Deregulation of translational control of the 65-kDa regulatory subunit (PR65 alpha) of protein phosphatase 2A leads to multinucleated cells.

Efficient translation of the mRNA encoding the 65-kDa regulatory subunit (PR65α) of protein phosphatase 2A (PP2A) is prevented by an out of frame upstream AUG and a stable stem-loop structure (ΔG = −55.9 kcal/mol) in the 5′-untranslated region (5′-UTR). Deletion of the 5′-UTR allows efficient translation of the PR65α message in vitro and overexpression in COS-1 cells. Insertion of the 5′-UTR into the β-galactosidase leader sequence dramatically inhibits translation of the β-galactosidase message in vitro and in vivo, confirming that this sequence functions as a potent translation regulatory sequence. Cells transfected or microinjected with a PR65α expression vector lacking the 5′-UTR, express high levels of PR65α, accumulating in both nucleus and cytoplasm. PR65α overexpressing rat embryo fibroblasts (REF-52 cells) become multinucleated. These data and previous results (Mayer-Jaekel, R. E., Ohkura, H., Gomes, R., Sunkel, C. E., Baumgartner, S., Hemmings, B. A., and Glover, D. M.(1993) Cell 72, 621-633) suggest that PP2A participates in the regulation of both mitosis and cytokinesis.

Protein phosphorylation/dephosphorylation is a crucial intracellular control mechanism. The phosphorylation state of a protein is the net result of the antagonistic activities of protein kinases and protein phosphatases. Protein phosphatase 2A (PP2A), 1 one of the major Ser/Thr protein phosphatases, consists of a catalytic (C) subunit of 36 kDa, complexed to a "constant" regulatory subunit of 65 kDa (PR65). This core heterodimer further associates with a single "variable" third subunit to form multiple trimeric holoenzymes. To date several classes of variable subunits with molecular sizes of 55 kDa (PR55), 72 kDa (PR72), 130 kDa (PR130), 54 kDa or 74 kDa have been identified (for reviews, see Refs. [1][2][3]. Given this complexity of regulatory subunits it can be anticipated that many different control mechanisms are likely to operate to coordinate the synthesis and assembly of the different holoenzymes. These controls will most probably act at both transcriptional and post-transcriptional levels. It appears that the level of PP2A catalytic subunit is tightly regulated because attempts to overexpress this protein have been unsuccessful even though it is possible to overexpress the mRNA (4,5). Recently, Wadzinski et al. (6) succeeded in expressing an amino-terminal tagged form of PP2A catalytic subunit, but this expression probably occurs at the expense of the genomically derived catalytic subunit. Taken together these results indicate that some form of translational or post-translational control mechanism operates to control the amount of the catalytic subunit. Whether such mechanisms operate to coordinate the translation of the PP2A regulatory subunits has not been investigated.
Translational control is often exerted at the level of translation initiation. According to the scanning hypothesis of mRNA translation initiation (reviewed in Ref. 7), the complex between initiator Met-tRNA and the 40 S ribosomal subunit, binds at or near the m 7 GpppG cap structure at the 5Ј end of the mRNA and scans in the 3Ј direction until an AUG codon is reached. The 60 S ribosomal subunit is subsequently recruited to the complex and translation starts. These steps are catalyzed by eukaryotic initiation factors. Two elements within the 5Ј-UTR of a mRNA are known to inhibit translation initiation. First, an AUG located upstream of the authentic start codon can inhibit translation initiation. If this upstream AUG is followed by a stop codon, the ribosome can resume scanning and reinitiate at the correct AUG start codon. The longer the distance between this upstream stop codon and the authentic start codon, the smaller the inhibitory effect of the upstream AUG. If the stop codon is located within the coding sequence (thus down stream of the authentic start AUG), inhibitory effects of the upstream AUG are maximal, since only ribosomes that ignore the upstream AUG will initiate at the correct start codon (8). Second, stable secondary structures in the 5Ј-UTR interfere with scanning of the ribosome and lead to inhibition of translation. Kozak (9) showed that secondary structures in the 5Ј-UTR must have a free energy of more than Ϫ30 kcal/mol to inhibit translation in COS-1 cells, although Sagliocco et al. (10) reported that structures with a free energy of Ϫ20 kcal/mol are able to inhibit translation in yeast. Besides these inhibitory mechanisms, other mechanisms for control of translation initiation exist, such as phosphorylation/dephosphorylation of initiation factors (7), but it is assumed that these have a more general effect on translation and are less messenger-specific.
