NIPP1-mediated Interaction of Protein Phosphatase-1 with CDC5L, a Regulator of Pre-mRNA Splicing and Mitotic Entry*

NIPP1 is a regulatory subunit of a species of protein phosphatase-1 (PP1) that co-localizes with splicing factors in nuclear speckles. We report that the N-terminal third of NIPP1 largely consists of a Forkhead-associated (FHA) protein interaction domain, a known phosphopeptide interaction module. A yeast two-hybrid screening revealed an interaction between this domain and a human homolog (CDC5L) of the fission yeast protein cdc5, which is required for G2/M progression and pre-mRNA splicing. CDC5L and NIPP1 co-localized in nuclear speckles in COS-1 cells. Furthermore, an interaction between CDC5L, NIPP1, and PP1 in rat liver nuclear extracts could be demonstrated by co-immunoprecipitation and/or co-purification experiments. The binding of the FHA domain of NIPP1 to CDC5L was dependent on the phosphorylation of CDC5L, e.g.by cyclin E-Cdk2. When expressed in COS-1 or HeLa cells, the FHA domain of NIPP1 did not affect the number of cells in the G2/M transition. However, the FHA domain blocked β-globin pre-mRNA splicing in nuclear extracts. A mutation in the FHA domain that abolished its interaction with CDC5L also canceled its anti-splicing effects. We suggest that NIPP1 either targets CDC5L or an associated protein for dephosphorylation by PP1 or serves as an anchor for both PP1 and CDC5L.

Type 1 protein phosphatases (PP1) 1 belong to the PPP family of Ser/Thr protein phosphatases and regulate diverse cellular processes such as transcription, pre-mRNA splicing, intracellular transport, and metabolism (1)(2)(3). They consist of a single catalytic subunit (PP1 C ) and one or two regulatory subunits. The regulatory subunits act as substrate specifiers and anchor the holoenzymes in specific cell compartments in close vicinity to their substrates. In addition, the regulatory subunits mediate the control of the holoenzymes by hormones and growth factors through interaction with allosteric effectors or through phosphorylation by specific protein kinases. It has been estimated that mammalian cells contain tens of different regulatory proteins of PP1 (4). Altogether about 20 of these have already been characterized and cloned, including the glycogenbinding G-subunits, the myosin-binding M-subunits, the cytosolic regulator inhibitor-1, and the nuclear RNA-binding protein NIPP1 (1)(2)(3). Recent investigations have revealed that these regulatory proteins have multiple points of interaction with PP1 C , including a common phosphatase binding motif with the consensus sequence RVXF (5)(6)(7)(8)(9)(10). In addition, most regulatory subunits contain domains that mediate the binding to substrates (e.g. myosin for the M-subunits) and/or a subcellular structure (e.g. glycogen for the G-subunits) to which the substrates are bound.
In nuclear extracts, NIPP1 (39 kDa) is present as an inactive complex with PP1 C , termed PP1N NIPP1 (11). This heterodimer can be activated by phosphorylation of up to 4 Ser/Thr residues in the central domain of NIPP1 by protein kinases A and CK2 (12), which disrupts its interaction with PP1 C via the RVXF motif without dissociation of the two polypeptides (10,12). The C terminus of NIPP1 binds in vitro to A/U-rich RNA sequences (13)(14)(15). Since the binding of NIPP1 to RNA and PP1 C are not mutually exclusive, we have proposed that NIPP1 may function as an RNA-targeting subunit of PP1 (15). NIPP1 also co-localizes with splicing factors in the nuclear speckles, which largely correspond to the "interchromatin granule clusters" and represent storage or recycling sites for pre-mRNA splicing factors (16).
