Preferential translation of internal ribosome entry site-containing mRNAs during the mitotic cycle in mammalian cells.

A cell synchronization protocol was established in which global and individual mRNA translational efficiencies could be examined. While global translational efficiency was reduced in mitotic cells, approximately 3% of mRNAs remained predominantly associated with large polysomes during mitosis, as determined by cDNA microarray analyses. The 5'-non-coding regions of six mRNAs were shown to contain internal ribosome entry sites (IRES). However, not all known mRNAs that contain IRES elements were actively translated during mitosis, arguing that specific IRES sequences are differentially regulated during mitosis.

A cell synchronization protocol was established in which global and individual mRNA translational efficiencies could be examined. While global translational efficiency was reduced in mitotic cells, ϳ3% of mRNAs remained predominantly associated with large polysomes during mitosis, as determined by cDNA microarray analyses. The 5-non-coding regions of six mRNAs were shown to contain internal ribosome entry sites (IRES). However, not all known mRNAs that contain IRES elements were actively translated during mitosis, arguing that specific IRES sequences are differentially regulated during mitosis.
Cells of higher organisms can regulate gene expression at the translational level in response to a wide variety of stimuli and conditions. While much is known about the control of translational initiation and elongation in response to nutritional deprivation (1,2) and environmental stress (2), the role of translational control in mammalian cell cycle progression is not well understood. It has long been known that specific proteins are needed at specific times during the cell cycle to ensure cell cycle progression. For the most part, the expression of these cell cycle-specific proteins is thought to be regulated at the transcriptional or posttranslational level (3).
In cultured mammalian cells arrested at G 2 and M phases, the rate of total protein synthesis was markedly decreased to about 25% of the rate in non-arrested, cycling cells (4). This reduction was shown to be, at least partly, due to inhibition of the initiation step of polypeptide synthesis (4 -6). Subsequently, several eukaryotic initiation factors that regulate the assembly of 40 S ribosomes at the 5Ј ends of capped mRNAs were observed to change their phosphorylation status (7)(8)(9). Therefore, it seemed likely that inhibition of translation initiation was caused by diminishing ribosome recruitment to capped mRNAs during mitosis. Viral mRNAs whose translation is initiated by an internal ribosome entry site (IRES) 1 mechanism (10), such as polioviral and hepatitis C viral mRNAs (4,(11)(12)(13)(14) or mRNAs which lack significant structures in their 5Ј-non-coding regions (5Ј-NCRs), such as the mRNAs containing the late leader of adenovirus (15), were found to be selectively translated during mitosis due to their lessened requirement for eIF4F. More recently, additional IREScontaining mRNAs, those encoding ornithine decarboxylase and kinase p58 PITSLRE , have been reported to be selectively translated during G 2 /M of the cell cycle in cultured cells (11,12). These findings raise the question whether all IREScontaining genes are preferentially translated during mitosis and whether any of these encoded products play roles in cell cycle progression. We have begun to address these questions by genomic analysis of cellular mRNAs that are associated with mitotic polysomes and are, thus, predicted to be translated during the overall translation repression in mitosis. We determined that many, but not all, IRES elements are present in mRNAs which are selectively translated during mitosis.

EXPERIMENTAL PROCEDURES
Cell Synchronization and Transfection Protocols-HeLa cells were maintained in Dulbecco modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (Invitrogen). To synchronize cells at the G 1 /S boundary, cells were treated at 50% confluence with 2 mM thymidine (Sigma) or 5 g/ml aphidicolin (Sigma). After 16 h, cells were washed and fed with fresh medium for 6 h before the addition of either 0.6 g/ml colcemid (Invitrogen) or nocodazole (Sigma; Ref. 16). Synchronization was monitored by flow cytometry analysis after DNA staining with propidium iodide. To measure protein synthesis rates in nonsynchronized and mitotic arrested cells, cells were incubated in DMEM with 100 Ci/ml [ 35 S]methionine/cysteine for 10 min. Extracts were prepared and processed as described previously (17).
DNA plasmids were transfected into 50 -60% confluent cells with the Lipofectin reagent (Invitrogen). After a 24-h incubation, luciferase assays were performed or RNA samples for Northern analyses were prepared.
