PKR and PKR-like Endoplasmic Reticulum Kinase Induce the Proteasome-dependent Degradation of Cyclin D1 via a Mechanism Requiring Eukaryotic Initiation Factor 2α Phosphorylation*

Cyclin D1 plays a critical role in controlling the G1/S transition via the regulation of cyclin-dependent kinase activity. Several studies have indicated that cyclin D1 translation is decreased upon activation of the eukaryotic initiation factor 2α (eIF2α) kinases. We examined the effect of activation of the eIF2α kinases PKR and PKR-like endoplasmic reticulum kinase (PERK) on cyclin D1 protein levels and translation and determined that cyclin D1 protein levels decrease upon the induction of PKR and PERK catalytic activity but that this decrease is not due to translation. Inhibition of the 26 S proteasome with MG132 rescued cyclin D1 protein levels, indicating that rather than inhibiting translation, PKR and PERK act to increase cyclin D1 degradation. Interestingly, this effect still requires eIF2α phosphorylation at serine 51, as cyclin D1 remains unaffected in cells containing a non-phosphorylatable form of the protein. This proteasome-dependent degradation of cyclin D1 requires an intact ubiquitination pathway, although the ubiquitination of cyclin D1 is not itself affected. Furthermore, this degradation is independent of phosphorylation of cyclin D1 at threonine 286, which is mediated by the glycogen synthase kinase 3β and mitogen-activated protein kinase pathways as described in previous studies. Our study reveals a novel functional cross-talk between eIF2α phosphorylation and the proteasomal degradation of cyclin D1 and that this degradation is dependent upon eIF2α phosphorylation during short, but not prolonged, periods of stress.

Originally cloned as a break point rearrangement in parathyroid adenoma (1), cyclin D1 serves as a key sensor and integrator of extracellular signals of cells in early to mid-G 1 phase (2). It primarily functions as the regulatory subunit of the holoenzymes that regulate progression through the G 1 phase of the cell cycle. By working in conjunction with the cyclin E-cyclin-dependent kinase 2 (CDK2) complex, cyclin D1-CDK complexes phosphorylate and inactivate the cellcycle inhibitory function of the retinoblastoma protein (2). Via this inhibition, cyclin D1 can release the repression of E2F transcriptional activity by retinoblastoma protein and the associated histone deacetylases and facilitate the transcription of genes active during the S phase of the cell cycle. Cyclin D1 has been demonstrated to have both cyclindependent kinase-dependent and -independent functions. Cyclin-dependent kinase (CDK)-dependent functions include the ability to regulate the function of CDKs 4 and 6, leading to the phosphorylation of retinoblastoma protein, and the indirect activation of the cyclin E-CDK2 complex, whereas cyclin D1 can promote the dissociation of histone deacetylase 1 from the retinoblastoma protein repressor complex in a CDK-independent manner. The ability of cyclin D1 to regulate histone acetylases and deacetylases as well as to directly interact with numerous transcription factors (peroxisome proliferator-activated receptor ␥ (3), C/EBP (4), Stat3 (5), and B-Myb (6) allows it to also function as a transcriptional co-regulator.
Transcription of the cyclin D1 gene is increased upon many different types of stimuli, including insulin-like growth factor-1 (IGF-1 (7)) and IGF-II (8), amino acids (9), androgens (10), and retinoic acid (11), depending on the cell type. Cyclin D1 gene expression is also induced by many oncogenic signaling pathways, including Ras (12), Src (13), Her2/neu (14), ␤-catenin (15,16), and members of the signal transducer and activator of transcription (STAT) family (17,18). At the translational level, cyclin D1 is regulated primarily through the phosphatidylinositol 3-kinase/Akt pathway. Serum stimulation (19) and the co-operation of fibroblast growth factor receptor-4 and ErbB2 (20) induce cyclin D1 translation through this pathway via activation of the mammalian target of rapamycin (mTOR) S6 kinase 1 (S6K1) (20). Another target of mTOR, the eukaryotic initiation factor 4E (eIF4E) 7 -binding protein (4E-BP1), also regulates the capdependent translation of cyclin D1 (21). eIF4E binds to the cap structure of mRNA, and this function is inhibited by 4E-BP1. A separate study indicated that the 5Ј-untranslated region of cyclin D1 contains an internal ribosome entry site that is negatively regulated by Akt activity (22) but enhanced after exposure to rapamycin in a manner dependent upon signaling through the p38 MAPK and RAF/MEK/ERK pathways.
