Cyclin D1 and c-myc internal ribosome entry site (IRES)-dependent translation is regulated by AKT activity and enhanced by rapamycin through a p38 MAPK- and ERK-dependent pathway.

The macrolide antibiotic rapamycin inhibits the mammalian target of rapamycin protein (mTOR) kinase resulting in the global inhibition of cap-dependent protein synthesis, a blockade in ribosome component biosynthesis, and G1 cell cycle arrest. G1 arrest may occur by inhibiting the protein synthesis of critical factors required for cell cycle progression. Hypersensitivity to mTOR inhibitors has been demonstrated in cells having elevated levels of AKT kinase activity, whereas cells containing quiescent AKT activity are relatively resistant. Our previous data suggest that low AKT activity induces resistance by allowing continued cap-independent protein synthesis of cyclin D1 and c-Myc proteins. In support of this notion, the current study demonstrates that the human cyclin D1 mRNA 5' untranslated region contains an internal ribosome entry site (IRES) and that both this IRES and the c-myc IRES are negatively regulated by AKT activity. Furthermore, we show that cyclin D1 and c-myc IRES function is enhanced following exposure to rapamycin and requires both p38 MAPK and RAF/MEK/ERK signaling, as specific inhibitors of these pathways reduce IRES-mediated translation and protein levels under conditions of quiescent AKT activity. Thus, continued IRES-mediated translation initiation may permit cell cycle progression upon mTOR inactivation in cells in which AKT kinase activity is relatively low.

The macrolide antibiotic rapamycin inhibits the mammalian target of rapamycin protein (mTOR) kinase resulting in the global inhibition of cap-dependent protein synthesis, a blockade in ribosome component biosynthesis, and G 1 cell cycle arrest. G 1 arrest may occur by inhibiting the protein synthesis of critical factors required for cell cycle progression. Hypersensitivity to mTOR inhibitors has been demonstrated in cells having elevated levels of AKT kinase activity, whereas cells containing quiescent AKT activity are relatively resistant. Our previous data suggest that low AKT activity induces resistance by allowing continued cap-independent protein synthesis of cyclin D1 and c-Myc proteins. In support of this notion, the current study demonstrates that the human cyclin D1 mRNA 5 untranslated region contains an internal ribosome entry site (IRES) and that both this IRES and the c-myc IRES are negatively regulated by AKT activity. Furthermore, we show that cyclin D1 and c-myc IRES function is enhanced following exposure to rapamycin and requires both p38 MAPK and RAF/MEK/ERK signaling, as specific inhibitors of these pathways reduce IRES-mediated translation and protein levels under conditions of quiescent AKT activity. Thus, continued IRES-mediated translation initiation may permit cell cycle progression upon mTOR inactivation in cells in which AKT kinase activity is relatively low.
The global regulation of cap-dependent translation is mediated via the mTOR 1 signaling cascade (1)(2)(3). Activation of mTOR results in phosphorylation of the p70 S6 kinase and the translation repressor 4E-BP1, allowing the formation of func-tional eIF-4F complexes resulting in cap-dependent mRNA translation initiation and ribosomal component biogenesis (4,5). The efficiency with which a mRNA can initiate cap-dependent translation is a function of the length and degree of secondary structure present within the 5Ј-UTR as well as the sequence context of the initiation codon (6). Most eukaryotic mRNAs contain 5Ј-UTRs with relatively short and unstructured 5Ј-UTRs (Ͻ100 nucleotides), which allow efficient capdependent ribosomal scanning (6). However, some key regulators of cell proliferation and apoptosis have leaders that are quite long, highly structured, and contain many upstream AUG or CUG codons and, as a result, are inhibitory to scanning ribosomes (7). Translation initiation in a number of these mRNAs is achieved via IRES-mediated mechanisms (8). Protein synthesis via this alternative form of initiation is typically favored under conditions when the default cap-dependent pathway is inhibited (9 -12).
The ability of AKT to regulate cap-dependent initiation is mediated via its inhibitory effects on the mTOR inhibitor complex TSC1/TSC2 (13)(14)(15). A direct linkage between AKT and the mTOR kinase has also been described. AKT can phosphorylate mTOR, and studies in Drosophila have demonstrated that dTOR (Drosophila target of rapamycin) is downstream and epistatic to the phosphatidylinositol 3-kinase/AKT pathway (5,16,17). However, AKT has recently been shown to negatively regulate the translation of the Elk-1 and Sap1a mRNAs independently of mTOR (18), and our prior work (19) also suggests a role for AKT in the regulation of cap-independent translation.