Indirect Immunofluorescence Analysis-Cells were grown on 12-mm coverslips, fixed for 5 min in 100% methanol at Ϫ20°C, and permeabilized in phosphate-buffered saline with 0.2% Triton X-100. Following fixation and permeabilization, cells were washed with PBS and incubated in PBS containing 3% bovine serum albumin, 0.1% Triton X-100, and 5% calf serum for 40 min at 25°C. Coverslips were then incubated with diluted rat antibodies against ␣-tubulin (a gift from P. Jackson, Stanford University) in IF buffer (PBS with 3% bovine serum albumin and 0.1% Triton X-100) for 1 h at 25°C. After washing, coverslips were inverted into the IF buffer with Texas Red-conjugated donkey anti-rat antibody for 30 min at 25°C. Cover slips were washed twice with PBS, and once with PBS containing 0.2 g/ml Hoechst 33258, drained, and mounted in vectashield mounting medium (Vector Laboratories) onto glass slides. Cells were viewed using an Olympus BX-60 fluorescent microscope. Pictures were acquired using a Hamamatsu Orca digital camera and Image Pro plus software with 40ϫ or 60ϫ objectives.
5Ј RACE and Dicistronic Plasmid Constructions-The 5Ј NCRs of selected mRNA species were converted to cDNA copies using the First-Choice RLM-RACE Kit (Ambion). The DNA fragments were amplified by the Advantage-GC PCR kit (Clontech) and directly cloned into PCR2.1-topo vector (Invitrogen). The nucleotide sequences of all inserted fragments were obtained. The 5Ј NCRs were amplified by the PCR with 5Ј primer containing EcoRI sequences and 3Ј primers con- taining NcoI sequences. These fragments were inserted into dicistronic SV40/T7 plasmid, c-53, between EcoRI and NcoI sites, or a T7 promotercontaining dicistronic vector, c-84, as described previously (18).
Northern Blot Analyses-Total RNA was isolated using the TRIzol Reagent (Invitrogen) and analyzed as described previously (17). Radiolabeled DNA hybridization probes were generated using the RadPrime Kit (Invitrogen).
Polysomal Analysis-Polysomal mRNA was prepared as described previously (17). Briefly, cells were incubated with 0.1 mg/ml cycloheximide for 3 min at 37°C before being harvested. Non-synchronized control cells were washed with PBS, harvested directly on the plate in lysis buffer (15 mM Tris-Hcl, pH 7.4, 15 mM MgCl 2 , 0.3 M NaCl, 1% Triton X-100, 0.1 mg/ml cycloheximide, 1 mg/ml heparin), and transfered to Eppendorf tubes. The mitotic cells were shaken off from plates and pelleted at 1000 rpm for 3 min. Mitotic cells were washed in PBS and lysed in lysis buffer for 10 min on ice. The nuclei and debris were removed by centrifugation at 12,000 ϫ g for 10 min, and the supernatants were loaded onto 10 -50% sucrose gradients composed of same extraction buffer lacking Triton X-100. The gradients were sedimented at 35,000 rpm for 160 min in a SW41 rotor at 4°C. Fractions of equal volumes were collected from the top using an ISCO fraction collector system. RNAs were collected in 4 M guanidine HCl, precipitated after addtion of ethanol, and resuspended in 1 mM Tris-HCl, pH 8.
Western Blot and Immunoprecipitation Analyses-Proteins were separated by SDS-PAGE and transferred to Immobilon-P membrane (Millipore). Poly(ADP-ribose) polymerase (PARP) was detected with antibody number SC7150 (Santa Cruz Biotechnology) as described previously (19). To examine newly synthesized proteins, cells were grown in DMEM lacking methionine and cysteine (Invitrogen) with 10% fetal bovine serum for 1 h and metabolically labeled with 100 Ci/ml [ 35 S]methionine-cysteine in DMEM (lacking methionine-cysteine) with 10% fetal bovine serum for 30 min. Labeled cells were lysed in a buffer containing 20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, 0.5% Nonidet P-40, and protease inhibitor mixture. Equal amount of protein extracts, determined with the Bio-Rad protein reagent, were incubated with mouse anti-nucleophosmin (Zymed Laboratories Inc.) and anti-vimentin (Santa Cruz Biotechnology) antibodies and agarose-protein A/G plus beads according to the procedure recommended by Santa Cruz Biotechnology. The agarose beads-attached proteins were recovered by boiling in Laemmli sample buffer and separated on SDS gels. The gels were dried and enhanced with Amplify solution (Amersham Biosciences) before exposed to a phosphoimager screen. Quantification was performed using a STORM phosphoimager and ImageQuant software (Amersham Biosciences).
