Inhibition of Spliceosome Assembly by the Cell Cycle-regulated Protein Kinase MELK and Involvement of Splicing Factor NIPP1*

NIPP1 is a ubiquitous nuclear protein that is required for spliceosome assembly. We report here that the phosphothreonine-binding Forkhead-associated domain of NIPP1 interacts with the cell cycle-regulated protein Ser/Thr kinase MELK (maternal embryonic leucine zipper kinase). The NIPP1-MELK interaction was critically dependent on the phosphorylaton of Thr-478 of MELK and was increased in lysates from mitotically arrested cells. Recombinant MELK was a potent inhibitor of an early step of spliceosome assembly in nuclear extracts. This splicing defect was also seen with a kinase-dead mutant but was absent after mutation (T478A) of the NIPP1 binding site of MELK, indicating a mediatory role for NIPP1. Our data suggest that MELK has a role in the cell cycle-regulated control of pre-mRNA splicing.

More recently, we have demonstrated that NIPP1 is also implicated in transcription as well as in pre-mRNA splicing by mechanisms that do not involve PP1 (8,9). In transient transfection experiments, NIPP1 acted as a transcriptional repressor, which may be accounted for by the binding of the central and C-terminal domains of NIPP1 to the Polycomb protein, EED (embryonic ectoderm development) (9). The latter promotes transcriptional repression by the recruitment of a histone methyltransferase and histone deacetylases. NIPP1 also appears to be required for the assembly of spliceosomes, the protein-RNA complexes that catalyze pre-mRNA splicing (8). The spliceosomal function of NIPP1 requires its C-terminal domain as well as its N-terminal Forkhead-associated (FHA) domain, an established phosphothreonine-binding module. The FHA domain of NIPP1 mediates targeting to both the spliceosomes and the nuclear storage sites for splicing factors, known as "speckles" (8,10). The targeting function of the FHA domain of NIPP1 is likely explained by its ability to bind to phosphorylated forms of the essential splicing factors CDC5L (11) and SAP155 (12).
Here we show that the protein kinase MELK, which is structurally related to the AMP-activated protein kinases, also interacts in a phosphorylation-dependent manner with the FHA domain of NIPP1 and that this interaction is increased during mitosis. Furthermore, we demonstrate that recombinant MELK blocks spliceosome assembly by a mechanism that involves NIPP1. Our data suggest a novel link between pre-mRNA processing and cell cycle progression.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Screening-NIPP1- , cloned into the pEG202 vector in-frame with the LexA DNA-binding domain, was used as bait for the screening of a HeLa cell cDNA library (11). In this library, the cDNAs are subcloned behind a galactose-inducible promoter in the pJG4 -5 vector in-frame with the B42 activation domain. Interacting proteins were identified by growth of the yeast strain EGY188 in a Ϫleucine/ϩ galactose medium. The use of a 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. For the mapping of the NIPP1-MELK interaction site, MELK-(1-651), MELK-(1-291/590 -651), and 10 different Thr to Ala mutants of MELK-(326 -651) were subcloned in the pJG4 -5 plasmid. MELK-(1-651), kindly provided as clone KIAA0175 by Dr. Nagase (Kazusa DNA Research Institute, Chiba, Japan), was subcloned by PCR with primers CTGGAATTCATGAAAGATTATGAT-GAACTTCTCA and TCTGCTCGAGTTATACCTTGCAGCTAGATA-GGAT into the EcoRI/XhoI sites of pJG4 -5. MELK-(1-291/590 -651) was obtained by cutting pJG4 -5-MELK-(1-651) with Bsp1407I and religation. All constructs were verified by DNA sequence analysis. The interaction between NIPP1-(1-142) and the MELK variants was quantified by measuring ␤-galactosidase activity in a liquid culture assay using 2-nitrophenyl-␤-D-galactopyranoside as substrate (Clontech yeast protocols handbook).
