Chemopreventive and Bioenergetic Signaling Effects of PDK1/Akt Pathway Inhibition in a Transgenic Mouse Model of Prostate Cancer

Abstract

The phosphoinositide-dependent kinase 1 (PDK1)/Akt pathway is an important regulator of multiple biological processes including cell growth, survival, and glucose metabolism. In light of the mechanistic link between Akt signaling and prostate tumorigenesis, we evaluated the chemopreventive relevance of inhibiting this pathway in the transgenic adenocarcinoma of the mouse prostate (TRAMP) model with OSU03012, a celecoxib-derived, but COX-2-inactive, PDK1 inhibitor. Beginning at ten weeks of age when prostatic intraepithelial neoplasia (PIN) lesions are well developed, TRAMP mice received OSU03012 via daily oral gavage for 8 weeks. The drug treatment significantly decreased the weight of all 4 prostate lobes as well as the grade of epithelial proliferation in the dorsal and lateral lobes compared to vehicle-treated control mice. The incidences of carcinoma and metastasis were decreased, although not to statistically significant levels. Treated mice lost body fat and failed to gain weight independent of food intake. This change and periportal hepatocellular atrophy can be linked to sustained PDK1 inhibition through downstream inactivation of glycogen synthase. Centrilobular hepatocellular hypertrophy and necrosis of Type II skeletal myofibers were also compound-related effects. We conclude that targeting of the PDK1/Akt pathway has chemopreventive relevance in prostate cancer and causes other in vivo effects mediated in part by an alteration of bioenergetic signaling.

Introduction

The 3-phosphoinositide-dependent protein kinase-1 (PDK1)/Akt signaling pathway is an important regulator of multiple biological cellular processes and is widely conserved across species. In addition to influencing cell proliferation and survival, and energy metabolism in normal cells, the role of Akt in carcinogenesis and tumor progression, including that of the prostate, has been firmly established (Li et al., 2005). Accordingly, a number of small-molecule inhibitors of Akt signaling have emerged in recent years with promising efficacy against prostate, breast, hematopoietic, and ovarian cancer among others (Komander et al., 2003; Yang et al., 2004; Zhu et al., 2004; Ihle et al., 2005; Kucab et al., 2005; Mimeault et al., 2006; Xie et al., 2006). While these studies demonstrate the cancer therapeutic success of Akt-inhibiting agents, either alone or in combination with currently used drugs, few conclude about their cancer preventability. Regarding prostate cancer, Akt is an especially attractive target since deregulation of the PI3-K/PDK1/Akt cascade correlates with an increased Gleason score and androgen-independent phenotype with a poor prognosis (Edwards et al., 2003; Liao et al., 2003). Using Akt antagonists to prevent or delay the onset of prostate cancer is logical considering the implication of this protein in the progression of prostatic intraepithelial neoplasia (PIN) to carcinoma (Narayanan et al., 2004), together with the long initial latency of prostate cancer in men (Gupta, 2004).

Among the animal models available to study the efficacy of prostate chemopreventive candidates, the transgenic adenocarcinoma of the mouse prostate (TRAMP) model has received much attention because it involves defined stages of spontaneous tumor progression similar to that in man (Klein, 2005). Paramount to this multistage escalation from simple hyperplasia to carcinoma is the development of PIN lesions, which represent a key intervention point at which a molecularly targeted agent could potentially prevent or slow the progression of prostate epithelial cells to a malignant phenotype.

The term “chemoprevention” here is applied to the chemotherapy of these precancerous PIN lesions (Gupta, 2004). The TRAMP model has been used extensively in chemoprevention studies, including the evaluation of celecoxib, flutamide, R-flurbiprofen, toremifene, decitabine, and green tea polyphenols (Raghow et al., 2000; Wechter et al., 2000; Gupta et al., 2001; Raghow et al., 2002; Gupta, 2004; Narayanan et al., 2004; McCabe et al., 2006).

