Myeloid‐derived cells in prostate cancer progression: phenotype and prospective therapies

Short abstract

Review on the potential contribution of myeloid‐derived cells to prostate cancer, mechanisms for myeloid cell recruitment, and emerging myeloid‐cell targeted therapies in the clinic.

Abstract

Prostate cancer is the second most common cause of cancer mortality in men in the United States. As is the case for other tumor types, accumulating evidence suggests an important role for myeloid‐derived cells in the promotion and progression of prostate cancer. Here, we briefly describe myeloid‐derived cells that interact with tumor cells and what is known about their immune suppressive function. We next discuss new evidence for tumor cell–mediated myeloid infiltration via the PI3K/PTEN/AKT signaling pathway and an alternative mechanism for immune evasion that may be regulated by an endoplasmic reticulum stress response. Finally, we discuss several interventions that target myeloid‐derived cells to treat prostate cancer.

Abbreviations

ATF6

activating transcription factor 6

BMDC

bone marrow–derived dendritic cell

cDC

classical dendritic cell

CSF1R

colony stimulation factor 1 receptor

DR5

death receptor 5

ER

endoplasmic reticulum

FA

fatty acid

FDA

U.S. Food and Drug Administration

IRE‐1α

inositol‐requiring protein 1α

mCRPC

metastatic castration‐resistant prostate cancer

M‐MDSC

monocytic myeloid‐derived suppressor cell

MDSC

myeloid‐derived suppressor cell

NLR

neutrophil‐to‐lymphocyte ratio

OS

overall survival

PERK

protein kinase RNA‐like endoplasmic reticulum kinase

PMN‐MDSC

polymorphonuclear myeloid‐derived suppressor cell

PTEN

phosphatase and tensin homolog

RANK

receptor activator of NF‐κB

RANKL

receptor activator of NF‐κB ligand

TAM

tumor‐associated macrophage

TME

tumor microenvironment

T_reg

regulatory T cells

UPR

unfolded protein response

XBP1

X‐box binding protein 1

YAP

yes‐associated protein

Introduction

Prostate cancer is the most common noncutaneous malignancy in men in the United States, accounting for 1 in 5 new cancer diagnoses and ranking second in mortality among all cancers in men [1]. The standard primary treatment for patients with localized prostate cancer includes surgery and/or radiation therapy, followed by androgen ablation—chemical or surgical castration—if disease returns. Alternatively, patients may be monitored closely if the cancer is thought to be of sufficiently low risk [2] and treated only upon signs of progression. Prostate cancer that presents in the metastatic state is generally treated with androgen ablation, but most patients eventually become refractory to this treatment, developing castration‐resistant disease (i.e, mCRPC). Although there are a number FDA‐approved therapeutic agents for the treatment of mCRPC—docetaxel, cabazitaxel, abiraterone, enzalutamide, radium‐223, and sipuleucel‐T—that have shown a positive impact on OS, currently there is no treatment that can cure mCRPC [3].

Cancer immunotherapy is a rapidly evolving treatment option for multiple cancers; here, patients’ immune systems are likely directed to activate a cytotoxic CD8⁺ T cell response against tumor cells [4]. Of interest, prostate cancer is the only solid tumor type for which there exists an FDA‐approved therapeutic vaccine [5]. Conversely, immune checkpoint blockade with anti–CTLA‐4 has shown relatively limited success [6, 7–8]. This may be partially explained by the preexistence of an immunosuppressive TME [9]. Myeloid‐derived cells are major players during cancer progression and might contribute to the experience of treatment failure in patients with prostate cancer. Recent clinical studies have shown that increased Mϕ infiltration in the primary tumor at baseline correlates with the failure of androgen ablation, known as biochemical recurrence [10]. In addition, a recent study reported that circulating MDSCs correlate with a failure to respond to cancer vaccines and immune checkpoint blockade [9]. Taken together, these data highlight the potential of targeting myeloid‐derived cells or the mechanism(s) that regulate their recruitment to improve response to immunotherapy in men with prostate cancer. In this review, we first briefly introduce myeloid‐derived cells that may contribute to prostate cancer progression and evaluate the evidence that supports their contribution. We next discuss a potential role for PI3K/PTEN/AKT activation in tumor cells and an ER stress response in myeloid‐derived cells in modulating immune infiltration and cancer immune evasion. Finally, we outline some of the approaches that have been studied in preclinical models and clinical trials, with an emphasis on therapies that target the immunosuppressive myeloid compartment in prostate cancer.

MYELOID‐DERIVED CELLS IN THE TME

Myeloid cells may orchestrate various aspects of cancer progression, including establishing an immunosuppressive TME, promoting tumor cell growth, promoting angiogenesis, establishing a metastatic niche, and facilitating metastasis. As several in‐depth reviews of myeloid cells and their role in cancer have been published [11, 12], we will focus on the myeloid‐derived cells that have been described as playing a role in prostate cancer progression.

Dendritic cells

BMDCs are a heterogeneous group of APCs that can be classified into two basic subtypes: plasmacytoid dendritic cells, which accumulate in blood and lymphoid tissue, and cDCs, which infiltrate lymphoid and nonlymphoid tissues [13]. In humans, BMDCs are defined as cells that are negative for the lineage markers CD3, CD19, CD20, CD56, and CD14, express MHC class II (HLA‐DR⁺), and that are low or intermediate for CD11c (integrin subunit α x) [14]. Whereas plasmacytoid dendritic cells can be characterized by their production of IFN‐α, cDCs express either CD1c (BDCA1) or CD141 (BDCA3) [14]. cDCs that express CD141 are a noteworthy subset of cDCs superior at cross‐presenting soluble Ag [15]. In rodents, this subset of cross‐presenting cDCs can be further subclassified into two populations depending on their expression of surface markers and their localization; CD8α in lymphoid tissue and CD103 in nonlymphoid tissue [16]. Both types of cDCs differentiate from CD11c⁺ precursors [14]. For a more complete review of BMDCs and other dendritic cell subsets, see Merad et al. [14].

