MYC Cooperates with AKT in Prostate Tumorigenesis and Alters Sensitivity to mTOR Inhibitors

MYC and phosphoinositide 3-kinase (PI3K)-pathway deregulation are common in human prostate cancer. Through examination of 194 human prostate tumors, we observed statistically significant co-occurrence of MYC amplification and PI3K-pathway alteration, raising the possibility that these two lesions cooperate in prostate cancer progression. To investigate this, we generated bigenic mice in which both activated human AKT1 and human MYC are expressed in the prostate (MPAKT/Hi-MYC model). In contrast to mice expressing AKT1 alone (MPAKT model) or MYC alone (Hi-MYC model), the bigenic phenotype demonstrates accelerated progression of mouse prostate intraepithelial neoplasia (mPIN) to microinvasive disease with disruption of basement membrane, significant stromal remodeling and infiltration of macrophages, B- and T-lymphocytes, similar to inflammation observed in human prostate tumors. In contrast to the reversibility of mPIN lesions in young MPAKT mice after treatment with mTOR inhibitors, Hi-MYC and bigenic MPAKT/Hi-MYC mice were resistant. Additionally, older MPAKT mice showed reduced sensitivity to mTOR inhibition, suggesting that additional genetic events may dampen mTOR dependence. Since increased MYC expression is an early feature of many human prostate cancers, these data have implications for treatment of human prostate cancers with PI3K-pathway alterations using mTOR inhibitors.


Introduction
Prostate cancer is the second most common cause of cancer-related deaths in American men, who carry a 16% lifetime risk of developing invasive prostate cancer. Effective treatment of early-stage localized disease involves active surveillance, surgery (radical prostatectomy) or radiation therapy; however, recurrent and/or metastatic disease is incurable and androgen deprivation therapy is the primary treatment modality [1,2]. The predominant genetic and cellular changes in human prostate cancer include presence of the TMPRSS2-ERG gene fusion [3]; loss of the phosphatase and tensin homolog (PTEN) tumor suppressor gene leading to accumulation of its substrate phosphatidylinositol 3,4,5-triphosphate (PIP 3 ) and constitutive PI3K-pathway up-regulation [4]; amplification, over-expression or mutation of the androgen receptor (AR) [2]; and amplification of the MYC oncogene [5,6]. Activating mutations in some signaling pathways can lead to tumor cell 'addiction' to that same pathway, providing an Achilles heel for clinical intervention.
The PI3K-pathway activates multiple targets including AKT and its downstream effector mammalian target of rapamycin (mTOR) [7,8], thus promoting cell growth and survival by suppression of apoptosis and modulation of glucose uptake and cellular metabolism [9]. mTOR function is governed by its participation in the mTORC1 (mTOR complex 1) and mTORC2 (mTOR complex 2) multiprotein complexes [7,8]. AKT is one of several mTORC2 kinase substrates, whereas activated mTORC1 phosphorylates two key effectors: i) eukaryotic initiation factor 4E-binding protein 1 (4EBP1) that regulates cap-dependent protein translation; and ii) ribosomal protein S6 kinase 1 (S6K1) that in turn phosphorylates 40S ribosomal protein S6, leading to protein synthesis.
PI3K-pathway inhibitors are undergoing clinical evaluation in multiple tumor types [10] including prostate cancer [11]. Despite promising preclinical efficacy in PI3K-pathway-dependent prostate cancer models [12,13,14], there have been only sporadic clinical responses in single-agent trials with rapamycin analogs (rapalogs, eg CCI-779, RAD001) targeting the PI3K-pathway via allosteric inhibition of mTORC1 [15,16]. One reason for the limited clinical efficacy of mTOR inhibitors could be a compensatory upregulation of PI3K signaling to mitigate the inhibitory block placed on the rapamycin-sensitive mTORC1 complex, either via release of the negative feedback on AKT that is potentiated by activated S6K in the absence of rapamycin, or via mTORC2 signaling, which is largely insensitive to rapamycin [17,18,19]. Additionally, mTORC1 inhibition can lead to feedback activation of mitogen-activated protein kinase (MAPK) signaling via an S6K-PI3K-Ras-dependent pathway [20]. Furthermore, rapamycin does not fully inhibit mTORC1, as demonstrated by comparison with ATP-competitive mTOR kinase inhibitors [16]. Another explanation for rapalog failure in the clinic is that tumorigenesis depends on accumulation of more than one genetic aberration in pathways regulating cell proliferation and survival [21]. Elucidation of these cooperating lesions is essential to development of effective therapeutic strategies.