Although the anti-cancer activity of signaling pathway inhibitors is often well characterized, less is known about mechanism-based, in vivo sequelae of chronic pathway inhibition. Identification and characterization of these effects are appropriate prior to human use since their overall significance may be considered greater in a chemopreventive (vs. cancer therapeutic) context due to an increased patient risk-to-benefit ratio. An opportune time to realize many of these mechanism-based or off-target effects is early in the drug discovery process by thorough phenotyping of animals used in efficacy studies. Albeit ultimately designed to evaluate cancer endpoints, these studies can provide toxicologic information that allows researchers to predict the impact of chronic therapy.

Here we exemplify an approach of combining the disciplines of medicinal chemistry, biochemistry, and veterinary pathology to thoroughly evaluate the prostate chemopreventive ability of a novel compound concurrent with an investigation of whole-body ramificiations. Specifically, TRAMP mice were treated with OSU03012, an orally bioavailable, celecoxib-derived PDK1 inhibitor, to determine its suppressive effects in the progression of PIN to carcinoma and analyzed with respect to prostate and phenotype endpoints. Retardation of prostate lesions was anticipated in light of the mechanistic link between PDK1/Akt signaling and prostate tumorigenesis.

Systemic effects were difficult to predict given the limited safety information of similarly targeted agents in the literature, but were suspected to be minimal based on shorter-duration efficacy studies completed in our laboratory. It is noteworthy that OSU03012 is currently undergoing preclinical evaluations in the Rapid Access to Intervention Development (RAID) program at NCI. The chemopreventive activity of OSU03012 is described here in addition to an alteration of bioenergetic signaling linked to PDK1/Akt pathway inhibition. Our results show evidence of a negative effect on prostate epithelial lesion development and reveal systemic non-cancer parameters that should be considered for monitoring in future preclinical/clinical evaluation of this and similarly targeted agents.

Materials and Methods

Animals

Hemizygous C57BL/6 TRAMP mice from our own colony were cross-bred with FVB/n mice to generate F1 litters. Tail tips from all F1 male offspring were genotyped by polymerase chain reaction (PCR) to identify mice carrying the transgene as described (Greenberg et al., 1995). These TRAMP mice (C57BL/6TRAMPxFVB) were housed in cages with corncob bedding under a constant photoperiod (12 hours light: 12 hours dark) in a temperature- and humidity-controlled room (68°F–72°F and 45–55% respectively) with ad libitum access to pelleted rodent chow and water. The procedures performed were in accordance with protocols approved by the Institutional Laboratory Animal Care and Use Committee of The Ohio State University.

Reagents

OSU03012, a novel celecoxib-derived, COX-2-inactive PDK-1 inhibitor, was synthesized in our laboratory as described previously (Zhu et al., 2004). This compound is currently undergoing preclinical evaluation through the National Cancer Institute’s RAID Program. OSU03012 was prepared as a suspension in vehicle consisting of 0.5% methylcellulose (w/v) and 0.1% Tween 80 (v/v) in sterile water for oral administration. Treatment volumes remained constant at 0.01 ml per gram of body weight. Rabbit polyclonal antibodies against β-Actin, Akt, phospho-Thr308-Akt, glycogen synthase (GS) kinase 3 beta (GSK3β), phospho-Ser9-GSK3β, and phospho-Ser641-GS were purchased from Cell Signaling Technology (Beverly, MA).

Study Design and Experimental Procedure

Ten-week-old TRAMP mice were randomly assigned to experimental (n = 31) and control (n = 30) groups and treated once daily by oral gavage with either 200 mg/kg OSU03012 or vehicle alone under isoflurane anesthesia using a flexible gavage needle. Treatment was initiated at 10 weeks of age when 100% of animals have PIN lesions, and continued through 18 weeks of age when a high percentage (~60%) have developed prostate carcinoma (Kaplan-Lefko et al., 2003). The mice were sacrificed at 18 weeks of age by CO₂ inhalation, which was performed approximately 8 hours posttreatment corresponding to the time of peak plasma concentration.1 Throughout the study, all mice were weighed once per week and daily food consumption was monitored in a representative group (n = 9 control, 14 treated animals). Food consumption was measured as the difference in the mass of food per mouse in 24 hours. The age at which palpable tumors developed was determined for each mouse by daily palpation at the time of gavage. At sacrifice following 8 weeks of treatment, the dorsal (DP), lateral (LP), ventral (VP), and anterior (AP) prostate lobes were microdissected and weighed, with one lobe of each pair saved in formalin and the other frozen in liquid nitrogen.