Regardless of their subtype, cDCs in nonlymphoid tissues process captured Ag and, after activation and up‐regulation of CCR7 and l‐selectin, migrate to T cell zones in lymph nodes where they present Ag to induce either a T cell response or tolerance. Cross‐presenting cDCs are the most efficient APCs known and have the potential to activate CD4⁺ helper T cells and CD8⁺ CTLs [17]. In the presence of cancer, a subset of immature dendritic cells migrates to the tumor‐draining lymph node where they may stimulate the expansion of naturally occurring T_regs by secreting TGF‐β [18]. IL‐10 and other factors in the TME may hinder dendritic cell maturation [19, 20] and, consequently, facilitate Ag presentation in a suppressive context. Moreover, disrupting the ability of cDCs to present Ag to T cells may facilitate tumor immune evasion, as will be discussed in the subsection on ER stress in myeloid differentiation and Ag presentation. In addition, the protein, CTLA‐4, on T_regs may interact with the normally costimulatory molecules CD80 and CD86 on a subset of BMDCs (suppressive DCs), inducing them to express IDO, which further contributes to the immunosuppressive microenvironment [21] ( Fig. 1 ). This occurs when IDO catabolizes tryptophan, an essential amino acid for T cell function, in the TME [23]. As will be discussed in the section on therapeutic modulation of myeloid components in the TME, several inhibitors of IDO are in various stages of clinical development. For a more complete review of the immunosuppressive mechanisms of dendritic cells, see Shurin et al. [24].

Figure 1.

Contribution of myeloid‐derived cells to prostate cancer progression. M1‐like Mϕ may contribute to tumor initiation after an inflammatory insult to the prostate gland [22], likely by inducing accumulation of IL‐17 and tumor‐derived factors in the TME. Other myeloid‐derived cells—M2‐like Mϕ, MDSCs, and suppressive dendritic cells—likely infiltrate the TME early during tumor progression and suppress an antitumor response by inducing T_regs and preventing CD8 T cell infiltration and function. Bone resident Mϕ (osteoclasts) and PMN‐MDSCs further contribute to tumor dissemination to bone by increasing osteoclastic bone resorption and regulating tumor cell homing (via CXCR4 expression), respectively. RNS, reactive nitrogen species; ROS, reactive oxygen species.

MDSCs

During an inflammatory response, bone marrow–derived monocytes migrate into the tissues and replenish the resident pool of Mϕ and dendritic cells [25]. In cancer, this steady supply of mature leukocytes may be perturbed by factors that promote myelopoiesis, which leads to an accumulation of immature myeloid cells. Additional factors prevent these cells from differentiating into mature cells of the myeloid lineage—Mϕ, dendritic cells, and mature neutrophils—and promoting their pathologic function [12]. Accordingly, monocytes from peripheral blood of patients with prostate cancer may not develop into mature dendritic cells as efficiently as those derived from healthy donor blood samples [26], although a number of studies have shown that ex vivo culture of monocytes from patients with cancer can result in fully functional cDCs for cancer vaccines [17].

Two main types of MDSCs have been identified, PMN‐MDSCs and M‐MDSCs. PMN‐MDSCs are phenotypically similar to neutrophils, whereas M‐MDSCs are similar to monocytes. In humans, PMN‐MDSCs are defined as CD11b⁺ (integrin subunit α m⁺) CD14⁻ CD15⁺ or CD11b⁺ CD14⁻ CD66b⁺ (in some studies, CD33^dim is used instead of CD11b⁺) [27]. In mice, these cells are defined as CD11b⁺ Ly6G⁺ Ly6C^low [27]. M‐MDSCs are defined as CD11b⁺ CD14⁺ HLA‐DR^−/low CD15⁻ in humans and CD11b⁺ Ly6G⁻ Ly6C^hi in mice [27]. These two types of MDSCs use different mechanisms of immunosuppression (reviewed previously [28]). In brief, M‐MDSCs produce high amounts of NO (mediated by the Nos2 gene), decrease T cell nutrient availability (l‐arginine, l‐cysteine, and tryptophan) in the microenvironment, and induce T_reg differentiation via the production of IL‐10 and TGF‐β [29]. Conversely, PMN‐MDSCs produce ROS and reactive nitrogen species that facilitate CXCR4‐mediated tumor dissemination [30], lead to the down‐regulation of Ag presentation in cDCs [31], impede CD8⁺ T cell infiltration [32], and induce Ag‐specific CD8⁺ T cell tolerance [33]. In addition, both types of MDSCs prevent CD4 and CD8 T cell homing to lymph nodes via cleavage of l‐selectin from the plasma membranes of T cells [34]. Supporting a potential role for PMN‐MDSCs in prostate cancer, protein nitration (i.e., 3‐nitro‐tyrosine formation) was found to be associated with prostate cancer but not benign prostatic hyperplasia [35].

Mϕ

Inflammatory monocytes that enter into tissues from the bloodstream have been suggested to be the major source of Mϕ in the TME (i.e., TAMs) [36]; however, the contribution of in situ expansion of tissue‐resident Mϕ to TAMs in prostate cancer remains to be addressed. Inflammatory monocytes are defined as CD14^hi CD16⁻ CX3CR1^low CCR2^hi in humans and Ly6C^hi CX3CR1^low CCR2^hi in mice. The phenotype of these cells changes upon tumor infiltration; they mature into CD14^low CD16⁺ CX3CR1⁺ CCR2^low cells in humans and Ly6C^low CX3CR1⁺ CCR2^low Mϕ in mice [37, 38]. Mature Mϕ are subsequently polarized into distinct phenotypes depending on the cytokines present in the TME. In vitro, Mϕ can be polarized toward two distinct phenotypes (M1 and M2), but in vivo, these cells show a wide spectrum of polarization between those canonical states [39]. Mature Mϕ can be identified by the markers CD68 in humans and F4/80 (adhesion g protein‐coupled receptor e1) in mice [40]. In mice, MHC‐II^hi Mϕ have been shown to express M1 genes (IL1b, IL12b, and Ptgs2) and to produce NO more efficiently when stimulated with INF‐γ or LPS than MHC‐II^low Mϕ, which are characterized by the expression of M2 genes (IL10, Arg1, IL4Ra, and Mrc1) [41]. In general, M1 or classically activated Mϕ are considered antitumorigenic, whereas M2 or alternatively activated Mϕ are considered protumorigenic. TAMs may promote tumor growth by increasing tumor angiogenesis via the expression of angiogenic growth factors and matrix metalloproteinases [38] and by inducing T_reg differentiation in the TME via production of IL‐10 [39]. Of interest, the TAM phenotype may flux between the M1 and M2 phenotype according to the cytokines present in the TME during tumor progression [38] (Fig. 1) by a mechanism that may be dependent on the NF‐κB subunit 1, also known as the p50 subunit of NF‐κB [42]. For a more complete review of the skewing of Mϕ function in disease, see Franklin et al. [36].