The MYC transcription factor directly regulates expression of the translational machinery for protein synthesis, as well as genes controlling cell cycle progression, metabolism, mitochondrial number and function and stem cell self renewal [22]. A potential cooperative role for PI3K-pathway activation and the MYC oncogene has not yet been documented in human prostate cancer, although pathway-interaction has been suggested by several in vitro and in vivo models [23,24,25].
We identified an association between PI3K-pathway alteration and MYC amplification in a cohort of primary and metastatic human prostate cancer samples. To explore a cooperative role for the PI3K-pathway with the MYC oncogene in human prostate cancer, we used existing murine models of human prostate cancer harboring prostate-specific homozygous deletion of PTEN (PTEN pc2/2 model) [26,27], or over-expression of either human MYC (Hi-MYC model) [28] or the downstream PI3K-pathway active allele of AKT1 (MPAKT model) [29] and studied the combinatorial effect of these pathways on tumorigenesis. Initial generation of a PTEN pc2/2 /Hi-MYC bigenic cross was used to validate results of a related study [24] that demonstrated an interaction between PTEN and MYC signaling using prostatespecific deletion of PTEN with concurrent Cre-induced focal MYC expression to induce high-grade mPIN (HG-mPIN) lesions and invasive adenocarcinoma. To address whether AKT downstream of PTEN might be the key mediator, we further explored the cooperation between these pathways using a bigenic mouse cross, MPAKT/Hi-MYC. Treatment with an mTOR inhibitor allowed direct assessment of the impact of MYC expression on the welldocumented sensitivity of prostate lesions in the activated AKT model [14]. Our results suggest the disappointing clinical activity of single-agent rapamycin analogs in PTEN-deficient human cancers, as compared to single-lesion transgenic mouse models, may arise from secondary genetic alterations in human tumors.

Materials and Methods
Detailed methods are provided as supplemental information (Text S1).

Ethics statement
Human prostate tissues analyzed in this study were from patients treated at Memorial Sloan-Kettering Cancer Center (MSKCC), all of whom provided written informed consent. The study was approved by the MSKCC Institutional Review Board and the MSKCC Human Tissue Utilization Committee. Animal studies were carried out under protocol 06-07-012 approved by the MSKCC Institutional Animal Care and Use Committee. Institutional guidelines for the proper, humane use of animals in research were followed.

Comparative Genomic Hybridization Analysis of human tumors
Copy number data from 194 high-quality primary and metastatic tumors were generated using the Agilent 244K aCGH array, and tumors assessed for genomic gain or amplification in MYC, PIK3CA, AKT1, AKT2 and AKT3, and for PTEN-loss (Table  S1). The complete aCGH dataset is reported separately [30] and available online at http://cbio.mskcc.org/prostate-portal/.

MYC amplification co-occurs with PI3K-pathway activation in human prostate tumors
Activation of the PI3K signaling pathway, often via PTEN inactivation, and amplification of MYC are common genetic alterations in prostate cancer that correlate with high histological grade and poor prognosis [5,32]. To evaluate whether PI3Kpathway activation and MYC oncogene amplification co-occur in human prostate cancer, we examined oligonucleotide array CGH data from 194 prostate tumors, including 37 metastases. PI3Kpathway activation rarely occurred through point mutation of PTEN or PIK3CA in this dataset: exon-resequencing of 80 tumors revealed only 2 tumors with PIK3CA mutation and none with PTEN mutation [30].
