The roles of proteases in prostate cancer

Since the proposition of the pro‐invasive activity of proteolytic enzymes over 70 years ago, several roles for proteases in cancer progression have been established. About half of the 473 active human proteases are expressed in the prostate and many of the most well‐characterized members of this enzyme family are regulated by androgens, hormones essential for development of prostate cancer. Most notably, several kallikrein‐related peptidases, including KLK3 (prostate‐specific antigen, PSA), the most well‐known prostate cancer marker, and type II transmembrane serine proteases, such as TMPRSS2 and matriptase, have been extensively studied and found to promote prostate cancer progression. Recent findings also suggest a critical role for proteases in the development of advanced and aggressive castration‐resistant prostate cancer (CRPC). Perhaps the most intriguing evidence for this role comes from studies showing that the protease‐activated transmembrane proteins, Notch and CDCP1, are associated with the development of CRPC. Here, we review the roles of proteases in prostate cancer, with a special focus on their regulation by androgens.


| INTRODUCTION
Proteases (also known as proteinases, peptidases, and proteolytic enzymes) play diverse roles in the settings of inflammatory diseases and cancer. [1][2][3] In particular, proteases have an important impact on the development of prostate cancer and its progression to aggressive and metastatic castration-resistant prostate cancer (CRPC), which represents a major treatment challenge. While several proteases are relevant for multiple cancers, they sometimes appear to have different functions in different cancers. 4,5 Moreover, the expression of several proteases is highly enriched in the prostate in comparison to other tissues, 6,7 and many prostate-derived proteases are regulated by androgens. 8,9 Thus, for this overview, we will focus principally on the roles of prostate-derived proteolytic enzymes in prostate cancer.

| Prostate cancer
Prostate cancer 10,11 is the second most frequent cancer and the fifth leading cause of cancer death in men globally. 12 While genetic predisposition has a role in prostate cancer development, environmental factors also contribute. 13,14 The vast majority of prostate cancers are adenocarcinomas, which arise from secretory glandular cells. Like cancers in general, prostate adenocarcinomas comprise many different molecular subtypes, which follow distinct pathways of progression. 15,16 Furthermore, localized prostate cancers often have several foci, which may be genetically different from each other. 10,11 When prostate cancer is suspected, magnetic resonance imaging is now generally recommended to be performed before diagnostic biopsies. 10,11,17 As prostate cancer usually progresses very slowly; low-risk patients, identified using several clinicopathological features, are often followed with active surveillance, in order to monitor whether the cancer progresses or is indolent. 11 Apart from those cancers classified as low risk, current clinicopathological evaluation or molecular biomarkers are inadequate for predicting patient outcome. 18 If detected early, most prostate cancers are curable by radical prostatectomy or radiotherapy. 11 However, about one third of treated cancers relapse, at least biochemically, based on measurements of serum prostate-specific antigen (PSA). Since the prostate is an androgendependent organ, relapsed cancers can be treated by androgen-deprivation therapy (ADT). Manipulation of the levels of androgenic hormones has been on the frontline of prostate cancer treatment since Huggins and Hodges published their seminal paper describing the androgen-dependency of malignant growth of prostate. 19 This publication is considered the first translational research study on a molecularly targeted cancer therapy, 20 based on which Huggins received a Nobel prize in 1966. While metastatic prostate cancers initially respond well to ADT, the disease eventually progresses to so-called CRPC, for which treatment options are limited. 11 Such cancers are not usually androgen-independent, but rather have acquired mechanisms that, irrespective of ADT, lead to restored androgen receptor (AR) signaling by various mechanisms, such as gain-offunction mutations in AR gene and increased sensitivity to residual concentrations of intra-tumoral androgens. 21,22 Less often CRPCs evade AR-targeted therapies through the development of the so-called lineage plasticity, where survival is independent of AR. 23,24 Of note, a body of evidence suggests that ADT, while suppressing cancer cell proliferation, may actually drive metastatic processes and development of CRPC, as described in more depth elsewhere. 25

| Proteases
Since the discovery of the digestive enzyme, pepsin, almost 200 years ago by Theodor Schwann, 27 proteases have been found to play roles in almost all biological signaling pathways and networks, performing many essential functions in all living organisms. [28][29][30] Thus, it is not surprising that more than 10% of the human protease genes are involved in human hereditary pathologies and several proteases are associated with more than one hereditary disease. 31 Proteases catalyze the hydrolytic cleavage of peptide bonds in their target proteins (substrates), either near the ends of polypeptide chains (exopeptidases) or at internal sites further away from the termini (endopeptidases). This cleavage leads to various activating or inactivating functional consequences in substrate proteins. 30,32 Some proteases exhibit highly selective substrate cleavage, while others have broader specificity. 33 To date, 473 active human proteases and 96 non-proteolytic homologs have been identified. 6,31,34 Collectively these represent $3% of the human protein-encoding genes and are called "the degradome". 31 Around half of these enzymes are secreted, whereas the rest are intracellular, with a minor fraction being membrane-bound, including an even smaller proportion that can be shed, that is, released from the cell surface, by other proteases. Based on their mechanism of action, involving specific amino acids or metal ions involved in catalysis, proteases can be divided into three major classes, metalloproteinases and cysteine and serine proteases, each containing around 150-200 members, and several smaller classes, including aspartic, threonine, glutamic, asparagine, and unclassified proteases.
Protease activity is controlled by several mechanisms. These include regulation of gene expression, activation of their inactive pro-forms (zymogens, such pro-proteases often have a suffix "-ogen" to distinguish them from active forms) either autocatalytically or by other proteases, other post-translational modifications, like glycosylation and phosphorylation, and regulation of their activity by endogenous protease inhibitors and cofactors. [35][36][37] Many proteases act in cascades or networks, which allow signal amplification and stringent regulation of their activity. Thus, mRNA expression or even protein levels of proteases do not necessarily correlate with activity. Furthermore, the physical context in which a protease is expressed, for example, a tumor and its microenvironment, is very important for understanding the functional roles of proteases, many of which have multiple substrates. Thus, analysis of in vitro cleavage of peptides or purified substrate proteins may not reveal the physiological role of a protease.