Protein translation is a well regulated process that involves three general steps: initiation, elongation, and termination. The majority of translation regulation is exerted at the initiation step, a complex and intricate event involving a large number of different polypeptides. Initiation requires the joining of the 40 S and 60 S ribosomal subunits to form a functional 80 S ribosome, and the correct positioning of the ribosome at the start cordon (AUG) (23). The first step of initiation is the formation of the ternary complex, composed of eIF2-GTP-tRNA Met . This complex binds to the 40 S subunit, facilitated by the initiation factors eIF1, eIF1A, and eIF3, to form the 43 S pre-initiation complex. Binding of the 43 S subunit to mRNA is facilitated by the eIF4F complex, which binds to the 7-methylguanosine cap at the 5Ј end of the mRNA. Once loaded onto the mRNA, the 43 S complex scans the sequence until it reaches the start cordon, at which point eIF2-GTP is hydrolyzed to GDP, and the resultant eIF2-GDP complex dissociates from the translational machinery. Following this step, the 60 S subunit joins the 40 S via eIF5B to yield the fully functional 80 S ribosome. There are a number of different steps at which regulation of initiation can occur, one of the best characterized being the phosphorylation of the eIF2 on the ␣-subunit, which prevents recycling of GDP bound to eIF2 for GTP by eIF2B (23). As a result, eIF2 remains bound to eIF2B, and global mRNA translation is inhibited. Four eIF2␣ kinases have been identified to date that are each activated under different conditions but which all act to inhibit global protein synthesis, including the protein kinase activated by double-stranded RNA (PKR), which reacts to virus infection (Kaufman, 2000) (24), PKR-like ER kinase (PERK) activated by endoplasmic endoplasmic reticulum (ER) stress (25), hemeregulated inhibitor activated by heme deficiency (HRI) (26), and general control non-derepressible-2 (GCN2), which responds to amino acid starvation (27). A number of studies in the past have implicated eIF2␣ kinases in regulating cyclin D1 protein levels (28 -31) and concluded that the eIF2 kinase activity after ER stress causes a decrease in the translation of cyclin D1 mRNA but does not affect transcription of the cyclin D1 gene (30).
In the following study we demonstrate that activation of two different eIF2␣ kinases, PKR and PERK, cause a decrease in cyclin D1 protein levels and that this down-regulation requires eIF2␣ phosphorylation at serine 51. We further demonstrate using polysome profiles that it is not translation of the cyclin D1 message that is affected by activation of the kinases, as the cyclin D1 message is translated equally well regardless of the state of eIF2␣ kinase activity or the inhibition of global protein synthesis. We determine that PKR and PERK function to induce the proteasomal degradation of cyclin D1 independent of any previously described mechanism but still requiring cyclin D1 ubiquitination. Furthermore, our data also shows that eIF2␣ kinases themselves may activate different pathways to regulate the "early" and "late" degradation of cyclin D1. Our study reveals a novel ability of PKR and PERK to promote the proteasome-dependent degradation of cyclin D1 in a manner that depends on eIF2␣ phosphorylation.

EXPERIMENTAL PROCEDURES
Cell Culture and Treatments-HT1080 cells expressing GyrB.PKR wild type (WT) were established as described (32). HT1080 parental and GyrB.PKR WT cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% calf serum (Invitrogen) and 100 units/ml of penicillin-streptomycin (Wisent). eIF2␣ S/S and A/A immortalized mouse embryonic fibroblasts (MEFs) were generated and maintained as described (33). SV-40 immortalized PERK ϩ/ϩ and PERK Ϫ/Ϫ MEFs (34) as well at temperature-sensitive ts20 and control E36 cells (35) were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, and the latter was maintained at the permissive temperature of 31°C. GyrB.PKR WT cells were treated with 100 ng/ml coumermycin (Sigma) to activate the fusion protein. Thapsigargin (Sigma) treatment was performed at a concentration of 1 M, and MG132 treatment was performed at a concentration of either 10 or 20 M as indicated in the figure legends. Treatment with Sal003 (gift from Dr. J. Pelletier) occurred at a concentration of 75 M , 1-azakenpaullone (Calbiochem) at a concentration of 5 M and U0126 (BIOSOURCE) at 10 M. Cells were treated with dicoumarol (Sigma) at a concentration of 200 M. dsRNA transfection occurred at a concentration of 10 g/ml for 4 h as previously described (36).
Vaccinia Virus Transfection-WT and T286A mutant cyclin D1 constructs in pcDNA3 were kindly provided by Dr. B. Law. 3 ϫ 10 5 HT1080 GyrB.PKR cells were plated in 60-mm plates and infected with vaccinia virus/T7 as described in Li and Koromilas (37). Cells were co-transfected with 1 g of cyclin D1 WT or T286A vector DNA and 1 g of HA-tagged ubiquitin. 24 h post-transfection cells were treated with coumermycin and MG132 as indicated.