Antibodies-Bacterially expressed polyhistidine-tagged MELK was used to raise antibodies in rabbits. The MELK antibodies were affinitypurified on His-MELK linked to CNBr-activated Sepharose 4B (Amersham Biosciences). Anti-GST antibodies (sc-459), Anti-His antibodies (sc-803), and Anti-LexA antibodies (Sc-7544) were obtained from Santa Cruz Biotechnology. Swine anti-rabbit and rabbit anti-mouse antibodies were purchased from Dako. Anti-FLAG antibodies were obtained from Stratagene. Mouse monoclonal anti-HA antibodies (12CA5) were delivered by Roche Diagnostics.
Preparation of Recombinant MELK (Mutants)-Wild-type MELK and MELK-(326 -651) were cloned into the pET16b vector in-frame with a polyhistidine tag. The His-tagged proteins were purified by chromatography on Ni 2ϩ -pentadentate-chelator-Sepharose (Affiland). For the expression of FLAG-tagged MELK (mutants) in COS-1 cells the cDNAs of human MELK-wild-type, MELK-T478A, MELK-D150A, and MELK-1-266 were cloned as BamHI fragments into the pSG5-Flag vector. The first three cDNAs were also cloned as XhoI-EcoRI fragments into the baculovirus transfer vector pBleubac4.5. Recombinant viruses expressing MELK variants were generated using the Bac-N-Blue TM transfection kit (Invitrogen). For the production of recombinant MELK proteins, BT1-TN-5B1-4 (High Five™) Trichoplusia ni cells were grown as monolayer cultures at 27°C in Express Five serum-free medium (Invitrogen). The cells were infected with the recombinant viruses at a multiplicity of infection of 10 and harvested after 48 h by low-speed centrifugation. As most of MELK and its mutants were expressed in insoluble inclusion bodies, we followed the procedure described by Berndt and Cohen (13) to dissolve and renature the insoluble species.
Cell Cultures-COS-1 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 100 units/ml of both penicillin and streptomycin. Cells were blocked in mitosis by the addition of colcemid (0.14 g/ml) 20 -24 h before lysis. 48 h after transfection, the cells were washed twice with ice-cold phosphate-buffered saline and lysed in 50 mM Tris/HCl, pH 7.5, 0.3 M NaCl, 0.5% Triton X-100, 0.5 mM dithiothreitol, 0.5 mM phenylmethanesulfonyl fluoride, 0.5 mM benzamidine, and 5 M leupeptin. After sonication, the lysates were cleared by centrifugation (10 min at 16,000 ϫ g), and the supernatants were used for GST pull-down assays.
GST Pull-down Assays-For the expression of NIPP1 as a GST fusion in mammalian cells, the cDNA of wild-type NIPP1 was cloned into the EcoRV-NotI sites of the pGMEX-T1 vector. Glutathione-agarose beads were pre-blocked with Tris-buffered saline plus bovine serum albumin (1 mg/ml) and 0.5% Triton X-100. After washing with Tris-buffered saline plus 0.1% Nonidet P-40, the beads were incubated with COS-1 cell lysates (Fig. 6) for 2 h at 10°C. Subsequently, the beads were sedimented (30 s at 1000 ϫ g) and washed twice with Tris-buffered saline plus 0.1% Nonidet P-40 and twice with Tris-buffered saline, and finally the associated proteins were analyzed by Tricine-SDS-polyacrylamide gel electrophoresis and immunoblotting. GST pull-down experiments with recombinant proteins and HeLa cell nuclear extracts were performed as described previously (11).