PI3K-pathway activation, representing combinatorial alterations in PTEN, PIK3CA, AKT1, AKT2 and AKT3 (single-or multi-copy), was found in 27% of all samples and 70% of metastases. MYC multi-copy gain was identified in 6% of all samples and 24% of metastases, increasing to 20% of all samples and 51% of metastases when both single-and/or multi-copy MYC gain are considered (Table S1) [5,6,32]. We examined whether tumors harboring PI3K-pathway alteration were enriched for MYC copy-number gain (single-copy or greater) and found a positive association (P = 0.0002, Fisher's two-tailed exact test; Figs. 1, S2A). Enrichment of PI3K-pathway copy-alterations and high-level MYC amplification was also observed (P = 0.0004; Figs. S1A, S2A). The subset of tumors with specific PIK3CA amplification also showed an association with MYC amplification (Figs. 1, S1A, S2A). There was also a statistically significant association between PI3K-pathway alterations and MYC multicopy gain in metastases (P = 0.036, Figs. S1B-C, S2B). These data establish that alterations in the PI3K-pathway are enriched with MYC amplification in human prostate tumors.
MYC and AKT cooperate to accelerate progression of mPIN to invasion in a murine prostate cancer model To assess the functional implications of the association between PI3K-pathway alteration and MYC amplification in human prostate tumors, we turned to genetically engineered mouse (GEM) models. The role of PI3K signaling in prostate cancer has been modeled in mice by deletion of PTEN or by transgenic expression of activated AKT, while the role of MYC has been investigated by transgenic expression of MYC. A recent study demonstrated interaction between PTEN and MYC signaling using prostate-specific hetero-or homozygous deletion of PTEN with concurrent focal probasin-Credriven MYC overexpression [24]. In order to validate this result in a model with widespread prostate-specific MYC expression, and provide rationale for more extensive studies of the role of AKT, we employed the Hi-MYC [28] transgenic model (FVB background) in a bigenic cross with the prostate-specific PTEN pc2/2 conditional knockout mouse (C57BL/6J strain) [26,27] to generate bigenic PTEN pc2/2 /Hi-MYC mice.
In the Hi-MYC model [28], the modified (ARR 2 PB) probasin promoter-driven expression of human MYC in the prostate results in murine prostate intraepithelial neoplasia (mPIN) in the lateral prostate (LP) by 4 weeks of age that progresses to adenocarcinoma in all mice by 6-9 months. The ventral prostate (VP), dorsal prostate (DP) and anterior prostate (AP) are affected to a lesser extent. The PTEN pc2/2 model expresses probasin-Cre4 (Pb-Cre4) [31] upon puberty, thereby inactivating the floxed PTEN alleles in the VP, LP, DP and AP. PTEN pc2/2 mice develop HG-mPIN that progresses to invasive adenocarcinoma after ,6 months of age [26,27]. PTEN pc2/2 /Hi-MYC bigenic mice have large prostatic adenocarcinomas at 3 months ( Fig. 2), well in advance of either of the wellestablished single lesion models, which at this stage harbor mPIN exclusively (Fig. 2). Assessment of expression patterns for pAKT (increased by inactivation of PTEN) and MYC in the PTEN pc2/2 / Hi-MYC prostatic epithelium revealed a subpopulation of cells expressing both proteins at high levels in areas of invasion (Fig. 2). Consistent with previous work [24], PI3K-pathway activation and MYC cooperate to accelerate progression of invasive prostate cancer, providing the rationale to characterize this cooperation more extensively and in a pure genetic background.
To address whether AKT, downstream of PTEN, might be responsible for the interaction between PI3K-pathway activation and MYC signaling, and whether mTOR is a key mediator, we selected the established MPAKT [29] and Hi-MYC [28] transgenic models, both in the FVB background strain, and cross-bred them to generate MPAKT/Hi-MYC mice with prostate-specific expression of both transgenes. In the MPAKT model, over-expression of myristoylated human AKT1, driven by a portion of the prostate-specific rat probasin promoter, leads to phospho-AKT expression in luminal epithelial cells of predominantly the VP and rarely the LP. Expression of activated AKT correlates with a highly-penetrant phenotype of mPIN in mice by 6-8-weeks-old [29].