Preparation of Fluorescently Labeled cDNA for Microarray Hybridization and Data Analysis-Equal volume amounts of RNAs from pooled polysomal fractions were used to generate fluorescently labeled DNA probes as described (www.cmgm.stanford.edu/pbrown). The human cDNA microarrays contained a set of ϳ43,000 human cDNA clones, manufactured at Stanford University as described (www.cmgm. stanford.edu/pbrown). The microarrays were scanned on a Axon laser scanner and analyzed using ScanAlyze (available at www.rana. lbl.gov/). The computed data were entered and further analyzed in Stanford Microarray Data Base (www.genome-www5.stanford.edu/ MicroArray/SMD/). Only genes for which the fluorescence signal in each channel exceeded certain cutoffs (including regression correlation, signal-noise ratio, signal intensity, and the tolerance of missing data among multiple measurements) were retrieved from the Stanford Microarray Data Base for the deterimation of the microarray distribution score S(x) (see Fig. 3) and significant analysis of microarray.
To identify the individual candidates, three interphase arrays and three mitotic (thymidine/colcemid) arrays were applied to significant analysis of microarray (significant analysis of microarray; available at www-stat.stanford.edu/ϳtibs/SAM/) analysis, which employs permutation of the repeated measurements to estimate the percentage of genes identified by chance, the false discovery rate.

Reduction of Protein Synthesis in G 2 /M-arrested HeLa
Cells-To investigate translational control during mitosis, HeLa cells were synchronized in mitosis by a sequential block with thymidine and colcemid or with thymidine and nocodazole. Mitotic cells were shaken off from the plates, and the efficiency of synchronization was monitored by flow cytometry after propidium iodide staining. About 80 -90% of cells treated with thymidine-colcemid or thymidine-nocadozole displayed a double content of DNA, confirming their arrest (Fig. 1A). To identify the specific mitotic phase in which cells were arrested, the microtubule structure of thymidine/colcemid-arrested cells and non-synchronized cells were visualized by immunofluorescence staining using antibodies directed against ␣-tubulin (Fig.  1B). Loss of functional spindle structures and condensed chromosomes in treated cells revealed that these mitotic cells were arrested at prophase or premetaphase of mitosis.
To compare the translational efficiencies in mitotically arrested and non-synchronized cells, cells were pulse-labeled with [ 35 S]methionine/cysteine, and the incorporation of radiolabeled amino acids was measured. As reported previously (4), both thymidine/colcemid-and thymidine/nocodazole-arrested cells showed a reduced incorporation of radiolabeled amino acids compared with non-synchronized cells (Fig. 1C). It has been recently reported that many synchronization protocols can induce apoptosis (20,21). To examine whether the synchronization protocols used here triggered cells to enter apoptosis, caspase-induced cleavage of PARP was measured. Fig. 1D shows that little cleavage of PARP occurred in either thymidine/colcemid-or thymidine/nocodazole-treated cells. In contrast, notable amount of PARP was cleaved in another synchronization protocol, i.e. treatment of cells with aphidicolin and colcemid. These findings suggest that the global reduction of protein biosynthesis that was observed in thymidine/ colcemide-and thymidine-nocodazole-treated cells was not a result of ongoing apoptosis.
Reduced Association of Cellular mRNAs with Mitotic Polysomes-To examine the translation status of both total and individual mRNAs, cytoplasmic extracts from interphasic-and mitotic-arrested cells were fractionated in sucrose gradients (17). Fig. 2 shows that, in non-synchronized cells, most ribosomes were associated with mRNAs, sedimenting with polysomal fractions 8 -10. In contrast, most ribosomes in mitotic extracts had dissociated from mRNAs, sedimenting as free 40 and 60 S subunits (Fig. 2). This finding argues for mitotic inhibition at the initiation step of translation.