Sucrose Gradient Polysome Profiles-Polysome profiles were performed as previously described (38). Briefly, after the indicated treatments, 100 g/ml cycloheximide was added directly to growth media and then immediately removed, and the plates kept on ice. Plates were washed 3 times with ice-cold 1ϫ phosphate-buffered saline plus 100 g/ml cycloheximide and lysed directly on the plate. After a 10-min incubation on ice, samples were cleared by centrifugation at 13,000 rpm for 10 min, and the lysate was loaded onto 10 -55% sucrose gradients prepared on the Teledyne ISCO Density Gradient System. Gradients  RNA Isolation and Reverse Transcriptase (RT)-PCR-RNA was isolated from collected polysome fractions as follows. 600 l of Trizol reagent (Invitrogen) and 120 l of chloroform (Sigma) were added to each sucrose fraction and centrifuged, and the aqueous phase was removed to a separate tube. 600 l of isopropanol was added to each fraction, and RNA was precipitated overnight at Ϫ20°C. Samples were centrifuged at 13,000 rpm for 30 min, and the pellets were washed with icecold 75% ethanol and resuspended in 10 l of Milli-Q H 2 O. 3 l of RNA was used as a template for RT reactions with the AncT primer (5Ј-TTTTTTTTTTTTTTTTTTVN-3Ј). cDNA samples were diluted 10-fold, and 2 l of template used to carry out PCR reactions for CCND1 (forward human primer, 5Ј-CGC GCC CTC GGT GTC CTA CTT-3Ј; reverse human primer, 5Ј-ACG CTC CCC GCT GCC ACC AT-3Ј; forward mouse primer, 5Ј-CTG GAG GTC TGT GAG GAG CA-3Ј; reverse mouse primer, 5Ј-GCG GTA GCA GGA GAG GAA GTT) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH; forward primer, 5Ј-CAT CAT CTC TGC CCC CTC TGC T-3Ј; reverse primer, 5Ј-CAG CGG TCG T TC ACC ACC TTC T-3Ј). It should be noted that the fractions collected from these polysome profiles were also used in a previous publication by our laboratory studying the effect of eIF2␣ kinases on p53 (38). In that case RT-PCR was performed for different mRNAs.
[ 35 S]Methionine Labeling-HT1080 cells were either treated with MG132 for 1 h followed by coumermycin treatment for 3 h or treated with coumermycin for 3 h with MG132 added at the 2-h mark. Both treatments were performed in Dulbecco's modified Eagle's medium lacking methionine and supplemented with 10% dialyzed fetal bovine serum. Tran 35 S-label (ICN) was then added to the cells at a concentration of 100 Ci/10 6 cells, and the culture was continued for an additional 30 min in the presence of MG132 and coumermycin. Precipitation of radiolabeled proteins was carried out as follows; 250 g of total protein was aliquot into microcentrifuge tubes in triplicate. Trichloroacetic acid was added to each sample to a final concentration of 10% w/v and incubated on ice for 3 h. Samples were centrifuged at 13,000 rpm for 15 min, washed with 300 l of ice-cold acetone, and centrifuged again. Pellets were re-suspended in 50 l of 1ϫ phosphate-buffered saline plus 2% SDS and spotted on filter paper. CytoScint ESTM liquid scintillation fluid (MP Bio) was added to each sample, and the counts/min were determined.

Cyclin D1 Protein Levels Are Decreased upon Endoplasmic
Reticulum Stress-Previous reports have indicated that translation of cyclin D1 is affected by the induction of ER stress and the unfolded protein response (28 -30). To further investigate the role of eIF2␣ phosphorylation in regulating cyclin D1 in our system, we treated HT1080 cells with the pharmacological agent thapsigargin, which induces the UPR by promoting calcium release from the ER. Treatment of these cells with thapsigargin induced a decrease in cyclin D1 levels (Fig. 1A, panel a) coincident with increased eIF2␣ phosphorylation (Fig. 1A, panel b). To investigate the specific effect of eIF2␣ phosphorylation on cyclin D1 regulation, we used the compound Sal003, which inhibits eIF2␣ dephosphorylation (40). Cyclin D1 levels decreased significantly upon Sal003 treatment in HT1080 cells (Fig. 1B, panel a), once again accompanied by an increase in eIF2␣ phosphorylation (Fig. 1B, panel b), indicating a strong relationship between eIF2␣ phosphorylation and the level of cyclin D1. To determine whether the observed effect on cyclin D1 was due to decreased translation as previously proposed, we examined translation of cyclin D1 mRNA using polysome profiles on HT1080 cells treated with either thapsigargin or Sal003. This technique involves the separation of large cellular components using a sucrose gradient and the monitoring of the A 254 across the gradient. In this manner the 40 S and 60 S ribosomal subunits as well as monosomes and polysomes can be isolated and studied. Although GAPDH mRNA underwent a shift from the translating polysomes to the stalled monosomes ( Fig. 1C, panel b), cyclin D1 mRNA remained associated with the polysome fractions (Fig. 1C, panel a). The observable shift in GAPDH, but not in cyclin D1, confirms that a general translation is being inhibiting but suggests that a mechanism other than translation regulation contributes to the down-regulation of cyclin D1.