| Roles of proteases in cancer
Several proteases and their inhibitors are differentially expressed in cancerous and benign cells in the tumor microenvironment and are implicated in cancer progression. 6,38,39 However, as indicated above, changes in expression levels may not reflect alterations in protease activity. Furthermore, expression changes do not necessarily imply any functional roles, but may rather be a consequence, for example, of poor differentiation of cancer cells.
The roles of proteases in cancer, as outlined in Figure 1, have been widely studied since the discovery of their role in tumor cell invasion and metastatic dissemination, first suggested by Albert Fischer in 1946. 40 Collectively, these studies have shown that proteases play significant roles in virtually all stages of tumor progression. 1,2,30,41 Pericellular (i.e., secreted or cell surface) proteases degrade extracellular matrix (ECM) proteins, and, thus, provide paths for cells to migrate and generate protein fragments that promote cell migration and invasion. 1,42 Proteases also regulate the activity of several adhesion molecules, kinases, 36 growth factors, and receptors, 3 and are involved in epithelial-mesenchymal transition (EMT) 43 and other cancer-related processes. 1,2,30,41 Cancer has been thought to be primarily associated with increased proteolytic activity, often due to disturbed balance between proteases and their cognate inhibitors. 44 While increased activity in cancer holds true for many proteases, the opposite has also been observed. Some proteases may act as tumor suppressors by suppressing angiogenesis, inhibiting proliferation, inducing apoptosis, and, perhaps less often, by other mechanisms. 5 While proteases play critical roles in cancer progression, mutations in proteases do not appear to commonly initiate or drive cancer development. The Cancer Gene Census has catalogued genes containing mutations causally implicated in cancer. 45 Among over 700 such cancerdriving genes, only seven have been classified as proteases/peptidases. The cancer-driving activity in three of these, ubiquitin-specific peptidase 9 X-linked (USP9X), transmembrane serine protease 2 (TMPRSS2), and kallikrein-related peptidase 2 (KLK2), is not related to protease activity per se, but gene fusions. In the case of TMPRSS2 and KLK2, which encode prostate cancerassociated androgen-regulated proteases, these gene fusions result in a gain of androgen-regulated expression of transcription factors that drive tumor development. 46

| PROTEASES IN PROSTATE CANCER
About half of the 473 active human proteases are expressed in the prostate. 6 Some of these, like the kallikrein-related peptidase (KLK) family members, KLK2 and KLK3 (PSA), are predominantly expressed in F I G U R E 1 Examples of the roles of proteases in prostate cancer development and progression. A simplified scheme of the roles of the major proteases discussed in this review is illustrated. The schematic representation of tumor growth and early steps of metastatic dissemination, shown in the center panel, include: (1) proliferative and EMT signaling (interaction of cancer and stromal cells and inflammatory responses); (2) cancer cell migration and invasion, involving ECM remodeling and signaling; (3) stimulation of (lymph) angiogenesis and intravasation of invasive cancer cells, involving ECM remodeling, degradation of anti-angiogenic factors, production of angiogenic peptides and growth factor activation; (4) stromal cells, including cancer-associated fibroblasts (CAFs), are also involved in tumorigenic processes; (5) a small fraction of circulating tumor cells (CTCs) survive and eventually form distant metastases. These processes often involve autocrine-paracrine loops in the prostate cancer microenvironment, which due to simplicity have been omitted here. this organ. 7 Many proteases are differentially expressed in prostate cancer, as compared with benign prostatic tissue; and prostate cancer is also associated with increased protease activity in general. 38 Here, we do not aim to cover all potential prostate cancer relevant proteases, but rather limit the discussion to pericellular proteases expressed in cancer cells, which have gained substantial attention during recent years due to their potential roles in prostate cancer development and metastatic dissemination ( Figure 1). In addition to the KLKs, these proteases include matrix metalloproteinases (MMPs), type II transmembrane serine proteases (TTSPs), urokinase-type plasminogen activator (uPA), and trypsin-3. Since the strong dependency on androgen signaling is a unique feature in prostate cancer, we emphasize proteases and proteolytic events, which are regulated by or associated with androgen signaling. Of note, expression of proteases in the prostate is also regulated by several other factors, including hypoxia, inflammation, cytokines, and growth factors secreted by stromal and epithelial cells. We will not deal with these factors in this overview. Furthermore, it is important to point out that we will not consider indepth intracellular proteases and proteases expressed in stromal cells, although these play significant roles in cancer.
Among 150 core genes, reported to be downregulated by castration in prostate cancer xenograft tumors and upregulated in CRPC tissue, as compared with untreated prostate cancer tissue, one can point to meprin A subunit alpha (MEP1A), TMPRSS2, ubiquitin-specific peptidase 54 (USP54), KLK3, pyroglutamyl-peptidase I (PGPEP1), and cathepsin H (CTSH) encoded proteases (that is, the corresponding proteins possess peptidase activity based on Gene Ontology knowledgebase). 8,51 These and other proteases regulated by androgens may be involved in the therapeutic response to androgen ablation and development of CRPC. Indeed, recent findings have identified two known protease-activated molecules, Notch and CUB domain-containing protein 1 (CDCP1), that are key contributors to the progression to metastatic CRPC. 52,53