To monitor the association of individual mRNAs with ribosomes, the distribution of individual mRNAs in the polysome gradient was visualized by Northern blot analysis using genespecific hybridization probes. mRNAs encoding ␤-actin and transforming growth factor-␤ were mostly associated with fractions 10 and 9, respectively, in non-synchronized cells. In contrast, these mRNAs shifted to the lower molecular weight fractions in mitotically arrested cells, indicating that these mRNAs were less efficiently translated in mitosis than in interphase.

Identification of Mammalian mRNAs Species That Are Preferentially Associated with Mitotic
Polysomes-To identify individual mRNA species which are selectively translated during mitosis, the polysomal distribution of mRNAs was examined using human cDNA microarray analyses (22,23). Specifically, the amount of individual cDNAs obtained from RNA in frac-tions 4 -7 was compared with that obtained from RNA in polysomal fractions 9 and 10 (Fig. 2). Statistical analyses indicated (Fig. 3) that the majority of mRNAs shifted toward the lower molecular weight polysomal fractions during mitosis, arguing that the mitotic sedimentation profiles of ␤-actin and transforming growth factor-␤ mRNAs (Fig. 2) reflected an overall diminished association with ribosomes of mRNAs. However, at least 3% of the mRNA species (49 genes out of 1494 genes) continued to be associated with similar or more numbers of ribosomes in mitotic extracts, determined both by statistical analysis (Fig. 3) and Northern blot analysis (Fig. 4). Selected positive candidate genes are listed in Table I. Approximately one-third of these genes encode nuclear proteins, many of which are predicted to play roles in RNA metabolism (Table I).
We further substantiated the polysomal distribution of some of the mRNAs identified in the microaray analyses by Northern blot analysis using gene-specific hybridization probes. In contrast to ␤-actin mRNAs, mRNAs encoding the glucose regulated 58-kDa protein (GRP58), nucleosome assembly protein 1-like 1 (NAP1-L1), nucleophosmin 1 (NPM1), and the autoantigen La sedimented with the same polysomal fractions in non-synchronized and arrested cells (Fig. 4, A and B). In contrast to the findings reported by Pyronnet et al. (12), ornithine decarboxylase (ODC) mRNA shifted, like ␤-actin mRNA, to lower molecular weight polysomal fractions during mitosis (Fig. 4, A and B). Specifically, the majority ODC mRNAs was associated with ϳ4 -5 ribosomes (fraction 9) in non-synchronized cells. However, the bulk of ODC mRNA associated with only 1-3 ribosomes (fractions 7 and 8) in mitotically arrested cells. While the reason for this discrepancy is not known, different cell lines or cell synchonization protocols may have accounted for the observed difference.
To determine whether mRNAs associated with polysomes during mitosis were translated, we measured the relative translational efficiencies of nucleophosmin, vimentin, and La mRNAs in non-synchronized and mitotis-arrested cells. Translation efficiency was defined as the amount of newly synthe- sized protein compared with total translational efficiency during mitosis. The amounts of total nucleophosmin, vimentin, and La mRNAs were very similar in non-synchronized and mitotic extracts. While the translational efficiencies of nucleophosmin and La mRNAs were somewhat enhanced during mitosis, the mitotic translational efficiency of vimentin mRNA was more pronounced (Fig. 5, A and B) compared with the bulk of cellular mRNAs. These findings argue that the polysomal association of these mRNAs reflected translation by ribosomes. mRNAs Preferentially Translated during Mitosis Contain IRES Elements-The cDNA microarray analyses identified two genes known to contain IRES elements that were preferentially translated during mitosis. Specifically, La and Cyr61 mRNAs were previously identified to be associated with polysomes during poliovirus infection (23). To examine whether the 5Ј-NCRs of identified mitotic polysomal-associated mRNAs could mediate cap-independent translation via an IRES-mediated mechanism, cDNA copies of the 5Ј-NCRs of several identified candidate mRNAs were obtained by 5Ј RACE (Table II). As expected (10), there were no strong primary consensus sequences among them by multiple or pairwise alignment analyses. Also, the 5Ј-NCRs were of the same length and similar GC nucleotide content than the average mammalian mRNA. For comparison, cDNA copies were obtained from six mRNAs that represented the majority of cellular mRNAs that were not associated with polysomes during mitosis (␤-actin, PMSA1, CSTB, HNRPA1, NUBS, and RPS5).