Previously, we identified many mRNAs the translation of which was unaffected or induced under conditions of mTOR inhibition following rapamycin treatment (19). Many of these transcripts remained on actively translated polysomes or shifted from monosomal to polysomal translational states following the global inhibition of cap-dependent translation. Interestingly, two of these mRNAs demonstrated remarkable differential translational states depending on the AKT activity status of the cell. In cells with relatively active AKT, the cyclin D1 and c-myc mRNAs were translationally repressed by mTOR inhibition, whereas in cells containing quiescent AKT, these transcripts were well translated and found in polysome structures following treatment with the drug. In this report we demonstrate that the human cyclin D1 mRNA can mediate internal translation initiation. Furthermore, we demonstrate that both cyclin D1 and c-myc IRES function is stimulated following rapamycin exposure in cells with quiescent AKT activity but not in cells with activated AKT. Lastly, we show that differential AKT-dependent cyclin D1 and c-myc IRES activity is dependent on p38 MAPK and RAF/MEK/ERK signaling.

EXPERIMENTAL PROCEDURES
Cell Lines and Plasmids-The U87, U87 PTEN , LAPC-4 puro , and LAPC-4 AKT cell lines have been described previously (kind gifts of I. Mellinghoff and C. Sawyers, UCLA) (20). The murine embryonic fibroblasts (MEFs) in which PTEN was deficient as well as the parental control have also been described previously (21). The cell lines were maintained in media supplemented with 10% fetal calf serum. We validated the expression of wild-type cyclin D1 and c-myc mRNAs in all of these cell lines by Northern analysis and sequencing of the 5Ј-UTRs of these transcripts, which were identical to previously published sequences (22). The parental construct utilized in these studies was pRF (kind gift of A. Willis, University of Leicester, UK) (23). The 5Ј-UTR of the human cyclin D1 mRNA (GenBank TM accession number NM053056) was amplified from total RNA. Subsequently it was inserted into the intercistronic region of pRF to generate pRCND1F. The 396-nucleotide c-myc (23) and 365-nucleotide p27 Kip1 IRESes (24) were amplified from IMAGE clones 4667496 and 4298338, respectively, and also cloned into the intercistronic region of pRF to generate pRmycF and pRp27F. The inserts were also cloned into a promoterless version of pRF, pRF(Ϫp) (kindly provided by J-T. Zhang, Indiana University). All constructs were confirmed by sequencing.
Sequence Analysis and Secondary Structure Predictions-BESTFIT (GCG Wisconsin Package TM , Accelrys) was used to compare the cyclin D1 leader to the 18 S rRNA sequences (GenBank TM accession number X03205). We used the MFOLD web server to predict secondary structures for the human cyclin D1 leader using the default settings and the temperature fixed at 37°C (25).
Transient Transfection Analysis of Dicistronic mRNA Reporters-The reporter constructs were transfected into cells using Lipofectamine Plus (Invitrogen) and normalized for transfection efficiency by co-transfection with pSV␤Gal (Promega). Cells were harvested 24 h following transfection and Renilla, firefly luciferase, and ␤-galactosidase activities were determined (Dual-Glo luciferase and ␤-galactosidase assay systems, Promega). Luminescence of extracts was determined using a microplate luminometer (Turner BioSystems, Sunnyvale, CA). Northern analysis was performed as described previously (19), utilizing a PCR probe specific for sequences within the firefly luciferase open reading frame. In experiments in which rapamycin (Calbiochem) was used, cells were transfected and subsequently exposed to 10 nM rapamycin for 18 h after which extracts were prepared and luciferase and ␤-galactosidase activities were determined.
In Vitro Translation of Dicistronic mRNA Reporters-The dicistronic plasmids were linearized using BamHI and capped RNA transcribed in vitro (mMessage T7 transcription kit, Ambion). Capped RNA transcripts were used to program extracts of the indicated cell lines as described previously (26). The translation reactions were performed with either the cap analog m 7 GpppG (Ambion) or GTP as indicated.
