Loss of function of the PTEN tumour suppressor, resulting in dysregulated activation of the phosphoinositide 3-kinase (PI3K) signalling network, is recognized as one of the most common driving events in prostate cancer development. The observed mechanisms of PTEN loss are diverse, but both homozygous and heterozygous genomic deletions including PTEN are frequent, and often accompanied by loss of detectable protein as assessed by immunohistochemistry (IHC). The occurrence of PTEN loss is highest in aggressive metastatic disease and this has driven the development of PTEN as a prognostic biomarker, either alone or in combination with other factors, to distinguish indolent tumours from those likely to progress. Here, we discuss these factors and the consequences of PTEN loss, in the context of its role as a lipid phosphatase, as well as current efforts to use available inhibitors of specific components of the PI3K/PTEN/TOR signalling network in prostate cancer treatment.
THE PI3K/PTEN SIGNALLING NETWORK
Elevated and/or uncontrolled activation of the signalling networks activated by growth factors, in particular signalling through the phosphoinositide 3-kinase (PI3K) group of lipid kinases, is a characteristic of most cancers including prostate cancer [1,2]. The signalling network (Figure 1) is named after a group of lipid kinases, the class I PI3Ks that are tightly regulated, with low basal activity and that are activated by diverse cell surface receptors. These activating receptors include those for many growth factors, cytokines and chemokines, as well as extracellular matrix components acting via integrins . Class I PI3Ks are heterodimeric enzymes, with four human genes encoding catalytic isoforms and five encoding regulatory subunits. Although examples of both catalytic (PIK3CA encoding the α catalytic subunit of PI3K) and regulatory (PIK3R1 and PIK3R2 encoded) subunits have been found to be frequently mutated in several forms of cancer [4,5], rates of mutation and copy number changes in genes encoding PI3K itself are generally lower in prostate cancer than in many other common carcinomas [1,6]. As part of their activation mechanism, the PI3K catalytic subunits, which are all of approximately 110 kDa, directly bind to small GTPases, which appear to be members of the RAS sub-family in the case of the p110α, p110γ and p110δ PI3K isoforms, but members of the RAC GTPase sub-family in the case of p110β [3,7]. When activated at the plasma membrane, the PI3Ks phosphorylate the relatively abundant phosphoinositide lipid, phosphatidylinositol 4,5-bisphosphate [PI(4,5)P2] on the available 3-position hydroxy group, to form small amounts of the signalling lipid phosphatidylinositol 3,4,5-trisphosphate [(PIP3)]. The role of PTEN in the pathway is to catalyse the opposing reaction, metabolizing PIP3 into PI(4,5)P2 as shown in Figure 1(A).
PI3K/PTEN pathway signalling
Does signalling through the β isoform of PI3K selectively drive prostate tumorigenesis?
In 2008, genetic analysis of the PI3K isoform dependence of prostate tumour formation in mice lacking PTEN function showed that deletion of Pik3ca had little effect on tumour formation, yet deletion of Pik3cb dramatically reduced tumour burden . Furthermore, similar conclusions were drawn from work with prostate  and breast cancer cell lines  although a mechanistic basis for this selective link has yet to be identified. This over-reliance on signalling through PI3K β in cases of PTEN loss may partly explain the low occurrence of mutations and copy number changes in PIK3CA in prostate cancer and, although PIK3CB, the gene encoding the β catalytic subunit of PI3K, is very rarely found to be activated by mutation, rates of PIK3CB gene amplification are significant in metastatic prostate cancer (approximately 5%) and particularly high in neuroendocrine prostate cancer (21%). This has influenced some efforts to therapeutically target PI3K/PTEN/TOR signalling in the area of prostate cancer, motivated by the potential to selectively target the β isoform of PI3K and therefore perhaps has lower drug toxicity with similar efficacy compared with inhibitors of all four PI3K isoforms. On the other hand, the links between loss of PTEN, PIP3 generated by the β isoform of PI3K and tumorigenesis seem inconsistent [11,12]. Additionally, more recently data have been presented showing that when one isoform of PI3K is inhibited, feedback loops can lead to increased activation of other expressed isoforms [13,14]. Regardless, the real value of these ideas and the drug programmes they have motivated should soon be clear, as drugs specifically targeting PI3K β have entered clinical prostate cancer trials (GSK2636771 and AZD8186 that inhibit both PI3K β and δ).