MYELOID‐DERIVED CELLS IN PROSTATE CANCER INITIATION AND PROGRESSION

Myeloid‐derived cells in cancer initiation/establishment of the TME

Chronic inflammation may play a role in prostate cancer initiation [22]. Although the precise initiators of inflammation are unknown, potential etiologies include infectious agents, chronic noninfectious inflammatory diseases, and/or other environmental factors [43]. Indeed, inflammation on biopsy cores of benign prostate tissue is associated with the presence of prostate cancer [44]. Supporting a potential role for an inflammatory insult in tumorigenic inflammation, a human prostatic isolate of Escherichia coli accelerated prostate cancer progression in a spontaneous murine model of prostate cancer (Hi‐Myc) [45].

Upon insult, inflammatory Mϕ (Ly6C^hi CX3CR1^low CCR2^hi) accumulate in damaged tissue where paracrine signaling directs their maturation [38]. Once in the TME, TAMs themselves become a major source of inflammatory mediators, such as cytokines, chemokines, and growth factors [38]. Among these mediators, IL‐6 is of particular interest in prostate cancer [46]. IL‐6 binds to either its membrane receptor or its soluble receptor to induce the formation of a functional complex that induces the homodimerization of IL‐6 signal transducer, also known as gp130, which leads to the activation of the JAK pathway [47]. JAK‐mediated phosphorylation then leads to the activation of multiple signaling pathways, in particular, STAT3, MAPK, and PI3K/AKT [48] ( Fig. 2 ).

Figure 2.

Effects of PI3K/PTEN/AKT pathway dysregulation in prostate tumor cells. The noncanonical activation of AKT via IL‐6 signaling, ROS accumulation, and ER stress response in prostate cancer tumor cells is illustrated. Increased PI3K/PTEN/AKT pathway activation leads to prostate tumor cell survival (i.e., increased angiogenesis/lipid biosynthesis and decreased apoptosis) and the recruitment of myeloid cells. Binding of IL‐6 to its receptor activates JAK, which leads to the phosphorylation of PI3K and, ultimately, to AKT signaling. Accumulation of ROS can also indirectly mediate AKT phosphorylation by down‐regulating PTEN, which leads to unregulated PI3K activity. Finally, the ER stress response may also increase AKT signaling via the dissociation of HSPA5 from the ER sensors (PERK, IRE‐1α, and ATF6), although the precise mechanism(s) by which this occurs are currently unclear. In addition, XBP1s, generated by IRE‐1α RNase activity, increases lipid biosynthesis (saturated FA), which may also activate ER stress and maintain AKT signaling. HSPA5, heat shock protein family A member 5; IL‐6R, IL‐6 receptor; IL6ST, IL‐6 signal transducer.

The downstream effects of IL‐6 signaling are cell‐type dependent. Whereas IL‐6 signaling has been suggested to promote cancer progression by regulating cell growth, differentiation, and survival in prostate tumor cells [47], it has become apparent that IL‐6 can also exert its protumorigenic effects by modulating the TME. In this regard, IL‐6 promotes monocyte differentiation into M2‐like Mϕ when cultured in vitro [49] and induces naiüve T cells to differentiate into a subtype that secretes high amounts of IL‐17 [50, 51]. Accumulation of IL‐17 in the TME leads to further up‐regulation of IL‐6, potentially generating an amplification loop [52]. In addition, paracrine IL‐17 signaling may prime prostate tumor cells to produce factors that favor an M2‐like phenotype within TAMs (Fig. 1). Indeed, when media from murine prostate tumor cells that are cultured in the presence of IL‐17 is used to culture Mϕ, IL‐10 expression is increased [53]. Li and colleagues also reported that in vitro stimulation of a murine prostate cancer cell line with IL‐17 induces the up‐regulation of prostaglandin‐endoperoxide synthase 2, also known as COX‐2 [53]. This prostaglandin‐endoperoxide synthase 2 activity then leads to the conversion of arachidonic acid into PGE₂ [54], which, in turn, promotes the differentiation of monocytes into suppressive TAMs and prevents dendritic cell differentiation [55]. These data suggest that the IL‐6–mediated promotion of IL‐17 secretion might play a pivotal role in the switch between M1 and M2 Mϕ phenotypes during prostate cancer initiation and progression.

Myeloid‐derived cells in cancer progression

There is a large body of evidence from clinical studies that suggests that myeloid cells—neutrophils, MDSCs, and TAMs—contribute to cancer progression [56]. In prostate cancer, the immunosuppressive microenvironment has been shown to hinder monocyte differentiation and dendritic cell maturation [26]. Accordingly, the percentage of M‐MDSCs is significantly increased in the blood of patients with prostate cancer compared with sex‐ and age‐matched controls [57, 58]. Mechanistically, the ability of these cells to suppress T cell proliferation and to express high levels of IL‐10 has been confirmed in vitro [59]. Although M‐MDSC levels return to normal after removal of the prostate gland [58], elevated levels of M‐MDSCs have been associated with a shorter median OS in mCRPC [60].

Similarly, accumulation of PMN‐MDSCs in the peripheral blood of patients with prostate cancer has been correlated with decreased OS, increased levels of CXCL8 and IL‐6 [61], and prostate cancer progression from localized to metastatic disease [62]. Accordingly, a retrospective analysis of patients with mCRPC who received personalized peptide vaccination showed that a PMN‐MDSCs gene signature in PBMCs postvaccination correlated with poor prognosis [63]. Other studies that have focused on data from complete blood counts showed that an elevated NLR correlates with a positive needle biopsy for prostate cancer [64] and that NLR is a strong correlate of poorer OS in patients with mCRPC [65]. In addition, patients with mCRPC with a low NLR at baseline showed a significantly longer OS after first‐ and second‐line chemotherapy [57, 66]. At this point, it is difficult to discern whether the prognostic value of the NLR is a result of the contribution of PMN‐MDSCs in peripheral blood of patients with prostate cancer or whether it reflects CXCL8‐mediated neutrophil expansion. In this regard, a recently identified marker—oxidized LDL receptor 1—that distinguishes human neutrophils from PMN‐MDSCs [67] may inform future studies that are aimed at differentiating the contribution of neutrophils and PMN‐MDSCs to prostate cancer progression.