Western Blotting and in Vitro Kinase Assays-Western blotting was performed as described previously (19). For in vitro kinase assays, MEFs treated with SB203580 (25 M) or PD98059 (25 M) for predetermined time points were harvested and lysed in a buffer containing 20 mM HEPES, pH 7.5, 130 mM NaCl, 25 mM ␤-glycerolphosphate, 2 mM NaPP i , 2 mM EDTA, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 2.5 g of pepstatin, 2.5 g of leupeptin, 2.5 g of antipain, 0.5 mM dithiothreitol, and 1% Triton X-100. Cleared supernatants were subsequently incubated with either anti-p38 or anti-ERK antibody and protein A-Sepharose overnight. Pellets were washed three times in lysis buffer and once in kinase buffer containing 25 mM HEPES, pH 7.4, 25 mM ␤-glycerolphosphate; 25 mM MgCl 2 ; 0.5 mM EDTA, and 0.5 mM dithiothreitol. P38 kinase reactions were performed in the presence of 100 M ATP and 2 g of ATF-2 as a substrate. Phosphorylation of ATF-2 at Thr-71 was measured by Western blot using a phosphospecific ATF-2 antibody. ERK kinase reactions were performed using ELK-1 as a substrate, and phosphorylation of ELK-1 was detected by immunoblotting with a phospho-ELK-1 (Ser-383) antibody. P38 kinase reactions were performed in the presence of 100 M ATP and 2 g of ATF-2 as a substrate. Phosphorylation of ATF-2 at Thr-71 was measured by Western blot using a phospho-specific ATF-2 antibody. ERK kinase reactions were performed using ELK-1 as a substrate, and phosphorylation of ELK-1 was detected by immunoblotting with a phospho-ELK-1 (Ser-383) antibody. P38 MAPK, ERK, phosphor-ATF-2, and phosphor-ELK-1 antibodies were from Cell Signaling. Statistical analysis was performed by using Student's t test and analysis of variance models using Sigma Stat 3.0 (Jandel Scientific).

The Cyclin D1 and c-myc Leaders Function as IRESes in
Transfected Cells in an AKT-dependent Manner-The glioblastoma U87MG cell line has a PTEN-null mutation with resulting heightened AKT activity. It was stably transfected with a wild-type PTEN construct, which markedly down-regulated AKT activity (20). Similarly, the LAPC-4 puro prostate cancer cell line with relatively quiescent AKT activity was stably transfected with a myristoylated AKT allele (or empty vector control). The differential AKT/mTOR cascade activation of these paired isogenic lines has already been described (19,20). In addition, for this study, we also utilized MEFs in which the PTEN gene had been disrupted (21). The AKT activity in these MEFs is markedly higher as compared with PTEN ϩ/ϩ MEFs (21,27).
To determine whether the human cyclin D1 leader could internally initiate translation, it was cloned into the intercistronic region of a dicistronic reporter mRNA and tested in transiently transfected cells. The c-myc leader, containing a known IRES (28), was also cloned into this region. The dicistronic mRNAs used in these studies contain the Renilla and firefly luciferases as the first and second cistrons, respectively (pRF). Because we had previously observed a differential pattern of cyclin D1 and c-myc polysome association in quiescent as compared with activated AKT-containing cells (19), we investigated the ability of the cyclin D1 and c-myc leaders to initiate translation internally in transiently transfected cell lines containing differential AKT activities. We transiently transfected these cell lines with the indicated dicistronic constructs shown in Fig. 1A. Expression of the downstream firefly luciferase in the constructs containing the cyclin D1 5Ј-UTR or the c-myc IRES as compared with pRF (empty vector) was enhanced dramatically in those cell lines containing quiescent AKT activity (U87 PTEN , LAPC-4 puro , and PTEN ϩ/ϩ MEF in Fig. 1, B, C, and D, respectively). For example, in the U87MG PTEN , LAPC-4 puro , and PTEN ϩ/ϩ MEF cell lines the presence of the 5Ј-UTR of cyclin D1 or the c-myc IRES resulted in an ϳ3-4-fold increase in firefly luciferase activity. In contrast, firefly luciferase activity was minimally affected in the relatively "active AKT" member cell lines of these isogenic pairs, indicating that AKT activity prevented cyclin D1 and c-myc IRES function. Only a modest ϳ1.5-2-fold increase was seen in myc-IRES activity, and there was no increase in cyclin D1 IRES activity. The presence of the human p27 Kip1 IREScontaining sequences (24,29) in the dicistronic reporter construct resulted in an ϳ10-fold increase in firefly luciferase activity in all of the cell lines tested irrespective of AKT activity. This was consistent with our earlier observations that the p27 Kip1 mRNA was well translated in the face of mTOR inhibition regardless of the AKT status of the cell (10) and indicates that p27 Kip1 IRES function is independent of AKT activity. It also provides a control confirming that reporter expression will occur in the "high-AKT" cell lines if an IRES is functional.