Down-regulation of Cyclin D1 Requires eIF2␣ Phosphorylation-Because phosphorylation of eIF2␣ is believed to be essential for the observed regulation of cyclin D1 (30), we sought to elucidate whether it alone can affect cyclin D1 translation. We treated eIF2␣ A/A MEFs in which Ser-51 was replaced with a non-phosphorylatable alanine (33) and their wild type eIF2␣ S/S counterparts with thapsigargin and immunoblotted for cyclin D1 ( Fig. 2A, panel a). Cyclin D1 was quickly and significantly decreased by more than 50% in eIF2␣ S/S MEFs (Fig. 2, A, lanes 1-5, and B) but remained steady in eIF2␣ A/A knock-in cells (Fig. 2, A, lanes 6 -10, and B). Once again we performed polysome profiles to examine the translation of cyclin D1 in these cells (Fig. 2C, panels i-iv). Treatment with thapsigargin induced inhibition of global translation initiation in eIF2␣ S/S, but not A/A cells (compare panels ii and iv), but cyclin D1 mRNA remained associated with the polysome fractions (Fig. 2C, panel a). It is interesting to note that although inhibition of general translation does not result in a shift of cyclin D1 mRNA to earlier polysome fractions, the CCND1 transcript is localized to earlier fractions in eIF2␣ S/S cells compared with their knock-in counterparts. This may be due to slight differences in the overall translational activity of these cell lines, as the profiles themselves also show slight variances. Together these data suggest that although the observed regulation of cyclin D1 protein levels is not a result of the ability of eIF2␣ to inhibit translation, phosphorylation of this factor is essential.
Activation of a Conditionally Active Form of PKR Decreases Cyclin D1 Protein Levels-The role of several eIF2␣ kinases in regulating cyclin D1 has been investigated in a number of previous studies. However, the treatments used to induce stress in these cases could potentially activate a large number of other cellular responses in addition to eIF2␣ kinase activity. Because eIF2␣ phosphorylation at Ser-51 is predominantly associated with inhibiting global translation initiation, we therefore investigated whether cyclin D1 was translationally regulated in a system containing a conditionally active form of PKR (41).
HT1080 cells stably expressing a chimeric protein in which the catalytic domain of PKR was fused to the first 220 amino acids of the Escherichia coli gyrase B protein were analyzed (32). Treatment of these cells with the antibiotic coumermycin caused dimerization of the GyrB domain and autophosphorylation and activation of PKR (32). Cells were treated with coumermycin, and the protein extract immunoblot was probed with an anti-cyclin D1 antibody (Fig. 3A, panel a). As early as 3 h after activation of GyrB.PKR, as shown by an induction of eIF2␣ phosphorylation at serine 51 (Fig. 3A, panel b), cyclin D1 protein levels decreased significantly. This down-regulation was maintained for at least 12 h after coumermycin treatment (data not shown). To determine whether this regulation occurred at the translational level, we performed polysome profiles as described above. HT1080 GyrB.PKR cells were left untreated (Fig. 3B, panel i) or treated with coumermycin for 4 h (Fig. 3B, panel ii). RT-PCR performed on cyclin D1 transcripts associated with monosomes and polysomes (panel a) revealed that upon coumermycin treatment cyclin D1 mRNA undergoes no shift and remains bound with the late polysomes. Translation of GAPDH was evaluated as a control (panel b) and undergoes a significant shift from fraction 15 in the polysomes to fraction 8 in the stalled monosomes upon activation of GyrB.PKR and subsequent eIF2␣ phosphorylation. These observations suggest that, contrary to the conclusions of previous studies (28 -30), the decrease in cyclin D1 protein levels upon activation of eIF2␣ kinases is due to a mechanism other than translational inhibition.
Activation of PKR and PERK Leads to Cyclin D1 Degradation-Because our studies, thus, far strongly indicated that a mechanism other than translational inhibition was responsible for the decrease in cyclin D1 protein levels, we investigated whether protein degradation may contribute to the observed effect. Transfection of HT1080 cells with dsRNA caused a significant decrease in cyclin D1 levels (Fig. 4A, panel a, lane 3), but this effect was absent when cells were pretreated with the proteasome inhibitor MG132 (Fig. 4A, panel a, lane 6). A similar effect was observed in HT1080 GyrB.PKR WT cells pretreated with MG132 before coumermycin treatment, as no decrease in cyclin D1 protein was observed (Fig. 4B, panel a,  lanes 6 -10) compared with coumermycin treatment alone, which caused a dramatic decrease in cyclin D1 levels. Cyclin D1 levels remained stable for the entire 8-h time course of GyrB. PKR activation (Fig. 4C). eIF2␣ phosphorylation was equally induced by coumermycin in the presence or absence of MG132 treatment (Fig. 4B, panel b), indicating that the lack of cyclin D1 down-regulation is not due to a lack of translation inhibition. Activation of PERK upon ER stress also appears to induce the degradation of cyclin D1 in an eIF2␣ phosphorylation-dependent manner, as pretreatment with MG132 prevented a cyclin D1 decrease upon thapsigargin treatment of eIF2␣ S/S cells (Fig. 4D, panel a, compare lanes 2 and 4). Cyclin D1 levels remained unchanged in eIF2␣ A/A cells regardless of treatment (Fig. 4, D, panel a, lanes 5-8, and E). Several pathways have been implicated in regulating cyclin D1 degradation by inducing phosphorylation of cyclin D1 on threonine 286, including the glycogen synthase kinase 3␤ (GSK-3␤) pathway (42), PTEN (43), activated Ras (44), and the mitogen-activated protein kinase (MAPK) pathway (45). Using 1-azakenpaullone, a specific inhibitor of GSK-3␤ (46), we determined that cyclin D1 degradation upon GyrB.PKR activation does not proceed through GSK-3␤ (supplemental Fig. 1A, panel a), as cyclin D1 levels were not rescued upon inhibition of GSK-3␤ activity. Although previous studies have demonstrated that GSK-3␤ can phosphorylate the ⑀-subunit of eIF2B (47), the guanine nucleotide exchange factor that targets eIF2, in our experiments eIF2␣ phosphorylation remained unaffected by 1-azakenpaullone treatment (supplemental Fig. 1A, panel b). The same lack of effect was observed when GSK-3␤ was inhibited by treatment with LiCl (data not shown) or when GSK-3␤ ϩ/ϩ and GSK- 3␤ Ϫ/Ϫ MEFs were used (data not shown). We also examined signaling through the MAPK pathway and observed that inhibition of MEK activity by the inhibitor U0126 did not rescue cyclin D1 levels after coumermycin treatment (supplemental Fig. 1B, panel a). Similar to our results obtained with 1-azakenpavllone, eIF2␣ phosphorylation was not affected by the MEK inhibitor (supplemental Fig. 1B, panel b). Combined, these data indicate that cyclin D1 degradation via the proteasome pathway is not mediated by one of the traditional pathways.

