life has been observed in patients with elevated numbers of PCa stem cells, which have been identified in bone. As metastatic PCa develops resistance to many of the current therapies, development of new therapies with novel approaches are needed. One approach is to target the bone microenvironment by focusing on signaling pathways which also govern PCa metastasis and CSC activities as those discussed in this review. Many studies have already been activated and several groups have progressed to preclinical and clinical stages of therapy development as summarized previously. This approach opens several new therapeutic possibilities in PCa and other tumor types. For example, breast cancer, which shares lineage-similarity with PCa, have been shown to be sensitive to selective pathway inhibitors identified by rapid high-throughput screening methods (Gupta et al., 2009). Furthermore, nano-therapy delivering siRNA (Zhou et al., 2013) or by targeting epigenetic events and/or genetic manipulation of these pathways have also shown early promise in breast cancer therapeutics (Eccles et al., 2013). These are a few of the many possible avenues not currently explored in the PCa field, but which, if applied in the context of targeting CSCs may be of great value in the future.