Histopathology

Individual prostate lobes, iliac lymph nodes, liver (left lobe), and lung from each TRAMP mouse were fixed overnight in 10% formalin, then transferred to 70% ethanol. Four μm-thick, paraffin-embedded tissue sections were stained with hematoxylin and eosin (H&E) by standard procedures. All tissues were evaluated microscopically by a veterinary anatomic pathology resident consulting with board-certified veterinary pathologists about relevant lesions. A TRAMP-specific grading scheme (Suttie et al., 2003) was employed to semiquantitatively compare proliferative prostate lesions between groups. The observer scoring the prostate slides was blinded to the treatment status.

The grading scheme, which assigns a single number (1–18) to each prostate lobe, accounts for the entire spectrum of proliferative lesions, from simple hyperplasia to poorly differentiated carcinoma, with adjustment for lesion distribution. In addition to these tissues, the remainder of each carcass excluding skin and vertebral column was saved in 10% formalin; tissues from representative animals (n = 7 control, 12 treated mice) were evaluated microscopically with respect to Society of Toxicologic Pathology-proposed guidelines for repeat-dose toxicity studies (Bregman et al., 2003).

Immunohistochemistry

Immunohistochemical detection of proteins using the antibodies indicated in Figures 2, 4, and 6 was performed on 4 μm-thick, paraffin-embedded prostate, liver, and skeletal muscle tissue sections as described (Tomita et al., 2006). For muscle fiber typing, paraffin sections of mouse muscle were deparaffinized with Ventana EZ Prep (Tucson, AZ) and treated with CC1 antigen retrieval on a Ventana Discovery autostainer. Slides were rinsed in diH₂O, placed on the DAKO autostainer plus (Carpinteria, CA), then dual stained with Novocastra Laboratories Ltd. Myosin Heavy Chain (MHC) slow and fast mouse monoclonal antibodies (Newcastle upon Tyne NE12 8EW, UK).

The DAKO ARK (Animal Research Kit) kit peroxidase was used to reduce reactivity of the 2 mouse antibodies with endogenous immunoglobulin. The volumes of primary antibody, biotinylation, DAKO diluent, and blocking reagents needed for the biotinylated primary antibody were calculated with ARKulator software and reagents included in the ARK kit. The procedure outlined in this kit was duplicated for use on the DAKO autostainer. The sections were first tested with a peroxidase block for 5 minutes then the biotinylated MHCs antibody (1:40) for 1 hour followed by streptavidin-HRP for 30 minutes and DAB+ for 15 minutes.

The slides were tested again with peroxidase block for 5 minutes, then the biotinylated MHCf antibody (1:40) for 1 hour followed by KPL streptavidin-phosphotase (Gaithersburg, MD) for thirty minutes. DAKO wash buffer was used for all rinse steps in the procedure. DAKO permanent red was manually applied to the sections for 6 minutes, rinsed and dipped in DAKO Hematoxylin for 20 seconds then air-dried overnight and cover-slipped.

Immunoblotting for Molecular Targets

Total protein was extracted from tissue homogenates of dorsal, lateral, ventral, and anterior prostate lobes from 18-week-old vehicle-control and treated TRAMP mice. Briefly, prostate tissues were mechanically homogenized in 16 μl of lysis buffer (20 mM tris pH 8, 20 mM EDTA, 0.5% NP-40) per mg of tissue, and left on ice for 30 minutes. Protein concentrations were determined in duplicate using a BCA assay (Pierce) with LC software and samples were prepared so that 20 μl containing 40 μg of protein per sample were loaded in each lane.

The samples were fractionated on 4–15% gradient SDS-PAGE gels (Criterion Precast Gel, BioRad) and transferred to nitrocellulose membranes. Western blotting was accomplished as done previously (Kulp et al., 2004) with the antibodies indicated in Figure 2, using β-Actin as an internal control. A chemiluminescence detection reagent (Western Lightning, Perkin Elmer Life Sciences) was used to visualize protein bands developed on Kodak Biomax light film.