| Kallikrein-related peptidases (KLKs)
Over the years, several members of the KLK family of 15 serine proteases have gained much interest as prostate cancer biomarkers and potential key contributors of cancer progression. 9,[54][55][56][57] KLKs are secreted proteases, which, after activation of zymogen forms, show trypsinand/or chymotrypsin-like proteolytic activities. 58 KLK2 and KLK3, in particular, are highly expressed in prostate due to regulation by androgens, and show highly prostate enriched expression. 7,9 While in blood circulation the majority of KLK3 is complexed with inhibitors, that is, inactivated, in the prostate, the active form is prevalent. 59,60 This is likely the case with other prostatic KLKs as well. KLK3, more widely known as "prostate-specific antigen" or PSA, is a well-known prostate cancer biomarker. 61 Its circulating levels are often increased in prostate cancer patients due to increased leakage from the disease-affected prostate into circulation (more in Section 5 "Proteases as prostate cancer markers"). 61,62 However, KLK3 expression tends to be lower in prostate cancer tissue than in the benign prostate and reduced KLK3 immunostaining is associated with poor prognosis. [62][63][64][65] KLK2 staining also appears to be reduced in advanced tumors, and, interestingly, loss of staining is associated with TMPRSS2:ERG gene fusion, which in turn is associated with prostate cancer development. 66,67 However, increased staining of KLK2 in prostate cancer tissue has also been reported. 68 The expression of several other KLKs, including KLK4 and KLK14, is also changed in prostate cancer. 39,69,70 Although these changes may partially reflect the loss of cell differentiation in highgrade cancer, rather than indicating a causal relationship of KLKs on cancer development, several studies have indicated a functional role for KLKs in prostate cancer progression. 9,[54][55][56][57] The physiological function of KLK3, and probably also of other major prostatic KLKs, is to promote sperm motility by cleaving semenogelins, thereby dissolving the seminal clot formed after ejaculation. 71 Several KLKs have also been suggested to promote or inhibit tumor growth and metastasis 9,54-57,72,73 ( Figure 1). These functions may be mediated via proteolytic cascades 74,75 or directly, for example, by activation of protease-activated receptors (PARs), which are G-protein coupled receptors, or other cell surface or secreted proteins. Of the four human PARs (PARs 1-4), PAR1 and PAR2 are known to be activated by KLKs. 3,76 PAR activation leads to a wide range of responses, which in many cases promote tumor growth and metastatic dissemination. 3,[76][77][78][79] It is noteworthy that the PAR activation by different proteases may lead to different responses. 3 KLKs are also able to cleave several other prostate cancer relevant substrates, at least in vitro. 9,57 Some of the KLKs, including KLK3, degrade ECM proteins and activate other ECM-degrading proteases, suggesting that KLKs may be involved in the early steps of metastatic dissemination. 9,57,72,75 KLK3 and KLK4 also induce the EMT of prostate cancer cells, 72,80 which is typically associated with the gain of migratory and invasive properties. Furthermore, enzymatically active KLK3 promotes the growth of prostate cancer cell lines and subcutaneous xenograft tumors. 77,81,82 KLK2 or KLK3 do not appear to be sufficient to induce cancer development, as studied using transgenic mice expressing KLK3 or both KLK2 and KLK3 in the prostate. 83 However, the KLK3 levels in these mice were several orders lower than those in the human prostate. Of note, mice do not have genes encoding KLK2 or KLK3, and, therefore, knock-out studies cannot be performed.
In endothelial cell models, KLK3 shows antiangiogenic activity, 73,84 which is dependent on its proteolytic activity. 85,86 As tumors need to establish blood circulation in order to grow beyond microscopic size, the slow growth of many prostate cancers has been proposed to be dependent on this antiangiogenic activity of KLK3. 54,73,84 However, KLK3 was recently found to activate vascular endothelial growth factor-C (VEGF-C) and VEGF-D, which may support angiogenesis and tumor growth. 73,77 These observations are not necessarily in conflict, since KLK3-activated VEGF-C is lymphangiogenic rather than angiogenic, and the overall effect of KLK3 is likely dependent on several different factors, including expression of different VEGFs and other (anti)angiogenic factors, along with their activating proteases. 73 Noteworthy, the rather artificial angiogenesis models used in KLK3 studies do not recapitulate different "non-orthodox" mechanisms that tumors use to establish a blood supply. 73