Although the N-terminal third of NIPP1 is not required for the binding of PP1 C or RNA, it represents by far the most conserved fragment of NIPP1, suggesting an essential role for this domain. We show here that the N-terminal third of NIPP1 largely consists of a "Forkhead-associated" (FHA) domain that is present in a large variety of mostly nuclear proteins and preferentially binds to specific phosphoproteins (17)(18)(19)(20)(21). The FHA domain of NIPP1 was found to interact with a human homolog (CDC5L) of Schizosaccharomyces pombe cdc5, a regulator of pre-mRNA splicing (22,23), and this interaction was dependent on the phosphorylation of CDC5L, e.g. by cyclin E-Cdk2. Combined with observations that the FHA domain of NIPP1 blocked pre-mRNA splicing in nuclear extracts, these data provide firm evidence for a regulatory role of PP1N NIPP1 in nuclear RNA processing.

EXPERIMENTAL PROCEDURES
Materials-PP1 C and PP2A C were prepared from rabbit skeletal muscle (12 (24) were produced by baculovirus expression in High Five TM cells (Invitrogen). Rabbit polyclonal antibodies against synthetic peptides comprising the C termini of CDC5L (YADLLLEKETLKSKF) and NIPP1 (PGKKPTPSLLI), coupled to keyhole limpet hemocyanin, were affinity-purified on the bovine serum albumin-coupled peptides linked to CNBr-activated Sepharose 4B (25). Rabbit polyclonal antibodies against a peptide (RTPLRDKLNINPEDG) from the central domain of CDC5L were prepared as described previously (26). Goat anti-rabbit and anti-mouse secondary antibodies were obtained from Dako. Texas Red-linked sheep anti-mouse secondary antibodies and CNBr-activated Sepharose 4B were delivered by Amersham Pharmacia Biotech. Monoclonal hemagglutinin (HA) antibodies (12CA5), restriction enzymes, and Pwo DNA-polymerase were purchased from Roche Diagnostics. Oligonucleotide primers for polymerase chain reaction and CY5-labeled fluoprimers were delivered by Eurogentec.
Preparation of Constructs and Recombinant Proteins-Polyhistidinetagged polypeptides (NIPP1 1-142 , NIPP1 225-351 , and CDC5L 260 -606 ) were produced in BL21(DE3)pLysS cells by transformation with the pET16b plasmid. The alanine mutation of residues 68 -71 in NIPP1  was obtained using the QuickChange site-directed mutagenesis kit (Stratagene). The proteins were purified on Ni 2ϩ -PDC columns (Affiland). A construct encoding NIPP1 1-351 C-terminally fused to the enhanced green fluorescent protein (EGFP) was made by cloning the coding sequence of bovine NIPP1 into the XhoI/SacII sites of pEGFP-N 1 (CLONTECH). GST fusions of CDC5L 272-606 , NIPP1  , and the alanine mutant of residues 68 -71 of NIPP1 1-142 were obtained by subcloning the corresponding cDNAs in the pGEX-2TK plasmid. Following transformation of BL21(DE3)pLysS cells, the expression of the fusion proteins was induced with 1 mM isopropyl-8-D-thiogalactopyranoside for 2 h. The cells were lysed by sonication in buffer A containing 50 mM Tris at pH 7.5, 1 mM dithiothreitol, 0.1% 2-mercaptoethanol, 5 M leupeptin, 0.5 mM phenylmethanesulfonyl fluoride, 0.5 mM benzamidine, and 450 mM NaCl. After centrifugation (15 min at 27,000 ϫ g), the supernatant was diluted with 2 volumes of buffer A without NaCl and applied to a glutathione-Sepharose column. The GST fusion proteins were eluted with a buffer containing 0.1 M Tris at pH 8 and 10 mM reduced glutathione. The pCDNA3 plasmid expressing CDC5L with a C-terminal HA tag has been described before (26). The sequence of all DNA constructs was verified by cycle sequencing on an ALFII sequencer (Amersham Pharmacia Biotech).