Signalling downstream of PI3K/PTEN
The significance in cellular transformation of signalling through PI3K is also highlighted by the very frequent identification in tumours of genetic changes in upstream signalling components that lead to PI3K activation. The clearest examples of such upstream PI3K activating oncogenes are the RAS GTPases and several receptor tyrosine kinases (RTKs), including the epidermal growth factor receptor family (EGFR, HER2/ERBB2/NEU etc.) and the PDGF and KIT receptors. However, the list of PI3K network components that appear to be oncogenic drivers extends further to include less central components such as the E3 ubiquitin ligase CBL, which promotes RTK degradation and INPP4B that metabolizes the secondary PI3K product lipid phosphatidylinositol 3,4-bisphosphate [PI(3,4)P2] . As with other cancers, integrated genomic studies of primary and metastatic prostate cancers have consistently identified changes leading to the aberrant activation of PI3K signalling in many of these tumours [1,15–17].
The PI3K product lipid, PIP3, promotes cell growth, proliferation, cellular polarization and other changes through a large and diverse group of downstream effector proteins endowed with a lipid binding domain capable of selectively binding to it. It is unclear how broad this PIP3-binding proteome is in humans, with estimates ranging from a few tens to over 100 [18–20] but the best understood examples by far are the AKT protein kinases, important oncogenes in their own right. Strong genetic experiments in mice, flies and worms have demonstrated the importance of the AKT kinases in mediating the effects of PI3K activation on cell growth and proliferation [21–25]. Additionally, the identification of numerous confirmed substrates for AKT has provided insight into the molecular mechanisms by which these effects are mediated [25,26]. Downstream of PI3K and AKT, a major regulator of cell growth that has been repeatedly implicated in tumorigenesis is mechanistic target of rapamycin (MTOR). MTOR is a protein kinase subject to multiple complex regulatory inputs in addition to activation downstream of AKT (e.g. nutrients and energy) and plays a key role in regulating cell growth and anabolic metabolism, as well as being involved in AKT activation .
On the other hand, the significance of the other regulatory mechanisms downstream of PI3K independent of AKT and MTOR remains largely unclear, particularly in cancer. Notably, gradients of PIP3 have been identified in the plasma membrane of several cell types. Particular attention has been paid to the dynamic enrichment of PIP3 seen in the leading edge of axonal growth cones, of mammalian neutrophils and fibroblasts and the slime mould Dictyostelium discoideum during directed cell migration. More relevant to the prostate, a stable enrichment of PIP3 has been seen in the basolateral membranes of some polarized epithelial cells [28–32].
The significance of the disturbance of such localized and/or AKT-independent PI3K signalling in prostate, or other, cancers is a key remaining question. Many lines of evidence point to such important AKT-independent oncogenic mechanisms, both through the identification of alternative mechanisms [33–35] and also the characterization of catalytically active mutants of PTEN found in human tumours that retain the ability to suppress AKT activity yet fail in other cell-based assays [36,37]. There are also numerous reports of discordance between the consequences of AKT activation and either activation of PI3K or loss of PTEN (e.g. [34,36,38,39]).