| Type II transmembrane serine proteases (TTSPs)
Several TTSPs have been implicated in cancer. 87 In humans, the TTSP family consists of 17 members, characterized by a short N-terminal cytoplasmic domain, a single-pass transmembrane domain, and an extracellular C-terminus comprising a serine protease domain, with some members also containing domains, generally of unknown function, between the transmembrane and protease domains. 88,89 Like many other proteases, TTSPs are synthesized as zymogens and require activation, which may also take place by autoactivation. [90][91][92] Proteolytic activity of TTSPs is also regulated by endogenous protease inhibitors, by shedding the protease domain from the cell surface and by internalization and lysosomal degradation. 88 Hepsin (TMPRSS1), TMPRSS2, and matriptase (TMPRSS14) have significant role(s) in prostate cancer and are upregulated in carcinoma cells (Figure 1). While hepsin is overexpressed in prostate cancer at levels up to 40-fold over benign epithelium, making it one of the most upregulated genes, [93][94][95] its decreased expression is associated with poor prognosis. 94 Androgen-regulated TMPRSS2 96,97 is consistently expressed at high levels in metastatic prostate cancers. 98 Overexpression of matriptase is also associated with advanced cancers. 99 The elevation of these proteases in prostate cancer tissue is accompanied by decreased levels of their cognate inhibitors, hepatocyte growth factor activator inhibitor (HAI) À1 and À2. [99][100][101][102][103] Hepsin overexpression in a mouse model of nonmetastasizing prostate cancer resulted in basement membrane disorganization and enhanced metastasis. 104 Thus, it has been proposed that hepsin upregulation promotes prostate cancer progression and metastasis. 104 However, overexpression of proteolytically active hepsin in tumor cell lines is associated with growth suppression and reduction in the invasiveness of the cells. 105,106 Such apparently opposite findings and a lack of hepsin expression in metastases have led to the so-called "hepsin paradox". 107 These conflicting findings have been partially explained by differences in microenvironmental factors, like ECM composition, and altered subcellular localization of hepsin. 106,108 Such considerations are likely to be critical for many other proteases, highlighting the importance not only of addressing the activity and localization of proteases in functional studies, but also the need for models that are able to recapitulate various subtypes and stages of prostate cancer development and progression in tumor microenvironment.
Like hepsin, TMPRSS2 contributes to prostate cancer invasion, tumor growth, and metastasis 97,98 and has been proposed as a key effector of androgen-driven prostate cancer progression. 98 TMPRSS2 activates matriptase, further promoting the invasion and migration of prostate cancer cells. 97,109 Matriptase may also autoactivate, a process that is enhanced by the acidity of the prostate cancer tumor microenvironment. 110,111 Furthermore, TMPRSS2 induces shedding of cell surface HAI-1 in prostate cancer cells, which may increase the activity of TTSPs in general. 97 Because of the important role of TMPRSS2 in prostate cancer progression, it is somewhat surprising that in about half of the prostate cancers TMPRSS2 expression is lost due to fusion of the TMPRSS2 gene with the gene encoding the ETS family transcription factor ERG or, less often, with other members of this family. 46,67,112 This is the most frequent genomic alteration in prostate cancers and leads to relocation of the androgen-responsive element of TMPRSS2 so that it drives expression of fusion partners. 46,113 The TMPRSS2:ETS fusion is likely an early carcinogenic event, potentially provoked by androgens. 21 However, the gene fusion has also been implicated in metastatic progression of prostate cancer. 67 Interestingly, in a subset of TMPRSS2:ETS fusion negative cancers, increased expression of the serine peptidase inhibitor, Kazal type 1 (SPINK1), has been observed. 114 Although SPINK1 is a protease inhibitor, inhibiting trypsin-1 and -2 and other proteases, its effects on prostate cancer appear to be mediated by other mechanism(s), perhaps acting via modulation of epidermal growth factor (EGF) receptor signaling pathways. 115,116 Both TMPRSS2 and the TMPRSS2:ETS gene fusion implicate the importance of AR signaling in prostate cancer progression. 21,98 Several TTSP substrates have also been implicated as direct contributors to tumor progression and TMPRSS2 may participate in prostate cancer relevant proteolytic cascades, including activation of KLKs. 98 As already noted, TMPRSS2 activates matriptase, which in turn is able to activate urokinase-type plasminogen activator (uPA), protease-activated receptor 2 (PAR-2), and CDCP1. 92,[117][118][119] Since all the TTSPs discussed above have been implicated in metastasis, it is not surprising that they also share some of the substrates and mechanisms mediating their functions. Indeed, they (and some other proteases) activate pro-HGF, which leads to increased c-Met receptor tyrosine kinase signaling and, thus, proliferative, anti-apoptotic, and pro-migratory signals in transformed cells. 98,[120][121][122][123] Overactivation of c-Met by elevated levels of matriptase, hepsin, and TMPRSS2 with concomitant decrease in HAIs has been found in localized prostate cancer. 93,99,124,125 While in most tumors Met activation has been suggested to be a late event that aggravates the intrinsic malignant properties of already transformed cells, 123 increased expression and phosphorylation of Met also are apparent in precancerous prostate. 126,127

