Consensus PP1 Binding Motifs Regulate Transcriptional Corepression and Alternative RNA Splicing Activities of the Steroid Receptor Coregulators, p54nrb and PSF

Abstract

Originally identified as essential pre-mRNA splicing factors, non-POU-domain-containing, octamer binding protein (p54nrb) and PTB-associated RNA splicing factor (PSF) are also steroid receptor corepressors. The mechanisms by which p54nrb and PSF regulate gene transcription remain unclear. Both p54nrb and PSF contain protein phosphatase 1 (PP1) consensus binding RVxF motifs, suggesting that PP1 may regulate phosphorylation status of p54nrb and PSF and thus their function in gene transcription. In this report, we demonstrated that PP1 forms a protein complex with both p54nrb and PSF. PP1 interacts directly with the RVxF motif only in p54nrb, but not in PSF. Association with PP1 results in dephosphorylation of both p54nrb and PSF in vivo and the loss of their transcriptional corepressor activities. Using the CD44 minigene as a reporter, we showed that PP1 regulates p54nrb and PSF alternative splicing activities that determine exon skipping vs. inclusion in the final mature RNA for translation. In addition, changes in transcriptional corepression and RNA splicing activities of p54nrb and PSF are correlated with alterations in protein interactions of p54nrb and PSF with transcriptional corepressors such as Sin3A and histone deacetylase 1, and RNA splicing factors such as U1A and U2AF. Furthermore, we demonstrated a novel function of the RVxF motif within PSF that enhances its corepression and RNA splicing activities independent of PP1. We conclude that the RVxF motifs play an important role in controlling the multifunctional properties of p54nrb and PSF in the regulation of gene transcription.

Steroid receptor-mediated gene transcription involves a series of consecutive and coordinate nuclear biochemical reactions starting from chromatin remodeling, transcription initiation, elongation, RNA splicing, and termination (1). Several RNA splicing factors were reported to participate in regulating transcription initiation through protein interactions with transcription factors and/or the core general transcriptional machinery (2–8), supporting the model of cotranscriptional RNA splicing (9, 10). Among these splicing factors, non-POU-domain-containing, octamer binding protein (p54nrb) and PTB-associated RNA splicing factor (PSF) were shown to modulate both transcription initiation and RNA splicing by several research groups, including ours (7, 8, 11–15). However, the mechanisms that regulate p54nrb and PSF function in the multiple steps of gene transcription remain poorly understood.

PSF was originally cloned as a pre-mRNA splicing factor associated with polypyrimidine tract-binding protein (16). Biochemically, PSF can bind both single- and double-stranded nucleotides and associate with many proteins in several protein complexes responsible for almost all steps of gene transcription (17). PSF interacts with the C-terminal domain (CTD) of RNA polymerase II (pol II) (7, 8), transcriptional factors (11–14, 18–22), coregulators (11, 13, 14) and, in some cases, binds directly to the DNA sequences in targeted promoters (15, 23) to regulate gene transcription initiation. As an auxiliary splicing factor, PSF forms complexes with several spliceosome components including U1A, U2AF, and all five small nuclear riboproteins (24–26). It is an essential RNA splicing factor that catalyzes both step I and II pre-mRNA splicing (27). In addition, PSF had been demonstrated to regulate alternative RNA splicing (28). Furthermore, PSF is also present in proteins complexes responsible for 3′-polyadenylation (29, 30), transcription termination, and RNA nuclear retention for proofreading (17). PSF forms a heterodimer with p54nrb, which was identified by an antibody against Saccharomyces cerevisiae splicing factor PRP18 (31). Subsequently cloned as a RNA splicing factor, p54nrb shares 71% identical amino acids with PSF in the RNA recognition motif region. Similar to PSF, p54nrb participates in several nuclear functions including transcription initiation, RNA processing (32), and DNA repair (33, 34).

The impact of p54nrb and PSF on gene transcription is complex because they can both positively and negatively regulate gene transcription. As components of the spliceosome complex, p54nrb and PSF facilitate both step I and II pre-mRNA splicing reactions (27, 35, 36). P54nrb and PSF associate with activated CTD of pol II to facilitate cotranscriptional pre-mRNA splicing, resulting in enhanced transcription (7, 8). The complex of p54nrb and PSF functions as a scaffold to link neuronal Wiskott-Aldrich syndrome protein with pol II-dependent transcriptional machinery (37). Moreover, p54nrb acts as a bridge to link cAMP response element binding protein/transducers of regulated cAMP response element binding protien 2 and pol II, which was demonstrated to be necessary for cAMP-dependent activation of cAMP response element binding protein target genes in vivo (11). These observations indicate that p54nrb and PSF can positively regulate gene transcription. However, several other studies, including ours, also demonstrate that p54nrb and PSF act as transcription corepressors of nuclear receptors and other transcriptional factors (14, 15, 20–22). Herbert Samuel's group (14) first demonstrated that PSF inhibits the thyroid receptor transactivation through interaction with Sin3A, a transcriptional corepressor. We further confirmed that p54nrb and PSF synergize to recruit Sin3A and histone deacetylase 1 (HDAC1) to the androgen receptor (AR) when recruited to the human prostate-specific antigen (PSA) gene promoter (13). Although p54nrb and PSF are recruited to both 5′- and 3′-splicing sites and facilitate intron excision during pre-mRNA splicing, they can also interact with exonic splicing silencer sequences to participate in alternative RNA splicing, which process determines exon inclusion or skipping in the final mRNA, modifying isoform expression (38). These variable impacts of p54nrb and PSF on gene transcription could result from their contributions to different functional protein complexes dependent on physiological conditions and gene context. Therefore, investigation of the protein complexes formed by PSF and p54nrb might provide better understanding of their roles in the regulation of gene transcription in various pathophysiological conditions such as cancer.