Transmission Electron Microscopy

Primary fixation of liver samples was done with 1.75% glutaraldehyde, 25 mM Na-phosphate buffer (pH 7.4), 5 mM sucrose, and 0.5 mM NaN₃. Post-fixation was accomplished with 1% OsO₄. Following dehydration, infiltration and embedding were carried out in an epon resin. Thin sections were cut for toluidine blue staining. For EM analysis, thin sections were double stained with 1 % uranylacetate in aqua and lead citrate, then observed by a Philips 300 Transmission EM at an accelerating voltage of 60 kV. Photographs were taken on Kodak Electron image films.

Statistical Analysis

A Student’s t-test was used to compare the means of 2 samples of independent, normally distributed observations. Specifically, this test was used to determine if certain responses (prostate lobe and epididymal fat pad weight, prostate lesion score, immunoblotting results) were influenced by drug treatment. A Chi-square contingency analysis was used to determine if drug treatment affected the incidences of carcinoma and metastases, which were evaluated as discrete, whole number values. Differences between groups were considered significant at p < 0.05.

Results

OSU03012 Decreases Prostate Lobe Weight and the Grade of Epithelial Proliferation

The weights of all 4 paired lobes of drug-treated animals were significantly less than those of control mice (Figure 1a) following 8 weeks of repeat dosing. Seminal vesicle weights were also significantly decreased in drug-treated TRAMP mice with an average weight of 279 ± 107 mg compared to 445 ± 185 mg in controls [p = 0.0001, mean ± standard deviation (SD)].

The TRAMP-specific grading scheme (Suttie et al., 2003) showed a significant reduction in lesion score in the DP and LP of drug-treated mice (Figure 1b), with an average score for the DP and LP of 9.0 ± 2.7 and 9.2 ± 3.4, respectively, in drug-treated mice compared to 10.5 ± 2.7 and 11.4 ± 4.2 in control mice (mean ± SD). Figures 2a and 2b demonstrate a representative histologic appearance of this suppressive effect. The grade of epithelial lesion development was also decreased in the VP and AP, although not to statistically significant levels.

OSU03012 Decreases the Incidences of Prostate Carcinoma and Metastasis, Although not to Statistically Significant Levels

After 8 weeks of daily oral treatment with 200 mg/kg OSU03012, carcinoma was detected in 8 of 31 (25.8%) mice compared to 12/30 (40%) of vehicle control mice. Metastasis was limited to the iliac lymph nodes and was decreased from 9 of 12 (75%) control mice to 3 of 8 (37.5%) drug-treated mice. Although the difference in tumor size was often striking at necropsy, neither of these differences was statistically significant (p = 0.24 and 0.09, respectively). No metastases were observed in the liver, lung, or other tissues of any TRAMP mice. There were no significant differences in age at development of palpable tumors. Prostate tumors could be palpated at approximately 8 mm-diameter or 400 mg weight. OSU03012-treated mice developed palpable tumors at 14.5 ± 2.3 weeks of age compared to 15.5 ± 4.8 weeks of age in control mice (mean ± SD).

OSU03012 Inhibits the Phosphorylation of Akt and the Downstream Effector GSK3β

Overexpression of p-Akt in proliferative prostate epithelium of TRAMP mice was demonstrated in a proof-of-principle study that preceded the longer-term drug treatment study. Figure 2c demonstrates this dramatic increase in p-Akt immunostaining in hyperplastic epithelial cells that are sharply demarcated from non-staining normal epithelium. After 8 weeks of treatment with OSU03012, immunostaining for p-Thr308-Akt revealed a sharp decrease in reactivity in the prostate epithelium compared to that of vehicle-treated control mice (Figure 2d and e).

Because Akt is a major regulator of GSK3β activity (Cross et al., 1995), we assessed the phosphorylation status of GSK3β, an Akt substrate, as a functional consequence of diminished Akt phosphorylation. As shown in Figure 2f, levels of phosphorylated GSK3β were markedly decreased in samples from OSU03012-treated mice without changes in the expression of total GSK3β protein. This treatment-induced activation of GSK3β correlates with the observed phosphorylating inactivation of GS, the downstream regulator of glycogen synthesis. In principle, OSU03012-mediated deactivation of Akt led to GSK3β dephosphorylation, thereby increasing the level of p-GS.