| Matrix metalloproteinases (MMPs)
The human MMP family consists of 17 soluble and 6 membrane-bound (MT-MMPs) zinc-dependent proteases. MMPs are well known for their prominent roles in tissue remodeling, promoting cancer invasion and metastasis, angiogenesis, and EMT. [128][129][130][131][132] Importantly, several MMPs also have activities that suppress tumor progression. 5,[133][134][135] By degrading ECM and other substrates, MMPs produce various bioactive protein fragments, release/activate growth factors and cytokines, and provide tracks for cells to move along. 128,129 Among the MMPs, MMP-14 (MT1-MMP) is strongly associated with cancer invasion and metastasis. 136,137 Since MMPs have been extensively studied in cancer in general, it is not surprising that they have also been evaluated in the setting of prostate cancer. [138][139][140] Overall, these studies have shown upregulation of several MMPs, including MMP2 and À 9, and MT1-MMP, which often, but not always, have been associated with downregulation of their tissue inhibitors, the TIMPs. 138,140,141 These upregulated MMPs are likely to promote several hallmarks in prostate cancer progression, including formation of bone metastasis 139,140,142 (Figure 1). MMP-7 and -9 were both upregulated in osteoclasts at the tumorbone interface in a rat model that mimics metastatic prostate cancer interactions with bone and the osteolytic and osteogenic responses to tumor colonization of bone. 143,144 MMP-7 promotes bone destruction by cleaving and solubilizing the TNF family member, RANKL, which is an essential mediator of osteoclast activation. 143 MMP-9 promotes prostate cancer growth in the bone microenvironment by processing VEGF-A to a bioavailable form and stimulating angiogenesis. 144 Thus, MMPs play critical roles in controlling cytokine activity and availability to drive prostate cancer progression.
Cellular context and the availability of cytokines and growth factors also in turn play a major role in regulating the expression and activity of MMPs and TIMPs, 128,138 but androgen regulation also appears to have a role. Studies with androgen-sensitive human prostate cancer LNCaP cells have demonstrated that androgen stimulates MMP-2 expression 145 and downregulates MMP inhibitor TIMP-2. 146 In the rat ventral prostate, androgen ablation by castration increases mRNA levels of Mmp-9, Mmp-2, Timp-1, and Timp-2. 147 MMP-9 upregulation by androgen deprivation or by down-regulation of AR expression has also been found in other animal and cell model studies. 25,[148][149][150] This phenomenon may facilitate the development of metastatic CRPC upon ADT. AR appears to regulate the expression of MMP-1, MMP-3, and MMP-7 negatively. 151 2.4 | Urokinase-type plasminogen activator (uPA) system Another key player in ECM remodeling is the uPA system, which is comprised of uPA itself, its receptor uPAR, plasminogen activator inhibitors PAI-1 and -2, and plasmin(ogen) 152 (Figure 1). uPA and uPAR-mediated signaling has been implicated in tumor cell migration, invasion, survival, metastasis, and angiogenesis in a variety of cancers, including prostate cancer, in which they are expressed at much higher levels than in benign tissues. 129,[152][153][154][155][156][157][158] Interestingly, overexpression of uPA, uPAR, and PAI-1 in prostate cancer tissue is associated with biochemical recurrence in patients treated by radical prostatectomy. 159 Perhaps, these proteases have facilitated, already in advance of radical prostatectomy, the formation of micrometastases that eventually cause recurrence. At least in part, tumorigenic activity of uPA is likely to be mediated by its proteolytic activation of CDCP1. 160 Proteolysis of CDCP1 by the plasminogen activation system was first noted from cultured human foreskin keratinocytes where plasmin-induced Src family kinase phosphorylation of cleaved CDCP1 provided the first insights into downstream intracellular signaling via this protease-activated receptor system. 161 The functional importance of this protease-initiated signaling cascade has been demonstrated in vivo. In chicken embryo and mouse models of vascular metastasis, plasmin-mediated proteolysis of CDCP1 in PC-3 prostate cancer cells initiates pro-survival signaling involving recruitment of Src and protein kinase C (PKC) δ, Src-mediated phosphorylation of cell-surface cleaved CDCP1, activation Akt and suppression of poly [ADP-ribose] polymerase 1 (PARP1)induced apoptosis. 162 Cleaved CDCP1 also complexes with activated β1 integrin-inducing signaling via FAK/-PI3K/Akt to facilitate the metastatic cascade. 163 Blockade of cleavage using a CDCP1-targeted monoclonal antibody, serine protease inhibitors, or mutation of the CDCP1 cleavage sites completely abrogates pro-survival signaling and successful tissue colonization by tumor cells. 162,163 The critical role of uPA as the initiator of CDCP1 proteolysis was elucidated using a novel "substrate-biased activity-based probe" approach that permits the unbiased identification of proteases that are responsible for a particular cleavage event in complex biological settings. 160 Using this approach, uPA was identified as the master regulator of CDCP1 proteolysis, acting directly as well as indirectly by converting plasminogen to CDCP1-cleaving plasmin. 160 As detailed in Section 3.2, at the protein level, CDCP1 is especially overexpressed in a subset of phosphatase and tensin homolog (PTEN) negative advanced and metastatic prostate cancers and is associated with poor outcome. 53 While it is clear that CDCP1 cooperates with the loss of the tumor suppressor PTEN to promote metastatic prostate cancer, 53 the role of proteolysis is yet to be explored and it will be interesting to define the contribution of uPA to CDCP1-regulated progression of PTEN-null prostate cancer. AR activation appears to downregulate the uPA system proteases, although this observation is somewhat controversial [164][165][166][167] ; and uPA and uPAR genes are notably upregulated in androgeninsensitive PC-3 prostate cancer cells in comparison with hormone-sensitive LNCaP cells. 168

| PRSS3
Three different genes (PRSS1, PRSS2, and PRSS3) encode human trypsin precursors, trypsinogen-1, À2, and À 3. These can be either autoactivated or activated by other proteases, like enteropeptidase, to trypsin-1, À2, and À 3 (also known as cationic and anionic trypsin, and mesotrypsin, respectively). 169,170 While the main physiological function of these trypsins is likely related to degradation of dietary proteins and activation of other digestive enzymes, 169 trypsins have proposed roles in promoting progression of various cancers, including prostate cancer. 170,171 Such activity may be, at least in part, mediated by activation of proMMPs, including proMT1-MMP and proMMP-9. 170,172 Trypsin-3 in particular is overexpressed in prostate cancer and mediates prostate cancer progression and metastasis. 173 Trypsin-1 and -2 are also overexpressed in prostate cancer, 174 but their roles have not been as well studied. The mechanisms mediating the effects of trypsin-3 on prostate cancer metastasis are still mostly speculative. All of the trypsins, including trypsin-3 are capable of regulating PARs-1, À2, and À 4, 175,176 which may mediate some of its pro-tumorigenic properties. Intriguingly, trypsin-3 is able to proteolytically inactivate most polypeptide trypsin inhibitors, including SPINK1 and HAI-2, the inhibitor of matriptase and hepsin. 115,177,178 Since these inhibitors inhibit trypsin-1 and -2, TTSPs and other proteases, trypsin-3 may increase the overall proteolytic activity in cancer, although it is possible that SPINK1 inactivation may also abolish its protease-independent activities promoting cancer progression. 115 While androgen decreases secretion of trypsinogen-1 and -2 in LNCaP cells, 174 it is not known whether it has an effect on trypsinogen-3 expression or secretion. However, trypsinogen-3 is expressed in androgen-independent PC-3 and PC3-M cells, 173,175 suggesting that this is not the case. Of note, PRSS3 is one of the genes most disproportionately overexpressed in PC-3 cells compared with LNCaP cells. 168

| NOTCH AND CDCP1 SIGNALING IN PROSTATE CANCER
Some of the most intriguing support for the role of proteases in the development of CRPC is indirect and comes from recent studies showing compelling evidence that Notch and CDCP1, which are known to be activated by proteases, are involved in the process.