Yeast Two-hybrid Screening-We used the dual reporter strategy introduced by Brent and co-workers (27). The FHA domain of NIPP1 (NIPP1 1-142 ), the FHA-mutant, and sds22  were cloned in frame with the LexA protein (which includes a DNA-binding domain) into the pEG202 plasmid. For the screening of interacting proteins, a HeLa cell library was used in which the cDNAs were subcloned behind a galactose-inducible promotor (pJG4 -5 plasmid) and in frame with the B42 activation domain. Interacting proteins were identified by growth of the yeast strain EGY188 in a Ϫleucine/ϩgalactose medium. The use of a second plasmid-borne lacZ reporter gene (pSH18 -34) with upstream LexA operators enabled a second, independent screening involving the expression of ␤-galactosidase in the presence of galactose. The plasmids of the positive clones were isolated, transformed into E. coli XL1-Blue cells, and sequenced.
Immunological and Biochemical Techniques-Rat liver nuclear extracts were prepared as described (11), except that the extraction buffer contained only 0.15 M NaCl. Immunoprecipitations and (far) Western blots were performed as detailed in Ref. 11, except for the presence of 20 mM NaF and 1 mM Na 3 VO 4 during the preparation of nuclei and nuclear extracts. For the assay of PP1 C in immunoprecipitates from nuclear extracts that had been prepared with protein phosphatase inhibitors, trypsinolysis was done in the presence of Mn 2ϩ (1 mM MnCl 2 ), which is known to release free catalytic subunit and to revert the inhibition by fluoride (2).
For affinity chromatography, 3 mg of recombinant polyhistidinetagged NIPP1   Phosphorylations were performed in a buffer containing 50 mM glycylglycine at pH 7.4, 0.5 mM dithiothreitol, 5 mM 2-mercaptoethanol, 0.1 mM ATP (with or without [␥-32 P]ATP), 2 mM magnesium acetate, and the indicated protein kinase(s). One unit of cyclin E-Cdk2 incorporates 1 nmol of phosphate/min into histone H1. Phosphorylation reactions were terminated by boiling the samples in SDS-sample buffer or by the addition of 4 mM EDTA.
For GST pull-down assays (Fig. 6), polyhistidine-tagged CDC5L 260 -606 was incubated overnight with phosphorylation buffer (see above) in the presence of 0.3 mM ATP and 6 mM magnesium acetate, with or without cyclin E-Cdk2 (50 milliunits/ml). The phosphorylation reaction was arrested by the addition of 12 mM EDTA. GST and the GST fusions of the FHA domain (NIPP1 1-142 ) or the FHA mutant were incubated for 30 min at 4°C with glutathione-agarose beads (Sigma) that had been preblocked with TBS plus bovine serum albumin (1 mg/ml), 0.5% Triton X-100, and 1 mM dithiothreitol and then equilibrated with TBS plus 1 mM dithiothreitol. After washing with TBS plus 1 mM dithiothreitol, the glutathione-agarose beads were incubated with (phosphorylated) CDC5L 260 -606 for 1 h at 4°C. The beads were washed once with TBS, 1 mM dithiothreitol, 0.1% Triton X-100 and twice with TBS plus 1 mM dithiothreitol. The associated proteins were separated by 10% Tricine-SDS-polyacrylamide gel electrophoresis and visualized with antibodies against a central synthetic fragment of CDC5L. The same GST pull-down procedure was also used for the analysis of the association between full-length NIPP1, or an in vitro reconstituted complex of NIPP1 and PP1 C , and Cdk2-phosphorylated GST-CDC5L 260 -606 . The co-precipitation of PP1N NIPP1 and CDC5L was evaluated by the assay of phosphorylase phosphatase activity after a preincubation with trypsin, which liberates PP1 C (12). Finally, the GST pull-down assay was also used to analyze the phosphorylation dependence of the interaction of CDC5L in nuclear extracts to GST-NIPP1 1-142 ( Fig. 7). For that purpose, nuclear extracts were prepared in the absence or presence of the protein phosphatase inhibitors NaF (10 mM) and Na 3 VO 4 (1 mM). An aliquot of the nuclear extract that was prepared without phosphatase inhibitors was additionally incubated with PP1 C (6 nM) plus PP2A C (50 nM) for 10 min at 30°C before the addition of GST-NIPP1  .