Several studies suggest a role for phosphorylation status of p54nrb and PSF in their regulation of gene transcription: 1) p54nrb and PSF interact with kinases (39–41) and protein phosphatases (42) and are phosphorylated at serine, threonine, and tyrosine residues (43, 44); 2) activation of protein kinase C phosphorylates p54nrb and PSF and alters their affinities for RNA and DNA (45); 3) PSF is N-terminally hyperphosphorylated during cell apoptosis, resulting in changes in its conformation and protein interactions (44); and 4) cAMP activates unknown protein phosphatases, which dephosphorylate p54nrb and PSF and subsequently cause the dissociation of Sin3A from the p54nrb-PSF complex and the transcriptional activation of the steroidogenic factor 1 (also a member of the nuclear receptor family) (20, 46). These observations suggest that the phosphorylation status of p54nrb and PSF determines their protein associations and subsequently alters their roles in gene transcription. This hypothesis is supported by the findings from several research groups that RNA splicing factors such as Tra2, steroidogenic factor 2/ASF, nuclear inhibitor of protein phosphatase-1 (NIPP), and splicing factor that interacts with PQBP-1 and protein phosphatase 1 (PP1) contain consensus PP1 binding RVxF motifs, which mediate protein interaction with PP1 (47–49). PP1 uses its catalytic subunit to bind a spectrum of substrates and catalyze the removal of the phosphoresidues at serine and threonine. Dephosphorylation of Tra-β2 alters its alternative RNA splicing activity (47).

Both p54nrb and PSF contain the RVxF motif, implying that they are potential substrates of PP1. In this report, we show that PP1 forms a protein complex with the p54nrb and PSF heterodimer through the RVxF motif only with p54nrb, but not PSF. However, PP1 can dephosphorylate both p54nrb and PSF in vivo and modulate their protein associations with transcriptional corepressors as well as RNA splicing factors, resulting in changes in their functions in gene transcription.

Results

PP1 forms a protein complex with p54nrb and PSF heterodimer

We initiated our study to determine whether PP1 interacts with p54nrb and PSF by performing immunoprecipitation assays. First, human embryonic kidney cell line (293T) cells were transiently transfected with Flag-PSF (Fig. 1A) or Flag-p54nrb (Fig. 1B) in the presence or absence of hemagglutinin (HA)-tagged PP1. Cell lysates were precleared with control mouse IgG and then precipitated with the M2 Sepharose beads (Sigma), which are conjugated with Flag antibody. The associated proteins were immunoblotted with Flag or HA antibody. We observed that PP1 interacts with both p54nrb and PSF. To determine whether endogenously expressed PP1, p54nrb, and PSF can form a protein complex, 293T cell lysates were precipitated with PP1 antibody. Both p54nrb and PSF were coprecipitated, further confirming that p54nrb and PSF form a protein complex with PP1 (Fig. 1C). Using 293T cells overexpressing green fluorescent protein (GFP), GFP PP1γ, or GFP PP1γ(m), we observed that both PP1 and PP1(m) have similar affinity to p54nrb and PSF (Fig. 1D). PP1(m) is the catalytic-dead form of PP1γ with histidine 125 replaced by alanine (47). To determine whether PP1 directly interacts with p54nrb and PSF, and whether different PP1α-, β-, or γ-isoforms have different affinities to p54nrb and PSF, we performed glutathione-S-transferase (GST) pull-down assays. P54nrb, but not PSF, directly interacts with PP1 and there is no difference in association affinities of PP1 isoforms with p54nrb (Fig. 1E). Both p54nrb and PSF contain RVxF motifs. To determine the significance of these motifs on PP1 interactions with p54nrb and PSF, we performed site-directed mutagenesis, switching the valine and phenylalanine amino acids within the motifs into alanine. Two expression vectors encoding p54nrb and PSF mutants were created with p54nrb(m) defining the mutation of VF141,143AA and PSF(m) indicating the mutation of VF364,366AA. GST pull-down assays were repeated by incubating GST PP1γ with purified PSF, PSF(m), p54nrb, and p54nrb(m) (Fig. 1F). Only wild-type p54nrb interacted with PP1. Purified recombinant PP1, Flag-tagged PSF, PSF(m), p54nrb, and p54nrb(m) used in the GST pull-down assays were separated by sodium dodecyl sulfate (SDS) gel electrophoresis and stained with Coomassie blue (Fig. 1, G and H). Expression of Flag-p54nrb, -p54nrb(m), -PSF, and -PSF(m) in transfected 293T cells was also tested as shown (Supplemental Fig. 1 published on The Endocrine Society's Journals Online web site at http://mend.endojournals.org). Taken together, our data unequivocally demonstrated that PP1 forms a protein complex with the p54nrb and PSF heterodimer through a direct interaction with the RVxF motif in p54nrb. Although PSF also possesses the RVxF motif, it does not mediate direct protein interactions between PSF and PP1.

Fig. 1.

PP1 interacts with p54nrb and PSF. A and B, 293T cells were transiently transfected with vectors encoding HA PP1 and Flag p54nrb or Flag PSF. Cell lysates were immunoprecipitated with Flag antibody, and the associated proteins were immunoblotted with Flag and HA antibodies. C, Endogenous PP1/p54nrb/PSF complex was immunoprecipitated from 293T cell lysate with control IgG or PP1 antibody. Associations of p54nrb and PSF were detected with PSF, p54nrb, and PP1 antibodies. D, 293T cells were transfected with Flag-PSF, Flag-p54nrb together with GFP, GFP PP1γ, or GFP PP1γ(m) vector. Cell lysates were immunoprecipitated with GFP antibody, and the associated proteins were immunoblotted with Flag and GFP antibodies. E, GST, GST PP1α, -β, and -γ fusion proteins were immobilized on glutathione beads and incubated with purified Flag p54nrb or Flag PSF. Association of PSF and p54nrb was detected by Flag antibody. F, GST PP1γ fusion protein were immobilized on glutathione beads and incubated with purified Flag p54nrb, PSF, p54nrb(m), and PSF(m), and their associations with PP1γ were detected by Flag antibody. G, The amounts of GST fusion proteins used in the pull-down assay were separated on SDS gel and stained with Coomassie blue. H, Purified Flag p54nrb, Flag PSF, and their mutants were separated on SDS gel and stained with Coomassie blue. Note: PP1γ(m) defines the PP1γ mutation of H125A. P54nrb(m) defines the p54nrb mutation of VF₁₄₁,₁₄₃AA, whereas PSF(m) defines the mutation of VF₃₆₄,₃₆₆AA. IP, Immunoprecipitation; WB, Western blot.