Immunohistochemistry for phospho-AKT(Ser473) (pAKT) confirmed AKT activation in MPAKT and, at lower levels, in bigenic MPAKT/Hi-MYC mice (Figs. 3B, S3A, S4). Similarly, immunohistochemical staining of MYC confirmed expression of the MYC transgene in Hi-MYC and MPAKT/Hi-MYC mice (Figs. 3B, S3A). Bigenic animals expressed lower levels of transgenic mRNA than single transgenic mice (Fig. S3B). By 5-9 weeks, all three strains (MPAKT, Hi-MYC and bigenic MPAKT/Hi-MYC mice) had mPIN as expected (largely affecting VP in MPAKT mice, DP and LP in Hi-MYC mice, but DP, LP and VP in bigenic mice) (Fig. 3A). Although the growth pattern of mPIN lesions in Hi-MYC and MPAKT/Hi-MYC mice were similar and often cribriform, nuclear atypia was more pronounced in bigenic mice ( Fig. S5). At this early time-point, the key distinguishing feature in MPAKT/Hi-MYC mice was significant stromal proliferation, inflammation and remodeling in VP and LP with disruption of the basement membrane and smooth muscle layer surrounding glands affected by mPIN, and presence of epithelial cell clusters within adjacent stroma (Figs. 3A, 4). This stromal remodeling phenotype was further investigated by immunohistochemistry for smooth muscle actin (SMA) and collagen IV, which revealed progressive disruption and loss of the smooth muscle layer and basal laminae in focal points around the proliferative glands suggesting early microinvasion in ,70% of bigenic mice (Figs. 3C, 4, S6) [33]. In summary, the histopathological features of mPIN lesions in the bigenic mice were similar to those of Hi-MYC mice; however, the stromal remodeling and inflammation, particularly severe in the VP and LP, together with the nuclear atypia of proliferative cells within areas of mPIN, were unique features of the bigenic mice (Figs. 3, 4, S5, S6).
Progression to adenocarcinoma was accelerated in the MPAKT/Hi-MYC model with evidence of invasion in 8% of mice at 5-9 weeks, and in 67% mice at 16-20 weeks, compared respectively with 0% and 25% of Hi-MYC mice (Figs. 3A, 3D, 4; sample size precluded statistical significance). Of note is that pAKT expression was occasionally evident in populations of cells near the invasive regions (Fig. S4). In more advanced disease beyond 6 months of age, the acceleration in disease progression conferred by AKT activation in presence of MYC overexpression was no longer evident ( Fig. 4) (due to the highly penetrant MYC alone phenotype), although the unique stromal reaction persisted in the bigenic phenotype.

The MPAKT/Hi-MYC prostate lesions are accompanied by infiltration of immune cells
The tumor microenvironment can significantly influence tumorigenesis, and cells from the stromal compartment such as fibroblasts and inflammatory cells can exert effects on adjacent epithelial cells via paracrine signals and extracellular matrix components [34]. To characterize the intense stromal remodeling and inflammatory infiltrate surrounding mPIN and prostate tumors in MPAKT/Hi-MYC mice, we performed immunohistochemistry for T-lymphocytes (CD-3), B-lymphocytes (B220) and macrophages (Mac-2) on prostate tissues from mice aged 5-9 weeks (Fig. 5). All three classes of immune cells were present at high concentrations in the stromal infiltrate and in lesser amounts within the epithelial compartment of mPIN lesions and tumors of the MPAKT/Hi-MYC prostates. In contrast, only occasional macrophages and T-cells were found surrounding mPIN lesions in Hi-MYC prostates, and rare or no inflammatory cells were noted in MPAKT or WT prostates. Thus, the unique stromal remodeling and early invasive phenotype resulting from cooperation between AKT1 and MYC in the mouse prostate is associated with an infiltration of T-and B-lymphocytes, as well as macrophages.

AKT does not rescue MYC-induced apoptosis in the prostate
To explore the cellular mechanism of AKT-MYC cooperativity, we examined the prostates of bigenic mice and their littermates, using markers of proliferation and apoptosis. As expected [28], elevated levels of both proliferation (Ki67 staining) and apoptosis (measured by TUNEL staining of DNA fragments) were seen in Hi-MYC mPIN lesions (Fig. S7), consistent with the wellestablished fact that MYC can induce both cell-proliferation and apoptosis [22]. In contrast, Ki67 and TUNEL ratios were only modestly elevated in MPAKT mice compared with WT ( Fig. S7) [14,29]. Ki67 staining in VP and LP of MPAKT/Hi-MYC was comparable to Hi-MYC littermates, with highest proliferative rates occurring in mPIN lesions. Previous reports using different model systems and tissue-types have suggested PI3K-pathway activation can rescue the proapoptotic phenotype of MYC overexpression [35], providing a potential mechanism for cooperativity. However, apoptotic rates remained high in mPIN lesions of MPAKT/Hi-MYC mice and were not obviously different from Hi-MYC littermates.