In the present work we investigated whether expression of the PR65 subunit is controlled at the translational level. This subunit is common to all forms of PP2A characterized and therefore forms the scaffold for the assembly of the different trimeric holoenzymes. The PR65 subunit is encoded by two genes (11), termed PR65␣ and PR65␤. The ␣-isoform seems to be the most abundant in all tissues and cells examined, except in Xenopus oocytes, where the ␤-isoform predominates (12). We show that the 5Ј-UTR of PR65␣ acts as a translational repressor, due to an upstream AUG and to a stable stem-loop structure. We further show that release of this translational inhibition leads to overexpression of PR65␣ and that this apparently causes defects in cytokinesis resulting in multinucleated cells.
The 5Ј-UTR regions of C␣, PR55␣, PR55␤, and PR65␣ were deleted by introducing (using PCR) a unique SalI site upstream of the ATG start codon (at Ϫ28 bp for C␣, Ϫ24 bp for PR65␣), or a XbaI site at Ϫ18 for PR55␣ or a HindIII site at Ϫ29 for PR55␤. The resulting constructs were subcloned into either pBluescript (C␣, PR55␣, PR55␤) or pGEM (PR65␣) for in vitro transcriptions, or into pECE for transfections.
The 5Ј-UTR of PR65␣ was isolated by PCR using a full-length cDNA (11) as template and the following primers (HindIII sites underlined): ACAAGCTTCCGGTTCTCACTCTT (primer 1, sense), ATAAGCTTCAT-GGGG GAGTCA (primer 2, sense), and ATAAGCTTGGCTCCGTC-CCTTT (primer 3, antisense), resulting in 138-and 62-bp fragments, respectively. In the 62-bp fragment, the upstream ATG was mutated to ATT by PCR with primer 4 (ATAAGCTTCATTGGGGAGTCA, sense) and primer 3 (antisense). To introduce an ATG to ATT mutation in the 138-bp fragment, we performed PCR with primer 1 and primer 5 (AG-ATGACTCCCCAATGGA, antisense, HinfI site underlined), and introduced the product into the 138-bp 5Ј-UTR using an internal HinfI site. The PCR products are schematically presented in Fig. 4A. The resulting PR65␣ 5Ј-UTR constructs were subcloned in the 5Ј-UTR of the pSV-␤galactosidase vector (Promega) using an HindIII site. An in-frame stop codon is present 11 residues downstream of the HindIII site. The orientation of the 5Ј-UTR inserts was verified by dideoxy sequencing. Part of the coding sequence of ␤-galactosidase was excised from the pSV-␤-galactosidase vector by digestion with HindIII and EcoRV (thus encoding a polypeptide with a predicted molecular size of 49.8 kDa), and subcloned into the corresponding pBluescript sites to create pBS.⌬␤galactosidase. The PR65␣ 5Ј-UTR PCR fragments described above were then subcloned into the 5Ј-UTR using the HindIII site of pBS.⌬␤galactosidase.
In Vitro Transcription and Translation-pGEM or pBluescript constructs were linearized using a unique restriction site located close to the 3Ј end of the cDNA inserts, and used as templates for in vitro transcription from the T3, T7, or SP6 promoters. The reactions were performed in the presence of m 7 GpppG cap analogue (New England Biolabs) using the Riboprobe kit (Promega) according to the manufacturer's instructions. RNA was analyzed and quantitated on 1% agaroseformaldehyde gels.
Equal amounts of RNA (1 g), 10 Ci of [ 35 S]methionine (Amersham), and 10 l of cell-free rabbit reticulocyte lysate (Stratagene) were incubated for 1 h at 30°C with gentle shaking every 15 min. Protein translation products were analyzed on 10% SDS-PAGE followed by fluorography.