RVxF motif from p54nrb determines PP1 colocalization with p54nrb and PSF

Next we investigated whether the RVxF motif from p54nrb mediates PP1 colocalization with p54nrb and PSF. We used the pDsRed C1 vector (CLONTECH Laboratories, Inc., Palo Alto, CA) as a backbone to construct vectors encoding red fluorescent protein (RFP) chimeras with PSF, p54nrb, PSF(m), and p54nrb(m). When transfected into COS-7 cells, these vectors reveal the regulation of cellular localization of p54nrb and PSF by the RVxF motifs. Immunofluorescence assay with PP1γ antibody was used to determine PP1γ cellular localization. Although all three PP1 isoforms exhibit dynamic localization depending on their interacting partners, PP1γ is mainly expressed in the nucleus as p54nrb and PSF.

When the mock pDsRed vector was transfected into cells, RFP was predominantly expressed in cytoplasmic compartment with some nuclear staining (Fig. 2A). In contrast, endogenous PP1γ was concentrated in the nucleus (Fig. 2B), indicating that PP1γ does not colocalize with RFP. Fusion of PSF into RFP relocated RFP into nucleus with a speckle pattern in the nucleoplasm (20–50 foci absence from nucleoli) (Fig. 2E), the same pattern as endogenous PSF (50). In addition, overexpression of PSF attracted all PP1γ into nucleus, forming the same nuclear distribution as PSF (Fig. 2F). Mutation of the RVxF motif in PSF did not change its nuclear location or its ability to attract PP1γ to colocalize with PSF (Fig. 2, I–L). RFP fusion protein with p54nrb presented the same nuclear localization as endogenous p54nrb (Fig. 2M) (51). Overexpression of p54nrb also attracted PP1γ into the nucleus and induced PP1γ distribution overlapping with p54nrb (Fig. 2N). However, mutation of the RVxF motif in p54nrb dramatically changed its nuclear localization, with well-separated and highly concentrated foci outside of nucleoli (Fig. 2Q). Overexpression of this p54nrb mutant did not change the cellular distribution of endogenous PP1γ (Fig. 2R). Immunoblotting assays indicated that mutant PSF and p54nrb in these transfection assays were expressed at the same protein size and at approximately the same levels with wild-type proteins (Supplemental Fig. 2). Together, these data demonstrated that the RVxF motif from p54nrb determined the localization of PP1 with p54nrb. In addition, these data further supported our protein interaction assay that the RVxF motif within p54nrb, but not PSF, mediated direct interaction between PSF and PP1.

Fig. 2.

The RVxF motif from p54nrb determines PP1 colocalization with p54nrb and PSF. COS-7 cells were transfected with pDsRed C1 mock vector (A–D), pDsRed PSF vector (E–H), pDsRed PSF(m) (I–L), pDsRed p54 (M–P), or pDsRed p54nrb(m) (Q–T), respectively, for 24 h. Cells were then fixed with 4% paraformaldehyde and immunoblotted with PP1γ antibody (1:100 dilution) at 4 C overnight. The cells were washed and incubated with an FITC-conjugated anti-goat antibody (1:50 in 1% BSA in PBST) for 1 h. Cells were mounted with mounting medium containing 4′,6-diamidino-2-phenylindole (DAPI), and examined with a Zeiss fluorescent microscope to monitor RFP for p54nrb and PSF localization and green fluorescence for endogenous PP1. Note: p54nrb(m) defines the p54nrb mutation of VF₁₄₁,₁₄₃AA, whereas PSF(m) defines the mutation of VF₃₆₄,₃₆₆AA.

PP1 dephosphorylates p54nrb and PSF

We next investigated whether PP1 dephosphorylates p54nrb and PSF. In in vivo dephosphorylation assays, 293T cells were transfected with PP1, Flag-p54nrb or Flag-p54nrb(m) in the presence or absence of the PP1 specific inhibitor, NIPP (49) (Fig. 3A). Currently, there is no antibody that recognizes phosphorylated p54nrb. To detect the levels of phosphorylated forms of p54nrb, we immunoblotted the purified Flag-p54nrb or Flag-p54nrb(m) with antibodies against phosphoserine and -threonine. Phosphotyrosine antibody was also used as a negative control. PP1 overexpression significantly decreased the serine/threonine-phosphorylated form of p54nrb, an effect antagonized by NIPP. The degree of serine/threonine phosphorylation of p54nrb(m) was affected neither by PP1 nor by NIPP. Tyrosine phosphorylation was not changed. Overexpression of PP1 but not PP1(m), the PP1 catalytic-dead mutant, significantly reduced the serine/threonine-phosphorylated p54nrb (Fig. 3B).

Fig. 3.

PP1 dephosphorylates p54nrb and PSF. In vivo dephosphorylation assay: 293T cells were transfected with p54nrb or p54nrb(m) (A and B), or PSF or PSF(m) (D and E), PP1, PP1(m), and/or NIPP expression vectors for 24 h. Flag-tagged p54nrb or PSF was purified and immunoblotted with an antibody mixture of phosphoserine/threonine antibodies or with antibodies against phosphotyrosine or Flag tag. In vitro dephosphorylation assay: Flag p54nrb and p54nrb(m) (C) or Flag PSF and PSF(m) (F) were immobilized on M2 Sepharose beads and incubated with 2 U of PP1 and/or 500 nm Tautomycetin (Tau) for 30 min at 30 C. The beads were washed and the Flag-tagged proteins were immunoblotted with an antibody of phosphoserine/threonine antibodies or with antibodies against phosphotyrosine or Flag tag. Note: PP1γ(m) defines the PP1γ mutation of H125A. IP, Immunoprecipitation; WB, Western blot.