Transgenic MYC expression abrogates the mTORdependence of the AKT-induced mPIN phenotype
The AKT-induced mPIN phenotype in young MPAKT mice is dependent on mTOR [14]. We confirmed this in a cohort of 5week-old MPAKT mice treated with RAD001 or placebo (4-5 mice/treatment group) for 2 weeks (Fig. 6, left panels). As expected, mPIN lesions in a cohort of 5-week-old Hi-MYC mice did not revert after two weeks of RAD001 treatment and were histologically indistinguishable from the lesions in control mice (Fig. 6, center panels) confirming that mPIN in Hi-MYC mice does not depend on mTOR signaling. We next examined the mTOR dependence of mPIN lesions in bigenic MPAKT/Hi-MYC mice by treatment of 5-week-old animals with either RAD001 or placebo for 2 weeks. No reversion of the mPIN phenotype upon RAD001 treatment was observed in the VP and LP of the MPAKT/Hi-MYC mice, and the lesions were identical to those of vehicle-treated mice (Fig. 6, right panels).
To confirm that mTOR was inhibited in RAD001-treated mice, we examined the phosphorylation status of the downstream mTOR substrate ribosomal-S6 protein by immunohistochemistry with a widely-used phosphospecific antibody to Ser235/236 (pS6). In all vehicle-treated MPAKT mice, pS6 in the regions of mPIN was similarly high, and treatment with RAD001 led to dramatically reduced pS6 staining (Fig. 6), indicating that RAD001 effectively inhibited mTOR. pAKT expression was retained, confirming continued transgene expression (Fig. 6). pS6 staining was also decreased by RAD001 treatment in MPAKT/ Hi-MYC and Hi-MYC mice, with some tissues showing residual weak pS6 staining (Fig. 6). S235/236 of S6 is also the site for phosphorylation by p90 ribosomal kinase (RSK), raising the possibility of mTORC1-independent (therefore RAD001-resistant) phosphorylation of S6 [36].
In summary, mPIN lesions in young MPAKT mice were fully reverted upon RAD001-treatment; however, mPIN lesions in Hi-MYC and MPAKT/Hi-MYC bigenic mice did not respond to RAD001 despite effective mTORC1 inhibition. We conclude that transgenic MYC expression is sufficient to override the mTOR dependence of lesions arising from constitutive AKT activation. RAD001 treatment did not affect intensity or composition of the inflammatory infiltrate in prostates of bigenic mice.
The mTOR dependence of the activated AKT-driven mPIN phenotype has been demonstrated only in young MPAKT mice [14]. Having demonstrated that MYC can rescue the mTOR dependence of AKT-driven mPIN lesions, we asked if the mPIN lesions of older MPAKT mice would remain dependent on mTOR, or whether additional genetic lesions potentially accumulated with aging might render the prostate lesions insensitive to RAD001 treatment. In contrast to young MPAKT mice, the response of older MPAKT mice (28-32weeks-old) to mTOR inhibition was incomplete and variable (Figs. 7, S8). Of seven mice treated with RAD001 for two weeks, five had residual mPIN, whereas two had no evidence of mPIN. As expected, mPIN was detected in the VP of all 6 placebo-treated mice. pAKT was expressed in mPIN of vehicle-treated MPAKT mice and in both RAD001-sensitive and RAD001-resistant mice, whereas loss of pS6 staining in all RAD001-treated animals confirmed mTOR inhibition (Fig. 7). Strong p27 expression, a documented marker of mPIN in MPAKT mice [37], was observed in mPIN of the vehicletreated and RAD001-resistant MPAKT mice, but absent in WT animals and in the reverted lesions of RAD001-sensitive mice, providing additional evidence for RAD001-resistance (Fig. 7). Therefore, the mPIN phenotype of MPAKT mice becomes progressively independent of mTOR with age.