Induction of eIF2␣ Phosphorylation Does Not Significantly Affect Global Proteasome-dependent Protein Degradation-
The effect we observed regarding the ability of eIF2␣ kinases to mediate the proteasome-mediated degradation of cyclin D1 led us to investigate whether there were other proteins regulated in a similar manner. [ 35 S]Methionine labeling revealed that a significant amount of labeled proteins were up-regulated in cells pretreated with MG132 (Fig. 5A, panel a, lanes 3 and 4) compared with MG132-untreated cells (lanes 1 and 2) either before (lanes 1 and 3) or after activation of GyrB.PKR with coumermycin (lanes 2 and 4). However, GyrB.PKR activation was still capable of decreasing the amount of 35 S-labeled protein even when MG132 was present, as assessed by autoradiography and scintillation counting (Fig. 5A, panels a and c). We obtained a similar result when the effects of MG132 were examined after GyrB.PKR activation by coumermycin (Fig. 5A, panel a, lanes 5  and 6). These data indicated that the eIF2␣ kinases exert their inhibitory effects on protein synthesis mainly at the level of translation. It is important to note, however, that a few radio-  active bands retained their original intensity upon GyrB.PKR activation in the presence of MG132 (panel a), suggesting that in some cases eIF2␣ kinases can also regulate the degradation of specific proteins. The resistance of specific labeled proteins to degradation in the presence of MG132 as well as the ability of MG132 pretreatment to prevent cyclin D1 degradation upon eIF2␣ phosphorylation prompted us to examine whether MG132 post-treatment could similarly rescue cyclin D1 protein levels. HT1080 GyrB.PKR cells were treated with coumermycin followed by MG132 treatment 2 h after induction of GyrB.PKR activity. Inhibition of proteasome-dependent degradation caused a substantial rescue of cyclin D1 protein levels in cells treated with coumermycin (Fig. 5B, panel a) but not in cells treated with cycloheximide (Fig. 5C, panel a), a compound that inhibits the elongation step of protein synthesis and is, therefore, independent of eIF2␣. To further confirm that induction of eIF2␣ kinase activity targeted the degradation and not translation of cyclin D1, we studied the two events separately and in conjunction. GyrB.PKR cells were treated either with coumermycin alone for 2 h, cycloheximide alone for short time points as controls, or both reagents together. These treatments were performed at shorter than normal intervals due to the fact that prolonged treatment with either reagent rapidly degrades cyclin D1 signal. Coumermycin treatment alone caused a gradual decrease in cyclin D1 protein levels as previously observed (Fig. 5D,  panel a, lanes 1-4), whereas cycloheximide caused a more rapid decline in cyclin D1 (lanes 5-8).
When cells were treated with cycloheximide after coumermycin treatment, the decrease in cyclin D1 protein levels was immediate and dramatic, falling below detectable levels after only 30 min of treatment (lanes 9 -12). Because cycloheximide inhibits translation by a mechanism independent of eIF2␣ phosphorylation, if PKR targeted the translation of cyclin D1 we would not expect to see the additive effect of the two reagents observed in this experiment. This indicates that the proteasomal degradation of cyclin D1 is dependent upon eIF2␣ phosphorylation but not on the inhibition of overall protein translation. It should also be noted that in both experiments treatment of GyrB. PKR cells with MG132 alone caused an increase in the basal levels of cyclin D1, confirming that cyclin D1 is a very labile protein subject to a high rate of turnover.