Next, we performed in vitro dephosphorylation assays, in which purified Flag-p54nrb or Flag-p54nrb(m) was first immobilized on M2 Sepharose beads and then incubated with recombinant PP1 (New England Biolabs Ltd., Pickering, Ontario, Canada) in the presence or absence of the specific PP1 inhibitor, Tautomycetin (Fig. 3C). We observed that PP1 decreased p54nrb phosphorylation at the serine/threonine residues, and this was inhibited by Tautomycetin. Phosphorylation of p54nrb(m) was not altered when PP1 and/or Tautomycetin were added to the reactions.

We used the same experimental design to study PSF dephosphorylation by PP1. Interestingly, the in vivo assay indicated that both PSF and PSF(m) could be dephosphorylated at serine/threonine residues by PP1, and these effects were inhibited by NIPP (Fig. 3D). There were slight increases in tyrosine phosphorylation by PP1, but no change when NIPP was cotransfected. Overexpression of PP1, but not PP1(m), reduced the serine/threonine-phosphorylated PSF (Fig. 3E). These observations suggested that both PSF and PSF(m) associate with PP1 indirectly through endogenous p54nrb. This hypothesis was confirmed by our in vitro dephosphorylation assay, in which purified PSF and PSF(m) could also be dephosphorylated by PP1 (Fig. 3F), as well as our finding that both PSF and PSF(m) can be immunoprecipitated with endogenous p54nrb (see Figs. 9 and 10 below). Taken together, these results indicate that PP1 can dephosphorylate both PSF and p54nrb in vivo at serine/threonine residues. This function of PP1 is dependent on the RVxF motif within p54nrb.

Fig. 9.

PP1 regulates p54nrb protein interactions with transcription corepressors and RNA splicing factors. A, 293T cells were transfected with Flag p54nrb together with either GFP mock or GFP-PP1 vector. B, 293T cells were transfected with Flag p54nrb and treated with vehicle or 500 nm Tautomycetin (Tau) for 18 h. C, Cells were transfected with either p54nrb or p54nrb(m) vector. Cell lysates were used, 24 h after transfection, for immunoprecipitation with Flag antibody. The associated proteins were fractioned on SDS gel and immunoblotted with antibodies as shown. IP, Immunoprecipitation.

Fig. 10.

PP1 regulates PSF protein interactions with transcription corepressors and RNA splicing factors. A, 293T cells were transfected with Flag PSF together with either GFP mock or GFP-PP1 vector. B, 293T cells were transfected with Flag PSF and treated with vehicle or 500 nm Tautomycetin (Tau) for 18 h. C, Cells were transfected with either PSF or PSF(m) vector. Cell lysates were used, 24 h after transfection, for immunoprecipitation with Flag antibody. The associated proteins were fractioned on SDS gel and immunoblotted with antibodies as shown. IP, Immunoprecipitation.

PP1 alleviates the transcriptional corepressor function of p54nrb and PSF

P54nrb and PSF are both reported to be corepressors of steroid receptors including the AR and the progesterone receptor (PR) (12, 13, 52). To determine whether PP1 modulates the corepressor activity of p54nrb, we performed luciferase reporter assays. First, human prostate cancer LNCaP cells were transfected with the PSA promoter linked to a luciferase reporter. P54nrb decreased AR-induced luciferase activity as reported (13, 52) (Fig. 4A). Cotransfection with PP1 blocked p54nrb corepressor activity to AR. This PP1 action requires the presence of the RVxF motif within p54nrb, because mutation of this motif rendered p54nrb nonresponsive to PP1. Second, because the AR expressed in LNCaP cells possesses a mutation that broadens its ligand specificity, we also studied p54nrb corepressor activity on wild-type AR using the probasin promoter in PC3 cells overexpressing AR wild-type vector (Fig. 4B). This repressor function of p54nrb was strengthened when cells were treated with the PP1-specific inhibitor, Tautomycetin. Because p54nrb(m) is not dephosphorylated by PP1, Tautomycetin did not change p54nrb(m) corepressor activity. Moreover, p54nrb(m) presented a stronger corepressor activity than p54nrb wild type, possibly because it is maintained in a hyperphosphorylated state. Third, the ability of p54nrb to inhibit PR was tested in T47D breast cancer cells using the mouse mammary tumor virus (MMTV) promoter (Fig. 4C). PP1 also alleviated p54nrb, but not p54nrb(m), corepressor activity to PR. P54nrb(m) showed a stronger corepressor activity than p54nrb wild type even in the absence of PP1 transfection. PP1 overexpression did not alter p54nrb or PSF expression in LNCaP, PC3, and T47D cells (Supplemental Fig. 3).

Fig. 4.

PP1 alleviates transcription corepressor activity of p54nrb and PSF. LNCaP, PC3, and T47D cells were transfected with PSA, Probasin, and MMTV luciferase reporter genes and treated with 10 nm dihydrotestosterone (DHT) or P4 to monitor AR and PR transcriptional activities as shown on top of each figure. Cells were also cotransfected with mock, p54nrb, p54nrb(m) vectors (A–C) or mock, PSF, PSF(m) vectors (D–F) together with −/+PP1 expression vector, followed by DHT or P4 treatment for 24 h. PC3 cells were also cotransfected with AR vector and treated with either vehicle or 500 nm Tautomycetin (Tau). Luciferase activities were measured and calibrated with Renilla luciferase activities. Values are shown as means ± sem from three independent experiments. *, P < 0.05.