We next asked whether 4EBP1, an mTORC1 target, plays a role in mediating the sensitivity to RAD001 in MPAKT mice, and the RAD001-resistance in the Hi-MYC and MPAKT/Hi-MYC models, as proposed by a study that used genetically engineered prostate epithelial cells to examine the affect of MYC expression on rapamycin sensitivity [23]. Surprisingly, immunohistochemical assessment of 4EBP1 phosphorylation in the VP of mice aged 7weeks showed no decline in p4EBP1 levels in MPAKT mice following 2 weeks of RAD001 treatment (Fig. 8), despite clear histologic regression of mPIN lesions (Figs. 6, 8). Similarly, expression of p4EBP1 in wild type, Hi-MYC and MPAKT/Hi-MYC mice was either unchanged or slightly increased by RAD001 treatment (Fig. 8). We confirmed this result by immunoblot of protein lysates from isolated ventral prostates, and verified the increased 4EBP1 phosphorylation in the VP of RAD001-treated mice, independent of total 4EBP1 expression (Fig. 8, S9). Abrogation of pS6 expression along with increased glycogen synthase kinase-3b (GSK3b) phosphorylation (a measure of AKT activation due to relief of mTORC1-mediated feedback inhibition of AKT) confirmed successful inhibition of mTOR (Figs. 8, S9). Therefore 4EBP1 phosphorylation in WT, MPAKT, Hi-MYC and MPAKT/Hi-MYC mice is not uniquely dependent on mTOR and cannot explain resistance to mTOR inhibition.
MYC expression may confer resistance to rapamycin by disrupting the balance between proliferation and apoptosis or senescence. Interestingly, prostate tumors from Hi-MYC (n = 5) and MPAKT/Hi-MYC (n = 5) mice all showed reduced TUNEL staining after 14 days of RAD001 treatment compared to prostates from vehicle-treated animals (n = 4-5/genotype) (Fig. 9, S10). The Ki67 staining in the same tissues was unaffected by RAD001 treatment (Fig. 9, S10). Therefore, MYC expression does not simply confer resistance to mTOR inhibition. The reduction in apoptosis may, in fact, reveal paradoxical effects of mTOR inhibitors on tumor progression.  Discussion PI3K-pathway upregulation in primary and metastatic prostate cancers provides the rationale for clinical evaluation of PI3Kpathway inhibitors (including rapalogs). Here we demonstrate a statistically significant co-occurrence of MYC amplification and PI3K-pathway disruption in 194 human prostate tumors, including 37 metastatic tumors. To investigate the potential functional interaction between the MYC and PI3K-pathways in the prostate, we first generated a PTEN pc2/2 /Hi-MYC bigenic mouse that confirmed a prior model of cooperativity between these two pathways [24]. Next, to further investigate the role of PI3K downstream mediators in the interaction with MYC, we crossbred previously characterized mice expressing activated human AKT1 (MPAKT model) [14,29] and human MYC (Hi-MYC model) [28]. In the resultant MPAKT/Hi-MYC model, AKT1 and MYC are expressed together in the prostate, recapitulating the co-incidence of the genetic lesions in human prostate tumor samples.
The prostate glands of MPAKT/Hi-MYC mice are characterized by significant stromal reaction and infiltration of B-and T-lymphocytes, as well as macrophages early in development of mPIN and persisting throughout tumorigenesis. This inflammatory response is of particular interest because of possible roles for the immune system in tumor growth regulation. In the prostate, inflammation is commonly observed in cancer precursor lesions [34,38]. In addition, recent work has implicated infiltrating T H 17 and/or T reg (FoxP3 + ) T-cells in development or progression of human prostate cancer [39]. Cytokines can confer survival to tumor cells in xenografts derived from the Hi-MYC model, facilitating prostate cancer progression [40]. Since it remains unclear to what extent the inflammatory cells in human samples play an active versus bystander role in cancer progression or suppression, the MPAKT/Hi-MYC model may help address this question. Indeed, genetically engineered mouse models of other tumor types (breast, pancreatic islet cells, etc) have firmly established both tumor-promoting and -suppressive actions for distinct subsets of inflammatory cells [41].