Pre-mRNA Splicing in Nuclear Extracts-A capped ␤-globin pre-mRNA fragment, comprising exon 1 through the BamHI site in exon 2, was synthesized in the presence of [␣-32 P]GTP. This primary transcript was used as a substrate for splicing in HeLa cell nuclear extracts (8). Spliceosomal complexes were separated by native gel electrophoresis as described previously (8). , which is essential for catalytic activity, Thr-167 (T167), which is a candidate substrate for phosphorylation/activation of MELK (see the Discussion), and Thr-478 (T478), which is essential for interaction with NIPP1. B, two-hybrid interaction of the indicated MELK and NIPP1 fragments. NIPP1-(1-142) mutant refers to the NIPP1 fragment with a mutated phosphate-binding loop (S68A/R69A/V70A/ H71A). The interactions were quantified by liquid ␤-galactosidase assays. The results are expressed as units Ϯ S.E. for five observations. C, immunoblot of yeast extracts showing expression of the bait and prey proteins with antibodies against LexA and HA, respectively.

MELK Is a Novel Interactor of NIPP1-A yeast two-hybrid
screening of a HeLa cell library with the N-terminal third of NIPP1 (residues 1-142) as bait yielded, in addition to clones encoding CDC5L (11) and SAP155 fragments (12), 16 prey clones that all encoded the C-terminal half of the poorly characterized protein kinase MELK (for maternal embryonic leucine zipper kinase) (14) (Fig. 1). Compared with this MELK fragment, the full-length protein was a much better interactor with NIPP1 in two-hybrid assays (Fig. 1B). However, a deletion mutant of MELK, lacking the TP-rich domain, did not interact with NIPP1-(1-142). The MELK-NIPP1 interaction was lost by mutation (S68A/R69A/V70A/H71A) of the phosphate-binding loop of the FHA domain of NIPP1 (Fig. 1B), indicating that MELK interacted with the FHA domain of NIPP1 via a phosphorylated residue(s). Fig. 1C shows that all bait and prey proteins were well expressed, further validating the two-hybrid data.
We have previously demonstrated that CDC5L (11) and SAP155 (12) bind to the FHA domain of NIPP1 via threoninephosphorylated TP dipeptide motifs. The fragment of MELK that binds to NIPP1 also contains numerous TP motifs (Fig.  1A). To examine whether one or several of these TP motifs are involved in the interaction with NIPP1 we performed twohybrid assays after mutation (Thr to Ala) of each of the 10 TP motifs in the NIPP1-interaction domain of MELK (Fig. 2). All mutants were expressed well (Fig. 2B). However, four mutants showed a significantly reduced interaction with NIPP1 in the two-hybrid assays. The T478A mutant in particular was severely impaired in binding NIPP1. These data implicate some TP motifs of MELK, in particular the TP motif comprising Thr-478 and Pro-479, in the interaction with NIPP1.
Interaction of Purified Components-To further explore whether MELK and NIPP1 interact directly and in a phosphorylation-dependent manner, we expressed the NIPP1-interacting C-terminal domain of MELK (residues 326 -651) as a Histagged protein in bacteria. Purified recombinant MELK-(326 -651) did not sediment with GST-tagged NIPP1-(1-142) (Fig. 3). However, we noted that a preincubation of MELK-(326 -651) with a HeLa cell nuclear extract in the presence of Mg-ATP enabled its interaction with GST-NIPP1-(1-142). Importantly, a co-sedimentation with GST-NIPP1-(1-142) was not seen after a preincubation of the T478A mutant with nuclear extracts under phosphorylation conditions (Fig. 3). The above data showed that MELK interacts directly with the FHA domain of NIPP1 and that this interaction is largely mediated by phosphorylated Thr-478. In further agreement with this interpretation, we found that the synthetic phosphopeptide MELK-(470 -487)-phospho-Thr-478 could be used as a competitor to disrupt the interaction between GST-NIPP1-(1-142) and its FHA ligand, CDC5L, in nuclear extracts (Fig. 3B). By contrast, the nonphosphorylated MELK peptide did not have this effect.