OSU03012 Decreases Body Fat and Prevents Body Weight Gain Independent of Food Intake, and Alters the Serum Chemistry Profile

The body weight of treated mice remained relatively constant throughout the 8-week treatment period compared to a gradual weight gain in control mice (Figure 3a). The overall reduction in body fat of drug-treated animals was quantified by weighing epididymal fat pads, which was significantly decreased compared to controls (Figure 3b). Food intake was estimated to be equal between the groups by serial measurements of daily food consumption. This observation suggested a shift towards higher bioenergetic metabolism in the drug-treated animals and encouraged continued phenotyping of the drug-treated and control mice.

Serum chemistry revealed a modest, but significant reduction in blood glucose and cholesterol, and a significant increase in aspartate aminotransferase (AST) in treated mice (Table 1). Elevations in creatine kinase (CK), alanine aminotransferase (ALT), and alkaline phosphatase (ALP) were also noted in OSU03012-treated TRAMP mice. Hemograms of animals in both groups were unremarkable with the exception of a relative lymphopenia in drug-treated mice (Table 2). The overall significance of the decreased lymphocyte count is uncertain since, although decreased compared to controls, the count lies within the normal reference range determined for various mouse strains and ages (Suckow et al., 2001).

OSU03012 Causes Hepatic Lesions that can be Linked in Part with Target Pathway Inhibition

A distinct accentuation of lobular architecture, characterized by atrophy of zone 1 (periportal) cells and hypertrophy of zone 3 (centrilobular) cells, was evident histologically in the livers of drug-treated mice (Figure 4b). Considering that the liver is a main site of glycogen synthesis and storage in the body, together with the role of PDK1/Akt signaling in glycogen metabolism, an effort was made to determine if these changes could be mechanism-based effects of drug adminstration. Accordingly, periodic acid-Schiff (PAS) staining of the liver was employed and revealed a dramatic reduction in glycogen storage in drug-treated TRAMP livers compared to controls (Figure 4d). Moreover, intense immunostaining for the inactive, phosphorylated form of GS was detected in the atrophic zone 1 cells in the livers of drug-treated mice (Figure 4f).

To further characterize centrilobular hypertrophy, and specifically to investigate if cytochrome P450 induction was responsible for hypertrophy, toluidine blue staining and transmission electron microscopy (EM) of livers from control and treated mice were performed. A finely stippled, ground glass-like appearance of hepatocellular cytoplasm of drug-treated TRAMP mice was evident in toluidine blue-stained sections (Figure 5b). Ultrastructurally, this finely granular material corresponds to focal clusters of tightly packed smooth endoplasmic reticulum (ER) mixed with dispersed arrangements of rough ER and mitochondria within a glycogen-poor cytoplasm (Figure 5d).

This pattern contrasts sharply with the deep blue, clumped foci observed in control TRAMP mice (Figure 5a) that correspond to large aggregations of mitochondria and rough ER ultrastructurally within a glycogen-rich cytoplasm (Figure 5c). EM also revealed an apparent increase in the number of canalicular microvilli in OSU03012-treated TRAMP, the cause of which is uncertain but is not an uncommon alteration in experimental or spontaneous disease states in rodents (Ghadially, 1982).

A Multifocal, Multiphasic Skeletal Myopathy Targeting Type II Fibers is Observed in Drug-Treated TRAMP Mice

Multifocal segmental myonecrosis was noted in various skeletal muscles of treated mice including biceps femoris, parasternal, masseter, and occipital muscles. Figure 6b demonstrates a spectrum of lesions from acute myofiber degeneration and necrosis to myofiber loss with macrophage infiltration and regeneration of myofibers. Intense immunostaining for p-GS was demonstrated in affected skeletal muscle of drug-treated TRAMP (Figure 6d) similar to that observed in periportal cells of the liver. Immunohistochemical typing of myofibers revealed targeting of Type II (glycolytic/fast) fibers with sparing of Type I (oxidative/slow) fibers (Figure 6e).

