AKT Activity Determines Sensitivity to Mammalian Target of Rapamycin (mTOR) Inhibitors by Regulating Cyclin D1 and c-myc Expression*

Prior work demonstrates that AKT activity regulates sensitivity of cells to G1 arrest induced by mammalian target of rapamycin (mTOR) inhibitors such as rapamycin and CCI-779. To investigate this, a novel high-throughput microarray polysome analysis was performed to identify genes whose mRNA translational efficiency was differentially affected following mTOR inhibition. The analysis also allowed the assessment of steady-state transcript levels. We identified two transcripts, cyclin D1 and c-myc, which exhibited differential expression in an AKT-dependent manner: High levels of activated AKT resulted in rapamycin-induced down-regulation of expression, whereas low levels resulted in up-regulation of expression. To ectopically express these proteins we exploited the finding that the p27kip1 mRNA was efficiently translated in the face of mTOR inhibition irrespective of AKT activity. Thus, the p27kip1 5′-untranslated region was fused to the cyclin D1 and c-myc coding regions and these constructs were expressed in cells. In transfected cells, expression of cyclin D1 or c-myc was not decreased by rapamycin. Most importantly, this completely converted sensitive cells to a phenotype resistant to G1 arrest. Furthermore, the AKT-dependent differential expression patterns of these two genes was also observed in a mouse xenograft model following in vivo treatment with CCI-779. These results identify two critical downstream molecular targets whose expression is regulated by AKT activity and whose down-regulation is required for rapamycin/CCI-779 sensitivity.

Drugs that specifically inhibit the mammalian target of rapamycin (mTOR) 1 are currently being developed as potential anti-tumor agents. mTOR is a critical protein that integrates signals that link the ability of cells to undergo cell cycle transit to the availability of nutrients in their immediate environment (1)(2)(3)(4). By inhibiting mTOR, these drugs essentially trick the cell into believing that conditions are not appropriate for cell cycle progression to ensue and induce G 1 arrest. Rapamycin is the prototype mTOR inhibitor and CCI-779, a recently developed analog of rapamycin, is currently in clinical trials for cancer patients. mTOR activation is mediated by upstream signals from the phosphatidylinositol-3Јkinase(PI3-K)/3-phosphoinositide dependent protein kinase-1 (PDK-1)/AKT cascade. mTOR activity, in turn, results in phosphorylation of the p70S6 kinase (p70) and 4E-BP1 translational repressor (5)(6)(7)(8). Phosphorylation of p70 is critical for ribosome biogenesis, and phosphorylation of 4E-BP1 disrupts its interaction with the eIF-4E translation initiation factor, allowing eIF-4E to participate in assembly of a translation initiation complex (eIF-4F). In this complex, eIF-4E binds to the cap structure at the 5Ј-end of mRNAs, which promotes ribosome recruitment to mRNAs and the initiation of translation. By up-regulating the components of the protein synthetic machinery and cap-dependent translation, both of these mTOR-dependent phosphorylation events lead to translation of proteins required for cell cycle transit.
Because malignant clones frequently demonstrate up-regulation of the PI3-K/PDK-1/AKT/mTOR pathway (9 -12), mTOR inhibitors may be particularly effective as they target this pathway. In keeping with this notion, high levels of AKT activity, whether due to PTEN mutation or introduction of activated AKT alleles, result in hypersensitivity to CCI-779-induced G 1 arrest (13). At first glance, this correlation is intuitive because AKT activity is an indicator of activation through the PI3-K/PDK-1/AKT pathway, and this pathway is an upstream stimulator of mTOR function. However, mTOR inhibition should result in inhibition of ribosome biogenesis and cap-dependent translation in all cells, irrespective of AKT activity, and, in fact, cells with lower levels of basal activity (i.e. those with less stimulation through AKT to mTOR) might be more susceptible to rapamycin-induced decreases in mTOR activity below a specific threshold.