Cyclin D1 Phosphorylated at Thr-286 Is Not Subject to Degradation upon eIF2␣ Kinase Activation-Although known upstream pathways are not affected in our system, we examined phosphorylation of cyclin D1 at a key threonine residue (Thr-286) in the carboxyl terminus of the protein. This phosphorylation causes a relocalization of cyclin D1 from the nucleus to the cytoplasm where it can be processed by the proteasome. Therefore, since we have demonstrated that the regulation of cyclin D1 in our system is at the level of degradation and not translation, we examined the effect of eIF2␣ kinase activity on cyclin D1 phosphorylation at this site. Using a phosphospecific antibody, we observed no increase in phosphorylation at Thr-286 (Fig. 6A,  panels a and b). In fact, the level of phosphorylated cyclin D1 remained stable, whereas that of total cyclin D1 decreased (compare panels a and b with panel c). Specificity of this antibody was verified in quality control experiments probing extracts of cells overexpressing either wild type or T286A mutant forms of cyclin D1 (supplemental Fig. 2). Because cyclin D1 is rapidly degraded upon treatment with coumermycin, stimulation was also performed in the presence of MG132 to prevent this degradation (Fig. 6A, lanes 5-8). No discernable induction of Thr-286 phosphorylation is observed up to 4 h of coumermycin in cells pretreated with MG132, indicating that phosphorylation at this site is not regulated by GyrB.PKR activity. To further determine whether this effect was common to multiple eIF2␣ kinases, cyclin D1 phosphorylation was examined in PERK ϩ/ϩ and PERK Ϫ/Ϫ MEFs treated with thapsigargin (Fig. 6B). As observed in the GyrB.PKR system, the level of threonine-phosphorylated cyclin D1 remained stable, whereas total cyclin D1 decreased in cells containing catalytically active PERK (Fig. 6B, lanes 1-4, compare panels a and b to c). No decrease in cyclin D1 levels or change in cyclin D1 phosphorylation was observed in PERK Ϫ/Ϫ cells (Fig. 6B, lanes 5-8).
Although Thr-286 phosphorylation is not induced by PKR or PERK, this site is still critical for the ubiquitin-dependent, proteasomal degradation of cyclinD1. To determine whether this site is required for the induction of cyclin D1 degradation by eIF2␣ kinases, we overexpressed cyclin D1 wild type and T286A mutant forms in GyrB.PKR cells (Fig. 6C). When cells were treated with coumermycin, we found that wild type cyclin D1 protein levels decreased (Fig. 6C, panel a, lanes 4 -6) at the same rate as the mutant cyclin D1 construct (Fig. 6C, panel a,   lanes 5-8). These observations indicate that in contrast to previously published reports, the phosphorylation of cyclin D1 at Thr-286 is not required for the eIF2␣ kinase-induced degradation of the protein and may actually contribute to the stability of the protein in this system. eIF2␣ Kinases Promote the Ubiquitin-dependent Degradation of Cyclin D1 but Do Not Directly Affect Its Ubiquitination-Although degradation of cyclin D1 was not affected by the lack of phosphorylation at Thr-286, it does not necessarily follow that the proteasomal degradation of cyclin D1 is not dependent on protein ubiquitination. Using temperature-sensitive cells defective in the ubiquitin-activating enzyme E1 (35), we examined the degradation of cyclin D1 in the presence and absence of an intact ubiquitination pathway. Upon TG treatment of temperature-sensitive ts20 cells, cyclin D1 levels decreased at the permissive temperature of 31°C (Fig. 7A, panel  a, lanes 1-5) but were stabilized at the restrictive temperature of 39°C (Fig. 7A, panel a, lanes 6 -10). In contrast, cyclin D1 levels were decreased upon TG treatment at both temperatures in E36 control cells (Fig. 7B, panel a). It should be noted that even at the permissive temperature of 31°C, cyclin D1 was degraded more rapidly in the E36 control cells than their temperature-sensitive counterparts. It is, therefore, possible that even at the lower temperature, the ubiquitin pathway may not be fully functional, and some residual defect exists. Interestingly, the degree of eIF2␣ phosphorylation also shows some variation between the permissive and restrictive temperatures (Fig. 7, compare lanes 1-5 and 6 -10 in both A and B). These results suggest that despite the lack of induction of Thr-286 phosphorylation by PERK and PKR, degradation of cyclin D1 upon eIF2␣ kinase activation proceeds through the ubiquitin pathway. We next examined whether these kinases could directly affect the ubiquitination of cyclin D1. We overexpressed HA-tagged ubiquitin along with either WT or T286A constructs of cyclin D1 in GyrB.PKR WT cells and examined the effect of coumermycin treatment on ubiquitination of the protein. Coumermycin treatment had no effect on the expression of HA-tagged ubiquitin (Fig. 7C, panel a) and, as determined in previous experiments, caused a down-regulation of both WT and T286A cyclin D1 (Fig. 7C, panel b). More importantly, the levels of ubiquitinated cyclin D1 were not affected by coumermycin treatment (Fig. 7C, panel b, upper section). Together, these data imply that although eIF2␣ kinases play a positive role in the ubiquitin-dependent proteasomal degradation of cyclin D1, it is not at the step of ubiquitination.