Discussion

The PI3-K pathway is known to play a major role in prostate tumorigenesis, owing much of its oncogenicity to activation of Akt through PDK1 (Li et al., 2005). We have previously reported potent anti-cancer effects of OSU03012, our novel, celecoxib-derived, PDK1/Akt-targeted agent, in a variety of human malignanices in vitro, including prostate, breast, glioblastoma, and hematopoietic cancers (Zhu et al., 2004; Johnson et al., 2005; Kucab et al., 2005; Tseng et al., 2005; McCubrey et al., 2006; Tseng et al., 2006). In this report, we describe onco-suppressive effects of PDK1/Akt inhibition by repeat oral dosing of TRAMP mice with OSU03012 on the progression of PIN to carcinoma. In addition to prostate endpoints, an alteration in energy metabolism was observed and linked to inactivation of GS in normal cells.

Evaluation of endpoints for prostate lobe weight and the severity of epithelial proliferative lesions revealed that OSU03012 is able to suppress the overall grade of lesion development in TRAMP mice when treatment is initiated after marked hyperplasia (PIN) is well established. Significant decreases of these parameters in the dorsal and lateral lobes are especially noteworthy since this part of the rodent prostate is believed to mimic the peripheral zone of the human prostate in its disposition to develop malignancy (Abate-Shen and Shen, 2002). Although food consumption was considered equal between the groups, the influence of decreased body weight on prostate lesion development in drug-treated mice cannot be discounted.

Diet restriction in this model has been shown to retard lesion development (Suttie et al., 2003, 2005). However, TRAMP mice in the present study exhibit at most a 6% decrease in body weight, the effect of which on tumorigenesis is unclear. Despite large numbers of experimental animals (n = 30 per group), we were unable to demonstrate strong evidence for a decrease in the incidence of carcinoma in the treatment group. The data indicate that, considering all alveoli that comprise a prostate lobe, OSU03012 has a general negative effect on epithelial lesion development (the decreased lobe weights in treated animals correlate with the reduction in microscopic lesion score), but the transformation of certain cells to neoplasia is not substantially affected. It appears that once a clonal population of cells is established, the tumors grow quickly and aggressively regardless of treatment.

Two explanations of this failure to stop the progression to malignancy are possible. First, the TRAMP model is characterized by a strong oncogenic stimulus that drives the rapid progression from noninvasive, low-grade lesions, which were present at the time treatment was started, to aggressive carcinoma.

This rapid tumorigenic progression is a considerable drawback of the model (Gupta, 2004) and may counter chemopreventive effects of single agent treatment with OSU03012 that would be evident in an in vivo system that more closely models the slower progression of the human disease. Secondly, although we demonstrated Akt activation in hyperplastic prostate epithelium in TRAMP mice prior to this study (Figure 2c), the exact transition point in the prostate at which Akt activation leads to tumor growth remains unclear (Narayanan et al., 2004). The promotion of cell proliferation by viral T antigens, and subsequent inhibition of retinoblastoma and p53 (Greenberg et al., 1995), may override the necessity of Akt signaling in progressed lesions in this model.

This apparent time-dependency of Akt modulation in TRAMP is further demonstrated by Natarayanan et al. in which Akt suppression was correlated with reduced PIN lesions when TRAMP mice were treated with celecoxib beginning at 6 weeks of age (Narayanan et al., 2004). This study also reflects the COX-2-independent anti-tumor activity of COX-2 antagonists, which have been previously described (Hsu et al., 2000), as does our study since OSU03012 is a celecoxib analogue in which COX-2 targeting has been removed (Zhu et al., 2004).