Cell Culture, (Immuno)fluorescence Microscopy, and FACS Analysis-COS-1 cells or HeLa cells were grown on sterile glass coverslips (Nunc) in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. After 1 day, the cells were transfected with the plasmids encoding NIPP1 1-351 -EGFP and CDC5L-HA using the FuGene TM 6 Transfection reagent (Roche Diagnostics). One day later, cells were washed with PBS and fixed at Ϫ20°C with a solution consisting of 90% methanol, 10 mM MES at pH 6.9, 1 mM MgCl 2 , and 1 mM EDTA. The cells were washed again with PBS and permeabilized with PBS containing 1% Nonidet P-40, followed by washes with PBS containing 0.1% Tween 20. Subsequently, the permeabilized cells were blocked in PBS containing 0.1% Tween 20, 3% bovine serum albumin, followed by an incubation with HA antibodies (16). NIPP1 1-351 -EGFP was visualized by direct green fluorescence, and CDC5L-HA was visualized by red fluorescence using Texas Red-labeled secondary antibodies (26).
FACS analysis was performed on COS-1 cells or HeLa cells that had been transfected with NIPP1 1-351 -EGFP or NIPP1 1-142 -EGFP. More than 60% of the transfected cells expressed the fusion proteins, as determined by fluorescence microscopy. The cells were fixed with ethanol, washed with PBS and stained with propidium iodide as fluorochrome, and analyzed with a FACScan flow cytometer using the Mod-FIT program of Becton Dickinson (28).
In Vitro Splicing-HeLa cell nuclear extracts were prepared as described (29). Splicing assays were performed using a uniformly labeled, capped ␤-globin pre-mRNA fragment comprising exon 1 through the BamHI site in exon 2, according to the procedure of Mayeda and Krainer (30).

RESULTS
FHA Families-A ProfileScan analysis 2 showed that the N terminus of NIPP1 contains a previously unrecognized FHA domain (Fig. 1). Sequence alignments of various FHA-containing proteins revealed that there actually exist at least two families of FHA domains. NIPP1 contained an FHA domain highly similar to that found in the putative proline isomerase pinA (GenBank TM accession no. U78757) and in kanadaptin (GenBank TM accession no. AF035526), a protein that binds to a Cl Ϫ :HC0 3 Ϫ exchanger (31). Surprisingly, the FHA domain of kanadaptin was mapped upstream of the suggested open reading frame, indicating that kanadaptin may be a larger protein than previously predicted.
Interaction of NIPP1 and CDC5L-NIPP1 1-142 , encompassing the FHA domain of NIPP1, was used as a bait in a yeast two-hybrid screen to search for interacting proteins expressed from a HeLa cell library. Six positive clones encoded a fragment of the human homolog of fission yeast cdc5 (Fig. 2B), known as CDC5L (for "cdc5-like") (26). No interaction was detected with empty bait or prey plasmids or with a prey plasmid that contained the cDNA for sds22, a structurally unrelated regulatory subunit of PP1 ( Fig. 2A). The interaction with CDC5L was abolished when residues 68 -71 of the FHA domain were simultaneously mutated to alanine. These residues correspond to a conserved structural feature of FHA motifs that was shown to be essential for the interaction with phosphopeptides (21).
Further evidence for a NIPP1-CDC5L interaction came from observations that CDC5L could be immunoprecipitated from liver nuclear extracts with antibodies against either NIPP1 or CDC5L (Fig. 3A). Both immunoprecipitates also contained a PP1 C -binding protein with the size of NIPP1, as detected by far Western blotting with digoxygenin-labeled PP1 C (Fig. 3B). Furthermore, the NIPP1 and CDC5L immunoprecipitates also contained PP1 C (Fig. 3C), as indicated by the detection of phosphorylase phosphatase activity that could be inhibited by inhibitor-2 (not shown). While the NIPP1 antibodies immunoprecipitated a large fraction of the CDC5L that could be precipitated with the CDC5L antibodies (Fig. 3A), the latter antibodies immunoprecipitated only a small fraction of NIPP1 and PP1 C (Fig. 3, B and C). Collectively, these data suggest that a large fraction of CDC5L in nuclear extracts was associated with NIPP1 but that NIPP1 and its associated PP1 C were present in a large molar excess.