| Notch
There are four different Notch receptors in mammals, each being a large single-pass type I transmembrane protein activated through a sequence of proteolytic events. 179 A prerequisite for canonical Notch activation is binding of the Notch ligand, which is often a type I transmembrane protein in an adjacent cell. This binding leads to a conformational change in Notch that allows subsequent proteolytic ectodomain shedding, which is the key to the regulation of Notch activation and translocation of the Notch intracellular domain (NICD) into the nucleus. NICD participates in the regulation of gene expression, leading to activation of several signaling pathways. ADAM10 metalloprotease is essential for such ligandinduced activation. 180 While Notch signaling has a fundamental role in development of tissues, including normal prostate, 181 its aberrant activation has been linked to initiation and progression of several malignancies. [182][183][184] However, the functions of Notch appear to be dependent on the tissue and cellular context. Thus, Notch also acts as a tumor suppressor. 185 In prostate cancer, Notch1 activation (or overexpression) triggered by cell-cell communication promotes cell proliferation, differentiation including EMT, and invasion. [186][187][188][189] Activation of the Notch pathway promotes development and progression of different subtypes of prostate cancer, including TMPRSS2:ERG rearrangement positive cancers, 52 neuroendocrine cancers, 190,191 and PTEN-deficient advanced prostate cancers. 192,193 Oncogenic alterations in the Notch pathway are enriched in aggressive metastatic CRPCs. 194 Importantly, Notch signaling promotes progression to metastatic CRPC and resistance to androgen deprivation, 186,195,196 while inhibition of the Notch pathway restores the response to ADT. [197][198][199][200][201] While in benign prostate tissue the Notch-activator ADAM10 is predominantly found at the cell membrane, where Notch activation takes place, in prostate cancer, ADAM10 is more often detected in the nucleus. 202 Furthermore, androgens upregulate the expression of ADAM10, but at the same time direct it to the nucleus. 202,203 The precise mechanisms whereby proteases are involved in the regulation of Notch activation and its downstream effects in prostate cancer cells are somewhat unclear. Thus, further research is warranted.

| CDCP1
Like Notch1, CDCP1 is a proteolytically activated type I transmembrane glycoprotein that is upregulated in several cancers, including prostate cancer. 204 CDCP1 has a central role in regulation of multiple oncogenic signaling pathways. 204,205 While the serine proteases trypsin and matriptase are capable of cleaving CDCP1 to induce downstream signaling via Src and integrins, 119,163,206 as detailed in Section 2.4, uPA was recently identified as the master regulator of CDCP1 proteolysis in various cancer cell lines, including PC-3 prostate cancer cells. 160 At the protein level, CDCP1 is overexpressed, especially in a subset of CRPCs and metastatic prostate cancers, and high expression is associated with loss of tumor suppressor gene PTEN. 53 In a series of experiments, for example, using mouse models with conditional expression of CDCP1 in prostate and Pten-null prostate conditional mice, it was shown that CDCP1 drives prostate cancer development and, together with the loss of PTEN, further drives prostate cancer progression and metastasis. 53 Importantly, androgen deprivation with enzalutamide, as is often used for therapy of CRPC patients, induced CDCP1 expression in PTEN-deficient cells and led to activation of the SRC/MAPK pathway, 53 one of the key pathways involved in CDCP1-mediated cancer progression and metastatic dissemination of cancer cells. 204 This activation may be one of the mechanisms explaining the emergence of treatment-resistant and metastasis-prone CRPC after initial ADT. Thus, targeting CDCP1 is a potential novel therapeutic strategy for treating metastatic prostate cancers. 53 Since proteolytic activation of CDCP1 was not addressed in the above study 53 and since the functions of CDCP1 may be mediated by a mechanism not involving proteolytic activation, 204 the involvement of proteases in CDCP1-driven prostate cancer progression is still open to further study.

| PROTEASES AS PROSTATE CANCER DRUG TARGETS
The first clinically useful protease inhibitor approved by the U.S. Food and Drug Administration (FDA) was the AIDS drug, saquinavir (Invirase) approved in 1995. 207 Thereafter, major efforts to develop protease inhibitors have been related to viral diseases, including TMPRSS2 inhibitors to reduce cellular infection by SARS-CoV-2. 208,209 Some of these inhibitors also have anticancer activities. 207,210 Thus, such efforts may also facilitate the development of prostate cancer drugs. Protease inhibitors have also proven to be effective as proteasome inhibitors for treatment of multiple myeloma. 211 Several protease inhibitors have shown promising effects in experimental, preclinical cancer models, including those for prostate cancer. 57,98,[212][213][214] Still, there are no FDA-approved protease inhibitory drugs for prostate cancer. In part, this outcome is likely due to the lack of selectivity of inhibitors toward highly similar proteases and the broad range of protease activities that are crucial for normal physiology. 215 For example, specificity has been one of the major reasons for the failure of all early MMP inhibitors to treat cancer, as they also inhibited MMPs that are needed for normal tissue function or to act as tumor suppressors. 5,133,134 Nevertheless, efforts to develop more selective MMP inhibitors are continuing. [216][217][218] Instead of targeting active sites, targeting exosites and allosteric interactions may offer a better solution for inhibiting some proteases. 32,218 In pursuit of selectivity, RNA-interference (RNAi)-based therapeutics, inhibitory antibodies and peptides, modified versions of endogenous protease inhibitors, and complex natural molecules may serve as potential alternatives for traditional small molecule enzyme inhibitors. 212,[219][220][221][222][223] Proteases can also serve as targets for therapeutic vaccines. In particular, KLK2 and KLK3, which show high expression primarily in the prostate, are attractive targets in this respect. 57 Another approach for protease-related therapeutics is to utilize their activity, for example, several pro-drugs have been developed that are activated in the prostate by KLK3. 57,224 Furthermore, proteolysis can be used for regulation of the AR. 225 Adaptive responses of AR signaling in CRPC development seem to be, at least partially, due to increased AR protein stability, mediated by mutations in the AR or the machinery regulating ubiquitin-mediated proteasomal degradation of the AR. 225,226 AR degradation has been induced by proteolysis-targeting chimeras, PROTACs, which are heterobifunctional small molecules that link a proteinbinding ligand, in this case targeting AR, with a second ligand that binds to E3 ubiquitin ligase for recruitment of the ubiquitin-proteasome system for target degradation. 227 Such molecules have been found to overcome the resistance mechanisms developed during prostate cancer treatment. 228 Drugs that mediate conformational changes in the AR have likewise been used to promote AR degradation via the ubiquitin-proteasome system. 229 To our knowledge, none of these approaches have yet reached clinical use for treating prostate cancer. However, some drug studies are ongoing or under evaluation, including BXCL701, a small molecule inhibitor of dipeptidyl peptidases (DPP), and radioactively labeled or bispecific antibodies targeting KLK2 (ClinicalTrials.gov). 57 We foresee that protease-based therapies would be useful for prostate cancer patients, especially when combined with existing treatment modalities.