Given the significance in oncogenesis of components that activate signalling through the PI3K network, attention has also fallen on the balancing inhibitory components that act to suppress PI3K signalling. The clearest examples of these inhibitory components are the lipid phosphatases that metabolize the growth promoting signals synthesized by the PI3Ks. SHIP and SHIP2 are phosphatidylinositol 5-phosphatases (PI5Ps) that convert PIP3 into the secondary PI3K lipid product PI(3,4)P2. However, the evidence that the SHIPs represent bona fide tumour suppressors is limited, perhaps due to the existence of redundancy between these and other related enzymes, perhaps because PI(3,4)P2 also activates downstream PI3K effectors including AKT [19,40]. Related to this point, the phosphatidylinositol 4-phosphate (PI4P) encoded by INPP4B, which converts PI(3,4)P2 to phosphatidylinositol 3-phosphate (PI3P), has been shown to act as a tumour suppressor in several tissues by removing PI(3,4)P2 and suppressing AKT activity, yet also plays conflicting oncogenic roles in some tumours including melanomas, colon cancers and leukaemias possibly by promoting PI3P-dependent activation of SGK3 [41,42]. These results highlight the contrast with the best studied phosphoinositide phosphatase however, PTEN, which acts as a PIP3 3-phosphatase and has consistently been shown to act as a significant tumour suppressor in many organ systems, particularly the prostate.
Notably, multiple genetic changes leading to the activation of the broad PI3K pathway within individual prostate tumours are commonly observed. For example, there seems to be a significant association between PTEN loss and PIK3CA mutation or amplification in prostate adenocarcinoma tumour sets (www.cbioportal.org), with both changes being commonly identified in individual tumours. This is in contrast with the frequently observed mutual exclusivity of different mutational events within other specific functional networks that have been observed in previous studies [43,44].
PTEN: A HAPLOINSUFFICIENT TUMOUR SUPPRESSOR FUNCTIONALLY OPPOSING PI3K
In 1995, Gray et al.  showed that loss of region q23–24 on chromosome 10 was a frequent occurrence in prostate cancer (62% of tumours studied) contributing to a body of data showing loss of this region in a number of other tumours [46–48]. In 1997, two groups isolated Phosphatase and Tensin homologue deleted on chromosome 10 (PTEN–also referred to as MMAC1 and TEP1) as the tumour suppressor gene on Chr10, identifying coding mutations in this gene in samples from prostate cancer patients as well as glioblastoma, breast and kidney cancer derived samples [49,50]. Steck et al.  screened chromosome fragments for their ability to suppress tumorigenic phenotypes and were subsequently able to clone PTEN, whereas Li et al.  focused on identifying PTEN mutations in cancer cell lines and tumours. This tumour suppressor status was further confirmed by the identification of PTEN mutations in patients suffering from the familial cancer-syndromes Cowden disease and Bannayan–Riley–Ruvalcaba syndrome [51,52], although notably, prostate cancers have rarely been described in sufferers of these syndromes , which are now often considered together under the umbrella term PTEN hamartoma tumour syndrome (PHTS).
Both groups that identified PTEN commented on its homology to the family of protein tyrosine phosphatases [49,50], and further publications showed that PTEN had weak dual specificity protein tyrosine and serine/threonine phosphatase activity, with a preference for highly acidic substrates [54,55]. However, it was demonstrated soon after that PTEN dephosphorylates the 3-position on the inositol ring of the PI3K product lipid, PIP3  to regenerate PI(4,5)P2. This reaction is illustrated in Figure 1(A). Recombinant PTEN expressed in Escherichia coli also dephosphorylates the 3-position on the inositol ring of the related substrates, inositol 1,3,4,5-tetrakisphosphate [Ins(1,3,4,5)P4] (which represents the soluble headgroup of PIP3) and PI3P and PI(3,4)P2, but at reduced rate [56,57]. Significantly, strong evidence that it is the lipid phosphatase activity that is critical for the tumour suppressor functions of PTEN was provided by the demonstration that a missense mutation, PTEN G129E, identified in two families suffering from Cowden disease, retains protein phosphatase activity but has greatly reduced lipid phosphatase activity towards PIP3 and was unable to suppress AKT phosphorylation or cell growth in culture [57,58]. Intense research since these early studies of PTEN has revealed many details regarding how it fulfils its roles in regulating PI3K signalling and blocking tumorigenesis in many tissues (see reviews e.g. [59,60]). However, major questions remain about the downstream mechanisms by which elevated and mislocalized PIP3 increases the chances of tumour formation (as described above) and also whether PTEN has additional unrelated functions, such as dephosphorylating protein substrates or non-catalytic functions. On this point, the consistent finding that heterozygous mice expressing the stable PTEN G129E mutant develop a worse spectrum of tumours than null allele heterozygous mice, implies that if PTEN fulfils such PIP3-independent functions, these additional functions have little effect on their own to suppress tumours in most tissues [61,62].