TAMs in the prostate cancer TME are also associated with poor prognosis and outcome. An elevated density of CD68⁺ TAMs that infiltrate the prostate at the time of radical prostatectomy correlates with extracapsular extension, which is a marker of poor prognosis [68]. Similarly, CD68⁺ TAM infiltration in prostate needle biopsy specimens correlates with prostate cancer progression as determined by prostate‐specific Ag recurrence or prostate‐specific Ag failure in patients who are treated with hormonal therapy as the first line of treatment [69]. In addition, an elevated density of CD68⁺ TAMs in tissue from patients who underwent androgen‐deprivation therapy before radical prostatectomy was found to be associated with an increased risk of biochemical recurrence [10]. In line with these observations, cell culture experiments have shown that IL‐1β—a cytokine that is highly produced by Mϕ—in the TME may mediate androgen independence by inducing degradation of the androgen receptor signaling complex [70, 71]. In animal models, the reduction in tumor growth after depletion of Mϕ using clodronate‐encapsulated liposomes or a pharmacologic inhibitor of CSF1R signaling (GW2580 and PLX3397) supports a role for TAMs in progression [72, 73] and biochemical recurrence [71].

cDCs, which, in general, promote an adaptive CTL response, are associated with an improved prognosis in patients with prostate cancer [74]. In addition, elevated numbers of BMDCs in the peripheral blood of patients are associated with a less aggressive phenotype [75]. Accordingly, patients with prostate cancer who received a cDC‐based vaccine and successfully mounted a specific T cell response against the vaccine Ag showed a significantly increased OS compared with those who received the vaccine but did not mount a specific T cell response [76]. These data highlight the potential importance of cDC maturation for an effective antitumor response and suggest that this process might be impeded in a subset of patients. Mechanistically, IL‐10 inhibits BMDC differentiation from monocytes in culture [19, 20] and thus may also inhibit the maturation of cDCs in the TME. It should be noted, however, that the ex vivo generation of BMDCs is feasible in most patients with prostate cancer. Some of the strategies to induce in vitro and in vivo cDC maturation for the treatment of prostate cancer have been reviewed elsewhere [4, 17].

Myeloid‐derived cells and cancer cell metastasis

Over time, prostate tumor cells accumulate genetic alterations that allow them to emigrate from the primary tumor and seed metastases at different anatomic sites [77]. These alterations may also allow them to recruit myeloid cells to the TME [78] and reprogram these cells in ways that may facilitate the escape from the prostate gland, survival in the circulation, and the establishment of a premetastatic niche in distant sites [11]. Early evidence of the importance of TAMs in the metastatic process came from a breast cancer model wherein primary tumors developed similarly in Mϕ‐deficient mice but were unable to form pulmonary metastases [79]. Although there is currently no preclinical model that accurately recapitulates the metastatic processes that are involved in human prostate cancer, M2‐like Mϕ have been reported to infiltrate metastatic prostate cancer lesions in rapid autopsy samples in higher concentrations than in adjacent normal tissues [49]. In addition, clinical data report the accumulation of PMN‐MDSCs and neutrophils in the peripheral blood of patients with mCRPC [61, 65].

Bone is the major metastatic site for human prostate cancer [80] and is a rich reservoir of MDSCs in animal models [29]. Accordingly, prostate tumor cells may travel to this premetastatic niche in response to a chemokine gradient that involves CXCR4 [11] (Fig. 1). This mechanism resembles a normal physiologic process in which immature myeloid cells in the bone marrow are maintained by CXCR4/CXCL12‐dependent chemokine signaling [81]. Indeed, increased CXCR4 expression on prostate tumor cells is associated with the presence of bone metastasis in patients with prostate cancer [82], and its role in bone homing has been highlighted by preclinical models that have shown that CXCR4 blockade significantly reduced the total metastatic load in tumor‐involved bone [83]. Of interest, CXCR4 has been reported to be up‐regulated by ROS‐mediated PTEN loss [30], which suggests that the premetastatic niche might be driven, to some degree, by PMN‐MDSC–derived ROS.

Another subset of myeloid‐derived cells that are fundamental for the successful establishment of cancer metastasis in the bone are bone‐resident Mϕ or osteoclasts. Osteoclasts are responsible for breaking down bone via a process of bone resorption that has the untoward effect of enabling tumor‐cell seeding [11] (Fig. 1). This process is highly dependent on RANK, recently renamed as TNFRSF11A, and its ligand RANKL or TNFSF11 [84], which has made the RANK/RANKL pathway a target of interest for the treatment of metastatic prostate cancer. In addition to its role in osteoclast activity, the RANK/RANKL pathway can also be activated in tumor cells. By using a transgenic mouse model that expresses the SV40 large T Ag in prostate epithelium, Luo and colleagues showed that metastasis to lymph node, liver, and lung was dependent on NF‐κB signaling [85].

A POTENTIAL ROLE FOR PI3K/PTEN/AKT PATHWAY ACTIVATION AND ER STRESS IN THE SUPPRESSIVE TME

Recent preclinical data indicate that dysregulation of intrinsic pathways in tumor cells results in the production of inflammatory mediators and, ultimately, in immune infiltration [86, 87–88]. In addition, preclinical models in other cancer types support the notion that these tumor‐cell intrinsic pathways may promote resistance to immune checkpoint blockade [89]. This may help to explain why only a small fraction of patients with prostate cancer respond to checkpoint blockade [6, 7, 90, 91] as well as to guide the search for new therapeutic agents for use in combination with current immunotherapies.