| PROTEASES AS PROSTATE CANCER MARKERS
Proteases that are overexpressed in prostate cancer and contribute to tumor progression are potential diagnostic and prognostic biomarkers that can be of potential utility for facilitating personalized medicine. Whereas some tissue-localized proteases show an association with prostate cancer prognosis, it is unclear whether they have any real clinical value as predictive biomarkers. Thus, further research to identify tissue biomarkers is warranted, using standardized and, perhaps, multiparametric approaches. 230 For functional biomarkers, one approach could be to monitor the protease activity, either in diagnostic samples or using in vivo imaging. This can be achieved using activitybased probes or substrates. 75,214 The latter has been implemented with a protease activity-based nanosensor library, which was used to evaluate protease activities in vivo in a prostate cancer mouse model. 38 The readout of the proteolytically liberated reporters was done using urine samples.
Since the concentrations of prostatic proteins in the circulation are dependent both on leakage/transfer from tumors into blood and on their expression in the prostate, the circulating levels may provide information beyond expression and localization of proteases in tissue. Indeed, while the expression of KLK3 (PSA) is decreased in malignant prostate tissue, as compared with normal prostatic epithelium, and loss of apical staining is associated with poor prognosis, 65 KLK3 levels in the circulation are often increased in prostate cancer. During the 1980s, the PSA/KLK3 test was introduced for prostate cancer monitoring and detection. 231,232 About two decades later, a massive randomized study confirmed that population-based prostate cancer screening utilizing circulating KLK3 reduced prostate cancer-specific mortality. [233][234][235] However, this result came at the cost of extensive overdiagnosis, that is, identification of indolent, slowly growing cancers that are clinically insignificant. 236,237 Furthermore, KLK3 levels are often increased in cancer-free men with benign prostatic hyperplasia (BPH) and prostatitis. 54,61,62 Therefore, recommendations against population-based KLK3 screening and opportunistic KLK3 tests have been issued. 231,238,239 The current trend is again favoring KLK3 tests when used as a first step in a risk-adapted approach and together with other risk stratification tools. 240 Selective KLK3-only testing is also useful in some cases, both for detection and monitoring of prostate cancers. 231 For instance, systematic KLK3 screening has been recommended for men with a BRCA2 mutation. 241 Since KLK3 is very sensitive to regulation by androgen, it can be used as a marker for monitoring response to ADT. 242 Decreased KLK3 levels indicate a response to therapy, while the subsequent development of resistance to ADT is associated with increasing KLK3 levels. This rise in KLK3 levels, which precedes clinically observable disease recurrence, is used to define the so-called biochemical recurrence after prostatectomy or radiotherapy. The accuracy of the KLK3 test to identify clinically significant cancers can be increased by analyzing, in addition to total KLK3, different KLK3 forms, such as complexed and free KLK3, and other markers, like KLK2. 59,243 Monitoring the different KLK3 glycoforms may also prove useful to this end, 244 since protein glycosylation is often changed in cancer 245 and, furthermore, prostate cancer risk associated with SNP has been found to introduce an additional glycosylation site in KLK3. 82 However, the technology to detect specific glycoforms of proteins is complicated. To date, there are no widely accepted biomarkers that are sufficiently specific for the identification of clinically significant prostate cancers. Thus, novel prostate cancer markers are urgently needed.
High circulating tumor cell (CTC) number in liquid biopsy from the blood has been found to correlate with high tumor burden, aggressive disease, and shorter time to relapse. 246 Thus, the US FDA has approved a technology for determining CTCs for the prognosis of patients with advanced breast, colorectal, and prostate cancers. 246,247 Perhaps molecular analysis of CTCs could provide additional accuracy when monitoring cancer progression and advising on individualized treatment decisions. While there are not many studies to address such markers in CTCs, and even fewer relating to proteases and prostate cancer, it is interesting that a recent small-scale study reported MMP activity in CTCs of prostate cancer patients. 248 This activity often exceeded that of the leukocytes of the same patients. Another such study reported KLK3 mRNA in CTCs as a potential prognostic marker in metastatic CRPC patients treated with docetaxel. 249 Proteases are also useful for cancer imaging and as theranostic targets. Among such proteases, prostatespecific membrane antigen (PSMA) is well established and a prime example of a theranostic approach based on tumor-specific overexpression of a cell-surface proteases. 250,251 PSMA is a cell-surface carboxypeptidase that is strongly expressed by prostate cancer cells, also at metastatic sites. Various specific ligands have been developed to utilize this specific expression profile for radiologic imaging and targeted drug delivery. 250,251 Recently the type II transmembrane protease fibroblast activation protein (FAP) has emerged strongly as a biomarker and putative target for detection and potential treatment of a range of malignancies. 252 In contrast to other proteins that are enriched on malignant cells, FAP is enriched on the surface of fibroblasts in the tumor stroma, the socalled cancer-associated fibroblasts (CAFs). 253 FAP targeting ligands (generally low molecular weight inhibitors) labeled with radionuclide 68 Ga and delivered intravenously home to FAP that is enriched on the surface of CAFs. Positrons emitted by 68 Ga payloads are detectable by positron emission tomography-computed tomography (PET-CT) imaging to define tumor burden. For patients who carry tumors that are sufficiently avid for the 68 Ga-FAP ligand, the payload can be switched to a high energy emitting radionuclide, such as 177 Lu, converting the FAP-directed agent to a cancer therapy. 252,253 Recently, several highly prevalent cancers, including prostate cancer, have demonstrated high uptake and image contrast in patients by 68 Ga-FAP-targeted PET-CT 254 and FAP-directed ligands carrying therapeutic payloads have shown promising results in cancer patient trials. [255][256][257] FAP is elevated in the stroma of prostate cancer compared with benign prostate tissue, 258,259 suggesting that imaging agents directed against it may have utility for selecting patients who will benefit from FAP-targeted therapy.