The Ubiquitin-independent NAD(P)H Quinone Oxireductase 1 (NQO1) Pathway Also Contributes to the Degradation of Cyclin D1-Although the data obtained from the temperaturesensitive cells strongly indicates that eIF2␣ kinases induce the degradation of cyclin D1 in a manner dependent on an intact ubiquitin pathway, the fact that ubiquitination itself is not affected by these proteins led us to investigate whether other, ubiquitin-independent pathways may also contribute to the down-regulation of cyclin D1. The protein NQO1 has been shown to regulate the degradation of a number of proteins by the 20 S proteasome in an ubiquitin-independent fashion (48 -50). NQO1 prevents the degradation of its target proteins by physically binding to, and stabilizing them. We therefore exam- ined the regulation of cyclin D1 in cells treated with dicoumarol, a competitive inhibitor of NQO1. We determined that in GyrB.PKR WT cells dicoumarol could indeed induce cyclin D1 degradation but with kinetics much slower that of coumermycin (Fig. 8A, panel a, compare lanes 1-5 with lanes 6 -10).
Coumermycin caused a decrease in cyclin D1 levels as early as 2 h post-treatment, whereas it took 8 h for dicoumarol to take effect. This suggests that eIF2␣ kinase activation and inhibition of NQO1 binding do not function through related pathways and that the two mechanisms may be responsible for early and late cyclin D1, respectively. To confirm this and to investigate whether eIF2␣ phosphorylation affects the response to dicou-marol, we treated eIF2␣ S/S and A/A cells with either thapsigargin or dicoumarol for prolonged periods (Fig. 8B). As previously observed, both treatments induced the decrease of cyclin D1 protein levels but at different rates (Fig. 8B, panel  a, compare lanes 2-4 with lanes  5-7 and lanes 9 -10 with 12-14). Furthermore, the efficacy of dicoumarol treatment was independent of eIF2␣ phosphorylation, and dicoumarol itself did not induce eIF2␣ phosphorylation (Fig. 8B,  panel b). Interestingly, we also observed that prolonged thapsigargin treatment could induce cyclin D1 degradation in eIF2␣ A/A as well as S/S cells (Fig. 8B, compare lanes 1-4 with 8 -11). This is in contrast to short thapsigargin treatments in the same cell lines, in which no decrease in cyclin D1 protein levels is detected (Fig. 2A, panel a). To determine whether this late regulation is dependent upon eIF2␣ kinase activity, we performed thapsigargin and dicoumarol treatments in PERK ϩ/ϩ and PERK Ϫ/Ϫ cells for 24 h (Fig. 8C). In PERK wild type cells, as in eIF2␣ S/S cells, both thapsigargin and dicoumarol induced the down-regulation of cyclin D1 (Fig. 8C, panel a, lanes 2  and 3). Prolonged treatment with both reagents also induced the degradation of cyclin D1 in PERK Ϫ/Ϫ cells (Fig. 8C, panel a, lanes 5 and 6). Together this data indicates that the degradation of cyclin D1 upon prolonged thapsigargin treatment occurs independently of both eIF2␣ phosphorylation and kinase activity.

DISCUSSION
In this study we observed a similar negative regulation of cyclin D1 protein levels after eIF2␣ kinase activation as noted previously (28 -30, 51) and further revealed a novel property of eIF2␣ kinases in promoting the proteasome-dependent degradation of cyclin D1. Our data provide evidence that the decrease in cyclin D1 levels observed upon activation of PKR and PERK is not due to their ability to inhibit translation initiation but via their ability to induce degradation of the protein. Results obtained from both a conditionally active PKR system and human fibrosarcoma cells demonstrate that whereas cyclin D1 protein levels decrease upon activation of the eIF2␣ kinases (Figs. 1, A and B, and 3A), translation of the CCND1 transcript does not differ once overall  translation is inhibited (Figs. 1C and 3B). We further demonstrated that although cyclin D1 is not inhibited at the translational level after PKR or PERK activation, eIF2␣ phosphorylation is still required for the down-regulation of cyclin D1 protein ( Fig. 2A). How cyclin D1 mRNA bypasses the global inhibitory effects of eIF2␣ phosphorylation on translation is not immediately clear. Previous work demonstrated an important role of eIF4E in stimulating cyclin D1 mRNA translation (52). We recently reported that activation of eIF2␣ kinases can also lead to the phosphorylation of 4E-BP1 (53). Inasmuch as 4E-BP1 phosphorylation results in the dissociation of eIF4E/ 4E-BP1 complex (54), the possibility remains that free eIF4E selectively maintains cyclin D1 mRNA translation under conditions of increased eIF2␣ phosphorylation. The presence of the proteasome inhibitor MG132 ablated the down-regulation of cyclin D1 upon induction of PKR and PERK activity, strongly supporting the hypothesis that eIF2␣ kinases induce degradation of the protein rather than inhibiting its synthesis (Fig. 4, A,  B, and E, and 5B). Cyclin D1 levels also decreased upon treatment of cells with the translation inhibitor cycloheximide (Fig.  5C, panel a, lanes 1-3), but this down-regulation could not be rescued by MG132 (Fig. 5C, panel a, lanes 4 and 5). Furthermore, when the same cells were treated with both coumermycin and cycloheximide in conjunction, cyclin D1 levels decreased more rapidly than when either agent used alone (Fig.  5D). This suggested that activated GyrB.PKR promotes the proteasomal degradation of cyclin D1 that remains in the cells after the blockade of mRNA translation by cycloheximide. Our study also indicated that this function of eIF2␣ kinases is limited to a small number of specific proteins. That is, activation of GyrB.PKR did not exhibit a global effect on proteasome-dependent degradation of cellular proteins (Fig.  5A), strongly suggesting that the effect of eIF2␣ kinases in mediating the degradation of cyclin D1 is specific rather than general.