MELK Is an Inhibitor of Spliceosome Assembly-To explore the role of the NIPP1-MELK interaction, we prepared purified bacterially expressed NIPP1 and baculovirusexpressed MELK. In addition to wild-type MELK we generated a NIPP1-binding mutant (MELK-T478A) as well as a kinasedead mutant (MELK-D150A). The latter is mutated in the essential "DFG" triplet in kinase subdomain VII, which is required for the binding of Mg-ATP (15). All three proteins were expressed in Sf9 insect cells as insoluble inclusion bodies and were inactive. However, the inclusion bodies could be dissolved in 6 M guanidinium chloride, and MELK could be renatured by extensive and rapid dilution with a buffer containing Mn 2ϩ , as described previously for the catalytic subunit of PP1 (13). The renatured wild-type MELK was capable of phosphorylating exogenous substrates such as myelin basic protein (Fig. 4) as well as histones H1, 2A, and 3 (not shown). In addition, autophosphorylation of MELK was detected (Fig. 4). As expected, the D150A mutant was inactive, but surprisingly, at equal protein concentrations the T478A mutant was much more active than the wild-type protein.
Recombinant NIPP1 did not affect the ability of baculovirusexpressed MELK to autophosphorylate or to phosphorylate myelin basic protein (not shown). Because NIPP1 is an established pre-mRNA splicing factor, we subsequently examined whether the interactor MELK also has a role in pre-mRNA splicing. Fig. 5A shows that the addition of 1 M wild-type MELK completely blocked the splicing of a fragment of ␤-globin pre-mRNA by HeLa cell nuclear extracts. Interestingly, the kinase-dead mutant of MELK had the same effect, indicating that the splicing inhibition was independent of phosphoryla- tion by MELK. On the other hand, no splicing inhibition was seen with the NIPP1-binding mutant, indicating a mediatory role for NIPP1. No splicing intermediates were detected in the presence of MELK or MELK-D150A, suggesting that MELK interfered with spliceosome assembly rather than with splicing catalysis per se. In further agreement with this conclusion we found that MELK did not affect pre-mRNA splicing when added after spliceosome assembly, i.e. after a preincubation of the splicing extracts for 20 min (not shown). Spliceosome assembly occurs stepwise (Complex E 3 Complex A 3 Complex B 3 Complex C) and involves the recruitment of small nuclear ribonucleoproteins and splicing factors as well as intraspliceosomal rearrangements (16). Native gel electrophoresis revealed that MELK (Fig. 5B) and MELK-D150A (not shown) blocked an early step of spliceosome assembly, i.e. a step before the assembly of the A-complex. This implies that MELK inhibits either the assembly of the E-complex or the transition from the Ecomplex to the A-complex.
The MELK-NIPP1 Interaction Is Increased during Mitosis-The previous data suggested that the MELK-NIPP1 interaction results in an inhibition of pre-mRNA splicing. One of the cellular processes associated with a complete block of pre-mRNA splicing is mitosis. Because MELK appears to be maximally phosphorylated during mitosis (17,18) and only interacts with NIPP1 when phosphorylated (this study), we reasoned that the MELK-NIPP1 interaction might be increased during mitosis. To explore this hypothesis, we co-expressed GST-tagged NIPP1 and FLAG-tagged MELK in COS-1 cells. The amount of MELK that co-precipitated with GST-NIPP1 from the cell lysates was indeed dramatically increased when the cells were blocked in mitosis by the addition of either colcemid (Fig. 6A) or nocodazole (not shown) at 20 -24 h before cell lysis; this increase could not be explained by different expression levels of the GST-NIPP1 or FLAG-MELK fusions (Fig. 6, B and C). The same results were obtained with the kinase-dead mutant MELK-D150A, showing that MELK kinase activity is not required for the phosphorylation of MELK on a site(s) that mediates its interaction with NIPP1. Importantly, MELK-T478A was a weak interactor of NIPP1 in cell lysates from both asynchronized and mitotically arrested cells, in accordance with our two-hybrid (Fig. 1) and GST pull-down experiments (Fig. 3). MELK-(1-266), which lacks the NIPP1 interaction site but contains the kinase catalytic domain (Fig.  1A), did not bind to NIPP1 in either condition. The above results demonstrated that NIPP1 and MELK interact with each other in vivo and that their interaction is considerably increased during mitosis, when pre-mRNA splicing is abolished.