Still another line of evidence for an interaction between NIPP1 and CDC5L was the finding that CDC5L in nuclear extracts was retained by NIPP1 1-142 -Sepharose and could be largely eluted with 0.3 M NaCl (Fig. 4). By contrast, no CDC5L was bound to NIPP1 225-351 -Sepharose, in further agreement with a specific interaction of CDC5L with the N terminus of NIPP1.
NIPP1 and CDC5L Co-localize in the Nuclear Speckles-Both NIPP1 (16) and CDC5L (23) have been reported to localize in nuclear speckles, together with splicing factors. To determine whether NIPP1 and CDC5L co-localize, EGFP-fused NIPP1 and HA-tagged CDC5L were transiently co-expressed in COS-1 cells. Both proteins gave a speckled nuclear distribution (Figs. 5, A and B). Remarkably, in the larger speckles CDC5L was sometimes more prevalent peripherally. An overlay showed a partial co-localization of CDC5L and NIPP1 (Fig. 5C). However, some speckles that contained NIPP1 did not contain detectable CDC5L. Moreover, NIPP1 was also abundant in the core of the larger speckles that were relatively deficient in CDC5L.
The NIPP1-CDC5L Interaction Is Controlled by Phosphorylation-While the above data show that NIPP1 and CDC5L are part of the same complex in the nucleus, they do not prove a direct interaction between both components. A direct interaction between recombinant NIPP1 and CDC5L 260 -606 could not be demonstrated, however, by co-precipitation (see below) and far Western analysis (not shown). In view of recent findings FIG. 2. NIPP1 interacts with CDC5L in a yeast two-hybrid assay. A shows the results of a ␤-galactosidase staining in yeast cells that were transformed with the indicated bait and prey plasmids as well as with the lacZ reporter gene, as detailed under "Experimental Procedures." B shows the sequence of the CDC5L fragment that interacted with NIPP1 1-142 in the two-hybrid assay. Also indicated are the consensus phosphorylation sites for proline-directed protein kinases (thick underline). that FHA domains represent phosphopeptide-interacting motifs (20, 21), we explored the role of the phosphorylation of CDC5L on its interaction with the FHA domain of NIPP1. CDC5L 258 -614 contains numerous consensus sites ((S/T)PX 0,1 -(R/K)) for phosphorylation by proline-directed kinases (Fig.  2B). CDC5L 260 -606 was indeed an excellent substrate for in vitro phosphorylation by cyclin E-Cdk2, reaching final stoichiometries of phosphorylation of 4.3 mol of phosphate/mol of CDC5L 260 -606 (not shown). Moreover, following its phosphorylation by cyclin E-Cdk2, CDC5L 260 -606 interacted with a GST fusion of the FHA domain of NIPP1, as determined by coprecipitation analysis (Fig. 6). On the other hand, phosphorylated CDC5L 260 -606 did not co-precipitate with a GST fusion of an FHA domain in which residues 68 -71 were substituted for an alanine. Collectively, these data indicate that the NIPP1-CDC5L interaction required both a functional FHA domain and phosphorylation of CDC5L. We have also found that full-length NIPP1 or an in vitro reconstituted complex between NIPP1 and PP1 C specifically co-precipitated with phosphorylated GST-CDC5L 272-606 (not illustrated). This last finding indicates that the binding of PP1 C and CDC5L to NIPP1 are not mutually exclusive.
The pull-down assays with GST-NIPP1 1-142 were also used to investigate whether the binding of NIPP1 1-142 to CDC5L from hepatic nuclear extracts was also phosphorylation-de-pendent (Fig. 7). We found that CDC5L from nuclear extracts co-sedimented with GST-NIPP1 1-142 when the extracts were prepared in the presence of phosphatase inhibitors, and this interaction was not seen with GST-NIPP1  in which residues 68 -71 were mutated into alanine. By contrast, a much smaller fraction of CDC5L was retained by GST-NIPP1  when the extracts were prepared in the absence of phosphatase inhibitors. Finally, no CDC5L could be shown to bind to GST-NIPP1 1-142 when the extracts were in addition preincubated with PP1 C and PP2A C . Thus, these data indicate that at least part of CDC5L is phosphorylated in hepatic nuclei and that this phosphorylation is required for its interaction with NIPP1.