Antibodies were directed against recombinant PR65␣ (Ab65 recomb ) or against synthetic peptides of PR55␣ (Ab55␣ 1/19 ) and PR65␣ (Ab65 177/196 ) (12,13). All of these antisera have been extensively characterized and found to be specific for the appropriate antigen (12,13). For chloramphenicol acetyltransferase and ␤-galactosidase assays, cells were scraped into 40 mM Tris-HCl buffer (pH 7.5) containing 1 mM EDTA and 150 mM NaCl, centrifuged for 10 min at 800 ϫ g and resuspended in 150 l of 0.25 M Tris-HCl (pH 8). Cell pellets were freeze-thawed 3 times and centrifuged for 10 min at 12,000 ϫ g at 4°C. ␤-Galactosidase assays were carried out on the supernatant following the instructions of the manufacturer (Promega). Prior to chloramphenicol acetyltransferase assays (18), the extracts were heated for 10 min at 60°C. Protein concentrations were measured with the Protein Assay kit (Bio-Rad) using bovine serum albumin as standard.
For Mono-Q FPLC analysis, cells were homogenized in 50 mM Tris-HCl (pH 7.4), 1 mM EDTA, 10% glycerol, 0.5 mM phenylmethylsulfonyl fluoride, 0.5 mM benzamidine and centrifuged for 5 min at 12,000 ϫ g at 4°C. 1 mg of soluble total protein was loaded on a Mono-Q column and developed with a linear gradient from 100 to 600 mM NaCl in the same buffer. PP2A was assayed with a peptide substrate (Leu-Arg-Arg-Ala-Ser-Val-Ala), phosphorylated with cAMP-dependent protein kinase (19).
Rat embryo fibroblasts (REF-52) cells were cultured in Dulbecco's modified Eagle's medium, supplemented with 10% fetal calf serum as described previously (20). Prior to microinjection, cells were subcultured onto acid-washed glass coverslips. Microinjection was performed with a normal Leitz micromanipulator, using 0.1 M inner diameter micropipettes pulled on a linear horizontal puller. The pECE.PR65␣ expression plasmids were injected at a concentration of 0.5 mg/ml in a buffer containing 100 mM potassium glutamate, 40 mM potassium citrate, 1 mM MgCl 2 (pH 7.2), and 1 mg/ml mouse IgG (to act as a marker for injected cells). Cells were incubated for a further 24 -48 h and then processed for immunofluorescence. The results presented are the mean of five independent microinjection experiments involving 15-30 cells per experiment.
For immunofluorescence, cells were fixed in 3.7% formaldehyde and further treated as described previously (13,20). To mark the injected cells, cells were stained for the presence of mouse IgG using a fluorescein-conjugated anti-mouse IgG antibody (Organon Teknika).
Analysis of RNA Structure-Probing of the secondary structure of RNA was performed as described previously (21,22). The full-length PR65␣ cDNA clone in pGEM2 was digested with RsaI and used as template for in vitro transcription from the SP6 promoter. The resulting RNA (5 g) was 5Ј end labeled to a specific activity of 0.2 Ci/g with polynucleotide kinase and [␥-32 P]ATP (Amersham, specific activity 3000 Ci/mmol), gel-purified, and dissolved in water (10 6 cpm/25 l).
For renaturation, 15 l of RNA was mixed with 45 l of renaturation buffer (15 mM Tris-HCl (pH 7), 80 mM NaCl, 10 mM MgCl 2 , and 2 g/l tRNA), heated to 65°C and allowed to cool slowly to room temperature. The mixture (3 l) was incubated for 15 min at room temperature with distilled water (control) or with the indicated concentrations of enzyme (see below). Reactions were stopped by the addition of 7 l of stop solution, heated for 20 s at 80°C, and frozen on dry ice. Reaction products were analyzed on a denaturing 6% polyacrylamide sequencing gel. Concentrations of the enzymes were as following: ribonuclease A, 0.2 microunits per reaction; nuclease S 1 (Pharmacia), 2.5, 5, 10 units per reaction; ribonuclease V 1 (Pharmacia), 2, 4, 8 milliunits per reaction; ribonuclease T 1 , 18, 37, 75 milliunits per reaction. Lead acetate (pH 5.5) was used at a final concentration of 0.1, 0.25, or 0.5 mM.