The ability of PP1 to modulate corepressor function of PSF was also investigated. In LNCaP cells, PSF inhibited PSA promoter activity driven by AR (Fig. 4D). Similarly to p54nrb, PSF repressor function was abolished when PP1 was overexpressed. In contrast to p54nrb, PP1 also abolished PSF(m) repressor activity. Interestingly, although PP1 dephosphorylated both PSF and PSF(m), PSF(m) still presented stronger repressor activity than the wild-type PSF. This novel observation suggests that the RVxF motif does not mediate the association of PSF with PP1, but functions to restrain PSF corepressor activity independent of PP1. This was confirmed in PC3 cells transfected with AR wild-type vector and the probasin promoter, in which PSF(m) showed a greater corepressor activity than PSF (Fig. 4E). Tautomycetin enhanced both PSF and PSF(m) repressor activity. Experiments were repeated in T47D cells to monitor PSF repression of PR transactivation, and similar results were obtained (Fig. 4F).

To further confirm that PP1 enzymatic activity regulates p54nrb and PSF corepressor function, PP1 or PP1(m) was transfected with p54nrb, PSF, and their mutants in LNCaP cells (Fig. 5). PP1, but not PP1(m), alleviated p54nrb, PSF, and PSF(m) repression on AR, whereas p54nrb(m) corepressor activity was unchanged. Similar results were observed in 293T cells cotransfected with AR and probasin reporter vector (Fig. 5B). Scramble or small interfering RNA (siRNA) against PP1 was also used (Supplemental Fig. 4). PP1 knockdown strengthened both p54nrb and PSF corepressor activity on AR in the LNCaP cells (Supplemental Fig. 4C). PP1 knockdown also enhanced p54nrb corepression on PR in the MMTV promoter context (Supplemental Fig. 4D). In summary, we conclude that PP1 abolishes the corepressor activity of p54nrb on steroid receptors, a function that is dependent on the RVxF motif in p54nrb. Blockade of endogenous PP1 dephosphatase activity by Tautomycetin/siRNA or mutation of the p54nrb RVxF motif maintained p54nrb in a hyperphosphorylated state and enhanced its corepressor activity. PSF corepressor activity can also be abolished by PP1. The RVxF motif within PSF functions as a controller that restrains PSF corepressor activity, an action independent of PP1.

Fig. 5.

PP1 alleviates transcription corepressor activity of p54nrb and PSF. LNCaP (A) and 293T (B) cells were transfected with PSA or Probasin luciferase reporter genes. Cells were also cotransfected with mock, p54nrb, p54nrb(m), PSF, PSF(m) vectors together with PP1 or PP1(m) expression vector and then treated with 10 nm dihydrotestosterone (DHT) for 24 h. Luciferase activities were measured and calibrated with Renilla luciferase activities. Values are shown as means ± sem from three independent experiments. Note: PP1γ(m) defines the PP1γ mutation of H125A. *, P < 0.05.

PP1 regulates the RNA splicing activity of p54nrb and PSF

Alternative RNA splicing is recognized as a major contributor to proteomic complexity in human cells (53). Aberrant alternative splicing is associated with a broad range of human diseases (54). Although p54nrb and PSF play roles in constitutive RNA splicing, we focused on understanding whether PP1 impacts alternative RNA splicing by p54nrb and PSF. To study this function, we used the CD44 minigene driven by the HSV promoter, established by Auboeuf et al. (55). The ratio of skipping to inclusion of CD44 transcripts is used to measure p54nrb and PSF alternative splicing activities. P54nrb decreased skipping-inclusion ratio, whereas PSF increased skipping-inclusion ratio of CD44 transcripts (Fig. 6, A and B). PP1 decreased both p54nrb and PSF alternative splicing activity on CD44 minigene, when compared with PP1(m) (Fig. 6C). In contrast, PP1 knockdown strengthened only p54nrb splicing activity (Supplemental Fig. 4, E and F).

Fig. 6.

PP1 regulates p54nrb and PSF alternative RNA splicing activity. A, Schematic representation of the construct of the CD44 minigene driving by HSV promoter used in the alternative splicing assay (55). 293T cells were transfected with 2 μg HSV-CD44 minigene vector together with (B) 3 μg of mock, p54nrb, or PSF vector (panel B) or 2 μg of mock, p54nrb, or PSF vector plus 1 μg of either PP1γ or PP1γ(m) vector (panel C). Total RNA was extracted to perform RT-PCR, and the CD44 splice variant ratio of skipping-inclusion was calculated as described in Materials and Methods (55). To eliminate the generic impacts of PP1 and PP1(m) on the testing system, and then compare the p54nrb and PSF splicing activity in panel C, the values from mock transfections in the presence of PP1 or PP1(m) were calibrated to 1, respectively. *, P < 0.05. Note: PP1γ(m) defines the PP1γ mutation of H125A.

Because alternative splicing involves several splicing factors, we transfected cells with increasing doses (0, 1, and 2 μg) of p54nrb or PSF expression vectors to monitor the dose-dependent splicing activities by p54nrb and PSF (Figs. 7 and 8). To test the impact of PP1, we either overexpressed PP1 to achieve gain-of-function or treated cells with Tautomycetin to specifically block endogenous PP1 activity. Because PP1 can also regulate splicing factors other than p54nrb and PSF, we eliminated the impacts of PP1 on other splicing factors in the system by normalizing the skipping-inclusion ratios to those generated from 0 μg of transfected p54nrb or PSF to 1. Therefore, the changing degrees in skipping-inclusion ratio of CD44 transcript in the presence or absence of PP1 gain or loss of function reflects PP1 regulation on p54nrb and PSF splicing activities. We observed that PP1 overexpression up-regulated, whereas Tautomycetin down-regulated, the p54nrb-dependent skipping-inclusion ratio (Fig. 7, A and B). P54nrb(m), which does not associate with PP1, showed a stronger alternative splicing activity [decreased skipping-inclusion ratio (Fig. 7C)], an effect that was not altered by PP1 (Fig. 7D). These data indicated that PP1 regulates the alternative splicing activity of p54nrb on the CD44 minigene, depending on the presence of the RVxF motif within p54nrb. Similar to p54nrb, PP1 also decreased PSF alternative splicing function (Fig. 8A), whereas inhibition of endogenous PP1 activity by Tautomycetin strengthened PSF splicing function (Fig. 8B). PSF(m) showed a stronger alternative splicing activity than PSF wild type (Fig. 8C). Cotransfection of PP1 prevented PSF(m) splicing activity from up-regulating the skipping-inclusion ratio (Fig. 8D). These results demonstrate PP1 regulates PSF splicing activity independently of the RVxF motif within PSF. Instead, the PSF RVxF motif functions as a regulator of PSF splicing activity, possibly through protein interactions with other splicing factors.