The cDNAs encoding the various 5Ј-NCRs were inserted between two cistrons in dicistronic plasmids, which can direct the expression of dicistronic mRNAs that contain the coding regions for Renilla and firefly luciferase as the first and second cistrons, respectively (Fig. 6A) (23). The first, Renilla luciferase, cistron is predicted to be translated by a cap-dependent initiation mechanism, while the second, firefly luciferase, cistron should be translated if preceded by an IRES element. The various dicistronic plasmids were transfected into HeLa cells and luciferase activities were measured. The six 5Ј-NCRs of mRNAs that were associated with fewer ribosomes in mitosis than in non-synchronized cells expressed the second cistron firefly luciferase either similarly or in 2-3-fold higher amounts as the dicistronic control mRNA construct that lacked an intergenic inserted region (Fig. 6B). Therefore, these 5Ј-NCRs did not contain significant IRES activities. In contrast, six out of seven 5Ј-NCRs from preferentially mitotic polysome-associated mRNAs significantly promoted the expression of firefly luciferase encoded in the second cistron, yielding 7-30-fold higher rates than seen in cells transfected with dicistronic actin 5Ј-NCR-containing mRNAs (Fig. 6B). To examine whether expression of the second cistron resulted from functionally monocistronic transcripts generated by nuclease activity, cryptic splicing, or expression of an unknown promoter in the insert 5Ј-NCR-encoding cDNA, the size and integrity of expressed dicistronic mRNAs from transfected cells were inspected by  5. Measurement of relative translational efficiency of mRNAs encoding vimentin, nucleophosmin, and the La autoantigen. A, total mRNAs were visualized by Northern analysis using gene-specific hybridization probes. Newly synthesized nucleophosmin, vimentin, and La proteins were visualized after labeling with [ 35 S]methionine/cysteine for 30 min followed by immunoprecipitation. Phosphoimages of the gels are shown. B, relative translational efficiencies of nucleophosmin, vimentin, and La mRNAs. Incorporation of radioactive amino acids was quantified by phosphoimaging and normalized to the incorporation of radioactive amino acids into total protein in mitotic extracts (which 30% compared with non-synchronized extracts; see Fig. 1C). Each bar represents the average of three independent measurements.
Northern blot analysis. Fig. 6C shows that each plasmid directed the synthesis of predominantely full-length mRNAs, arguing that full-length dicistronic mRNAs were the major templates for the second cistron translation. To test this hypothesis more rigorously, the SV40 promoters were deleted in all plasmids and luciferase activities were measured after transfection of these promoterless plasmids into cells. All but one of the constructs expressed little to no luciferase activity (data not shown). Surprisingly, the NAP1L1 leader cDNA sequences could mediate second cistron translation at the level of 10-fold higher than background (data not shown), demonstrating that smaller mRNAs may have contributed to the translation of the second firefly luciferase cistron in this case. This finding exemplifies that by Northern analysis an undetectable amount of functional monocistronic RNA can be generated in the dicistronic assay system, and additional control experiments are warranted to claim IRES activity. To test whether the 5Ј-NCR of NAP1L1 mRNA carried a bona fide IRES activity, full-length dicistronic control, and NAP1L1-or NPM1-NCR-containing mRNAs were generated in vitro by T7 RNA polymerase, and mRNAs were translated in HeLa S10 translation extracts. Fig. 6D shows that relative expression level of the second cistron in dicistronic NAP1L1-containing mRNA was more than 100-fold higher than in mRNAs lacking intercistronic sequences. This finding suggests that the NAP1L1 leader does contain IRES activity that is significantly higher than the various control 5Ј-NCRs (Fig. 6B). Overall, these findings indicate that IRES activities can be active during mitosis when translation of the bulk of the cellular mRNAs has declined. DISCUSSION We have studied translational control in mitotically arrested HeLa cells using cell synchronization protocols that minimized the accumulation of apoptotic cells. Specifically, we examined whether selected individual mRNAs could be preferentially translated during the global translational repression during mitosis. To this end, polysomal mRNAs were analyzed by cDNA microarray analysis. The ribosome density on a particular