Upstream AUGs Are a Common Element in the mRNAs of
Various PP2A Subunits-Most PP2A subunit mRNAs carry one or more upstream AUG codons in their 5Ј-UTR (Fig. 1). This is the case for C␤, PR65␣, PR55␤, PR55␥, PR72, and PR130 subunits. In the case of the C␤ subunit, the AUG is located just upstream of the major transcription initiation site (23), thus most C␤ messages will not contain this codon. To function as a start codon, an AUG should be located in the appropriate sequence context (24). Most efficient AUGs have a purine (most often A) in position Ϫ3 and a G in position ϩ4 (relative to the A of AUG, which is ϩ1). Upstream AUG codons are rather rare in the messengers of eukaryotic proteins. They occur in only 10% of all mRNAs, but in 70% of the mRNAs that encode proto-oncogenes (24). In this respect it is thought that they serve to regulate protein levels by down-regulating translation.
We deleted the 5Ј-UTR of C␣, PR55␣, PR55␤, and PR65␣ and compared in vitro translation efficiency of the full-length and 5Ј-UTR deleted mRNAs. Deletion of the 5Ј-UTR had no significant effect on the in vitro translation efficiency of C␣ and PR55␣ messages because they lack upstream AUGs. In contrast, the translation efficiency increased dramatically after deletion of the 5Ј-UTR of PR65␣ (see below) and PR55␤. These messages carry one and two upstream AUGs in their 5Ј-UTR, respectively. These data prompted us to further investigate the translational regulation of the PR65␣ subunit mRNA.
Analysis of the Secondary Structure of the PR65␣ 5Ј-UTR-Besides an upstream AUG, the 5Ј-UTR of the PR65␣ mRNA has another potential inhibitory element. As predicted by an RNA-fold program (25,26), the 5Ј-UTR can form a stable (⌬G ϭ Ϫ55, 9 kcal/mol) stem-loop structure ( Fig. 2A) with the potential to interfere with mRNA scanning by the ribosomal machinery. The predicted structure was obtained by analyzing the sequence spanning nucleotides Ϫ140 to ϩ1, covering the 5Ј-UTR of the PR65␣ messenger, using the RNA-fold program. If the first 36 nucleotides of the coding sequence are also included, the predicted structure of the 5Ј-UTR remains unchanged. This indicates that the predicted stem-loop might exist as a stable entity within the messenger. In addition to the stem-loop structure, we also found that nucleotides Ϫ134 to Ϫ40 within the 5Ј-UTR of PR65␣ have the potential to base pair with nucleotides 517-612 within the coding sequence (see "Discussion").
We checked whether a stem-loop structure occurred in the 5Ј-UTR of the Drosophila PR65␣ mRNA (27). Residues Ϫ194 to Ϫ47 have indeed the potential to fold as a stem-loop, but this structure is less stable (⌬G ϭ -27.1 kcal/mol). The homology of this region with the stem-loop of human PR65␣ is 37.6% (gap  (25,26). A stem and two major loops (⌬G ϭ Ϫ55.9 kcal/mol) of more than 5 nucleotides (residues Ϫ76 to Ϫ71 and residues Ϫ52 to Ϫ46) are predicted. The upstream AUG is indicated with bold circles. Shaded symbols refer to residues that are cleaved with ribonuclease A, nuclease S 1 , or lead acetate. The arrows point to the cleavage sites of ribonuclease T 1 , and the open arrow is used to indicate a weak cleavage site. B, the first 175 nucleotides of PR65␣ mRNA were labeled at the 5Ј end and cleaved at single-stranded G residues with ribonuclease T 1  weight ϭ 5, length weight ϭ 0.3), as tested with the GAP program (28). In contrast, no upstream AUGs are present in the 5Ј-UTR of Drosophila PR65 (27).