The critical cytostatic effect of mTOR inhibitors is theoretically a reduction in cap-dependent translation of cell cycle proteins. Thus, a second possible explanation for the regulatory role of AKT might be that it can regulate the ability of cells to maintain levels of these proteins by controlling their translation through cap-independent mechanisms. To test this hypothesis, we screened the translational state of Ͼ5,000 mRNAs, using the methodology of Zong et al. (14). Our results identified two potentially critical mRNAs for cell cycle transit, that of cyclin D1 and c-myc, whose translational response to mTOR inhibitors both in vitro and in vivo were remarkably regulated by the degree of AKT activation. Further studies confirmed that the differential effects on the expression of these proteins determined sensitivity to mTOR inhibitors.

EXPERIMENTAL PROCEDURES
Cell Culture and Plasmids-U87 and LAPC-4 parental cell lines were originally obtained from ATCC and maintained as described previously (13). They were subsequently stably transfected using retroviral constructs with a wild-type PTEN gene to generate U87 PTEN and a constitutively active AKT allele to generate LAPC-4 AKT as described previously (13). Plasmids expressing p27-IRES-EGFP, p27-IRES-cyclin D1 and p27-IRES-myc mRNAs were constructed using pcDNA3 as the backbone vector. The 5Ј-UTR of p27 kip1 containing the 365 nucleotide IRES was PCR amplified from IMAGE clone 4298338 and inserted immediately upstream of the cyclin D1 open reading frame in pcDNA3. The c-myc open reading frame was amplified from pBABE puro (a gift from D. Felsher, Stanford University) and subcloned into pcDNA3. Subsequently the p27 kip1 5Ј-UTR was inserted immediately upstream of c-myc. Lastly, the EGFP open reading frame was subcloned into pcDNA3 and the p27 kip1 5Ј-UTR ligated immediately upstream. Plasmids were sequenced at all junction points to ensure that they were properly generated.
Polysome and Microarray Analysis-Extraction and display of polysomes was performed as previously described (15). Briefly, cells were lysed in buffer supplemented with 100 g/ml cycloheximide at 4°C. Following removal of nuclei and mitochondria, supernatants were layered onto 15-50% sucrose gradients and spun at 38,000 rpm for 2 h at 4°C in a SW 40 rotor (Beckman Instruments). Centrifuged gradients were fractionated into eleven 1-ml fractions using an ISCO Density Gradient Fractionator at a flow rate of 3 ml/min. The polysome profile of the gradient was monitored via UV absorbance at 260 nm. RNA from individual fractions was extracted using phenol/chloroform following incubation with proteinase K and precipitated. RNA was then either processed for Northern blots or used to generate labeled cDNA. Prior to hybridization to cDNA arrays, polysomal and monosomal RNA was reverse transcribed (SuperScript RTII, Invitrogen) in the presence of Cy3-dCTP or Cy5-dCTP, respectively. Custom spotted arrays containing Ͼ6000 array elements were hybridized with the labeled cDNA for 16 h at 42°C, washed, and then scanned using an Axon DNA array scanner (Axon Instruments, Union City, CA). Raw images were quantified using GenePix 3.2 (Axon Instruments) and the data analyzed using the GeneSpring (Silicon Genetics, Redwood City, CA) software package. Data were normalized using the Lowess method contained in GeneSpring. The change in translation state for a given mRNA was defined as the ratio of polysomal RNA to monosomal RNA signal intensities after rapamycin treatment divided by the ratio of polysomal RNA to monosomal RNA signal intensities under normal conditions (culture without rapamycin). A value of greater than 1 indicates that the mRNA moved from the monosomal to the polysomal fractions upon rapamycin treatment. If the ratio was less than 1, the value was inverted and a negative number was used to indicate a change in the reverse direction, indicating a shift for a given mRNA from polysomal to monosomal fractions. K-means cluster analysis of the data was done using algorithms included in GeneSpring (16).
Effects of Ectopic Cyclin D1/c-myc Expression-Transient transfections were performed using SuperFect reagent (Qiagen) or Lipo-fectAMINE 2000 (Invitrogen) using 5 g of plasmid containing p27-IRES-EGFP, 5.5 g of plasmid containing p27-IRES-cyclin D1, or 5.5 g of plasmid containing p27-IRES-myc. In groups where both IRES-containing cyclin D1 and c-myc plasmids were transfected only 2.75 g of each plasmid were used. To measure GFP expression and DNA content, cells were first fixed in 2% buffered formaldehyde (Polysciences) and then permeabilized with 70% ethanol at Ϫ20°C followed by propidium iodine staining. Texas Red-conjugated cyclin D1 antibody and Alexa 350-conjugated c-myc antibody were used for staining (Molecular Probes). Data were acquired and analyzed on FACScalibur flow cytometer using CellQuest software (BD Biosciences).