We have also explored whether Cdk2-phosphorylated CDC5L is a substrate for dephosphorylation by PP1 C (Table I).
Compared with the reference substrate phosphorylase a, CDC5L was dephosphorylated by PP1 C with a 9-fold lower V max and a 4-fold lower K m . Thus, CDC5L was dephosphorylated with an efficiency (V max /K m ) that was only 2-fold lower than that of phosphorylase a. As previously demonstrated for phosphorylase a (12), the dephosphorylation of CDC5L by PP1 C was completely blocked by the addition of NIPP1. Also, NIPP1 became a less potent inhibitor of CDC5L dephosphorylation following its phosphorylation by protein kinases A plus CK2 (not illustrated). We have also found that CDC5L could be dephosphorylated by PP2A C (Table I). Compared with phosphorylase a, the dephosphorylation of CDC5L by PP2A C occurred with a 2-fold higher V max and a 6-fold higher K m (i.e. with a 2.7-fold lower catalytic efficiency). Overall, PP1 C acted severalfold more efficiently on CDC5L than did PP2A C .
Role of the FHA Domain of NIPP1 in Pre-mRNA Splicing-Since CDC5L has been implicated in both the G 2 /M transition and in pre-mRNA splicing (22,23), we have investigated whether the FHA domain of NIPP1 perhaps also plays a role in these cellular processes. We found that the expression in COS-1 cells or HeLa cells of either full-length NIPP1 or its FHA domain (residues 1-142), C-terminally fused to the enhanced green fluorescent protein, did not affect the number of cells in the G 2 /M transition, as determined by FACS analysis (not illustrated). On the other hand, the N terminus of NIPP1 blocked the splicing of ␤-globin pre-mRNA in HeLa cell nuclear extracts (Fig. 8). This inhibition was dependent on the concentration of the FHA domain and was complete at 10 M. By contrast, the mutated FHA domain, which did not interact with CDC5L ( Figs. 2A, 6, and 7), did not show any anti-splicing effect (Fig. 8). DISCUSSION

FHA Domains Are Phosphopeptide-interacting Motifs-The
FHA domain of the PP2C-like protein phosphatase KAPP only recognizes the phosphorylated form of a receptor-like protein kinase domain (18). Likewise, the FHA domains of the RAD53 protein kinase are selectively recruited to phosphorylated Rad9, a component of the DNA damage checkpoint (19 -21). We report here that the FHA domain of NIPP1, which belongs to a different family of FHA-containing proteins (Fig. 1), also interacts with CDC5L in a phosphorylation-dependent manner. Moreover, the alanine mutation of residues 68 -71 in the FHA domain of NIPP1 abolished its interaction with CDC5L ( Figs.  2A, 6, and 7). Since residues 68 -71 of NIPP1 correspond to a loop region in the FHA2 domain of RAD53 that is essential for the binding to phosphopeptides (21), this effect of the mutation strongly indicates that the NIPP1-CDC5L interaction requires an intact FHA domain and is mediated by a phosphorylated fragment of CDC5L.