In order to investigate whether the predicted stem-loop structure of the 5Ј-UTR of human PR65␣ mRNA indeed occurred in solution, we probed the structure of the in vitro transcribed 5Ј-UTR with several ribonucleases and with lead acetate. The in vitro transcribed RNA was 175 nucleotides long and comprised the 5Ј-UTR plus 31 nucleotides of the coding sequence. By comparing the effect of ribonuclease T 1 (which cleaves single, but not double, stranded RNA at G residues) on denatured and renatured RNA (Fig. 2B), we were able to elucidate the secondary structure of the 5Ј-UTR in solution. A stretch of 5 G residues starting at nucleotide Ϫ60 allowed us to locate the observed structures in the sequence. The structure was then further analyzed (not shown) with reagents that specifically cleave single-stranded RNA (lead acetate, ribonuclease A, and S 1 nuclease) or double-stranded RNA (ribonuclease V 1 ). Our analysis confirmed the predicted loop at residues Ϫ51 to Ϫ45. The predicted loop at position Ϫ76 to Ϫ71 was not detected, but an additional loop at postion Ϫ64 to Ϫ59 was observed. Surprisingly, in the observed structure the upstream AUG (residue Ϫ62 to Ϫ60) is exposed in this loop. We used the RNA-fold program to calculate the stability (⌬G ϭ Ϫ39.0 kcal/ mol) of this observed structure (using the "prevent" command to insert a loop between residues Ϫ64 and Ϫ59). Taken together these results demonstrate that a stem-loop structure is indeed formed by the 5Ј-UTR of PR65␣, albeit somewhat less stable than predicted, but within the range that is reported to inhibit mRNA scanning (9).
Deletion of the 5Ј-UTR of PR65␣ Allows Efficient Translation in Vitro and Overexpression in COS-1 Cells-The cDNA encoding PR65␣ was used as a template for in vitro transcription and subsequent translation in rabbit reticulocyte lysates. No protein product was observed with the full-length mRNA as template (Fig. 3A). However, when the 5Ј-UTR was deleted at nucleotide Ϫ25 relative to the start ATG codon, and used as template for in vitro transcription, the resulting mRNA was efficiently translated into a protein product of about 65 kDa. This translation was not observed with the antisense construct.
The same full-length and 5Ј-UTR deleted cDNAs were subcloned in a mammalian expression vector (pECE) and used to transiently transfect COS-1 cells. Whereas transient transfection with both the full-length and the 5Ј-UTR deleted construct resulted in efficient overexpression of the PR65 message (analyzed by Northern blot, data not shown), only the 5Ј-UTR deleted construct resulted in efficient overexpression of PR65␣ protein as judged by immunoblotting (Fig. 3B). Overexpression was about 20-fold as judged from scanning of different exposures of the Western blot. (The data in Fig. 3B show a longer exposure which underestimates the level of overexpression relative to the control.) A small increase (approximately 2-fold) in the amount of PR65␣ was also observed with the full-length construct. , and with a 5Ј-UTR deleted PR65␣ expression construct (panels E and F). The images in panels C, D, and E were acquired at the same laser power, pinhole size, and photomultiplier sensitivity settings and are therefore directly comparable. Panel F shows the same field to that seen in panel E, but was collected with CLSM settings corresponding to a 10-fold reduced exposure. Cells in panels E and F with lower immunostaining (i.e. not overexpressing the PR65␣) are indicated by arrowheads. Bar, 5 m.
In a similar experiment COS-1 cells were transiently transfected with PR65␣ expression constructs with or without the 5Ј-UTR and subsequently analyzed by indirect immunofluorescence with PR65 specific antibodies as described under "Materials and Methods." When cells transfected with the 5Ј-UTR deleted PR65␣ were analyzed about one out of 5 cells displayed an intense staining for PR65 (Fig. 3, E and F) which was approximately 10 times stronger than in the surrounding cells. This low staining of the surrounding cells corresponded to that observed in non-transfected (not shown) and mock transfected cells (Fig. 3C). Therefore we concluded that the intensely stained cells actively overexpressed PR65. COS-1 cells transfected with the full-length PR65␣ construct were indistinguishable from untransfected cells (Fig. 3D), again confirming that the 5Ј-UTR allowed only very low levels of expression. The observed immunofluorescence signal was specific for PR65␣ as it could be competed with the appropriate antigen (data not shown; see also Ref. 13). In addition, we obtained identical results with an antisera raised against the recombinant protein (Ab65 recomb ). The distribution of the overexpressed PR65 was found to change slightly when different time points after transfection were analyzed. In about 30% of the transfected cells (harvested 48 h after transfection), the overexpressed PR65␣ was almost exclusively nuclear, whereas in about 70% of the transfected cells, PR65␣ was present in both nucleus and cytoplasm. The reasons for this almost exclusive nuclear staining remain to be further investigated. By 72 h most if not all transfected cells stained homogeneously for PR65␣ in the cytoplasm and the nuclear compartment. This distribution corresponds to that found for endogenous PR65 (Fig. 3C and Ref. 13).