Cytotoxicity and Cell Cycle Analysis-For testing cell survival, viable recovery was determined by MTT assays as previously described (17). Cell cycle phase distributions were determined by flow cytometry on propidium iodine stained cells.
Mouse Xenograft Studies-Male SCID mice were injected subcutaneously with single cell suspensions of the four cell lines as described previously (13). Tumor growth was measured daily, and mice were randomized to CCI-779 versus vehicle when tumors reached 200 mm 3 . Treatment was given by intraperitoneal injection for 5 consecutive days, and tumors were harvested for Western analysis 6 h following the last injection. Tumor growth was assessed on day 8 or 12 after initiation of CCI-79 treatment. IC 50 was determined by extrapolation of plots of percent growth inhibition by CCI-779 versus log concentration. Statistical analysis was performed by using Student's t test and ANOVA models using Sigma Stat 2.0 (Jandel Scientific).

RESULTS
AKT Activity Regulates Sensitivity to mTOR Inhibitors both in Vitro and in Vivo-To investigate the mechanism by which AKT activity regulates sensitivity to mTOR inhibitors, we studied two separate cell lines whose AKT activity was altered by transfection. The LAPC-4 line containing quiescent AKT was stably transfected with a constitutively active AKT allele and the U87MG glioblastoma cell line (termed U87 in this work), containing a mutated PTEN gene and resulting in heightened AKT activity, was transfected with a wild-type PTEN gene. LAPC-4 cells transfected with myristoylated AKT (termed LAPC-4 AKT ) had greatly increased expression of phosphorylated (activated) AKT compared with empty vector-transfected cells (LAPC-4 puro ), and U87 cells transfected with wild-type PTEN (U87 PTEN ) had significantly decreased expression compared with control U87 cells (Fig. 1A). In contrast, levels of phosphorylated ERK were unaffected by transfection in both cell lines, attesting to the specificity of the altered AKT activity.
The transfectants were then tested for sensitivity to the mTOR inhibitor CCI-779 by cell cycle analysis (Fig. 1B). As shown, a clear differential sensitivity was demonstrated between the "high AKT" cell lines (U87 and LAPC-4 AKT ) versus their "low AKT" counterparts (U87 PTEN and LAPC-4 puro ). The ID 50 for decrease in S-phase distribution was 7.5-10 nM of CCI-779 for the sensitive U87 and LAPC-4 AKT cell lines, whereas the U87 PTEN and LAPC-4 puro cell lines were completely resistant with no effect on S-phase distribution at any dose. There was no induction of apoptosis as detected by sub-G 1 peak or annexin V staining (data not shown).
To determine whether AKT activity could regulate sensitivity to CCI-779 in vivo, we examined the growth of the four cell lines in SCID mice. Tumor cells were injected subcutaneously, and mice were randomly assigned to treatment with CCI-779 at different doses when tumors reached 200 mm 3 in size. Mice were treated with CCI-779 each day for five consecutive days, and at 8 and 12 days after initiation of treatment tumor growth was assessed. As shown in Fig. 1C, CCI-779 inhibited growth of all the cell lines in a dose-dependent fashion. However, the low AKT cell lines were relatively more resistant (IC 50 values for U87 PTEN and LAPC-4 puro were 1 and 3 mg/kg, respectively, when assessed at day 12) then their high AKT counterparts (IC 50 values for U87 and LAPC-4 AKT were 0.07 and 0.9 mg/kg, respectively). These results are consistent with previous work (13) demonstrating the increased sensitivity of tumors containing high AKT activity to CCI-779 both in vitro and in vivo.
Prior work with PTEN Ϫ/Ϫ and PTEN ϩ/ϩ mouse embryo fibroblasts demonstrated that AKT-dependent differential sensitivity to CCI-779 was not due to differences in the ability of the drug to inhibit the mTOR pathway (13). This was also true in our paired prostate and glioblastoma cell lines. p70 and 4E-BP1 phosphorylation were completely abrogated at 10 nM of CCI-779 in all four cell lines and the ED 50 values did not differ between the high AKT and low AKT lines.