In terms of cell biology, PTEN is a relatively small 50 kDa enzyme, which associates transiently with the plasma membrane to metabolize its lipid substrate [29,59,63]. Additional functions and sites of action for PTEN, such as in the nucleus and on endomembranes, have been proposed but these remain controversial and detailed mechanisms remain to be determined [59,60,64,65]. PTEN binds to many other proteins, several through its extreme C-terminal PDZ binding sequence and it is this binding mode that seems responsible for the localization of PTEN to adherens junctions in epithelial cells and to neuronal synapses [29,66–69]. The apical junctional localization of PTEN in several epithelial tissues has been described, including in the prostate although in most cases this localization appears as an enrichment, rather than a bold unique localization (described later) [67,68,70,71].
It seems significant and in contrast with some other tumour suppressors that the potential impact of even small changes in PTEN expression appears large, given the evidence from clinical data and murine models that PTEN is a haploinsufficient tumour suppressor, with function being strongly dose dependent. Mice with only a single functional Pten copy throughout the body develop tumours in a diverse range of organs [72,73]. Furthermore, elegant studies using a hypomorphic allele, which generates approximately a half dose of normal PTEN protein relative to a wild-type allele, have found that heterozygous mice carrying either a single copy of wild-type PTEN or one hypomorphic allele show prostate cancer initiation and progression that correlates closely with reducing PTEN dose  and even animals carrying one wild-type and one hypomorphic copy show increased rates of tumour formation in some organs, particularly the breast . Conversely, overexpression of PTEN in mice leads to a tumour resistant state [76,77].
PROSTATE CANCER AND PTEN
Prostate cancer is the second most common cancer in the U.K. (Cancer Research UK, 2013), with approximately 50000 new cases diagnosed in 2013 and approximately 10000 deaths in 2012, making it the second most common cause of death from cancer in men. Age is the most significant risk factor for prostate cancer, with >50% of the cases diagnosed between 2011 and 2013 in men >70 years of age, although family history is also a risk factor . Treatment of prostate cancer is challenging, because histologically similar tumours give rise to diverse clinical outcomes. At this time, there are no reliable prognostic markers to discriminate indolent tumours from those likely to cause aggressive metastatic disease, and this complicates treatment decisions. Genetically, prostate tumours show inter- and intra-tumoural heterogeneity, suggesting multiple genetic changes and environmental factors are involved. Along with surgery and radiotherapy, androgen deprivation therapy is a very common treatment for metastatic prostate cancer, but almost all such cancers eventually become resistant to this treatment and their disease will progress, at this stage being termed as castrate-resistant prostate cancer (CRPC).
Approximately half of primary prostate cancers carry driving gene fusions between androgen regulated promoters and members of the E26 transformation specific (ETS) family of transcription factors, most commonly the TMPRSS2–ERG fusion. However, primary tumours are otherwise genetically very diverse . Metastatic tumours show more recurrent changes, with four dominant events each being present in approximately half of these tumours: (1) amplifications of the androgen receptor (AR) gene; (2) ETS family gene fusions, again most commonly the TMPRSS2–ERG fusion; (3) disruption of the TP53 gene and (4) as will be discussed here, disruption of PTEN . The significance to pathology of these changes in PTEN has been confirmed in animal studies, with full genetic deletion of PTEN specifically from the developed prostate gland causing rapid high grade prostate intraepithelial neoplasia (PIN) and in some studies an invasive prostate carcinoma within 3–6 months of age [74,80].