DISCUSSION
The MELK-NIPP1 Interaction-We have shown here that MELK is an interactor of the FHA domain of NIPP1 and that this interaction involves the binding of a threonine-phosphorylated TP motif to the phosphate binding loop of the FHA domain. The TP motif of MELK that appears to be most critical for the interaction with NIPP1 in yeast (Fig. 2), in COS-1 cells (Fig. 6), and in vitro with recombinant proteins (Fig. 3) comprises Thr-478 and Pro-479. However, as we have previously already demonstrated for SAP155 (12), it is possible that other TP motifs of MELK, when phosphorylated, can also mediate binding to NIPP1. It will therefore be important to identify the protein kinase(s) that phosphorylates the TP motifs of MELK and to examine whether the activation of this kinase(s) results in an increased interaction with NIPP1. Interestingly, all three established ligands of the FHA domain of NIPP1, i.e. CDC5L (19), SAP155 (12), and MELK (17,18), are hyperphosphorylated during mitosis, and at least in the case of SAP155 (12) and MELK (this work), this phosphorylation includes threonine residues that bind directly to the FHA domain of NIPP1. Thus, one or more mitotic kinases promote the interaction between the splicing factor NIPP1 and its FHA ligands, providing a novel link between cell cycle progression and pre-mRNA processing. Additional evidence for such a link comes from observations that the overexpression of the splicing factor CDC5L shortens the G 2 phase (20), whereas the overexpression of MELK is associated with an accumulation of cells in G 2 /M (18). It will be interesting to delineate the contribution of NIPP1 to these effects of CDC5L and MELK. Our preliminary data suggest that the overexpression of MELK-T478A also causes a G 2 /M block (not illustrated), indicating that this effect of MELK does not depend on the recruitment of NIPP1. Along the same lines, it will be important to investigate whether the increased SAP155-NIPP1 and MELK-NIPP1 interactions contribute to the splicing arrest during mitosis or rather are implicated in the re-initiation of splicing in the early G 1 phase. It should be pointed out that other splicing factors have also been implicated in connecting RNA splicing with cell cycle progression (21). For example, it has recently been demonstrated that the SR protein SRp38 is required for splicing repression during mitosis (22).
The mitotic kinase that phosphorylates MELK on Thr-478 is unlikely to be MELK itself, since a kinase-dead mutant still showed an increased interaction with NIPP1 during mitosis (Fig. 6). We also found that recombinant MELK is not an in vitro substrate for cyclin-B/Cdk1 (not shown), indicating that Thr-478 is phosphorylated by a distinct mitotic kinase. It should also be pointed out that the increased phosphorylation of MELK on Thr-478 during mitosis is not necessarily the result of an increased MEL kinase kinase activity but could also be accounted for by a reduced MELK phosphatase activity.
NIPP1 and Pre-mRNA Splicing-The inhibition of spliceosome assembly by MELK does not require a functional catalytic site but depends on Thr-478 (Fig. 5), which is essential for the binding of NIPP1. One possible explanation for this inhibition of spliceosome assembly is that MELK competes with other splicing factors for binding to the FHA domain and thereby prevents the recruitment of NIPP1 to the spliceosomes. We have previously demonstrated that the C-terminal third of NIPP1 is required for a late step (B 3 C transition) of spliceosome assembly and that the formation of the B-complex proceeds normally in the presence of NIPP1-(1-142) (8). Our observation that wild-type MELK, but not the NIPP1-binding mutant, blocks an early step of spliceosome assembly (Fig. 5) could therefore suggest that NIPP1-(1-142) is needed for an early step of spliceosome assembly. Inhibition of spliceosome assembly by MELK in splicing extracts did not require a functional catalytic domain. However, this does not rule out the possibility that MELK also affects pre-mRNA processing in vivo by protein phosphorylation. For example, we have noted that CDC5L and SAP155, as well as NIPP1, are in vitro substrates for phosphorylation by MELK, which could contribute to the regulation of splicing by MELK.