AKT Activity Regulates the Translational State and Transcriptional Response of Specific mRNAs to mTOR Inhibition-By inhibiting p70 and 4E-BP1 phosphorylation, mTOR inhibitors prevent translation. Thus, another possible downstream determinant of sensitivity is differential effects on the translation of specific critical mRNAs. To investigate this hypothesis we utilized a high-throughput methodology whereby microarray analysis of mRNA translational state can be assessed. This technique is based on the observation that well translated transcripts are typically associated with polysomes, whereas poorly translated mRNAs are monosomal. Thus, the two pairs of LAPC-4 and U87 cell lines were treated with or without the mTOR inhibitor, rapamycin, for 6, 24, or 72 h and extracts prepared for polysomal analysis. Polysomes were then separated from monosomal material on sucrose gradients and the associated RNAs were extracted from gradient fractions. Gradient fractions corresponding to monosomal (fractions 1-4) and polysomal material (fractions 5-11) were pooled, and the RNA from these two groups was reverse transcribed separately to generate fluorescently labeled probes and hybridized to microarrays. The mRNA translation state for a given transcript was defined as the ratio of polysomal to monosomal signal intensity. The differences between these ratios in rapamycintreated versus untreated cells gave an estimate of change in translational state (see "Experimental Procedures").
Transcripts for ϳ3000 genes (48%) from the Ͼ6100 genes assessed were reproducibly detected in the four cell lines. We were able to identify 181 transcripts whose translational state changed dramatically and consistently over the course of the experiment. We identified a subset of 107 mRNAs whose translation was markedly inhibited (Ͼ2.5-fold inhibition) in both pairs of cell lines, irrespective of AKT activity, at all time points tested. When we analyzed the rapamycin-induced decreases in mRNA translation in more detail, we found that they were time-dependent with modest inhibition after 6 h of exposure and greater inhibition at 24 and 48 h (not shown). Transcripts whose translation is known to be inhibited by rapamycin were also inhibited in our analysis (18). This subset included several ribosomal components and translation elongation factors. Furthermore, the recently reported inhibition of proteasomal subunit translation by rapamycin was also seen (19). In contrast, there were 74 genes whose translational states significantly increased (Ͼ2.5-fold increase in mRNA translational state) upon exposure to rapamycin in all the cell lines tested. Grolleau et al. (19) also reported that rapamycin increased the translation of a subset of mRNAs. This previously reported subset includes genes also identified in this analysis, such as cyclin F, Ets2 repressor factor, and Syk (see supplemental data). Fig. 2A demonstrates that the global changes in mRNA translational state induced by rapamycin was very similar in high AKT versus low AKT cell counterparts. The false-color images show the relative changes in translational state of the 181 genes (107 whose translation was inhibited and 74 whose translation was increased) clustered according to their expression profiles over time (6 -48 h) following rapamycin treatment. It is clear that the general pattern of mRNA translational state changes due to rapamycin is quite similar among the four cell lines irrespective of AKT activity. However, in contrast to these comparable global alterations, we identified two mRNAs, cyclin D1, and c-myc, whose translation was markedly differentially regulated. These mRNAs were of particular interest because the degree to which they were differentially translated was more marked than any other interrogated mRNA, and they are also known to have prominent roles in cell cycle transit. When all detectable mRNAs were ranked by greatest degree of rapamycin-induced decrease in translation state in the high AKT lines, cyclin D1 and c-myc were first and second, respectively. Similarly, when all mRNAs were ranked in order of the greatest degree of rapamycin-induced increase in translational state in the low AKT lines, again cyclin D1 and c-myc mRNAs were first and second (see supplemental data). Fig. 2B shows the prominence of the differential alterations in translational state of these two mRNAs very clearly. The data represent changes obtained following a 48 h exposure to rapamycin, although a similar pattern was present following 24  In addition to translational state, a measure of total steadystate level of any mRNA could be obtained from these microarrays by summing the signal intensities of the polysomal and monosomal fractions. By this analysis, rapamycin had significant effects on mRNA expression in a small subset (63 genes) of the detectable genes (greater that 2.5-fold change), inhibiting transcription in 34 genes (54%) and enhancing transcription in 29 genes (31%). Again, our analysis confirmed the published finding (20,21) that the transcripts of many ribosomal components are inhibited by rapamycin. In general, these effects on transcription were comparable between high AKT and low AKT paired cell lines. However, in similar fashion to translation, rapamycin differentially regulated cyclin D1 and c-myc transcription, inhibiting it in high AKT lines and enhancing it in low AKT lines. These major alterations in translation and transcription of cyclin D1 and c-myc mRNAs are summarized in Fig. 2C.