PTEN MUTATIONS IN PROSTATE CANCER
Recent detailed molecular analyses using exome and genome sequencing, array comparative genomic hybridization (aCGH), in situ hybridization and immunohistochemistry (IHC) have confirmed early observations that PTEN function is frequently lost in prostate cancer. These studies found copy number alterations, structural rearrangements and/or point mutations to the PTEN gene in between 16% and 41% of tumour samples [1,16,17,79,81–83]. The difference in mutation frequency is likely to reflect, at least in part, the stage of cancers analysed–studies looking at CRPC showed higher levels of mutation, with homozygous loss of PTEN being the most common alteration [16,17]. PTEN appears more frequently subject to deletion than point mutation in prostate cancer, and in all studies point mutations (nonsense, missense or insertion/deletion changes) were in the minority. For example, detailed genetic analyses of >300 primary prostate tumours, found 2.4% had missense or nonsense mutations, compared with 15% with copy number alterations . Similarly, consultation of the TCGA Research Network showed an overall alteration rate of 20.6% for PTEN from nine studies, with approximately three quarters of these changes representing PTEN deletions. These data are summarized in Figures 2(A) and 2(B). In other studies, point mutations were found only combined with heterozygous gene loss [16,17]. One reason why copy number alteration may be more common is that loss of PTEN can occur through a series of ‘shuffling’ events dubbed chromoplexy, where multiple distant loci undergo concurrent rearrangement [15,81,84]. This is in keeping with the emerging theme that relative to other cancer types, prostate cancer is driven more by large-scale genomic structural rearrangements than by point mutations [84,85]. Baca et al.  proposed that PTEN mutations may be a ‘gating’ point for the development of aggressive prostate cancer, based on the subclonal nature of PTEN loss in their tumour set. However, data such as the detection of shared PTEN alterations in multifocal prostate cancer  support a role for PTEN loss in prostate cancer initiation, at least in some cases (see below).
Classes and locations of PTEN alterations in prostate cancer
DETECTING PTEN CHANGES IN PROSTATE CANCER AND ASSOCIATION WITH DISEASE SEVERITY
In a clinical setting, fluorescence in situ hybridization (FISH) and IHC are commonly used to assess patient samples for changes in the abundance of PTEN mRNA and protein respectively. Substantial effort has been expended to optimize IHC methods in particular in light of evidence that PTEN expression is regulated post-transcriptionally .
in situ hybridization
Estimates of PTEN gene loss in prostate cancer by FISH are quite variable, and likely reflect differences in sample cohorts, intra-tumoural heterogeneity, methodology and analyses. For example, two studies using the same FISH probes but different sample sets found loss of PTEN in 68% and 44% of primary prostate carcinomas [88,89]. More recent four colour FISH protocols have been developed–now incorporating probes for regions flanking the PTEN locus as well as for the centromere of ch10 and for PTEN [90–92]. This provides information about the size of Ch10 deletions and allows the identification of false positive results due to sectioning artefacts. In an initial study of 330 samples, almost all from patients who had undergone radical prostatectomies, 40% showed PTEN loss by FISH . This number was in good agreement with Choucair et al.  who showed that 41% of primary tumours had heterozygous loss of PTEN and that this was associated with a reduced time to recurrence and decreased AR signalling. Studies of more advanced disease samples have found higher rates of PTEN loss, for example 77% of CRPC samples studied by Sircar et al. , supporting the hypothesis that PTEN loss is associated with these more aggressive tumours. In support of this, in a later study, from 612 patient tumour samples from radical prostatectomies, approximately 20% had lost one or both copies of PTEN, this was very strongly correlated with increased Gleason score and poorer outcome . Consistent with this picture, a recent meta-analysis and review using data from seven previous published studies, most discussed here, found a strong correlation of PTEN genomic deletion with both higher Gleason score and increased capsular penetration .