Due to growing interest in evaluating PI3K-pathway inhibitors in prostate cancer patients, we explored the activity of the rapamycin analog RAD001 in the MPAKT/Hi-MYC model. In contrast to the exquisite sensitivity of young MPAKT mice to this compound    [14], MPAKT/Hi-MYC as well as older MPAKT mice were completely or partially resistant, respectively. The mechanism of resistance remains to be determined but we can likely exclude pharmacologic explanations such as incomplete target inhibition. Because recent evidence suggests perturbations in levels of the eukaryotic elongation factor 4E (eIF-4E) or its inhibitor 4EBP1, a translational regulator acting downstream of AKT and mTOR, could mediate resistance [42] [43] [23], we considered this as a potential mechanism for RAD001-resistance in the MPAKT/Hi-MYC mice. However, bioinformatic mining of published transcriptome data [28,29] revealed no significant changes in levels of 4EBP1 or eIF-4E in prostate tissues from Hi-MYC or MPAKT mice. Furthermore, phosphorylation of 4EBP1 was unimpaired by mTOR inhibition in these mice (but rather increased, possibly due to feedback activation of the MEK/ERK-pathway [44]). Thus 4EBP1 is not a predictor of response to rapalog therapy in these mice.
Rapalogs, which selectively inhibit the TORC1 complex, can paradoxically activate AKT through loss of S6 kinase-mediated Prostates from mice aged 7-weeks, treated with either vehicle or 10 mg/kg RAD001. Upper panels, lower left panel: IHC of VP from 7-week-old mice. 14 d RAD001 treatment did not block p4EBP1 expression in prostates of any genotype (n = 4-6/treatment group, /genotype). Scale-bars: 100 mm. Lower right panel: Cropped immunoblots after SDS-PAGE gel electrophoresis of protein lysates from VP of 7-week-old mice treated 2 d with vehicle or RAD001 (full-length blots in Fig. S9). RAD001 did not lower (instead increased) p4EBP1 expression, despite successful inhibition of mTORC1, indicated by decreased expression of downstream target pS6 and feedback upregulation of pGSK3b levels. doi:10.1371/journal.pone.0017449.g008 negative feedback at the level of PI3K [17]. While RAD001 resistance could be theoretically mediated through AKT activation that results from TORC1 blockade, it is difficult to envision why this would occur selectively in the MPAKT/Hi-MYC mice and not in the young MPAKT mice, which are RAD001-sensitive. Indeed, our analysis of phospho-AKT levels in RAD001 treated animals revealed similar effects in both strains. Interestingly, the rapamycin-resistant PrEC cells expressing activated PI3K and MYC were sensitive to the dual PI3K/mTOR inhibitor BEZ235 [23], raising the possibility that reduced AKT activity is critical for response.
Another potential mechanism for rapalog-resistance may be the documented mitigation of cellular senescence upon mTOR inhibition in tumors with activated senescence programs [45]. We observed no consistent changes in expression of the senescence-marker p27 by immunohistochemistry in MPAKT/ Hi-MYC and Hi-MYC prostates following RAD001 treatment (not shown); however, we did observe a reduction in TUNEL staining in RAD001-treated tumors. The mechanism of this prosurvival effect of RAD001 treatment in the context of MYC expression could be mediated through relief of mTOR-mediated feedback (eg AKT activation) or other mechanisms requiring further study.