PI3K/PTEN/AKT pathway activation and recruitment of myeloid cells

The mutational landscape of mCRPC reveals that the most frequently altered genes include AR (62.7%), TMPRSS2‐ERG and other ETS fusions (56.7%), TP53 (53.3%), and PTEN (40.7%) [92]. Beyond these more common mutations, somatic alterations in genes that are involved in cell survival pathways (PI3K, Rb, RAF, and CDK), genome maintenance (BRCA2, BRCA1, and ATM), and cell fate (WNT) are also common in advanced disease [92]. Recent studies have shown that T cell and dendritic cell infiltration may be mediated by PTEN loss and WNT/B‐catenin activation in other tumor types [88]. These findings raise the possibility that the mutational signatures that are present in mCRPC could potentially underlie the suppressive microenvironment associated with disease.

In prostate cancer, PTEN loss and the subsequent PI3K/PTEN/AKT pathway activation [93, 94], has been described by multiple groups. Activation of this pathway correlates with a more aggressive phenotype [95] and a decreased recurrence‐free survival in patients with low‐risk prostate cancer (Gleason score 3+3 and 3+4) [96]. In addition to PTEN loss as a result of genetic or epigenetic factors, PTEN activity can also be lost as a consequence of a highly oxidative TME without overt loss of the gene itself [30]. These data suggest that abundant ROS‐producing cells, such as PMN‐MDSCs, could play an indirect role in sustained PTEN down‐regulation (Fig. 2).

Additional preclinical evidence supports the notion that myeloid‐derived cells can be recruited via PI3K/PTEN/AKT signaling in prostate tumor cells (Fig. 2). By using a conditional PTEN knockout mouse model, Garcia and colleagues showed that myeloid infiltration into the prostate gland is increased early during tumor development by epithelial PTEN loss [97]. In addition, in vitro experiments showed that myeloid recruitment may be mediated by CXCL8/CXCR2 signaling in human prostate cancer cell lines in which PTEN is lost [98, 99–100]. It will be interesting to determine whether similar events occur in men with prostate cancer (i.e., whether PTEN loss/PI3K activation correlates with the extent or function of an immunosuppressive microenvironment).

Additional mechanisms for PI3K/PTEN/AKT pathway activation

The cytokine IL‐6 may play a key role in linking PI3K/PTEN/AKT in tumor cells with myeloid recruitment (Fig. 2). Paracrine IL‐6 signaling has been shown to induce i.p. accumulation of monocytes in IL‐6 knockout mice that are treated i.p. with a fusion protein that contains IL‐6 fused to soluble IL6R [101]. This signaling has been shown to activate the PI3K/PTEN/AKT signaling pathway in prostate tumor cells [48]. Together, these data suggest that paracrine IL‐6 signaling could indirectly lead to PI3K/PTEN/AKT‐mediated myeloid recruitment (Fig. 2); however, other cytokines in the TME are also likely to contribute to the accumulation of myeloid‐derived cells in prostate tumors [86].

One process that may lie upstream of the PI3K/PTEN/AKT activation and subsequent recruitment of MDSCs is prolonged ER stress, which ultimately leads to UPR. This adaptive response increases tumor cell viability in the TME and is characterized by increased levels of the ER chaperone protein, HSPA5, also known as glucose regulated protein 78 [102, 103]. The ER sensors, PERK (also known as eukaryotic translation initiation factor 2 α kinase 3), IRE‐1α (also known as ER‐to‐nucleus signaling 1), and ATF6, restore cell homeostasis in cancer cells in conditions of ER stress [104, 105]. Activation of both PERK and IRE‐1α involves their dimerization, oligomerization, and trans‐autophosphorylation. Activated PERK leads to the inhibition of protein synthesis, whereas activated IRE‐1α exposes its RNase domain, which splices the mRNA of XBP1 to generate a transcription factor (XBP1s) that up‐regulates a subset of chaperone proteins. ATF6 translocates from the ER to the Golgi apparatus during stress conditions and is processed into ATF6f, a transcription factor that controls the expression of selected UPR target genes. UPR response is complex and is well‐discussed in a recent review [106]. Under resting conditions, HSPA5 binds to these ER sensors—PERK, IRE‐1α, and ATF6—and maintains them in an inactive state [107], thereby preventing a UPR response. A recent study in which persistent ER stress was modeled via multiple administrations of the toxin thapsigargin showed that ER stress induces accumulation of MDSCs in the spleens of mice that bear s.c. colon tumors [108]. Although this study did not directly address a role for AKT activity in tumor cells, experiments by Fu and colleagues showed that in vitro stimulation of human prostate tumor cells with thapsigargin resulted in ER stress and AKT activation [109]. Furthermore, PI3K/PTEN/AKT activation was found to be abrogated by HSPA5 deletion in a PTEN conditional knockout mouse model [109]. Taken together, these data raise the intriguing possibility that, in prostate cancer, myeloid recruitment after ER stress could possibly be regulated by PI3K/PTEN/AKT activation (Fig. 2); however, these observations must be tempered by the experimental models employed, as it is not clear how well prolonged treatment with thapsigargin models physiologic stressors in the human prostate TME.

Although multiple mechanisms may be responsible for an ER stress response, one physiologic mediator of increased ER stress in prostate cancer could be saturated FA. Accumulation of saturated FAs in the ER membrane activates the ER sensors, IRE‐1α and PERK, by enhancing their dimerization via their transmembrane domains [110]. Although not directly tested in prostate cancer specimens, PTEN inactivation in a human prostate cancer cell line led to increased expression of FA synthase, a lipogenic enzyme that catalyzes the terminal steps in the synthesis of long‐chain saturated FAs [111]. In addition, de novo lipogenesis has been shown to promote membrane lipid saturation in prostate tumor cells [112] and was found to be associated with an increased risk of prostate cancer in a nested case‐control study [113]. Together, these data suggest that saturated FA might induce ER stress and the subsequent PI3K/PTEN/AKT pathway activation as a possible step in the recruitment of myeloid cells in prostate cancer (Fig. 2).