Although some initial analyses of PTEN by IHC suffered from problems with poor methods and antibodies limiting sensitivity and PTEN specificity, optimization of method and materials have supported the generation of strong data, much strengthened by controls demonstrating a good correlation of PTEN loss by IHC with reduced PTEN signals at the DNA and mRNA levels [95–98]. By IHC analyses, both in published studies and online resources such as the Human Protein Atlas (www.proteinatlas.org), staining for the PTEN protein shows a cytoplasmic and nuclear pattern in the luminal and basal epithelial cells of the prostate, as well as being expressed in other cell types . Reported rates of loss of PTEN staining in prostate cancers have been observed in a range of 15–30% in unselected primary tumours, either from biopsy samples or from radical prostatectomies [82,99–102]. One common feature to come from this work has been frequently observed intra-tumoural heterogeneity, with regions positive and negative for PTEN staining [82,103,104] indicating that PTEN loss is often not an early founding event in prostate cancer formation. The lack of evidence for disturbed PTEN function in high grade PIN also argues that PTEN loss may be a later event in prostate cancer development .
A very important feature that has been observed is a correlation of PTEN loss as judged by IHC with higher grade tumours as judged by Gleason score [82,99,103,104] and with stage, with higher rates of PTEN loss being consistently observed in metastatic and in castration resistant cancers [1,82,101,103,106]. Accordingly, loss of PTEN assessed by IHC has also been independently associated with a variety of poor outcomes, including PSA-measured biochemical recurrence , poor clinical outcome  and tumours that scored as Gleason grade 6 at tumour biopsy but were upgraded to Gleason 7 or higher when the prostate gland was removed and analysed . The confirmation of the association of PTEN with lethal progression in a recently published prospective study implies that the association of PTEN loss with aggressive disease is not simple because these tumours have a higher mutational burden and more disrupted genome, but further supports the hypothesis that loss of PTEN is not usually an initiating event in prostate cancer, but rather is most frequently a later event that worsens patient outcome .
Is PTEN protein expression often reduced in tumours retaining normal
Where reported together, assessment of the PTEN gene by FISH and protein by IHC from tumour biopsies has shown good correlation in many but not all cases and this finding has been used to provide confidence in the presented data. For example, Han et al.  reported that all but 1 of the 21 tumours that had PTEN deletion by FISH showed reduced PTEN staining by IHC. More recently, similar results were reported by Lotan et al.  and Murphy et al. . However, a strong correlation between gene and protein analysis should not always be expected and a clear possibility, with some support, is that PTEN protein levels are suppressed in prostate tumours retaining wild-type copies of the PTEN gene, due to changes at one or more steps in the gene expression pipeline. Accordingly, it is common for samples to show a positive FISH result and a negative/reduced PTEN IHC, and much less common to have IHC positivity reported in cells carrying PTEN gene loss as assessed by FISH [99,109,111,112].