To confirm the results of the translational state microarray analysis, we performed Northern blot analysis on mRNAs associated with polysomes (well translated) versus monosomes (poorly translated) as separated by sucrose gradients. At the top of each series of Northern blots in Fig. 3 is shown the individual polysome profile obtained from each cell line during sucrose gradient fractionation. Northern blot analysis for cyclin D1, c-myc, and actin mRNA is shown below the profiles on total RNA isolated from the corresponding fractions of the sucrose gradient. Densitometric analysis of the signals obtained from the monosomal fractions (fractions 1-4) versus the polysomal fractions (fractions 5-11) allowed a calculation of percent mRNA found in polysome fractions (well translated) as shown to the right of the Northern blots. The differential regulation is very clear; in rapamycin-sensitive, high AKT cell lines (LAPC AKT or U87), rapamycin produces a marked shift in cyclin D1 and c-myc mRNA from polysome fractions (fractions 5-11) to monosome fractions (fractions 1-4), indicating a significant decrease in translational efficiency. In rapamycin-resistant, low AKT cell lines (LAPC4 puro or U87 PTEN ), rapamycin induces an opposite response, increasing the percentage of cyclin D1 and c-myc mRNA found in polysome fractions versus monosome fractions. For example, the percent of cyclin D1 message associated with polysomes in U87 cells decreases from 45 to 8% after treatment with rapamycin. Likewise, the percent of c-myc message in polysomal fractions decreases from 39 to 4%. In contrast, rapamycin increases the percent of cyclin D1 mRNA associated with polysomes in U87 PTEN cells from 27 to 67% and the percent of c-myc mRNA from 34 to 78%. As expected, in all cell lines regardless of AKT activity, rapamycin inhibits translation of actin (i.e. shift of actin mRNA from well translated polysomes to poorly translated monosomes). These data confirm the results of microarray analysis, namely that the level of AKT activity regulates the translational state response to rapamycin of cyclin D1 and c-myc.
The Northern blots further confirm the assessment of total steady-state mRNA transcription from the microarrays. The relative amounts were determined by densitometry on equally exposed autoradiographs and summing the signals from monosome and polysome fractions (data not shown). Although the total amount of c-myc and cyclin D1 mRNA in all fractions (monosome ϩ polysome fractions) is significantly decreased by rapamycin in LAPC4 AKT and U87 cell lines (decreased by 35.3% for c-myc and 27.2% for cyclin D1 in LAPC-4 AKT ; decreased by 25.7% for c-myc and 31.3% for cyclin D1 in U87) the total c-myc and cyclin D1 mRNA is increased by rapamycin in resistant LAPC-4 puro and U87 PTEN cell lines (increased by 27.4% for c-myc and 32.8% for cyclin D1 in LAPC-4 puro ; increased by 31.7% for c-myc and 27.6% for cyclin D1 in U87 PTEN ).
Western analysis of protein extracts obtained from the high AKT and low AKT cell lines prior to and following rapamycin exposure also demonstrated differential expression (Fig. 4A). Both cyclin D1 and c-myc protein levels decreased or became undetectable in the rapamycin-sensitive cell lines following exposure, whereas, in the resistant cell lines, the relative levels of both proteins increased upon rapamycin treatment. By densitometric analysis, cyclin D1 levels increased 3.5-fold in LAPC-4 puro cells (3.5 Ϯ 0.7, mean Ϯ S.D. of three experiments) and 5.1-fold in U87 PTEN cells. C-myc levels increased 6.2-fold in LAPC-4 puro cells (mean Ϯ S.D. of three experiments) and 5.2fold in U87 PTEN cells.