The interaction of the FHA domain of NIPP1 with its ligands is correlated with both splicing catalysis and splicing arrest. On the one hand, a functional FHA domain is required for spliceosome assembly (8), and ligands of the FHA domain, such as  4). After 120 min the RNA was phenolextracted and subjected to denaturing PAGE (9%). The splicing products are indicated schematically. B, splicing of the ␤-globin pre-mRNA fragment was stopped at the indicated time points by the addition of heparin, and the spliceosomal complexes were separated by native gel electrophoresis. Radioactively labeled RNA was detected by autoradiography. The position of the H ϩ E complexes and A ϩ Bϩ C complexes are indicated. H refers to nonspecific preribonucleoprotein complexes.
FIG. 6. The MELK-NIPP1 interaction is increased during mitosis. A, FLAG-tagged MELK (mutants) and GST-tagged NIPP1 were co-expressed in COS-1 cells. 20 -24 h before lysis the cells were either kept dividing asynchronically or were mitotically arrested by the addition of colcemid. GST-NIPP1 was pulled down from the cell lysates with glutathione-agarose, and the co-sedimented MELK (mutants) were visualized by immunoblot analysis with FLAG antibodies. GST-NIPP1 in the pellet (B) was visualized with GST antibodies. The expression of the MELK variants in the cell lysates was detected by immunoblot analysis with FLAG antibodies (C). The arrows indicate the positions of MELK variants.
MELK, competitively block spliceosome assembly (Fig. 5). On the other hand, the FHA domain of NIPP1 is also required for the targeting of NIPP1 to the nuclear storage sites of splicing factors (10), and NIPP1 shows an increased interaction with SAP155 (12) and MELK (Fig. 6) during mitosis, when splicing is arrested. One explanation for these paradoxical findings is that the binding of NIPP1 to some FHA ligands is associated with splicing arrest, whereas the association with other ligands promotes spliceosome assembly. Alternatively, the FHA-mediated targeting of NIPP1 to the nuclear speckles during interphase and to the mitotic storage sites of splicing factors serves as a "priming" mechanism that enables a rapid resumption of pre-mRNA splicing by the recruitment of pre-assembled complexes of splicing factors during spliceosome assembly. Nevertheless, the tight regulation of the interaction between NIPP1 and its ligands by reversible phosphorylation and the fact that NIPP1 is associated with both a protein kinase (MELK) and a protein phosphatase (PP1) suggest a key "integrator" function for NIPP1 in signal transduction to splicing.
Regulation of MELK-Based on the structural conservation of its catalytic domain (23), MELK belongs to the family of CAMK-like protein kinases that also includes protein kinases MARK (24), PASK (25), as well as the AMP-activated protein kinases (26). It is currently not known how the activity of MELK is controlled but this is likely to be complex. MELK is capable of autophosphorylation (Fig. 4), but it is not yet clear whether this also affects its activity. Thr-167 of MELK corresponds to a phosphorylatable Thr within the activation loop of related protein kinases (26,27). For example, activation of the AMP-activated kinase requires phosphorylation of this Thr in the activation loop by the tumor suppressor LKB1 (27). It also seems likely that the large C-terminal domain of MELK, which includes the TP-rich NIPP1-binding domain, has a regulatory function. Our preliminary data indicate that this domain is inhibitory because the T478A mutant is more active than wildtype MELK (Fig. 3) and because the recombinant C-terminal half of MELK is a strong inhibitor of the kinase activity of MELK. 2 A further exploration of the mechanisms that control the activity of MELK will be essential to elucidate the signaling pathways that impinge on MELK and its interactor NIPP1.