PTEN function is known to be regulated at the transcriptional, post-transcriptional and post-translational levels through diverse mechanisms both physiologically and in disease . In many cases, the significance of these regulatory mechanisms to the loss of PTEN function during carcinogenesis is still emerging. Examples of reduced PTEN transcription in prostate cancer by changes in transcription factor networks have been described [113,114] and in other cancers by promoter methylation [115,116], although there is little evidence for this in prostate cancer. However, the most recent attention in this area has fallen on changes in the miRNA networks that are able to suppress PTEN expression. Good evidence for a role of miRNA in regulating PTEN in prostate cancer exists [117–120] and a number of miRNAs, representing candidate oncogenic miRNAs, or oncomiRs, were shown to bind the 3′-UTR of PTEN [117,119,121]. An added layer of complexity in the post-transcriptional regulation of PTEN has come with its use as an example in work establishing a novel mechanism controlling gene expression. Several mRNAs, including the PTEN pseudogene PTENP1, contain similar miRNA binding sites to PTEN and these have been referred to as competing endogenous RNA . Depletion of PTENP1 or other competing transcripts is proposed to free-up more miRNAs, allowing it to bind to the PTEN mRNA and experimentally led to a reduction in PTEN protein levels [120,121]. Similarly and in support of this model, PTENP1 has been found to undergo genomic copy number loss in a small percentage of sporadic colon cancers . A more recent study also identified a number of long non-coding RNAs from clinical data as candidate tumour suppressors in prostate cancer, which were shown to be able to act as such miRNA-competing ‘sponges’ and the knockdown of which reduced PTEN expression and increased proliferation in the DU145 prostate cell line . These studies add to the great complexity found in mRNA gene expression data and leave the challenge of dissecting the relative significance of individual changes observed in tumour miRNA signatures and telling oncogenic drivers from passengers.
It seems important that in these and other studies, functional loss of PTEN would not be detected if only FISH and/or sequencing are performed, without IHC. Therefore, strong well-validated IHC seems likely to provide the best assessment of PTEN function. PTEN IHC appears reliable and is a promising way to identify patients incorrectly presumed as low risk who may be prone to disease progression, a conclusion recently strengthened by a large prospective study . The fact that PTEN is regulated at many levels beyond transcription may also mean that studies that have used DNA analyses to look at PTEN loss in prostate tumours may have underestimated the frequency at which PTEN is lost in prostate cancer.
In Figure 3 and Table 1, we present a summary showing mean percentage frequencies of reported PTEN loss assessed by different technologies in a large number of independent studies that contain data addressing the frequencies of PTEN loss and reduction in prostate cancer cohorts. This diagram is clearly a simplification, as side-by-side it presents data from many studies using somewhat different methodologies and patient groups, ignoring many specific features of the studies included. However, the overall pattern of data is compelling, illustrating the higher rates of loss observed in metastatic/CRPC patients. Also, the occurrence of undetectable PTEN expression by IHC appears consistently higher than the apparent rates of homozygous loss by FISH and other genetic assessments, giving further support to the proposal (see above) that PTEN protein levels in prostate cancer are often reduced in tumour cells retaining healthy levels of wild-type PTEN mRNA . This is in notable contrast with some other cancers where loss is usually at the level of mRNA, e.g. breast cancer .
A comparison of PTEN alteration frequencies in primary tumours and CRPC/metastatic prostate cancer
|.||Primary tumours .||Metastatic/CRPC .||.|
|Method .||% .||n .||het/ (%) .||hom/ (%) .||Reference .||% .||N .||het/ (%) .||hom/ (%) .||Reference .||P-value .|
|.||Primary tumours .||Metastatic/CRPC .||.|
|Method .||% .||n .||het/ (%) .||hom/ (%) .||Reference .||% .||N .||het/ (%) .||hom/ (%) .||Reference .||P-value .|
As an additional note in consideration of methods to detect genetic changes in PTEN, progress is being made in detecting tumour specific gene copy number changes and mutations in cell-free DNA in blood and urine, which is a challenge given the large background of unmutated DNA from other tissues [124,125]. The possibility for prostate cancer patients of tumour characterization and prognosis from non-invasive liquid biopsy is appealing, yet the detection sensitivities for free variant DNA must be improved before these techniques can be applied to most patients.
In addition to the studies discussed above that have selected PTEN to test its utility as a prognostic biomarker, other studies have undertaken to identify markers associated with disease severity taking a global unbiased approach and in some cases have identified PTEN loss as a component of aggressive disease signatures (e.g. [1,126,127]). Of these global studies, the majority look for signatures of aggressive disease–one notable exception was the work of Irshad et al. , which identified an ‘indolence signature’.