ER stress in myeloid differentiation and Ag presentation

Recent studies have shown that ER stress can be transmitted from tumor cells to myeloid cells. In these experiments, Mϕ that were cultured in conditioned medium from ER‐stressed tumor cells evidenced an ER stress response themselves, with the up‐regulation of Hspa5 and Xbp1s [114]. The exact mechanism by which this transmission is achieved is still a subject of investigation; however, it is possible that TLR4 receptors on myeloid cells are activated as a consequence of immunogenic cancer cell death via release of high‐mobility group protein B1 [115] or lipids [116]. The notion that ER stress can be transmitted from tumor cells to myeloid cells suggests that ER stress may not only be involved in myeloid cell recruitment, but also in hindering myeloid differentiation and contributing to immune evasion in the TME. Indeed, Gabrilovich and colleagues demonstrated that ER stress was increased in MDSCs that were isolated from both tumor‐bearing mice and patients with non–small‐cell lung cancer and head and neck cancer [117]. Specifically, they showed clear up‐regulation of XBP1s [117]. XBP1s has been reported to regulate cell proliferation in prostate cancer cell lines [118], but its role in MDSCs requires further investigation. In addition, recent data from preclinical models that involved induced ER stress in cancer cell stress suggest that XBP1s might regulate the immunosuppressive function of MDSCs by up‐regulating Arg1 and Nos2 [108]. As ER stress plays a pivotal role in human prostate cancer, it is tempting to speculate that similar mechanisms might regulate the immunosuppressive phenotype of MDSCs in prostate cancer [105].

The immunologically suppressive role of ER stress extends to other cell types. For example, in intratumoral cDCs, ER stress leads to the down‐regulation of their Ag cross‐presentation capacity and, subsequently, to decreased priming of CTLs. This was demonstrated in animal models in which XBP1s up‐regulation led to the accumulation of intracellular lipids in cDCs, as well as in vitro in BMDCs that were cultured with the ER stressor, tunicamycin [31]. Lipid accumulation in cDCs may be mediated by the macrophage scavenger receptor 1 on the plasma membrane of dendritic cells [119] or by tumor‐derived factors [120] that could induce ER stress in a HSPA5‐dependent manner [121]. Cubillos‐Ruiz and colleagues showed that cDCs accumulate lipids in a process that is mediated by triglyceride biosynthesis, rather than by lipid intake, and that is dependent on XBP1s [31]. By using a mouse model with XPB1 specifically knocked out in CD11c⁺ cells, the prevention of ER stress in cDCs increased their function and enhanced their ability to prevent tumor progression as demonstrated by studies in which XPB1^−/− cDCs were adoptively transferred to wild‐type mice that were challenged with ovarian tumors [31]. Thus, it is intriguing to further speculate that the dysregulated lipid environment in prostate cancer may also mediate immunosuppression by altering the function of APCs, such as cDCs [122].

THERAPEUTIC MODULATION OF MYELOID COMPONENTS IN THE TME

Although immune checkpoint blockade has clear activity in multiple tumor types [123], studies in prostate cancer with such agents have generally been disappointing [6, 7, 90]. One possible explanation for this lack of activity could be related to the underlying myeloid components of the prostate cancer microenvironment. Because multiple interventions to target myeloid cells and their effects on the TME have been recently reviewed [124, 125–126], we focus here on novel targets that are relevant to the treatment of prostate cancer.

One interesting agent in this regard is tasquinimod (Active Biotech AB, Lund, Sweden), a second‐generation quinoline‐3‐carboxamide derivative [127]. Although its mechanism of action is still under investigation, tasquinimod has been shown to target S100A9, which leads to the reduction of MDSCs in the TME in a breast cancer model [128]. In addition, tasquinimod was also reported to inhibit M2‐like polarization and increase CTL infiltration when used in combination with a vaccine in a prostate cancer model [129]. Recently, tasquinimod was found to not improve OS in men with mCRPC in a phase III study [130, 131]. Still, the preclinical data suggest that tasquinimod could potentially be repurposed in combination with immunotherapeutic approaches to prostate cancer.

Targeting of the Hippo/YAP pathway is a novel approach to modulating the myeloid compartment in prostate cancer. In this tumor suppressor pathway, Hippo restricts organ size in mammals by antagonizing the oncoprotein YAP [132]. Consistent with a critical role for Hippo‐YAP signaling in normal tissue homeostasis, the YAP oncoprotein is activated in a wide spectrum of human cancers [133]. In prostate cancer, recent studies have shown that the common TMPRSS2‐ERG genomic fusion may lead to YAP activation [134]. In addition, YAP has recently been shown to bind the promoter of CXCL5—a homolog of CXCL8—and up‐regulate its expression in tumor cells to recruit MDSCs into the TME [135]. These data suggest that the YAP antagonist, verteporfin (Novartis Pharmaceuticals, Basel, Switzerland) [133], and other related compounds could be of clinical use for the treatment of prostate cancer. Indeed, a phase I clinical trial of verteporfin in patients with established bone metastases has been initiated (NCT02464761; Table 1 ).

Table 1.

Clinical trials targeting myeloid‐derived cells in prostate cancer

Agent	Target	Phase	Trial status	Design/description	Number of participants	Comments	NCCT identifier
Tasquinimod	S100A9	III	Completed	Randomized, double‐blind, placebo‐controlled study in men with mCRPC	1245	Primary end point = radiographic PFS (determined as time from random assignment to radiologic progression or death)	NCT01234311
Verteporfin	YAP	I	Ongoing	Open‐label study in patients with establish vertebral bone metastasis has been initiated	30	Primary end point = intraoperative dosimetry; secondary end points = Neurologic function (as determined by ASIA score)	NCT02464761
PLX3397	CSF1R/Kit/Flt3	II	Completed	Open‐label study in patients with advanced CRPC	6	Primary end point = CTCs; secondary end point = PFS (determined as PSA progression)	NCT01499043
Axitinib	VEGF/PDGF/CSF1R	II	Ongoing	Open‐label study in patients with untreated prostate cancer with known or suspected lymph node metastasis	72	Primary end point = PFS (determined as PSA progression)	NCT01409200
IMC‐CS4	CSF1R	I	Ongoing	Open‐label study in patients with advanced, refractory breast or prostate cancer	16	Primary end point = changes in peripheral blood immune cells subsets; secondary end point = pharmacokinetics	NCT02265536
Indoximod	IDO	II	Ongoing	Randomized, double‐blind, placebo‐controlled trial in patients with refractory metastatic prostate cancer receiving sipuleucel‐T	50	Primary end point = immune response to sipuleucel‐T; secondary end points = PFS (determined as PSA progression) and OS	NCT01560923
Denosumab	RANKL	III	Ongoing	Randomized, open‐label study in patients with metastatic breast cancer and metastatic prostate cancer	1380	Primary end point = time to first on‐trial SSE; secondary end point = OS	NCT02051218
AMD3100	CXCR4	I	Ongoing	Nonrandomized, open‐label study in men with metastatic prostate cancer	23	Primary end point = Mobilization of CTCs; secondary end point = PFS (determined as PSA progression)	NCT02478125