Fig. 7.

PP1 regulates alternative RNA splicing activity of p54nrb. A and B, 293T cells were transfected with 2 μg HSV-CD44 minigene vector together with increasing doses (0, 1, and 2 μg) of p54nrb vector. Cells were also cotransfected with 1 μg of GFP mock or GFP-PP1 vector (A), or treated with vehicle or 500 nm Tautomycetin (Tau) for 18 h (B). C, 293T cells were transfected with 2 μg HSV-CD44 minigene vector together with increasing doses (0, 1, and 2 μg) of p54nrb or p54nrb(m) vector. D, Cells were transfected with (0, 1, and 2 μg) of p54nrb(m) together with 1 μg of GFP mock or GFP-PP1 vector. The CD44 splice variant ratio of skipping-inclusion was measured and calculated as described in Materials and Methods (55). To eliminate the generic impacts of PP1 activity on the testing system, and only measure the p54nrb dose-dependent splicing activity by PP1, the values from 0 μg of p54nrb transfections were calibrated to 1. *, P < 0.05; **, P < 0.001. DMSO, Dimethylsulfoxide.

Fig. 8.

PP1 regulates alternative RNA splicing activity of PSF. A and B, 293T cells were transfected with 2 μg HSV-CD44 minigene vector together with increasing doses (0, 1, and 2 μg) of PSF vector. Cells were also cotransfected with 1 μg of GFP mock or GFP-PP1 vector (A) or treated with vehicle or 500 nm Tautomycetin (Tau)for 18 h (B). C, 293T cells were transfected with 2 μg HSV-CD44 minigene vector together with increasing doses (0, 1, and 2 μg) of PSF or PSF(m) vector. D, Cells were transfected with (0, 1, and 2 μg) of PSF(m) together with 1 μg of GFP mock or GFP-PP1 vector. The CD44 splice variant ratio of skipping-inclusion was measured and calculated as described in Materials and Methods (55). To eliminate the generic impacts of PP1 activity on the testing system and only measure the PSF dose-dependent splicing activity by PP1, the values from 0 μg of PSF transfections were calibrated to 1. *, P < 0.05; **, P < 0.001. DMSO, Dimethylsulfoxide.

Mechanisms by which PP1 regulates p54nrb and PSF function in transcription corepression and RNA splicing

The phosphorylation status of PSF was implicated in altering its protein associations (44). To confirm this hypothesis, we conducted immunoprecipitation assays to determine p54nrb and PSF protein interactions with the transcriptional corepressors (Sin3A and HDAC1) and RNA splicing factors (U1A and U2AF65). Recruitments of Sin3A and HDAC1 to p54nrb and PSF confer transcriptional corepressor activity to these proteins. U1A and U2AF65 are RNA splicing factors that bind 5′- and 3′-splicing sites, respectively, and define the exon boundary for alternative splicing. First, lysates from 293T cells transfected with Flag-p54nrb plus either mock or PP1 vector were precipitated with Flag antibody. The associated protein complexes were immunoblotted with antibodies as shown in Fig. 9A. PP1 dramatically enhanced dissociation of Sin3A and HDAC1 but strengthened association of U1A and U2AF65 to p54nrb. Tautomycetin increased the interaction of Sin3A and HDAC1 with p54nrb but decreased U1A and U2AF65 association with p54nrb (Fig. 9B). These data indicated that PP1 induces dephosphorylation of p54nrb and switches its association to protein complexes that regulate RNA splicing as opposed to transcriptional corepression splicing. While maintained in a hyper serine/threonine-phosphorylated form, p54nrb(m) showed stronger associations with Sin3A and HDAC1, but a weaker affinity for U1A (Fig. 9C). Interestingly, p54nrb(m) showed a stronger interaction with U2AF65, whereas the association of p54nrb with U2AF65 decreased in the presence of Tautomycetin, implying that the function of RVxF motif in p54nrb not only mediates dephosphorylation by PP1 but may also regulate p54nrb protein conformation independently of PP1.

We also investigated the regulation of PSF protein interactions with corepressors and RNA splicing factors by PP1. In both PP1 gain- and loss-of-function conditions, no significant alteration of PSF protein associations with Sin3A, HDAC1, U1A, and U2AF65 was observed (Fig. 10, A and B). The explanation could be that PP1 association with PSF is indirect, thus has minor impacts on PSF protein interactions. This was consistent with PSF splicing activity that was unchanged by PP1γ silencing by siRNA (Supplemental Fig. 4). This ineffectiveness of PP1γ silencing on PSF splicing activity may be due to: 1) indirect PSF interaction with PP1 through p54nrb; 2) PSF is already highly phosphorylated in vivo and therefore cannot be further phosphorylated by PP1γ silencing; and 3) compensation of other PP1 isoforms. However, PSF(m) showed stronger interaction with Sin3A, HDAC1, and U1A. The association of PSF with U2AF65 was unchanged, even when the RVxF motif was mutated (Fig. 10C). These studies demonstrate that PP1 regulates PSF function in transcription initiation and alternative RNA splicing, which may not be through alteration of PSF protein interactions. Rather, it is possibly mediated indirectly through PSF heterodimer partner p54nrb. However, the RVxF motif within PSF functions as a controller of PSF protein interactions that is independent of PP1.