| ABOUT PROSTATE CANCER MODELS
For functional and drug studies, selection of a proper biologically relevant model is critical. 260,261 This applies particularly to oncology, where the majority of new drugs have failed in clinical trials, partly because the model systems did not mimic the human conditions and tumor complexity. 133,262 Tumor progression is not dependent solely on the cancer cells because the tumor microenvironment also plays a key role in this process, affecting drug responses, invasive properties, and protease activity in cancers. [263][264][265][266][267] Consequently, the models need to address, in addition to cancer cells, the tumor microenvironment, and metastatic dissemination. Since the pioneering work by Mina Bissell and colleagues, 268,269 the context of the cells has become widely appreciated in functional studies and many approaches have been developed to enable examination of proteases in the context of the tumor microenvironment. Unfortunately, the prostate cancer field suffers from a limited number of human cell lines and xenograft models, most of which fail to recapitulate the human disease adequately, while more faithful models are usually difficult to establish. 270 Furthermore, current cell lines and transgenic models recapitulate, even at their best, only some subtypes of phenotypically and genotypically diverse prostate cancers. Additional limitations derive from differences in proteases between man and common experimental animals, like the mouse, which lacks genes encoding the major human prostatic proteases KLK2 and KLK3. Different cell lines may show different responses, and cells isolated from their natural environment behave very differently from those grown within tumors and in contact with ECM and stromal cells, that is, the tumor microenvironment. For example, the cellular context has a profound effect on the regulation of genes by AR. 8 The majority of prostate cancer cell lines show poor invasive capability and, likewise, the prostate cancer patientderived organoids thus far developed also lack invasive capability. 271 Thus, although the vast majority of prostate cancer associated deaths are caused by metastasis, the cellular models available to study the early stages of metastatic dissemination of prostate cancer cells are limited. Furthermore, potential functions in vivo, based on observations with purified proteins or isolated cell lines in vitro, should be interpreted with great caution.
Taken together, one has to be wary when interpreting the results of functional studies performed in culture model systems. That said, several functions have been solidly established for proteases in prostate cancer using carefully selected orthogonal models that collectively strongly implicate important roles for proteases in prostate cancer progression.

| CONCLUDING REMARKS AND FUTURE PERSPECTIVES
This brief overview of key prostate cancer-relevant proteases points to the potential importance of members of the large enzyme class, especially those that are androgenregulated, in prostate cancer progression, including development of aggressive CRPC. At the same time, there are many additional proteases, we have not highlighted, that may also have significant functional roles in prostate cancer. It is important to recognize that proteases do not work in isolation, but that the context in which they act dictates their functions at the tissue level. Many proteases are activated by other proteases, sometimes in proteolytic cascades with homologous family members, and their activity is also modulated at multiple other levels, including the impact of post-translational modifications and endogenous inhibitors. Furthermore, there are different forms of proteases due to polymorphisms/mutations and alternative splicing of pre-mRNA. 82,272 The protease must meet its substrate to have an effect, and therefore, the subcellular and tissue localization of both the enzyme and its substrates are crucial. Moreover, many proteases have highly similar or overlapping proteolytic activities and may cleave the same substrates, often with identical or similar consequences. Thus, pathological outcomes are dictated by an interconnected complex network of all these factors. Due to differences in availability of substrates, proteases may have different functions in different cancers or even different molecular subtypes of prostate cancer, and at different stages of cancer development. All of these considerations have important implications for studying the functions of proteases, and their roles as cancer markers and drug targets. Therapeutic targeting of individual proteases is further challenging due to specificity issues of inhibitors toward structurally and functionally highly similar proteases with common catalytic mechanisms. Still, we believe that the specificity needed for targeting proteases is achievable in most cases, if not with traditional small molecule catalytic site inhibitors, then with currently evolving alternative approaches.
While studies elucidating the effects of individual proteases are capable of giving important information, we call for degradome-wide functional approaches that can account for the existence of complicated proteolytic networks. Indeed, recent technological developments provide new biochemical tools for tracking proteolytic events in a number of settings. 34,273 Furthermore, we emphasize the importance of developing improved models for functional studies that better recapitulate the human tumor microenvironment. We foresee that more holistic approaches, that is, analyzing the protease web in whole or part, can result in better understanding of the roles of proteases. Expression studies in human cancer cohorts provide valuable support for the findings of functional studies, especially when able to monitor the enzymatic activities of the studied proteases in human tissues or liquid biopsies. To this end, activity-based probes, such as those used for analysis of the KLK network 75 and for elucidation of uPA as a major regulator of CDCP1 activation, 160 will be useful. We look forward to future studies that will clarify the roles of proteases in prostate cancer further and reveal novel functional biomarkers and drug targets, which eventually would improve the treatment outcomes and quality of life of prostate cancer patients. edistämissäätiö (HK), the Academy of Finland (TM) and State funding for university level health research (TM). ESR acknowledges support from US National Institutes of Health grants R01 GM132100, R01GM144393, and R01 CA258274. MDH acknowledges support for protease-