Another agent that targets the myeloid compartment is DS‐8273a (Daiichi Sankyo, Tokyo, Japan), a second‐generation mAb that targets DR5, also known as TNFRSF10B. Preclinical data from a lymphoma model and from DR5 knockout mice showed that DR5 promotes the accumulation of MDSCs such that DR5 inhibition facilitates the expansion and function of CTLs [117]. These data suggest that DR5 is a potential target for the attenuation of MDSC infiltration into the TME. An earlier mAb against DR5, CS‐1008 (Daiichi Sankyo), was demonstrated to be well tolerated in a phase I trial and to induce stable disease in 8 of 19 patients with metastatic colorectal cancer [136]. Taken together, these data support the clinical potential of targeting DR5. The safety and tolerability of DS‐8273a is currently being evaluated in an open‐label clinical trial for patients with advanced solid tumors or lymphomas (NCT02076451) and, as with the YAP inhibitor discussed above, DR5 inhibitors may eventually have utility in prostate cancer.

Alternatively, the immunosuppressive microenvironment in prostate cancer may be modulated by decreasing TAM recruitment to the tumor site. Because this recruitment is at least in part mediated by CSF1R signaling on inflammatory Mϕ, its inhibition has been proposed as a potential treatment for several tumor types, including prostate cancer. In animal models of prostate cancer, CSF1R blockade with small molecules was shown to delay tumor growth [71, 73]. Accordingly, two tyrosine kinase inhibitors, PLX3397 (Plexxikon, Berkeley, CA, USA) and axitinib (Pfizer, New York City, NY, USA), are now being tested in patients with advanced CRPC (NCT01499043; Table 1) and in patients with prostate cancer who are undergoing androgen ablation therapy (NCT01409200; Table 1), respectively. In addition to small molecules, mAb can be used to target CSF1R, which results in the inhibition of the recruitment of inflammatory Mϕ to the TME. Demonstrating clear enthusiasm for this approach, a total of 4 mAbs against CSF1R are currently in clinical trials for the treatment of various tumor types: FPA008 (NCT02526017; FivePrime, San Francisco, CA, USA), IMC‐CS4 (NCT01346358 and NCT02265536; Lilly, Indianapolis, IN, USA), AMG 820 (NCT01444404 and NCT02713529; Amgen, Thousand Oaks, CA, USA), and RG7155 (NCT01494688, NCT02760797, and NCT02323191; Hoffmann‐La Roche, Basel, Switzerland). Of interest, RG7155 was shown to induce objective clinical responses in 74% of patients with extra‐articular pigmented villonodular tenosynovitis [137] and is now in clinical trials for the treatment of several solid tumors in combination with chemotherapy (NCT01494688) and programmed death‐ligand 1 blockade, atezolizumab (NCT02323191; Hoffmann‐La Roche). Although the majority of these clinical studies are not specifically recruiting patients with prostate cancer, the activity of IMC‐CS4 is being evaluated in a phase I study for advanced breast cancer and CRPC (NCT02265536; Table 1).

The phenotype of TAMs could potentially be modulated by the neutralization of IL‐6/IL6R signaling. Current inhibitors of this pathway include mAbs against IL‐6, siltuximab (Janssen Pharmaceuticals, Beerse, Belgium), and IL6R, tocilizumab (Hoffmann‐La Roche). Preclinical studies have demonstrated a therapeutic effect for siltuximab in prostate cancer [138]. This observation was supported in a multicenter phase II study that tested siltuximab in patients with CRPC that were pretreated with one prior chemotherapy [139]. Although none of the patients achieved a response to treatment, 23% had stable disease [139]. Less encouraging results were observed when siltuximab was administered in combination with chemotherapy agents to patients with mCRPC [140], which suggested that IL‐6/IL6R blockade may be more effective earlier during disease progression when the contribution of TAM to the development of CRPC is likely to take place. Tocilizumab is currently FDA‐approved for the treatment of rheumatoid arthritis and was found to induce and maintain complete remission in patients with giant cell arteritis [141], a disease for which Mϕ are the major drivers [142]. The maximal tolerated dose of tocilizumab in patients with hepatocellular carcinoma will be evaluated in a phase IB study that will be followed by the phase II design for which the primary end point will be median progression‐free survival (NCT02997956). Although there are currently no trials underway, it is possible that tocilizumab could eventually be evaluated in patients with prostate cancer.

An effective CTL response against prostate cancer may also hinge upon addressing the suppressive phenotype of not only MDSCs, but also a subset of BMDCs in the TME. In this regard, IDO inhibition has been shown to decrease host‐mediated immunosuppression and enhance antitumor immunity in multiple preclinical models [143]. As discussed in the section on myeloid‐derived cells in the TME, IDO is the rate‐limiting step in the catabolism of tryptophan [23] and is produced by MDSCs and suppressive DCs in the TME. Accordingly, IDO inhibition with 1‐d‐MT (a tryptophan racemic isoform, 1‐methyl‐l‐ tryptophan) was shown to improve response to CTLA‐4 blockade in a preclinical model of melanoma [144], which suggests that IDO inhibition can reverse the immunosuppressive microenvironment in cancer. Three IDO inhibitors are currently in clinical trials for the treatment of various solid tumors: indoximod (NCT01560923 and NCT02077881; NewLink Genetics Corporation, Ames, IA, USA), GDC‐0919 (NCT02048709; Genentech, South San Francisco, CA, USA), and epacadostat (NCT02752074 and NCT02327078; Incyte, Wilmington, DE, USA). Recently, data from an ongoing phase I and II study that were presented at the Society for Immunotherapy of Cancer annual meeting suggest interesting antitumor activity in several tumor types when epacadostat is used in combination with programmed death‐1 blockade, pembrolizumab [145]. This combination is currently being tested in a phase III trial for advanced melanoma (NCT02752074). Furthermore, epacadostat is being tested in combination with a second programmed death‐1 blocking Ab, nivolumab (Bristol‐Myers Squibb, New York City NY, USA), in a phase I and II study for lymphomas and several solid tumors (NCT02327078). Although patients with prostate cancer are not being included in these studies—presumably because of the lack of activity of programmed death‐1 and programmed death‐ligand 1 blockade as a single agent—it is possible that IDO inhibitors may reverse the immunosuppressive TME and prime T cells for checkpoint blockade in prostate cancer. The ongoing phase II trial that combined indoximod with Sipuleucel‐T (Valeant Pharmaceuticals, Laval, QC, Canada) for the treatment of mCRPC is likely to shed some light on the effect of modulating the immunosuppressive microenvironment in prostate cancer (NCT01560923; Table 1).