To examine whether in vivo treatment with CCI-779 induced AKT-dependent differential effects on cyclin D1 or c-myc protein expression, we performed Western analysis on tumor material removed from mice following treatment with the drug for 5 days. As shown, in vivo administration of CCI-779 resulted in a dose dependent decrease in both cyclin D1 and c-myc protein levels in "sensitive" U87 cells, (Fig. 4B, left panel). In "resistant" U87 PTEN cells, cyclin D1 and c-myc levels increased and accumulated in a dose dependent manner (by densitometry, cyclin D1 levels increased 3.5-fold and c-myc levels increased 2.5-fold following treatment with 4 mg/kg; Fig. 4B, right panel). Similarly, cyclin D1 and c-myc expression decreased in sensi-tive LAPC-4 AKT cells (Fig. 4C, left panel), whereas accumulating in resistant LAPC-4 puro cells (Fig. 4C, right panel).
Aberrant Levels of Cyclin D1 and c-myc Are Critical Determinants of Rapamycin Sensitivity-To determine if the downregulated expression of cyclin D1 or c-myc was specifically responsible for sensitivity to rapamycin-induced G 1 arrest, we attempted to revert the rapamycin-sensitive phenotype in the U87 cell line model by restoring expression of these genes. However a potential obstacle to this strategy was the ability of rapamycin to inhibit the translation of any transfected genes in rapamycin-sensitive cells. However, we observed from our array analysis that rapamycin did not significantly alter the translation of the p27 kip1 mRNA in either the sensitive high AKT or resistant low AKT cell lines. This suggested to us that the p27 kip1 mRNA contained sequences capable of directing translation in the face of mTOR inhibition. Indeed, recent work has demonstrated that a 365 nucleotide sequence in the human p27 kip1 mRNA leader is capable of directing cap-independent translation (22). Furthermore, the highly conserved IRES within the murine p27 kip1 mRNA (overall 78% sequence identity) has also been shown to mediate cap-independent translation that is resistant to rapamycin (23). Thus, we generated a series of constructs in which we replaced the native 5Ј-UTR of cyclin D1 and c-myc with the leader from the p27 kip1 mRNA containing the IRES sequences. The transcription of cyclin D1 and c-myc IRES-containing genes was driven by a cytomegalovirus promoter. One of these constructs also expressed the EGFP gene downstream of these 5Ј-UTR sequences, which allowed us to specifically examine successfully transfected cells. Rapamycin-sensitive high AKT U87 cells were transiently transfected. To ensure that the p27 kip1 -IRES-containing cyclin D1 and c-myc mRNAs were being faithfully transcribed (termed p27-IRES-cyclinD1 and p27-IRES-myc), we also performed Northern analysis on EGFP sorted cells using probes specific for these transcripts and were able to confirm mRNA expression (data not shown). To test effects of rapamycin on translation of the transfected genes, we next performed flow cytometry for cyclin D1 or c-myc expression of EGFP-gated, transiently transfected cells treated with or without rapamycin. Shown in Fig. 5A are the percent cells positively stained for cyclin D1 or c-myc expression (above a threshold fluorescence obtained from an isotype matched control antibody) and the respective mean fluorescence intensities (in parentheses). As expected, rapamycin-sensitive U87 cells, when either untransfected or transfected with an empty vector, demonstrated a marked decrease in endogenous cyclin D1 and c-myc expression (percent positive cells and fluorescence intensity) when exposed to rapamycin (compare lines 1-5 and 2-6 in Fig. 5A).
However, U87 cells expressing the p27-IRES-cyclin D1 mRNA (lines 3 and 7) or the p27-IRES-myc mRNA (lines 4 and 8) demonstrated little effect on cyclin D1 or c-myc expression, respectively, following rapamycin treatment. The specificity of this effect is demonstrated by the fact that in these same transiently transfected cells, rapamycin was still capable of down-regulating expression of the other protein (i.e. c-myc expression in p27-IRES-cyclin D1-transfected cells and cyclin D1 expression in p27-IRES-myc-transfected cells). These data demonstrated that the p27 kip1 leader was capable of directing translation of either cyclin D1 or c-myc in sensitive U87 cells, which was resistant rapamycin.