Discussion

The levels of gene transcription and isoform expression (through alternative RNA splicing) are fundamental regulators of cellular function and, ultimately, human health. Although p54nrb and PSF have been shown to play dual roles in each step of transcription initiation and RNA splicing, our study, for the first time, reveals that PP1-mediated dephosphorylation of p54nrb and PSF determines their regulatory roles during these steps of gene transcription, providing a remarkable switch point to the control of gene expression and cellular function. At the level of transcription initiation, hyperphosphorylated p54nrb and PSF act as transcriptional corepressors, whereas dephosphorylated p54nrb and PSF by PP1 favors the formation of a complex with activated pol II CTD to initiate gene transcription. At the level of RNA splicing, hyperphosphorylated p54nrb and PSF actively participate in alternative RNA splicing. Dephosphorylation of p54nrb and PSF by PP1 promotes their roles as facilitators of constitutive pre-mRNA splicing.

We demonstrated that hyperphosphorylation of p54nrb increases its association with Sin3A and HDAC1 corepressors, thereby inhibiting transcription initiation mediated by steroid receptors. This negative impact on gene transcription can be enhanced by treating cells with the PP1-specific inhibitor, Tautomycetin, or by dissociation of PP1 from p54nrb through mutation of the RVxF motif. Dephosphorylation of p54nrb by PP1 induces dissociation from the Sin3A and HDAC1, and subsequently alleviation of p54nrb corepressor function. These results demonstrate that PP1 modulates the corepressor activities of p54nrb. Once p54nrb dissociates from the Sin3A/HDAC complex, they are available to form a new complex with activated pol II CTD, promoting transcription initiation and elongation. As such, our data define the molecular mechanism by which p54nrb is dynamically assembled into different functional protein complexes to exert both positive and negative regulation of gene transcription, a process that is controlled by PP1.

In addition, PP1-induced dissociation of p54nrb from Sin3A/HDAC1 was concurrent with enhanced interaction of p54nrb with U1A and U2AF65 splicing factors, providing further evidence for a role of PP1 in regulating assembly of p54nrb into different protein complexes that regulate both transcription and RNA splicing. P54nrb and PSF participate in both constitutive pre-mRNA splicing as well as alternative RNA splicing of targeted genes. These dual roles of p54nrb and PSF can also be regulated by PP1. Gain of function of PP1 reduces the alternative splicing activity of p54nrb and PSF, and in turn enhances their contributions to constitutive pre-RNA splicing.

Our data suggest that the RVxF motif within p54nrb and PSF regulates important features of their functions in gene transcription and RNA splicing. Nevertheless, there are important differences in how this motif operates within the two proteins. Although the RVxF motif mediates dephosphorylation of p54nrb through binding of PP1, this motif does not support binding of PP1 to PSF. In addition, our data suggest that the RVxF motif mediates functions beyond dephosphorylation of p54nrb and PSF. Inhibition of endogenous PP1 activity induces a hyperphosphorylated form of p54nrb, which exhibits decreased association with U1A and U2AF65. In contrast, although mutation of RVxF also maintains hyperphosphorylation of p54nrb (because p54nrb can no longer bind PP1), this mutation decreases p54nrb association with U1A while increasing its association with U2AF65. The mechanism underlying this inconsistency remains unclear. It suggests that the RVxF motif not only mediates PP1 interaction and dephosphorylation of p54nrb but is also required for its interaction with other proteins (e.g. splicing factors, coregulators). This possibility may provide an explanation as to why the RVxF motif within PSF does not mediate the direct interaction of PSF with PP1, but still regulates PSF function through modulating its interactions with transcription corepressors and RNA splicing factors. Our data show that mutation of this motif within PSF increases its transcriptional repressive function but reduces CD44 minigene alternative splicing. These data are consistent with the RVxF motif being located within the RNA recognition motifs of p54nrb and PSF, which have been demonstrated to be responsible for p54nrb and PSF interactions with nucleotides and protein partners (18, 25).

In summary, our observations suggest that phosphorylation of p54nrb and PSF is a dynamic process, which is coordinated with protein phosphatases to ensure that p54nrb- and PSF-targeted genes are expressed under specific physiological conditions. PSF has been shown to be differentially phosphorylated throughout the cell cycle and during apoptosis (44). The phosphorylation status and function of p54nrb and PSF are dependent upon the RVxF motif within the RNA recognition domain of p54nrb. The multifunction role of the RVxF motif suggests that it might serve as a target to control the nuclear functions of p54nrb and PSF, especially in diseases with aberrant p54nrb expression/function.

Materials and Methods

Antibodies and reagents

Antibodies against Flag tag and PSF are from Sigma (St. Louis, MO). HA tag, PP1, mSin3A, HDAC1, U1A, U2AF65, mouse, and goat control IgG antibodies are from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Antibody against p54nrb is from BD Bioscience (Palo Alto, CA), and antibodies against phosphoserine, -threonine, and -tyrosine are from Invitrogen (Carlsbad, CA). Tautomycetin is from Tocris Bioscience (Ellisville, MO). γ³²P-ATP is from PerkinElmer Life Sciences (Wellesley, MA). Dual-Luciferase Reporter Assay Kit is from Promega Corp. (Madison, WI).

Plasmid construction

HA-PP1γ, GST PP1α, β, γ, EGFP PP1γ, and NIPP are from Drs. Stefan Stamm (Germany), James Manley (New York, NY), and Laura Trinkle-Mulcahy (Ottawa, Ontario, Canada) respectively. Flag p54 and Flag PSF were described previously (13). Site-directed mutagenesis was carried out by PCR amplification of the wild type of p54 and PSF constructs with the QuikChange Site-Directed Mutagenesis kit (Stratagene, La Jolla, CA). PSF(m) defines the mutation of VF₃₆₄,₃₆₆AA, whereas p54nrb(m) defines the mutation of VF₁₄₁,₁₄₃AA.