Elevated osteoclast activity is an important aspect of the pathophysiology of treatment‐related complications in prostate cancer. Inhibition of bone‐resident Mϕ (osteoclasts) with the RANKL‐directed mAb, denosumab (Amgen), or with zoledronic acid (Novartis Pharmaceuticals) effectively reduces the loss of bone mineral density that is associated with androgen‐deprivation therapy [146] and prolongs bone metastasis–free survival in patients with CRPC [80]. In phase III trials, denosumab was found to be superior at preventing skeletal complications compared with treatment with zoledronic acid [147]; however, treatment with zoledronic acid was also shown to decrease the number of MDSCs in the spleens of tumor‐bearing mice and to increase the induction of an Ag‐specific CTL response when combined with vaccination [148]. These data suggest zoledronic acid as a possible agent to reduce the accumulation of MDSCs in prostate cancer and support its combination with a cancer vaccine for the treatment of prostate cancer. Whether denosumab also decreases the number of MDSCs in patients with prostate cancer has yet to be examined.

In keeping with the data discussed in the subsection on myeloid‐derived cells and cancer cell metastasis that show that CXCR4/CXCL12 chemokine signaling may be involved in metastatic dissemination (Fig. 1), the inhibition of CXCR4/CXCL12 signaling was shown to inhibit tumor growth and reduce metastasis in preclinical models of prostate cancer [149, 150]. Both the synthetic peptide, CTCE‐9908 (British Canadian BioScience Corporation, Vancouver, BC, Canada) and AMD3100 (Sanofi, Paris, France), bind to CXCR4 on tumor cells and prevent CXCL12‐mediated recruitment to bone marrow. As discussed above, CXCR4 is also expressed on monocytes and MDSCs in the bone marrow, but the contribution of these cells to an anti‐CXCR4 treatment response has yet to be investigated. Clinical data show that CXCR4 blockade results in myeloid‐derived cells leaving the bone marrow. Indeed, AMD3100 is FDA‐approved to increase hematopoietic stem cell recruitment to peripheral blood in the treatment of multiple myeloma [151]. In addition, AMD3100 has been shown to decrease matrix metalloproteinase production in the TME of a breast cancer tumor model [152], and CTCE‐9908 was shown to reduce angiogenesis—by inhibiting VEGF production—and recruitment of MDSCs into the TME of a prostate cancer model [149]. CTCE‐9908 was well‐tolerated by patients with solid tumors in a phase I and II clinical trial [153]. Recently, another small molecule, X4P‐001 (X4 Pharmaceuticals, Cambridge, MA, USA), has been reported to block CXCR4 in peripheral blood of patients who are infected with HIV [154] and has now moved into a phase I and II study for the treatment of renal cell carcinoma in combination with the standard of care tyrosine kinase inhibitor, axitinib (NCT02667886; Pfizer), and in combination with programmed death‐1 blockade, nivolumab (NCT02923531; Bristol‐Myers Squibb). In addition, AMD3100 is now being evaluated in a phase I study for the treatment of mCRPC (NCT02478125; Table 1), and preliminary reports from clinical studies using MDX‐1338, an IgG4 Ab that binds CXCR4, showed encouraging clinical activity in the treatment of in hematologic malignancies [155] and may be considered for the treatment of solid tumors, such as prostate cancer, in the future.

CONCLUSION

Is now fairly well accepted that cancers that are detected clinically must have evaded an antitumor immune response [156]. Whereas CD8 T cells are effective in mediating tumor‐cell lysis, their infiltration to prostate tumors is modest compared with the infiltration of these tumors with myeloid‐derived cells in human and mouse specimens [10, 86, 97]. It seems likely that the immunosuppressive microenvironment that is established by myeloid‐derived cells (TAMs, PMN‐MDSCs/neutrophils, M‐MDSCs, and suppressive dendritic cells) hinders the antitumor response in prostate cancer. Involvement of myeloid‐derived cells in prostate cancer treatment failure is supported by clinical evidence [9, 10]. Disrupting the immunosuppressive microenvironment of prostate cancer by inhibiting myeloid‐derived cells or their products requires further exploration in the treatment of this malignancy; however, it is possible that the efficacy of these treatments may be limited as a single agent and that combination regimens may be required. Furthermore, understanding the genetic alterations that lead to the dysregulation of intrinsic signaling pathways in prostate tumor cells and the mechanisms by which they regulate the infiltration of myeloid‐derived cells may have a significant impact on prostate cancer treatment. These genetic alterations may not only serve as biomarkers to aid treatment selection, but also as targets to modulate the infiltration of myeloid‐derived cells and the immunosuppressive microenvironment. Recent preclinical data suggest a link between PTEN loss, myeloid recruitment, and an immunosuppressive microenvironment in prostate cancer. Although this link has not been comprehensively studied in patients with prostate cancer, the frequency of loss of PTEN in mCRPC suggests that this is a topic worthy of further exploration.

AUTHORSHIP

Z.L.‐B. performed the literature review, wrote and edited the manuscript, and generated the figures. C.G.D. wrote and edited the manuscript.

DISCLOSURES

The authors declare no conflicts of interest.