The Role of CDC5L and PP1N NIPP1 in Pre-mRNA Splicing-Various observations implicate both CDC5L and PP1N NIPP1 in nuclear pre-mRNA processing. First, both CDC5L (23, 33) and FIG. 3. Co-immunoprecipitation of CDC5L, NIPP1, and PP1 C from nuclear extracts. Liver nuclear extracts, prepared in the presence of protein phosphatase inhibitors, were used for immunoprecipitations (IP) with protein-A TSK and NIPP1 antibodies or CDC5L antibodies or buffer (none). The pellets were washed three times with TBS plus 0.1% Nonidet P-40. Aliquots of the washed pellets were used for Western analysis with antibodies against CDC5L (A), for far Western analysis with DIG-PP1 C (B), and for the assay of phosphorylase phosphatase activity after trypsinolysis of the fractions in the presence of MnCl 2 (C). The large band of 40 -50 kDa in A is due to cross-reaction of the secondary antibodies with the large IgG chain in the immunoprecipitates. The phosphorylase phosphatase activity in C was inhibited 95-97% when assayed in the presence of 1 M inhibitor-2, showing that the activity stems from PP1 C .  Fig. 5) co-localize with splicing factors. Cyclin E-Cdk2, which phosphorylates CDC5L and enables its interaction with NIPP1 (Figs. 6 and 7), is also associated with splicing factors in mammalian cells (34). Second, PP1 (35) and fission yeast cdc5 (23) are required for the first step in pre-mRNA splicing. Third, immunodepletion of NIPP1 or the addition of a dominant negative NIPP1 mutant inhibited pre-mRNA splicing in vitro (16). We report here that the FHA domain of NIPP1 also blocks splicing in nuclear extracts (Fig. 8). Since the FHA mutant did not show this anti-splicing effect, it is likely that the inhibition is mediated by the interaction of the FHA domain with CDC5L. A likely explanation for the inhibition of splicing by the FHA domain of NIPP1 is that it prevents the interaction of endogenous NIPP1 with CDC5L.
The exact role of CDC5L and PP1N NIPP1 in pre-mRNA splicing remains to be established. One view is that NIPP1 functions as a CDC5L-targeting subunit of PP1. In accordance with this hypothesis, we found that CDC5L can be dephosphorylated in vitro by PP1 C with an efficiency that is similar to that of the model substrate phosphorylase a (Table I). However, at variance with the function of other targeting subunits of PP1, such as the glycogen-binding and myosin-binding subunits, NIPP1 did not enhance the dephosphorylation of CDC5L by the associated PP1 C but, in contrast, potently inhibited its dephosphorylation. The possibility cannot be excluded, however, that the dephosphorylation of CDC5L is affected by other components of the CDC5L-PP1N NIPP1 complex or is controlled by additional (covalent) modifications. Thus, we found that NIPP1 became a less potent inhibitor of CDC5L dephosphorylation following its phosphorylation by protein kinases A and CK2, similar to what we have previously described for the dephosphorylation of other substrates of PP1 (12). An alternative view is that the physiological substrate(s) of PP1N NIPP1 is not (only) CDC5L but (also) CDC5L-associated protein(s). There is no information yet on CDC5L-associated proteins in higher eukaryotes, except that CDC5L is part of in vitro assembled spliceosomes (33) and is targeted to the nuclear speckles in COS-1 cells (Ref. 23, Fig. 5). However, fission yeast cdc5 (22) and its budding yeast homolog Cef1 (36) have been shown to be part of a multiprotein complex, which in the latter case was shown to bind tightly to the spliceosomes during splicing. Perhaps mammalian homolog(s) of Cef1-associated proteins represent the physiological substrate(s) of PP1N NIPP1 . PP1 has also been implicated in the recycling of splicing factors from sites of active splicing to the nuclear speckles, which for the most part represent storage and assembly sites for splicing factors (37). Thus, CDC5L and PP1N NIPP1 may be involved in both the splicing reaction itself and in spliceosome disassembly. NIPP1 was present in some nuclear speckles that contained little or no CDC5L (Fig. 5). Moreover, the CDC5L antibodies only precipitated a small fraction of NIPP1 (Fig. 3). Therefore, the association of NIPP1 with the speckles can only partly be accounted for by its interaction with CDC5L, and there may exist additional protein(s) in the speckles that bind to NIPP1. Alternatively, the association of CDC5L with the speckles is mediated by NIPP1. In the latter case, NIPP1 may turn out to be a targeting subunit of both PP1 C and CDC5L.