To evaluate whether the AKT-dependent enhanced translation of the downstream cistron in the cell lines tested was the result of initiation from shorter monocistronic transcripts or possibly from cryptic promoter activity, we analyzed the dicistronic mRNAs via Northern blot and luciferase activities in constructs where the SV40 promoter was absent (pRF(Ϫp), pRCD1F(Ϫp), and pRmycF(Ϫp)) (30). Fig. 2A shows a schematic diagram of the promoterless dicistronic constructs transfected into PTEN Ϫ/Ϫ and PTEN ϩ/ϩ MEFs. As shown in Fig. 2B, introduction of the promoterless constructs resulted in minimal luciferase activities in both the PTEN Ϫ/Ϫ and PTEN ϩ/ϩ MEFs, indicating that the AKT-dependent firefly luciferase expression from these constructs was not the result of internal promoter activities. Fur-thermore, Northern analysis of mRNAs from transfected PTEN Ϫ/Ϫ or PTEN ϩ/ϩ MEFs detected only the presence of the full-length dicistronic transcripts when probed for sequences within the downstream firefly luciferase open reading frame (Fig.  2C). The 5Ј-UTR of cyclin D1 is 208 nucleotides in length, whereas the c-myc IRES sequences are 396 nucleotides in length. These data further supported the notion that the cyclin D1 5Ј-UTR was capable of internal initiation and that both the IRES activities of the 5Ј-UTR of cyclin and the c-myc IRES were regulated by AKT activity.
The Cyclin D1 5Ј-UTR and c-myc IRES Mediate Internal Initiation in an AKT-dependent Fashion in Cell-free Extracts-To rule out that firefly luciferase reporter expression could be due to unusual cryptic splicing events, we analyzed the AKT-dependent IRES activities of the cyclin D1 and c-myc mRNAs in cell-free extracts. These extracts were prepared according to Carroll and Lucas-Lenard (26) and demonstrated to have high efficiencies in initiating protein synthesis. Capped dicistronic mRNAs that either lacked (pRF) or contained the cyclin D1 (pRCND1F) 5Ј-UTR, the c-myc IRES (pRmycF), or the p27 Kip1 IRES (pRp27F) were in vitro transcribed and subsequently used to program translation in lysates from U87MG, U87MG PTEN , LAPC-4 puro , LAPC-4 myrAKT , PTEN ϩ/ϩ MEFs, or PTEN Ϫ/Ϫ MEFs. Translation of the parent pRF mRNA yielded firefly luciferase activities, which were indistinguishable from the background obtained from control reaction mixtures that lacked the firefly luciferase reporter mRNAs. In contrast, as shown in Fig. 3, an equivalent amount of either pRCND1F or pRmycF mRNAs generated firefly luciferase activities, which were ϳ2-4-fold higher in extracts prepared from cell lines with relatively quiescent AKT levels. The firefly luciferase activities generated from the pRp27F mRNA were consistently ϳ6-fold higher in all cell extracts tested as compared with pRF mRNA. This again was consistent with the results from the dicistronic in vivo experiments, demonstrating that the IRES-dependent translation of p27 Kip1 was independent of AKT activity.