mRNA was used as the indicator of the relative translational efficiency of the mRNA. By separating the mRNAs into non-polysomal fractions and polysomal fractions and comparing their relative abundance by microarray analysis, we examined the translation state of individual mRNAs in non-synchronized cells and mitotically arrested cells. The 5Ј-NCRs of selected mitotic polysome-associated mRNAs were 100 -200 nucleotides in length (Table II). Six out of seven of those selected leader sequences functioned as IRES elements in dicistronic assays, suggesting that internal initiation is more efficient than cap-dependent initiation during mitosis. One possibility to this selective translational control is that cap-dependent and internal translation compete with each other for components of the translation apparatus. Many of the mRNAs which were associated with mitotic polysomes encoded nuclear RNA binding proteins, such as the La autoantigen or hnRNPA. Because certain nuclear RNA-binding proteins have been identified to stimulate IRES activity, such as La and hnRNPC (24 -26), selective up-regulation of IRES-containing mRNAs could be favored by the increased accessibility of these nuclear factors during mitosis. The target IRES-containing mRNAs may encode proteins that are important to aid in cell cycle progression and facilitate specific IRES-mediated translation during mitosis in an autoregulatory manner.
However, not all of the known cellular IRES-containing messages were associated with mitotic polysomes. For example, whereas IRES-contanining La (18) and Cyr61 (23) mRNAs were preferentially detected in polysomal fractions during mitosis, IRES-containing c-Myc (27, 28) (not shown) and ODC (12) mRNAs were not. These findings seem to be, at first glance, at odds with observations reported by Pyronnet and Sonenberg (8) who used a sequential thymidine and aphidiolin block to synchronize HeLa cells at S phase, following cell cycle progression through the G 2 /M phase after release of the drugs. Synthesis of FIG. 6. Measurements of IRES activity within the 5-NCRs of selected mitotically translated mRNAs. A, schematic representation of dicistronic constructs. Construct contain a SV40 promoter for in vivo expression studies and a T7 promoter for generation of transcripts in vitro (23). The Renilla luciferase gene serves as a first cistron and the firefly luciferase gene as a second cistron. B, IRES activity in the 5Ј-NCRs of selective mRNAs. HeLa cells were transiently transfected with the indicated plasmids and Renilla (Rluc) and firefly luciferase (Fluc) activities were measured. Fluc/Rluc ratios are displayed in reference to the construct with 5Ј-NCR of ␤-actin, whose Fluc/Rluc ratio was set to 1. Error bars indicate measurements from three independent experiments. Note, that the NPM1 cDNA comprises the predicted 97-bp 5Ј-NCR plus the first 9 codons, which contain 4 in-frame ATG codons. C, integrity of dicistronic mRNAs in transfected HeLa cells. Total RNA was isolated from transfected cells and analyzed by Northern blot hybridization using hybridization probes complementary to firefly luciferase sequences. The migrations of RNA markers are indicated. D, translation of selected dicistronic mRNAs in HeLa S10 lysates. Dicistronic RNA was in vitro transcribed with T7 RNA polymerase. A dicistronic construct lacking a 5Ј-NCR insert was used as a negative control. 20 ng/ml of the dicistronic RNAs were added to the lysate and incubated at 30°C for 1 h followed by dual luciferase assays (29). The ratios of Fluc/Rluc and error bars were determined by three independent measurements. ODC protein peaked 8 h after release from S phase, but dropped dramatically thereafter (12). It is possible that the up-regulation of ODC mRNA and protein occurred in G 2 , before cells entered mitosis. Because we focused in the present study on the translation activity of mRNAs in cells arrested in mitosis, up-regulation of ODC in G 2 would have not been scored in our assays. Nevertheless, it is clear from all studies (8,11) that IRES element can be regulated in a cell cycle-specific manner and that IRESs are preferentially used by the mitotic translation apparatus. The task will now be to identify the different regulatory elements in mitotically active IRES elements and the proteins and cellular structures that modulate their activity. The outcome of these lines of investigation will determine roles for the internal initiation mechanism in cell cycle progression and point to posttranscriptional circuits that regulate cell growth.