The 5Ј-UTR of PR65␣ Is a Translational Inhibitor-To directly demonstrate the importance of the 5Ј-UTR of PR65␣, we fused several different constructs of the 5Ј-UTR (Fig. 4A) to the coding sequence of a reporter protein (a fragment of ␤-galactosidase) and compared the in vitro rate of translation of the chimeric messages. Translation of reporter mRNA was dramatically inhibited (89% compared to the control, as judged by PhosphorImager quantitation) if the 5Ј-UTR of PR65␣ was present as a leader sequence (Fig. 4B). Truncation of the 5Ј-UTR at nucleotide Ϫ63, results in a construct that is devoid of the stem-loop structure but still contains the upstream AUG and inhibits translation of the reporter by 72%. This inhibition is mainly due to the upstream AUG, since mutation of this AUG to AUU partially relieves the inhibition (30% inhibition compared to the control). To further investigate the importance of the secondary structure, we made a construct of the entire 5Ј-UTR in which the upstream AUG was mutated to AUU. Since we have shown above that the upstream AUG is present in a loop structure, one might expect that this mutation would FIG. 4. The 5-UTR of PR65␣ is a translational inhibitor. A, schematic presentation of the various constructs of the 5Ј-UTR of PR65␣. The constructs correspond to either the entire 5Ј-UTR (stem/ loop with AUG), the entire 5Ј-UTR containing a point mutation in the upstream AUG (stem/loop with AUU), the 5Ј-UTR truncated at residue Ϫ63 and thus devoid of secondary structure (AUG), or the same construct containing a point mutation in the upstream AUG (AUU). The stability of each construct was calculated with the RNA-fold program (25,26). B, in vitro translation of chimeric mRNA encoding a fragment of ␤-galactosidase fused to various constructs of the 5Ј-UTR of PR65␣ (see panel A). As a negative control no RNA was added. C, COS-1 cells were transiently transfected with either pSV-␤-galactosidase or with the same vector containing various constructs of the 5Ј-UTR of PR65␣ (see panel A), inserted in the ␤-galactosidase leader sequence. Cells transfected with the pECE vector without insert were used as negative control. 48 h after transfection, cell extracts were assayed for ␤-galactosidase activity. The activity is expressed per microgram of protein in the cell extract. The figure shows means Ϯ S.E. (n ϭ 3), as indicated by the error bars. Transfection efficiency was tested by cotransfection of a pSV-chloramphenicol acetyltransferase plasmid. Variations in chloramphenicol acetyltransferase activity were less than 10%. have no influence on the secondary structure. When fused to the reporter message this construct inhibited translation by 55%.
The same set of PR65␣ 5Ј-UTR sequences (see Fig. 4A) were ligated into the 5Ј-UTR of the ␤-galactosidase gene (in the pSV vector) and used to transiently transfect COS-1 cells (Fig. 4C). We first demonstrated by Northern blotting that insertion of the 5Ј-UTR in the expression plasmid did not influence transcriptional efficiency (not shown). The 5Ј-UTR of PR65␣ inhibits translation of the ␤-galactosidase message dramatically (79 Ϯ 4%). The secondary structure alone is responsible for a moderate inhibition of translation (17 Ϯ 5%), whereas the AUG alone inhibits translation by 57% (Ϯ5%). In contrast to the in vitro data, mutation of the upstream AUG and disruption of the secondary structure, resulted in a slight stimulation of translation by 15% (Ϯ7%).
Taken together, the results demonstrate that the 5Ј-UTR of PR65␣ is a strong translational inhibitor, mainly due to the presence of an upstream AUG, and to a lesser (but significant) extent to the presence of a stem-loop structure. Apparently, the inhibitory effects of secondary structure and upstream AUG are additive.
Overexpression of PR65␣ Leads to Multinucleated Cells-The 5Ј-UTR of PR65␣ controls translation and prevents efficient overexpression of the protein. Deletion of the 5Ј-UTR overcomes this inhibition and allows overexpression of PR65␣. The PP2A activity in total cell extracts (as measured with a peptide substrate, and thus reflecting the amount of catalytic subunit) is unchanged in the overexpressing cells (61 milliunits/mg compared to 58 millliunits/mg in mock-transfected cells). We used the DEAE-dextran method followed by a dimethyl sulfoxide shock treatment to ensure high transfection efficiency, as reported by Sussman and Milman (29). We found that PR65␣ overexpression caused no differences in the elution pattern of PP2A holoenzymes, when total cell extracts are chromatographed on a Mono-Q ion exchange column (not shown). The overexpressed PR65␣ elutes at around 230 mM NaCl, which is exactly the position where recombinant PR65␣ elutes, 3 indicating that at least the majority of the overexpressed protein is apparently free and not sequestered to form PP2A holoenzymes.