To evaluate whether the translation of the cyclin 5Ј-UTR and the c-myc IRES in these reporter mRNAs was indeed cap-independent within these cell extracts, in vitro transcribed and capped mRNAs from pRCND1F and pRmycF were translated in the presence of increasing concentrations of the cap analog m 7 GTP. This analog blocks cap-dependent translation by binding to the initiation factor eIF-4E (31). Using mRNA in vitro transcribed from pRF and capped, translation of the Renilla luciferase cistron was blocked by ϳ90 -95% at 150 M of m 7 GpppG but was not affected by comparable concentrations of the non-methylated form of the analog GTP, and the translation of the firefly cistron was unaffected by m 7 GpppG (data not shown). As shown in Fig. 4, with both the in vitro transcribed and capped mRNAs from pRCND1F and pRmycF, the translation of the Renilla cistron was inhibited by ϳ80 -90% at concentrations of 100 M m 7 GpppG or higher in all cell extracts tested. However, the translation of the firefly cistron remained consistent in the extracts with relatively high levels of active AKT and increased ϳ1.5-2.5-fold in extracts from the cell lines with relatively quiescent AKT levels. This again supported the notion that sequences present within the 5Ј-UTR of cyclin D1 mRNA could confer cap-independent initiation and that both this sequence and the c-myc IRES activities were enhanced under conditions of low AKT activity.
AKT-dependent Cyclin D1 5Ј-UTR and c-myc IRES Activity Is Enhanced by Rapamycin-Because our previous studies (19) demonstrated a stimulation of cyclin D1 and c-myc polysome association and protein levels by rapamycin under conditions of quiescent AKT activity, we assessed whether cyclin D1 or c-myc IRES activity would also be enhanced following rapamycin exposure in cells with quiescent AKT activity. To address the AKT-dependent affects of rapamycin on cyclin D1 or c-myc IRES function, we transfected our dicistronic constructs into the cell lines shown in Fig. 5 and determined Renilla and firefly luciferase activities prior to and following rapamycin exposure. Renilla luciferase activity was reduced by rapamycin ϳ70 -90% as compared with values obtained in the absence of the drug for each construct tested (data not shown). However, the results show that in the cell lines with relatively quiescent AKT (U87 PTEN , LAPC-4 puro , and PTEN ϩ/ϩ MEFs), rapamycin treatment resulted in an ϳ3-6-fold stimulation of firefly luciferase activity as compared with values obtained for lines containing active AKT. Interestingly, rapamycin exposure also stimulated p27 Kip1 IRES activity ϳ4 -6-fold in all cell lines irrespective of AKT activity, again consistent with our previous data and results from others demonstrating the resistance of p27 IRES Kip1 activity following exposure to the phosphatidylinositol 3-kinase inhibitor LY294002 (24). Northern blot analysis further demonstrated that rapamycin had no effect on pRF steady-state mRNA levels prior to and following rapamycin treatment (data not shown).

Differential AKT-dependent Cyclin D1 and c-myc IRES Activity Is Dependent on Both p38 Mitogen-activated Protein Kinase and RAF/MEK/ERK Signaling-Because it has been
demonstrated previously that c-myc IRES function is dependent on p38 MAPK activity during apoptosis (28) and both p38 MAPK and ERK signaling following genotoxic stress (32), we investigated whether these cascades also contributed to the differential AKT-dependent cyclin D1 and c-myc IRES activity we had observed. An additional rationale was the known ability of AKT to down-regulate p38 (33) and ERK (34) activity. To determine whether differential p38 or ERK signaling could be correlated with AKT-dependent cyclin D1 or c-myc IRES activ- ity, we initially examined the activities of these kinases in the PTEN Ϫ/Ϫ and PTEN ϩ/ϩ MEFs prior to and following rapamycin exposure. As shown in Fig. 6A, the basal p38 and ERK kinase activities as determined by in vitro kinase assays was ϳ3-4-fold higher in the PTEN ϩ/ϩ MEFs as compared with the PTEN Ϫ/Ϫ MEFs. Additionally, treatment with rapamycin resulted in an 8-fold induction of p38 and 10-fold induction of ERK activity in PTEN ϩ/ϩ MEFs while only modestly increasing basal p38 and ERK activities (ϳ1-2-fold) in the PTEN Ϫ/Ϫ MEFs. Total p38 and ERK content in the samples was similar, demonstrating that equivalent amounts of material were immunoprecipitated (Fig. 6A). These data suggest that p38 and ERK signaling is activated by rapamycin exposure in an AKTdependent manner and is consistent with the known negative regulatory effects of AKT on p38 and on ERK (33,34).