TARGETED THERAPIES TO TREAT PROSTATE CANCERS LACKING PTEN
One notable feature of the PI3K/PTEN signalling pathway is its reliance on protein and lipid kinases, which are enzyme groups with a good track record of druggability. Detailed strategies and progress of drug programmes targeting the PI3K-pathway and components downstream including TOR have been reviewed elsewhere . To summarize briefly, a number of pan PI3K and allosteric or ATP competitive AKT and mTOR inhibitors are currently in phase I or II clinical trials for prostate cancer . Additionally dual PI3K–mTOR inhibitors BEZ235 and GDC-0980 are in phase I/II trials. To date, clinical trials of inhibitors of PI3K and TOR in prostate cancer patients have achieved very poor response rates as monotherapy (e.g. ). However, a key point is the demonstration that many of these drugs successfully and relatively selectively inhibit their target in vivo. This provides hope that successful applications for them in the treatment of prostate cancer may be identified. It is also noteworthy that in most of these trials, a small minority of patients did respond to treatment, yet no rationale and predictive biomarkers have yet emerged that would allow them to be selected in advance. Efforts are now focusing on stratification of patients prior to treatment following on from high response rates observed in a phase II trial of patients treated with the PARP inhibitor olaparib, who had existing mutations in DNA repair genes , and the use of combination therapy. For example, since a negative feedback loop exists between the PI3K–AKT pathway and AR signalling, it may be necessary to inhibit both pathways in patients. In support of this, AR and PI3K inhibition in a preclinical model using patient xenografts showed tumour regression , and clinical trials are underway (reviewed in ).
REGULATORY SYSTEMS INTERACTING WITH PI3K/PTEN
At the molecular level, several proteins that directly and robustly interact with PTEN are found to be altered themselves in prostate cancer. Amplification of the recognized PTEN inhibitor and RAC1 activator, PREX2, has been observed as well as (uncommon) deletion of the PTEN scaffold PAR3 ([135,136] and www.cbioportal.org). Additionally, potential cross-talk with other regulatory systems with established roles in oncogenesis is provided in reports that PTEN can directly interact with other major tumour suppressors and oncogenes, such as p53, β-catenin and SRC [137–139]. On the other hand, the significance of these regulatory influences on, or by, PTEN remains poorly understood.
At a higher level, signalling through the PI3K network influences major cellular processes such as cell-cycle entry and protein synthesis providing outputs that are integrated with those from other semi-independent functional networks. As discussed, examples of such pathways are also frequently dysregulated in prostate cancer, e.g. programmes of transcriptional regulation by the AR, TP53 and by ETS family transcription factors. The consequences of such pathway integration and cross-talk generally appear strongly context dependent and hard to study and predict. However, with strong relevance to prostate cancer development, multiple mechanisms of bidirectional negative feedback have been observed between AR and PI3K/PTEN signalling, the disruption of which is required for the strong activation of both PI3K and AR pathways observed in some aggressive tumours [14,140–142].
The role of PTEN as a suppressor of prostate cancer and the association of its loss with aggressive disease are well established and increasing numbers of drugs are available to target different points in the PTEN/PI3K/TOR signalling network. However, these agents have so far had limited success as monotherapies and progress has been slow to develop combination therapies and predictive biomarkers to match drugs with patients who will respond. The first significant impact of our understanding of signalling via PTEN/PI3K in terms of clinical impact may be the use of PTEN functional assessment as a prognostic biomarker.
This work was supported by the Prostate Cancer U.K. [grant number PG14-006]; the Chief Scientist Office [grant number ETM-433]; and the Brain Tumour Charity [grant number GN-000344].
castrate-resistant prostate cancer
E26 transformation specific
fluorescence in situ hybridization
mechanistic target of rapamycin
prostate intraepithelial neoplasia
4)P2, phosphatidylinositol 3,4-bisphosphate
5)P2, phosphatidylinositol 4,5-bisphosphate
receptor tyrosine kinase