Cell culture and transient transfection

Human embryonic kidney cell line (293T), human prostate cancer cell line (LNCaP and PC3), and human breast cancer cell line (T47D) were purchased from American Type Culture Collection (Manassas, VA). LNCaP cells were maintained in RPMI-1640, and all others were maintained in DMEM plus 5% fetal calf serum (Sigma) as described elsewhere (41, 42). For experiments involving steroid exposure, the medium was substituted with phenol red-free DMEM containing 5% charcoal-treated fetal bovine serum (Hyclone Laboratories, Logan, UT). Transfection was performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. In luciferase assays, cell lysates were collected in lysis buffer (Promega), and firefly luciferase activity was determined using the luciferin reagent (Promega). Transfection efficiency was normalized to Renilla luciferase activity.

Coimmunoprecipitation and immunoblotting

Cells were lysed in TNET buffer (50 mm Tris, pH 7.4; 150 mm NaCl; 1 mm EDTA; and 1% Triton) plus protease inhibitor cocktail. Cell lysates precleared with control IgG (Santa Cruz) were incubated with specific antibodies overnight at 4 C, followed by the addition of protein A/G (Santa Cruz) for another 2 h at 4 C. Resins were washed with TNET buffer (contains 500 mm NaCl) and eluted with Laemmli buffer, boiled, and centrifuged. The supernatant was separated by SDS-PAGE, transferred to polyvinylidene difluoride membrane, and blotted by specific antibodies.

GST pull-down assay

GST pull-down assay was performed according to modifications of previously published techniques (56). Flag-tagged PSF, p54nrb, and their mutants were purified from 293T cells as described elsewhere (13). GST or GST PP1 fusion proteins were induced in BL21(PLES3) cells with isopropyl-β-d-thiogalactopyranoside and immobilized onto glutathione-conjugated Sepharose beads (Amersham) incubated with 400 μg purified PSF, p54nrb, or their mutants in TNNT buffer. After extensive wash, the associated proteins were recovered by Laemmli buffer, separated on SDS-PAGE gels, and transferred to polyvinylidene difluoride membrane and Western blotted with Flag tag antibody.

Immunofluorescence

COS-7 cells were grown on coverslips and transfected with pDsRed C1 mock vector, or DsRed PSF, DsRed p54, DsRred PSF(m), and DsRed p54nrb(m), respectively. Cells were fixed with 4% paraformaldehyde at room temperature for 15 min. After being washed with PBS and permeabilized in PBS-Tween 20 (PBST) (0.25% Triton X-100), cells were preblotted with PBST plus 1% BSA at room temperature for 1 h and incubated with PP1γ antibody (1:100 dilution) at 4 C overnight. The cells were washed three times in PBS and incubated with an FITC-conjugated antigoat antibody (1:50 in 1% BSA in PBST) for 1 h. Cells were washed, mounted with mounting medium containing 4′,6-diamidino-2-phenylindole, and examined with a Zeiss fluorescent microscope (Carl Zeiss, Thornwood, NY).

Protein dephosphorylation assay

Dephosphorylation assays were performed in both in vivo and in vitro systems. In the in vivo system, vectors expressing Flag PSF, Flag p54nrb, and their mutants were transfected with PP1 and or NIPP vector in 293T cells. PSF or p54 was then immunoprecipitated by M2 Sepharose beads, and the associated proteins were immunoblotted with a mixture of phosphoserine and -threonine antibodies or antibody against phosphotyrosine, or Flag antibody to detect phosphorylated and total PSF and p54nrb proteins.

In the in vitro assays, 800 μg of Flag p54nrb, PSF, and their mutants were first immobilized on M2 Sepharose beads and then incubated with 2 U of PP1 (New England Biolabs) in the presence or absence of 500 nm Tautomycetin for 30 min at 30 C. After washing, PSF, p54nrb, and their mutants were immunoblotted with a mixture of phosphoserine and phosphothreonine antibodies or antibody against phosphotyrosine, or Flag antibody to detect phosphorylated and total PSF and p54nrb proteins

Alternative RNA splicing assay

HSV-CD44 splicing reporter gene was obtained from Dr. O'Malley (Baylor College of Medicine, Houston, TX) as described previously (55). 293T cells were transfected in triplicate in six-well plates with 2 μg/well of HSV-CD44 minigenes by Lipofectamine2000 (Invitrogen). When 500 nm Tautomycetin or vehicle was used, cells were treated for 18 h before RNA extraction. Total RNA was extracted by using RNA mini kit (Invitrogen) 24 h after transfection and deoxyribonuclease treated to digest plasmid contamination and any genomic DNA for 15 min at room temperature. Reverse transcription reaction was performed using hexamer and M-MLV reverse transcriptase (Invitrogen), after which it was used as template for PCR. In the PCR, primers were first radiolabeled with γ³²P-ATP (4500 Ci/mmol) by T4 nucleotide kinase (Invitrogen). CD44 Primers are as follows: sense, AGACACCATGCATGGTGCACC; and antisense, CCATAACAGCATCA GGAGTG. PCR is 35 cycles of 15 sec at 94 C, 45 sec at 48 C, and 1 min at 72 C. Radioactive RT-PCR products were fractionated on denaturing 5% polyacrylamide gels, dried and exposed to autoradiography and in phosphoimager cassettes to allow quantification using Phospho-Imaging Molecular Dynamics System (Molecular Dynamics, Sunnyvale, CA). Statistical analysis was carried out using SigmaStat version 1.01 (Jandel Corp., San Rafael, CA). One-way ANOVA followed by pairwise multiple comparison procedures (Student-Newman-Keuls method) were to determine differences between groups, with the level of significance for comparison set at P < 0.05.