A number of different cellular pathways have been identified that contribute to the ubiquitination and subsequent degradation of cyclin D1. As described above, phosphorylation at Thr-286 by GSK-3␤ is required for the ubiquitination of cyclin D1 (42). The SCF Fbx4/␣B-crystallin has been identified as an E3 ubiquitin ligase-targeting cyclin D1 (55). Our group recently published a study in which we demonstrated that GSK-3␤ is downstream of the eIF2␣ kinases (38). However, recent studies have called into question the significance of the contribution of GSK-3␤ to cyclin D1 regulation (56). Furthermore, data obtained in our study indicate that eIF2␣ kinases do not induce the degradation of cyclin D1 via this particular pathway (supplemental Fig. 1A).
Further studies have also implicated another well characterized signaling pathway, the MAPK pathway in regulating the degradation of cyclin D1. In cancer cells increased signaling through the Ras/Raf/MEK/MAPK pathway leads to increased phosphorylation at Thr-286 and the subsequent destabilization of cyclin D1 (44). A separate study also implicated another SCF E3 ubiquitin ligase, the S-phase kinase-associated protein (SKP1), and the scaffolding protein Cullin 7 (CUL7) in this regulation (57). SKP1 and CUL7 associate via the F-box protein FBXW8, and the co-operative action of these pathways regu-lates cyclin D1. Inhibition of the MAPK pathway using a specific inhibitor of MEK indicates that the action of eIF2␣ kinases is not exerted via MAPK activation (supplemental Fig. 1B). In combination, these data indicate that the regulation of cyclin D1 stability is not mediated by either of these previously described mechanisms. This hypothesis is further supported by the fact that although cyclin D1 is rapidly degraded upon induction of eIF2␣ phosphorylation, inhibition of 26 S proteasome activity after activation of PKR is sufficient to cause a nearly complete rescue of cyclin D1 levels (Fig. 5B). This indicates that the upstream pathways that contribute to cyclin D1 ubiquitination and degradation remain functional. Indeed, our study revealed that upon eIF2␣ kinase activation, ubiquitination of cyclin D1 occurs and remains unchanged. Based on this evidence, however, it cannot be concluded that PKR and PERK act in an ubiquitin-independent manner. On the contrary, PERK activation was unable to induce the degradation of cyclin D1 in temperature-sensitive cells defective in E1 ubiquitin ligase activity (Fig. 7A).
Although the ubiquitin-dependent degradation of cyclin D1 has been extensively examined, a number of other, ubiquitin-independent pathways leading to cyclin D1 degradation have been discovered. A 2004 study also identified an alternative mechanism of cyclin D1 degradation (58) in which the regulatory protein antizyme that predominantly functions to regulate the degradation of another enzyme, ornithine decarboxylase (59), binds to cyclin D1 and directs it to the 26 S proteasome. Because it does not depend on the presence of ubiquitin, this regulation is also independent of phosphorylation of cyclin D1 at Thr-286. The T286A mutant form of the protein is degraded in an identical manner as wild type cyclin D1 upon induction of antizyme expression (58). We postulated that due to the unique mechanism in which antizyme is translated (60), the eIF2␣ kinases may induce antizyme expression, but induction of PERK and GyrB.PKR activity caused no significant change in antizyme expression (data not shown). The ubiquitin-independent degradation of ornithine decarboxylase can be regulated by a separate mechanism involving NQO1 (48). In contrast to the action of antizyme, NQO1 binds to ornithine decarboxylase and consequently stabilizes the protein. NQO1 has also been demonstrated to regulate the tumor suppressor protein p53 via the same mechanism (49,50). We determined that in addition to ornithine decarboxylase and p53, NQO1 can also negatively regulate the degradation of cyclin D1 (Fig. 8,  A-C). Treatment of cells with dicoumarol caused a decrease in cyclin D1 protein levels, albeit at a later time than either coumermycin or thapsigargin treatment. This suggests that the two pathways are capable of independently regulating cyclin D1. The existence of this additional mechanism of regulation is supported by our observation that prolonged treatment with reagents inducing eIF2␣ kinase activity (in this case, thapsigargin) can induce cyclin D1 degradation independent of eIF2␣ phosphorylation and PERK activity (Fig. 8, B and C). It is possible, therefore, that early degradation of cyclin D1 depends on eIF2␣ phosphorylation, whereas late degradation proceeds through a mechanism independent of eIF2␣ altogether.