With no inhibitory effect of rapamycin on ectopic expression of transfected cyclin D1 or c-myc, we were able to test the role of these genes in G 1 arrest. For these experiments, rapamycinsensitive U87 and resistant U87 PTEN cells were similarly transfected with the indicated constructs, treated with or without rapamycin, and cell cycle phase distribution monitored on EGFP-gated cells (Fig. 5B). As expected, when the sensitive U87 cell line (transfected with empty vector) was exposed to rapamycin, the percent of cells in S-phase decreased from 54% in the untreated group to 11% (bars 1 and 2, Fig. 5B). In U87 cells transfected with the p27-IRES-cyclin D1 (bars 3 and 4) or the p27-IRES-myc constructs (bars 5 and 6), G 1 arrest induced by rapamycin was less marked but still significant (p Ͻ 0.05). However, when U87 cells were co-transfected with both constructs (bars 7 and 8, Fig. 5B), the sensitive phenotype was completely reversed as there was no decrease in S-phase distribution. Of note, transfection of these constructs into U87 PTEN , rapamycin-resistant cells, did not result in any increase in S-phase distribution, supporting the notion that enforced cyclin D1/c-myc expression specifically subverts the inhibitory effect of rapamycin rather than simply enhancing S-phase distribution in general. DISCUSSION mTOR inhibitors have demonstrated considerable potential as anti-cancer agents. Thus, our efforts to identify, in an unbiased analysis, how sensitivity to these drugs is regulated by AKT may have future clinical significance. Our study is particularly significant as it utilized isogenic cell lines and the differences in sensitivity to G 1 arrest were not artificially induced by in vitro selection with increasing concentration of drug. The study clearly indicates that AKT activity determines the ability of mTOR inhibitors to specifically down-regulate both transcription and translation of cyclin D1 and c-myc. Such down-regulation was necessary for cell sensitivity.
The importance of cyclin D1 and c-myc on the effects of mTOR inhibitors has been demonstrated previously in other systems (24 -26) and more recently by Nelsen et al. (27). Of particular importance, Barbet et al. (26), also found that the yeast homolog of cyclin D1, CLN3 under control of the rapamycin-insensitive UBI4 5Ј-UTR suppressed rapamycin-induced G 1 arrest. Our study utilized a similar strategy to translate rapamycin-sensitive mRNAs following exposure to the drug.
Rapamycin and CCI-779 classically block protein translation via their inhibitory effects on ribosome biogenesis, initiation complex formation and cap-dependent translation. However, our data clearly demonstrate important effects on transcription as well. This is especially true for cyclin D1 and c-myc, two critical genes that determine sensitivity. These results are consistent with prior work (18 -20) that also documented effects of mTOR inhibitors on transcription. We cannot discern from our experiments whether the inhibitory effects on transcription are mechanistically linked to inhibitory effects on translation.
Although it is unclear how AKT activity might regulate the transcriptional response to mTOR inhibitors, prior literature allows us to speculate on the regulation of translation. Inhibition of mTOR prevents phosphorylation of the p70S6 kinase and 4E-BP1 translational repressor. The dephosphorylation of these mTOR substrates results in the inhibition of translation. Our data clearly demonstrate that the AKT-dependent differential alterations in cyclin D1 and c-myc translation are not due to different effects on p70 or 4E-BP1. Both proteins were dephosphorylated equally and with the same sensitivity to rapamycin in high AKT versus low AKT cell lines. Furthermore, the mRNA translational state array analysis indicates that the global downstream effects were also comparably present irrespective of AKT activity. How then can one explain differential effects on cyclin D1/c-myc translation? One possibility is that, in the face of mTOR inhibition, 4E-BP1 dephosphorylation, and diminished cap-dependent translation, some transcripts can be translated via internal ribosome entry sites (IRESes), and IRES-dependent translation might be regulated by AKT activity for particular mRNAs. The c-myc mRNA is well known to contain an IRES within its 5Ј-UTR (28 -30). The human cyclin-D1 leader is relatively long, has high CG-content, and is predicted to be highly structured in addition to having limited rRNA-complementarity (31). These properties are common to other mRNAs known to contain IRES sequences (32,33). It is possible that the cyclin D1 leader also contains se-FIG. 6. AKT activity dependent regulation of cyclin D1/c-myc expression. Inhibition of mTOR results in reduced cap-dependent translation of cyclin D1/c-myc and activation of cap-independent translation as part of the stress/starvation response. In a low AKT activity setting, cap-independent translation of cyclin D1/c-myc is maintained. In a high AKT setting, cap-independent translation of cyclin D1/c-myc is prevented resulting in G 1 arrest. quences conferring IRES activity. Thus, cyclin D1and c-myc mRNAs may be capable of sufficient IRES-dependent translation in low AKT cells but not in high-AKT cells. AKT may regulate this activity by its known influence on IRES transacting factors (34).