COS-1 cells overexpressing PR65␣ show no apparent phenotype. This might be explained by the observation that the cells are essentially non-dividing in the 48 -72-h period after DEAEdextran transfection. (The presence of more than one nucleus or fragmentated nuclei in the COS-1 cell transfection experiment did not correlate with PR65 overexpression. Indeed nontransfected, mock transfected, and cells transfected with the full-length PR65␣ expression construct (Fig. 3D) showed the same proportion of abnormal nuclei compared to COS-1 cells overexpressing the PR65.) To investigate possible effects of PR65␣ overexpression on cell division, we microinjected REF-52 cells with PR65␣ expression constructs. When cells were microinjected with a 5Ј-UTR deleted PR65␣ expression construct this resulted in a dramatic increase of PR65␣ protein detected in both the nucleus and the cytoplasm (data not shown). As determined by indirect immunofluorescence and confocal laser scanning microscopy this increase of PR65 protein was 10 -20-fold compared to control uninjected cells. In contrast, microinjection of the full-length PR65␣ expression construct did not alter PR65 protein levels (data not shown). These results agree with those obtained using COS-1 cells (Fig. 3). Analysis of PR65␣ overexpressing cells 24 h after microinjection revealed major phenotypic changes during mitosis. The most marked effect observed was the formation of cells containing multiple nuclei (Fig. 5, A and  B). Whereas only 3% of the non-injected control cells were binucleated, 65% of the PR65␣ overexpressing cells contained two or more nuclei. The remaining 35% of the overexpressing cells did not divide in the 24 h after microinjection. REF-52 cells microinjected with a full-length PR65␣ expression plasmid did not show any phenotypic changes (Fig. 5, C and D). These results indicate that cytokinesis is blocked in PR65 overexpressing cells without apparently affecting nuclear division. DISCUSSION In the present work we show that deregulation of the translational control of PR65␣ mRNA leads to overexpression of PR65␣, which apparently disrupts cytokinesis and leads to biand multinucleated cells. This translational control is mediated by the 5Ј-UTR of PR65␣, which is a strong translational inhibitor, due to the presence of an upstream AUG and a stable stem-loop structure.
Translation of mRNAs with stable secondary structures in the 5Ј-UTR may be highly dependent on the helicase activities of eIF-4B and eIF-4F. The latter factor is a complex between eIF-4A, eIF-4E, and p220 (7). Phosphorylation of eIF-4E and eIF-4B increases after insulin treatment of fibroblasts (30) and this coincides with an increase of the translation of ornithine decarboxylase mRNA. Under basal conditions translation of this message is inhibited by a stable stem-loop structure (⌬G ϭ Ϫ68.2 kcal/mol) in the 5Ј-UTR. Moreover, overexpression of eIF-4E in NIH 3T3 cells overcomes the translational inhibition of cyclin D1 (31). These data suggest that phosphorylation of initiation factors may lead to a more efficient unwinding of 5Ј-UTRs, and thereby up-regulate the translation of certain mRNAs. Furthermore, an eIF-4E interacting protein (4E-BP1 or PHAS-I) has been identified, which inhibits translation (32,33). Phosphorylation of 4E-BP1 in response to insulin causes its dissociation from eIF-4E, and relieves the inhibition.
In the case of PR65␣, however, unwinding of the 5Ј-UTR may not be sufficient, since the upstream AUG alone inhibits translation (Fig. 4). The use of an alternative promoter, or alternative splicing, might produce messages that lack the inhibitory 5Ј-UTR and would therefore be efficiently translated (34). However, none of the PR65␣ cDNA clones isolated so far contain different 5Ј-UTR sequences, in fact the majority of the isolated cDNAs starts just downstream of the start ATG. 4 This probably reflects the inability of reverse transcriptase to proceed through a stable stem-loop structure.