A treatment-associated, food intake-independent reduction in body fat, observed in otherwise apparently healthy TRAMP mice prompted detailed phenotyping and comparison of experimental groups. This analysis revealed hepatic lesions suggestive of altered glycogen metabolism and of a link to target pathway inhibition. Glycogen synthase is a pivotal regulator of glucose storage (Plyte et al., 1992) and is controlled by PDK1/Akt/GSK3 signaling in liver, striated muscle, and fat (Mora et al., 2005). As GSK3β is an important inhibitor of GS activity, and inhibition of PDK1 promotes GSK3β function by inactivation of both Akt and p70 ribosomal S6 kinase (S6K) (Zhu et al., 2004), it not surprising that chronic, systemic PDK1 inhibition could influence basic energy metabolism by depleting liver and muscle glycogen stores. A subsequent increase in β-oxidation of fatty acids to maintain adenosine triphosphate production could, in principle, account for the observed decrease in body fat of the treated mice.

Supportive evidence of an effect on glycogen synthesis was determined by histopathology, immunohistochemistry, and electron microscopy of liver, and immunohistochemistry of skeletal muscle. Evidence of decreased hepatocellular glycogen content concurrent with inactivation of GS correlates well with the proposed pharmacologic effect of OSU03012 (Figure 7) and suggests a cause of the zonal pattern of hepatocellular atrophy. The increased intrahepatic level of p-GS in association with decreased glycogen content strongly suggest a pharmacological effect of OSU03012 treatment since GS activity and glycogen content exhibit an inverse relationship under normal physiologic conditions (Nielsen, 2001).

Inactivation of GS in skeletal muscle is supported by the intense immunohistochemical staining of p-GS and suggests that generalized suppression of GS occurred in mice subsequent to OSU03012 treatment. This is the first report, to our knowledge, of a systemic effect on glycogen synthase activity secondary to chronic PDK1 targeting by an orally administered compound. There are no reports of this effect by celecoxib, the parent compuond, which is fitting since PDK1 inhibition by celecoxib is modest in comparison to that caused by OSU03012.

Adverse effects by celecoxib are largely associated with its nonsteroidal anti-infammatory properties and include cardiovascular and gastrointestinal events (Food and Drug Administration, 2004). The 7-hydroxystaurosporine (UCN-01), a broad-acting kinase inhibitor of which PDK1/Akt is a target, has been reported to modulate bioenergetic signaling in Phase I clinical trials (Kondapaka et al., 2004). This effect, however, is one of insulin resistance and appears to be much different than that observed here. Similar phenotypic sequelae of chronic PI3-K inhibition, the upstream kinase, have been described and include hyperglycemia associated with perturbations of insulin signaling in adipocytes and myotubes. Effects on glucose metabolism in the liver, however, remain unexplored (Knight et al., 2006).

Considering that genetically engineered animals designed to predict the efficacy of therapeutic agents may provide ancillary information about toxicity (Bolon, 2004), conditional liver and skeletal muscle PDK1 knockout animals would be useful for further investigation of the causal relationship proposed here. Lawlor et. al. found that PDK1 is an essential regulator of cell size and is required for embryonic development. Although homozygous PDK1 deletion is embryonically lethal, hypomorphic mutant PDK1 mice are viable, fertile, and 40–50% smaller than control animals (Lawlor et al., 2002). In a conditional knockout model, Mora et. al. showed that insulin could not activate GS or cause its dephosphorylation at GSK3-phosphorylated residues in cardiac muscle from PDK1^−/− or double GSK3alpha/GSK3beta knock-in mice.

Perhaps surprisingly, normal levels of cardiac glycogen were detected in these mice suggesting a more complex regulation of GS activity in the heart (Mora et al., 2005). It would be interesting to know total hepatic and skeletal muscle glycogen content in animals from these studies. Regarding GS specifically, humans with congenital deficiencies of GS are reported to develop fasting hypoglycemia (Orho et al., 1998), and knockout mice lacking muscle GS are smaller than wild-type littermates and have less body fat (Pederson et al., 2005).