The Role of CDC5L and PP1N NIPP1 in the G 2 /M Transition-We did not find an effect of overexpressed NIPP1 in COS-1 cells and HeLa cells on the number of cells in the G 2 /M transition. On the other hand, fission yeast cdc5 was originally identified in a screen for "cell division cycle" mutants as a gene that is required for the G 2 /M transition (38). Likewise, Cef1 is essential for mitotic progression (22,36). Mammalian cells overexpressing CDC5L have a shortened G 2 and a reduced cell size, whereas cells overexpressing a carboxyl-terminal fragment of CDC5L display a slowed G 2 progression (39). Some investigations have focused on the function of cdc5 as a putative transcription factor to explain its role in the G 2 /M transition. The N terminus of cdc5 and its homologs contains two Myb-like repeats with evenly spaced hydrophobic residues that may make up the backbone of the DNA-binding motif (40). These repeats were found to be essential for the function of Cef1 (41) and to bind preferentially to the DNA sequence CT-CAGCG (42). The central region of CDC5L, which is included in the NIPP1 interacting domain (Fig. 2B), shows similarities with the hydrophilic, proline-rich transactivation domain of some transcription factors. Moreover, a chimeric molecule consisting of N-terminally nicked CDC5L fused to the GAL4 DNAbinding domain transactivated a reporter gene in COS cells (39). Thus, CDC5L may function both as a transcription and a splicing factor. These functions are not mutually exclusive, since there is increasing evidence for a link between cell cycle and splicing (22,34). One can speculate on the existence of a checkpoint that ensures that cells enter into mitosis only when there are no pre-mRNAs left. Alternatively, some transcripts Hepatic nuclear extracts were prepared with or without the protein phosphatase inhibitors Na 3 VO 4 (1 mM) and NaF (20 mM). Lanes 1 and 2 show the level of CDC5L (92 kDa) in these extracts, as revealed by Western blotting. Lanes 3-8 show the binding of CDC5L from these extracts to GST-NIPP1 1-142 and to GST-NIPP1 1-142 that was alaninemutated in residues 68 -71 (FHA-mutant), as determined by the same pull-down assay that was also used in Fig. 6. Lanes 5 and 6 show the absence of binding to the GST fusions of CDC5L from nuclear extracts that had been prepared without protein phosphatase inhibitors and had been incubated for 10 min at 30°C with 6 nM PP1 C and 50 nM PP2A C before addition to the GST fusions.  8. The FHA domain of NIPP1 blocks pre-mRNA splicing in nuclear extracts. The splicing of radioactively labeled ␤-globin pre-mRNA in HeLa nuclear extracts was carried out for 150 min at 30°C. The splicing assay was also supplemented with the indicated concentrations of NIPP1 1-142 or NIPP1 1-142 that was alanine-mutated in residues 68 -71 (FHA-mutant). The splicing products were separated by denaturing polyacrylamide (9%) gel electrophoresis and visualized by autoradiography. The diagrams to the left indicate the migration of the splicing substrate and products. encoding cell cycle regulators may be spliced in a cell cycle-dependent manner. The association of cyclin-dependent kinase(s) with the spliceosomes (34) is highly suggestive of cell cycleregulated splicing. It should also be noted that CDC5L was identified as a mitotic phosphoprotein (43) that can be phosphorylated in vitro by cyclin B-cdc2. Thus, both cyclin B-cdc2 (43) and cyclin E-Cdk2 (this work) may represent CDC5L kinases. However, we found that CDC5L is phosphorylated in nuclear extracts from the liver (Fig. 7), which mostly consists of cells in G 0 , when cyclin E-Cdk2 is inactive. Therefore, we suggest that there are other proline-directed protein kinases that can phosphorylate CDC5L during G 0 and control its interaction with NIPP1.
The identification of additional interacting proteins of CDC5L and NIPP1 is likely to result in a better understanding of their role in pre-mRNA splicing and cell cycle progression. Moreover, an analysis of the regulation of these proteins by phosphorylation may also lead us to the physiological substrates of PP1N NIPP1 .