Rapalogs have been explored in pilot studies in prostate cancer, and PI3K and mTORC1/2 kinase inhibitors are now in earlystage clinical trials across tumor types. In this context, our demonstration that MYC overexpression can convert AKTactivated mouse prostate tumors from rapalog-sensitive to rapalog-resistant has implications for clinical studies of PI3Kpathway inhibitors in men whose prostate cancers also harbor increased AKT signaling. As is clear with other tumor types such as glioblastoma and breast cancer, secondary genetic alterations such as PTEN loss can mitigate the response to EGFR or HER2 inhibitors [46,47]. In light of the relatively disappointing single agent activity of rapalogs in prostate cancer, it may be critical to assess the MYC status of prostate tumors (among several other markers) to guide the interpretation of response data in patients undergoing PI3K inhibitor therapy. Figure S1 MYC amplification co-occurs with AKT pathway activation in human prostate tumor samples. Contingency tables for co-occurrence of MYC and PI3K-pathway copy-number alterations (defined by aCGH) from (A) 157 primary and 37 metastatic prostate tumor samples (n = 194 total), or from (B, C) 37 metastatic samples alone. The tables indicate in red the proportion of tumors with PIK3CA amplification or general PI3K pathway gain, that also exhibit MYC amplification. (A), (B) The association between multi-copy MYC amplification and single-or multi-copy PI3K-pathway alterations is shown. (C) The association between single-or multi-copy MYC gain and PI3K pathway alterations is shown. The statistical significance of each association is reported as a P-value determined by 2-tailed Fisher's exact test. (TIF) Figure S2 MYC amplification co-occurs with AKT pathway activation in human prostate tumor samples. Graphic representation of contingency tables (Fig. 1, Fig. S1) for co-occurrence of MYC and PI3K pathway copy number alterations (as defined by aCGH in Methods) from (A) 157 primary and 37 metastatic prostate tumor samples (n = 194 total), or from (B) 37 metastatic samples alone. The graphs indicate the proportion of tumors with PIK3CA amplification or general PI3K pathway gain, that also exhibit MYC amplification. The statistical significance of each association is reported as a P-value determined by 2-tailed Fisher's exact test.  Fig. 3C) of disrupted and absent smooth muscle stroma and collagen IV surrounding glands affected by mPIN in the MPAKT/Hi-MYC prostate (middle and right columns). Note (left column), an example of minimally attenuated smooth muscle sheath and collagen IV around a gland with no evidence of microinvasion. Scale bars: 100 mm.

Supporting Information
(TIF) Figure S7 The MPAKT/Hi-MYC phenotype is characterized by an increase in cell proliferation and apoptosis compared to MPAKT mice, and the pro-apoptotic effects of MYC are not rescued by AKT expression. Mouse prostates from animals aged 5-9 weeks were stained with TUNEL or antibody to Ki67. Results were quantified as the mean percentage (6 SD) of positive cells, counting at least 300 cells from representative areas (ventral prostate for most mice, and mPIN lesions where present) from each of 9 mice per transgenic genotype and 4 wild-type mice. P  Figure 7) showing that 14d RAD001 treatment only variably reverts the mPIN phenotype in ventral prostates of 32-33 weekold MPAKT mice (H&E). Note that mPIN lesions are preserved in the RAD001-resistant examples, whereas the RAD001-sensitive MPAKT example shows normal glands similar to that of a vehicletreated wild-type mouse. Note also that even in hyperplastic areas of the wild-type glands, there is no nuclear atypia or cellular hypertrophy (6-7 mice per treatment group; only 2 of 7 RAD001treated MPAKT mice responded). Scale bars: 200 mm (black), 100 mm (red). (TIF) Figure S9 Uncropped immunoblots. Immunoblots from which portions were cropped for display in Figure 8. Immunoblots were prepared from 3 separate gels with matched loading of identical samples, after SDS-PAGE electrophoresis of lysates from ventral prostates of 7-week-old mice treated for 2 days with vehicle or RAD001. PVDF membranes were cut horizontally prior to probing with primary antibodies for p4EBP1 (Ser65), p4EBP1 (Thr37/46), total 4EBP1, pGSK3b (Ser9), pS6 (Ser240/244), and b-actin to control for protein loading. RAD001 did not lower (but instead increased) p4EBP1 expression, despite successful inhibition of mTORC1 as indicated by decreased expression of downstream target pS6 and feedback upregulation of pGSK3b levels. (TIF) Figure S10 Tumors in prostates of RAD001-treated Hi-MYC and MPAKT/Hi-MYC mice display lower levels of apoptosis and similar levels of proliferation compared with vehicle-treated mice. Lower magnification views of images displayed in Figure 9. Representative TUNEL and immunohistochemical staining using an antibody for Ki67, of prostate tumors from Hi-MYC and MPAKT/Hi-MYC mice aged 31-33 weeks. 14d RAD001 treatment reduces TUNEL staining in tumors, indicative of reduced apoptosis, but has minimal effect on the Ki67 proliferative marker. Images are representative of all HiMYC mice (4/4 vehicle-treated, 5/5 RAD001-treated) and MPAKT/Hi-MYC mice (5/5 vehicle-treated, 5/5 RAD001treated) examined. Scale bars: 200 mm (black), 500 mm (red). (TIF)