Additional in vivo studies are needed to further characterize cytochrome P450 induction associated with hypertrophy of zone 3 cells. For our purposes, smooth ER proliferation detected by EM is compatible with phenobarbitone-like P450 induction and rules out peroxisome proliferation as the cause of hypertrophy (Greaves, 2000). The observed focal clusters of tightly packed smooth ER may represent a late stage of hepatocellular drug tolerance by drug-handling enzymes (Ghadially, 1982). The enlarged zone 3 cells at least partly explain the elevated ALP detected in these mice via cholestasis. Both the centrilobular and periportal changes likely contributed to the increases in ALT and AST, however, overt hepatocellular necrosis was not observed. A similar pattern of centrilobular hypertrophy was associated with a dose-dependent increase in liver weight in nude mice treated with OSU03012 for 6 weeks at 100 and 200 mg/kg/day (unpublished research).

These results provide guidelines for future toxicity assays that may include specific time-course characterization, enzyme activity analysis, and profiling of P450 isosozymes. The latter will be important for inferring human risk (Watanabe et al., 1998; Greaves, 2000), especially since xenobiotic-induced activity of mixed function oxidases may affect the metabolism of other administered compounds, or may be a compensatory response with little clinical significance (Greaves, 2000).

Skeletal myonecrosis in treated TRAMP mice correlates with the elevated CK and AST levels detected in these animals. This lesion is intriguing considering that skeletal muscle diagnoses in general are very rare in National Toxicology Program studies in mice (Leininger, 1999), and that the literature contains very few reports of such lesions in the nonclinical and clinical safety evaluation of pharmaceutical candidates. Potential drug associations in our study include altered muscle homeostasis secondary to impaired glycogen metabolism, or a disturbance of Akt/GSK3 signaling, which regulates myofiber size and hypertrophic responses (Bodine et al., 2001; Rommel et al., 2001). As other Akt inhibitors have been reported to target glycolytic cells [in the context of cancer cells dependent on glucose (Elstrom et al., 2004; Plas and Thompson, 2005; Thompson, 2005)], it is also possible that Akt inhibition can affect glycolytic Type II myocytes, which we have shown to be targeted in our drug-treated mice.

It is noteworthy that the pattern of myonecrosis observed here is reminiscent of that induced by statins, the exact mechanism of which remains unknown but was recently shown to involve targeting of type IIB fibers (Westwood et al., 2005) and has been linked directly to cholesterol lowering, as detected in our mice, by an alteration of myocyte lipid/protein organization (Draeger et al., 2006). A mechanistic link to cholesterol lowering, shared by the statins and the present study of OSU03012, is also supported by reports of skeletal muscle necrosis with other hypocholesterolemic drugs including clofibrate in rats (McDonald and Hamilton, 1990).

We hope to demonstrate here that complete phenotyping of animals used in early efficacy studies may reveal unanticipated effects that could facilitate lead optimization, assist protocol design in toxicology, and provide additional biomarkers for further in vivo testing. This approach will undoubtedly improve the efficiency of preclinical evaluation of drug candidates, which remains a substantial roadblock in drug development partly due to the time and money required for performing both therapeutic and toxicologic studies in animals. Given that elucidating the exact mechanisms of the changes observed here requires toxicology-specific animal assays, we demonstrate evidence for association with target pathway inhibition. Fresh tissue obtained from animals in future studies will permit quantitative analysis of glycogen content, for example.

It is important to note that the 200 mg/kg/day dose used in this study is the highest dose of OSU03012 to be administered to xenograft-bearing mice in our work (unpublished data). Moreover, the reversibility and specific dose-dependency of the observed effects are unknown. Although OSU03012 potently inhibits PDK1 (Zhu et al., 2004), other effects likely contribute to its anti-cancer activity and must also be considered as potential influencers of phenotypic changes.

In conclusion, inhibition of the PDK1/Akt pathway during the progression of advanced PIN to carcinoma reduces the overall severity of epithelial lesion development as determined by lobe weight and lesion score endpoints, but is insufficient to significantly decrease the incidence of carcinoma and metastasis in TRAMP mice. In addition to effects on the prostate, chronic oral administration of OSU03012 caused phenotypic changes at least in part associated with inactivation of the enzyme responsible for glycogen synthesis. As the rational design and development of PDK1- and Akt-targeted small molecules is an active focus of research in the field of anticancer drug discovery, our study reveals important noncancer parameters, namely body fat/weight along with markers of liver and skeletal muscle health, that should be considered for monitoring in future preclinical and clinical evaluations of these agents.