We initially hypothesized that rapamycin would have little inhibitory effect on the translation of critical proteins in the low AKT-resistant cells due to the activity of the salvage pathway mediated via cap-independent translation. However, translation was, in fact, significantly increased in these cells. We do not have an obvious answer explaining the mechanism by which cap-independent translation can be stimulated by rapamycin. However, it is known that stress/starvation can elicit a response in which the translation of many mRNAs involved in various metabolic pathways is selectively stimulated (19,35,36). It is possible that the activated translation in the stress/ starvation response is due to a broad stimulation of cap-independent translation that could possibly include the cyclin D1 and c-myc mRNAs.
When mice carrying LAPC-4 or U87 xenografts were treated with CCI-779, AKT activity similarly affected the response. Although the in vivo growth of the xenograft pairs was comparable in untreated mice, the high AKT cell lines demonstrated a 10-50-fold increased sensitivity to CCI-779 in terms of suppressed tumor growth. Also in similar fashion to our in vitro studies, expression of cyclin D1 and c-myc, in response to in vivo treatment with CCI-779 was differentially regulated in an AKT activity-dependent fashion. These data suggest that relative resistance to CCI-779 in vivo is also due to a maintenance of cyclin D1/c-myc expression in the face of mTOR inhibition. However, there is still significant CCI-779-induced cytoreduction at the higher doses in the low AKT-resistant cell lines. In addition to this, cytoreduction occurs in these tumors in association with a CCI-779-induced increase in cyclin D1 and c-myc expression (see Fig. 6, B and C, 4 mg/kg lanes). It is certainly possible that at these high doses of CCI-779, an anti-tumor effect involves other mechanisms besides G 1 arrest, which cannot be prevented by up-regulated cyclin D1 and c-myc expression.
In summary, our work indicates a role for AKT in the regulation of cyclin D1 and c-myc translation during mTOR inhibition. In this speculative scenario (Fig. 6), mTOR inhibition inhibits cap-dependent translation but activates cap-independent translation that is meditated by IRES function. This latter mechanism of protein expression may have developed for the maintenance of basic cell homeostasis and survival during the starvation response. In cells with relatively low AKT activity, this activation of mRNA-specific translation during mTOR inhibition serves as a salvage pathway, effectively maintaining cyclin D1/c-myc expression and allowing for continual call cycle transit. In contrast, high AKT activity prevents this salvage pathway and cyclin D1/c-myc protein levels plummet, resulting in G 1 arrest. This additional level of regulation of the starvation response afforded by the level of AKT activity is interesting. AKT would be excellent placed monitor of the basal proliferative/metabolic activity of a cell and could function as an appropriate rheostat whereby a highly proliferative state would require a greater AKT-regulated effort to shutdown cell cycle transit.
Our findings that alterations in cyclin D1/c-myc expression are critical determinants of sensitivity to mTOR inhibitors like rapamycin or CCI-779 may have clinical implications. First, sampling tumor tissue soon after initiation of therapy for cyclin D1/c-myc expression may provide prognostic information concerning the likelihood of an ensuing response. This may be more informative then pre-therapy assessment of phosphorylated/activated AKT in tumor tissue as the latter would be difficult to evaluate as a relative marker of sensitivity. Second, it might be more difficult to treat tumors that contain unregulated expression of cyclin D1 or c-myc such as some non-Hodgkin's lymphomas. The excessive transcription of these genes in some tumors may prevent significant G 1 arrest induced by the mTOR inhibitors.