To investigate whether AKT-dependent cyclin D1 and c-myc IRES activity was regulated by p38 or ERK signaling we planned to transfect PTEN ϩ/ϩ and PTEN Ϫ/Ϫ MEFs with the indicated dicistronic constructs in Fig. 6B and subsequently to treat these cells with the p38 inhibitor SB203580 or the ERK inhibitor PD98058. Our preliminary experiments demonstrated almost complete inhibition of kinase activity using these inhibitors (data not shown). Treatment of either PTEN ϩ/ϩ or PTEN Ϫ/Ϫ MEFs with SB203580 inhibited basal p38 kinase activity by more than 95% within 2 h of treatment at a concentration of 25 M. Similarly, treatment of the MEFs with PD98058 inhibited ERK activity by more than 92% within 2 h of treatment at 25 M.
These inhibiting concentrations of SB203580 or PD98059 were found to significantly affect AKT-dependent IRES function. The experiments shown in Fig. 6, B and C, are assays performed in the absence or presence of rapamycin, respectively. As shown in Fig. 6B, PTEN Ϫ/Ϫ MEFs expressed relatively little firefly luciferase activity when transfected with the dicistronic constructs containing the cyclin D1 5Ј-UTR or the c-myc IRES within the intercistronic regions. However, as previously observed, firefly luciferase activity was markedly increased (ϳ3-4-fold) in the PTEN ϩ/ϩ MEFs transfected with pRCND1F or pRmycF. Treatment of these cells with either SB203580 or PD98059 resulted in more than 75% inhibition of firefly luciferase activity. The p38 inhibitor resulted in a modest decrease in Renilla luciferase activity (ϳ5-10% inhibition), whereas the ERK inhibitor reduced Renilla luciferase expression by 60 -65% in all of the constructs tested at the concentrations used in these experiments (data not shown). Interestingly, the p27 Kip1 IRES was unaffected by treatment with either of the inhibitors and did not demonstrate differential AKT-dependent activity as shown before (Fig. 1).
To address whether these signaling cascades contributed to the rapamycin-induced differential cyclin D1 and c-myc IRES activity we had previously observed, we performed the same assays in MEFs pretreated with SB203580 or PD98058, which had been transiently transfected with the indicated dicistronic constructs shown in Fig. 6C. As shown in the relatively quiescent AKT-containing PTEN ϩ/ϩ MEFs, rapamycin induced firefly luciferase expression by ϳ4-fold relative to control experiments without the drug. This enhancement of IRES activity was markedly inhibited by pretreatment with either SB203580 or PD98059. 1 h of preincubation with either of these inhibitors resulted in greater than 80% inhibition of the rapamycin-induced firefly luciferase expression in PTEN ϩ/ϩ MEFs. Renilla luciferase expression was reduced by rapamycin (ϳ65% inhibition) in all of the constructs tested, and pretreatment with SB203580 or PD98059 in combination with rapamycin did not significantly reduce Renilla luciferase expression further (data not shown). As before, p27 Kip1 IRES activity was enhanced by rapamycin irrespective of AKT activity but was not affected by inhibition of p38 or ERK signaling.
Differential cyclin D1 and c-Myc protein levels were also assessed in the PTEN ϩ/ϩ and PTEN Ϫ/Ϫ MEFs treated with rapamycin alone and in combination with either the p38 or ERK inhibitors. As shown in Fig. 7, treatment of PTEN Ϫ/Ϫ MEFs with rapamycin resulted in down-regulation of cyclin D1 and c-myc expression; however, PTEN ϩ/ϩ MEFs maintain or modestly increase expression in response to the drug. This differential response in cyclin D1 and c-myc expression is ablated by pretreatment of the cells with either the p38 or ERK inhibitors prior to exposure to rapamycin. As determined by densitometry, pretreatment with SB203580 inhibited cyclin D1 protein levels in rapamycin-treated PTEN ϩ/ϩ MEFs by 7.5fold (Fig. 7, compare lane 7 with 8), whereas pretreatment with the ERK inhibitor PD98059 reduced cyclin D1 expression by 9-fold in these cells (lanes 11 and 12). Pretreatment with either SB203580 or PD98059 reduced c-Myc protein expression to below detectable levels in quiescent AKT-containing PTEN ϩ/ϩ MEFs upon rapamycin exposure (Fig. 7, lanes 7 and 8, lanes 11  and 12).