In addition to a stem-loop structure, nucleotides Ϫ134 to Ϫ40 of the 5Ј-UTR of PR65␣ have the potential to base pair with nucleotides 517-612 in the coding sequence. This potential base pairing and the stem-loop structure in the 5Ј-UTR are mutually exclusive. The ability of the 5Ј-UTR to base pair with an internal coding sequence is not a unique characteristic of PR65␣, but is also found in the c-myc proto-oncogene transcripts. Saito et al. (35) suggest that a translocated c-myc gene, which lacks the exon encoding the 5Ј-UTR, is no longer under translational control and becomes oncogenic. Another example where abrogation of translation control leads to proto-oncogene activation is the lymphocyte-specific tyrosine kinase lck (36). The 5Ј-UTR of lck contains three upstream AUGs. Substitution of the 5Ј-UTR for retroviral sequences results in malignant transformation, as is observed in some murine lymphomas (36).
Another characteristic of certain proto-oncogenes transcripts, such as the c-myc transcript, is the presence of potential RNA-destabilizing sequences in the 3Ј-UTR (37). The AUUUA sequence, which is thought to mediate rapid mRNA turnover, is also found in most PP2A subunit messages. Five copies of this motif are present in the 3Ј-UTR of C␣, six in C␤, one in PR65␤, four in PR55␣, seven in PR55␤, one in PR72, and four in PR130. The striking exception is PR65␣ itself, which seems to be devoid of this motif. The absence of rapid turnover signals might explain why translation of the PR65␣ signal is tightly controlled. It therefore appears that the cell controls PP2A subunit transcripts in much the same way as mRNAs that encode proteins involved in growth control, such as proto-oncogenes.
Overexpression of PR65␣ leads to defects in cytokinesis and multinucleated cells. Cytokinesis is brought about by the contraction of actin-myosin fibers in the cleavage furrow. Injection of myosin antibodies specifically blocks cytokinesis, without affecting chromosome movement, and ultimately leads to multinucleated cells (38). Although speculative at the moment, one possible mechanism to explain our results is that overexpression of PR65␣ results in inefficient dephosphorylation and activation of myosin light chain kinase, a presumed PP2A substrate (1), resulting in hypophosphorylated myosin light chains which in turn could inhibit the contractile force in the cleavage furrow. The question that emerges from this study is, how does the overexpression of PR65␣ disrupt the normal regulation of PP2A? As judged from its elution on a Mono-Q column most of the overexpressed PR65␣ seems to be present as a free protein, i.e. not complexed with the catalytic or other regulatory subunits. Furthermore, neither the elution profile, nor the amount, of the PP2A trimer (C/PR65/PR55) is influenced by overexpression of PR65␣ in COS-1 cells. Also the total PP2A activity (as measured with a peptide substrate) remains unchanged. Two possibilities exist to explain the results described in this article. First, the overexpression of PR65␣ could disrupt a PP2A holoenzyme which plays a specific role in cytokinesis. PR65␣ could do so by sequestering the catalytic or variable subunit. Although we do not detect such dimers in a Mono-Q profile, we cannot exclude that a low abundance PP2A trimer is dissociated upon PR65␣ overexpression. Second, the excess PR65␣ could act as an inhibitor of the catalytic subunit of PP2A, as predicted from in vitro data (39). In this model it is necessary to suggest that the catalytic subunit (or the complex of the catalytic subunit with a variable subunit) is released from the PR65 subunit at a certain point in the cell cycle to dephosphorylate specific targets. Free PR65␣ in overexpressing cells would in this model immediately capture the released catalytic subunit and suppress its activity. We have recently obtained evidence that indicates that PP2A undergoes subunit rearrangements (12,13,40). Interestingly, a very similar phenotype (defects in cytokinesis and multinucleated cells) is observed in budding yeast when the TPD3 gene, encoding the PR65 homologue, is mutated (41).
In summary, the data presented in this paper, and in other recent publications (19, 40 -42) demonstrate the importance of PP2A in cell cycle regulation. In this context it should be pointed out that substrates of the cyclin-dependent protein kinase family need to be dephosphorylated prior to the subsequent round of cell division. This dephosphorylation is most likely not a stoichastic process, but stringently regulated by the action of PP2A and other protein phosphatases.