Sequence Analysis and Secondary Structure Prediction of the Cyclin D1 5Ј-UTR-The 5Ј-UTRs of the major human cyclin D1 and c-myc mRNAs consist of 209 and 400 nucleotides, respectively. Both leader sequences contain the hallmarks of 5Ј-UTRs from other mRNAs demonstrated to exhibit IRES activity. Both leaders are relatively long, highly structured, and contain upstream AUG or CUG initiation codons (22,35). Although a model of the c-myc IRES structure has been described (36), the leader of the human cyclin D1 mRNA has not been characterized in this regard.
Some mRNAs capable of internal translation initiation have been shown to contain sequence complementarity to 18 S ribosomal RNAs (37,38). It has been proposed that these regions of complementarity may serve as cis-acting elements involved in the direct recruitment of ribosomal 40 S subunits to mRNAs and possibly regulate cap-independent translation (39). To address whether the cyclin D1 leader contained regions of sequence complementarity to 18 S ribosomal RNAs, we performed sequence comparisons. Comparisons of 18 S ribosomal RNAs and the cyclin D1 5Ј-UTR identified several complementary sequence matches (see supplemental data). Seven of these regions ranged from 80 to 94% similarity to 18 S rRNA over 11-22 nucleotides for the cyclin D1 leader.
A secondary structural model of the cyclin D1 5Ј-UTR was derived by free energy calculations using the MFOLD algorithm (40). 13 structures were calculated, which ranged in initial free energies (dG) from Ϫ71.5 to Ϫ88.9 kcal/mol. The most stable predicted structure is shown as supplemental data. The structure is highly complex with several long stems, junctions, and higher order bifurcations. Taken together, these data support the notion that the cyclin D1 5Ј-UTR contains a bona fide IRES element. DISCUSSION Our previous studies (19) suggested that the cyclin D1 and c-myc mRNAs are transcripts that could be effectively translated under conditions of reduced cap-dependent initiation. In this report we have demonstrated that under specific signaling conditions, the leader of the human cyclin D1 mRNA can ini-FIG. 6. AKT-dependent cyclin D1 and c-myc IRES activity requires p38 MAPK and ERK signaling. A, p38 MAPK and ERK activities in PTEN Ϫ/Ϫ and PTEN ϩ/ϩ MEFs prior to and following rapamycin exposure. In vitro kinase reactions were immunoblotted for the indicated phosphorylated substrate and total p38 or ERK protein levels. B, PTEN Ϫ/Ϫ and PTEN ϩ/ϩ MEFs were transfected with the indicated dicistronic constructs and treated with either SB203580 or PD98059. Changes in firefly luciferase (cap-independent) expression are shown. Values were normalized to pRF without treatment and were performed in triplicate. C, PTEN Ϫ/Ϫ and PTEN ϩ/ϩ MEFs were transfected with the indicated dicistronic constructs, pretreated with either SB203580 or PD98059 for 1 h, and subsequently exposed to rapamycin (rapa) as shown. Relative -fold changes in firefly (cap-independent) luciferase activities are shown as compared with activities obtained in the absence of rapamycin. Data are representative of three independent experiments. tiate protein synthesis via an IRES. Although the cyclin encoded by the Kaposi sarcoma-associated herpes virus has also been reported to contain an IRES (41); to our knowledge, this is the first report of the ability of this mRNA to initiate translation internally. We have also shown that AKT activity regulates cyclin D1 and c-myc IRES function and demonstrated that rapamycin increases cyclin D1 and c-myc IRES activity in an AKT-dependent manner. These results are consistent with our previously observed AKT-dependent effects on cyclin D1 and c-Myc protein synthesis following rapamycin exposure. Furthermore, we have extended our studies by implicating the p38 MAPK and RAF/MEK/ERK signaling cascades in the regulation of AKT-dependent cyclin D1 and c-myc IRES activity. Our results support a working model in which the AKT-dependent control of cyclin D1 and c-myc IRES function in response to rapamycin may regulate the expression of these critical determinants resulting in either G 1 arrest or tumor cell survival. When AKT activity is relatively low, rapamycin treatment results in the inhibition of cap-dependent translation but stimulates the selective translation of cyclin D1 and c-myc via their IRESes mediated via p38 MAPK and RAF/MEK/ERK signaling, thus maintaining expression. However, when AKT is elevated, the rapamycin-induced inhibition of cap-dependent translation is not associated with enhanced IRES function, most likely because of the negative regulatory affects of AKT activity on the p38 MAPK and RAF/MEK/ERK pathways; thus, cap-independent translation is prevented, and protein levels fall. This differential regulation of cap-independent translation and overall cyclin D1/c-myc expression accounts for the differential sensitivity of "high versus low" AKT activity cell targets.
An interesting question arises; under what circumstances might there be a requirement for IRES-mediated translation initiation of cyclin D1 mRNA, particularly when cyclin D1 expression has been shown to be dependent on eIF-4E (42,43)? Recent data suggest that cyclin D1 also normally accumulates during the G 2 phase of the cell cycle and that synthesis during this phase may contribute to the rapid achievement of the levels of cyclin D1 required for the ensuing G 1 transit in actively proliferating cells (44). It has also been recently appreciated that there is a reduction in cap-dependent protein synthesis during the G 2 /M cell cycle transition (45,46), and interestingly, it is known that both AKT activity and protein levels transiently drop during the G 2 /M transition (47). Although it has been demonstrated that post-translational mechanisms contribute to the accumulation of cyclin D1 during G 2 (48,49), it is also possible that the IRES-mediated synthesis of cyclin D1 normally occurs during this phase of the cell cycle and supplements expression.
Our data imply that the factor(s) responsible for AKT-dependent cyclin D1 and c-myc IRES function are downstream of p38 and ERK. This is consistent with the results of others who have demonstrated roles for these effectors in regulating the IRES-mediated synthesis of c-myc during apoptosis or in response to genotoxic agents (28,32). It is possible that these effectors regulate cyclin D1 and c-myc IRES activity directly or indirectly via phosphomodulation of an IRES trans-acting factor(s). Recently, three members of the poly(rC)-RNA binding family, PCBP1, PCBP2, and hnRNPK, have been shown to be required for c-myc IRES activity and stimulate IRES-mediated translation when overexpressed and bound to the mRNA (50). Moreover, it is known that the activity of these proteins is regulated by phosphorylation (51)(52)(53). Alternatively, p38 or ERK activity may lead to changes in IRES trans-acting factor expression thereby affecting IRES function. Along these lines it has been demonstrated that the expression of PCBP1 under hypoxic conditions is dependent on p38 activity in cortical neurons (54). Experiments designed to address these questions are currently in progress.
The observation that p27 Kip1 IRES activity was not AKT-dependent following rapamycin exposure is interesting and suggests that the regulation of this IRES is similar to but distinct from the cyclin D1 and c-myc IRESes in this setting. P27 Kip1 IRES function may be regulated by a specific IRES trans-acting factor(s) that enhances its function following rapamycin exposure but is nonresponsive to changes in p38, ERK, or AKT activities. The p27 kip1 IRES has been demonstrated to be active under conditions of elevated cyclic AMP (29) and repressed by the neuronal ELAV HuD (24), whereas enhancement of c-myc IRES activity has been shown to be dependent on PCBP1, PCBP2, and hnRNPK (50). It is certainly possible that the factors mediating cap-independent translational control of p27 kip1 are distinct from those regulating other cellular IRESes.
Our data also suggest that the ability of tumor cells to respond to mTOR inhibitors by stimulating cap-independent mechanisms of initiation of critical cell cycle proteins may constitute a mechanism of cellular resistance to these drugs. In particular, tumors that have relatively little dependence on the phosphatidylinositol 3-kinase/AKT/mTOR signaling cascade (i.e. low AKT activity) appear to markedly increase the capindependent synthesis of cyclin D1/c-myc following mTOR inhibitor exposure. Further understanding the mechanisms regulating the expression of these determinants may assist in the development of compounds that function in a synthetically lethal manner with mTOR inhibitors. FIG. 7. Effects of p38 MAPK and ERK inhibitors on AKT-dependent cyclin D1 and c-Myc protein expression following exposure to rapamycin. Cyclin D1, c-myc, and actin protein levels were determined in PTEN Ϫ/Ϫ or PTEN ϩ/ϩ MEFs subsequent to rapamycin treatment alone or in combination with SB203580 or PD98059 as indicated.