Interactor-mediated nuclear translocation and retention of protein phosphatase-1.

Protein Ser/Thr phosphatase-1 (PP1) is a ubiquitous eukaryotic enzyme that controls numerous cellular processes by the dephosphorylation of key regulatory proteins. PP1 is expressed in various cellular compartments but is most abundant in the nucleus. We have examined the determinants for the nuclear localization of enhanced green fluorescent protein-tagged PP1 in COS1 cells. Our studies show that PP1gamma(1) does not contain a functional nuclear localization signal and that its nuclear accumulation does not require Sds22, which has previously been implicated in the nuclear accumulation of PP1 in yeast (Peggie, M. W., MacKelvie, S. H., Bloecher, A., Knatko, E. V., Tatchell, K., and Stark, M. J. R. (2002) J. Cell Sci. 115, 195-206). However, the nuclear targeting of PP1 isoforms was alleviated by the mutation of their binding sites for proteins that interact via an RVXF motif. Moreover, one of the mutants with a cytoplasmic accumulation and decreased affinity for RVXF motifs (PP1gamma(1)-F257A) could be re-targeted to the nucleus by the overexpression of nuclear interactors (NIPP1 (nuclear inhibitor of PP1) and PNUTS (PP1 nuclear targeting subunit)) with a functional RVXF motif. Also, the addition of a synthetic RVXF-containing peptide to permeabilized cells resulted in the loss of nuclear enhanced green fluorescent protein-PP1gamma(1). Finally, NIPP1(-/-) mouse embryos showed a nuclear hyperphosphorylation on threonine, consistent with a role for NIPP1 in the nuclear targeting and/or retention of PP1. Our data suggest that both the nuclear translocation and the nuclear retention of PP1 depend on its binding to interactors with an RVXF motif.

Protein Ser/Thr phosphatase-1 (PP1) is a ubiquitous eukaryotic enzyme that controls numerous cellular processes by the dephosphorylation of key regulatory proteins. PP1 is expressed in various cellular compartments but is most abundant in the nucleus. We have examined the determinants for the nuclear localization of enhanced green fluorescent protein-tagged PP1 in COS1 cells. Our studies show that PP1␥ 1 does not contain a functional nuclear localization signal and that its nuclear accumulation does not require Sds22, which has previously been implicated in the nuclear accumulation of PP1 in yeast ( Sci. 115, 195-206). However, the nuclear targeting of PP1 isoforms was alleviated by the mutation of their binding sites for proteins that interact via an RVXF motif. Moreover, one of the mutants with a cytoplasmic accumulation and decreased affinity for RVXF motifs (PP1␥ 1 -F257A) could be re-targeted to the nucleus by the overexpression of nuclear interactors (NIPP1 (nuclear inhibitor of PP1) and PNUTS (PP1 nuclear targeting subunit)) with a functional RVXF motif. Also, the addition of a synthetic RVXF-containing peptide to permeabilized cells resulted in the loss of nuclear enhanced green fluorescent protein-PP1␥ 1 . Finally, NIPP1 ؊/؊ mouse embryos showed a nuclear hyperphosphorylation on threonine, consistent with a role for NIPP1 in the nuclear targeting and/or retention of PP1. Our data suggest that both the nuclear translocation and the nuclear retention of PP1 depend on its binding to interactors with an RVXF motif.
Protein Ser/Thr phosphatase-1 (PP1) 1 is expressed in all eukaryotic cells and controls numerous cellular processes including metabolism, cell division, apoptosis, and protein synthesis (1)(2)(3)(4)(5). PP1 does not exist freely in the cell but is associated with a large variety of polypeptides that determine when and where the phosphatase acts. Currently, ϳ70 mammalian genes are already known to encode interactors of PP1 (4,5). Some of these function as targeting subunits and bring PP1 in close proximity to its substrates. Others (also) modulate the activity and substrate specificity of PP1 or are themselves substrates for associated PP1. The available information suggests that proteins interact with PP1 via multiple short sequence motifs. These PP1-binding motifs can be shared among PP1 interactors, accounting for the ability of PP1 to form stable complexes with a large number of structurally unrelated proteins. The best characterized and most common PP1-binding motif conforms to the consensus sequence RKX 0 -1 VI{P}FW in which X denotes any residue and {P} any residue except Pro; it is often referred to as the RVXF motif (6). The RVXF motif binds to a hydrophobic channel near the C terminus of PP1 (7). The binding of the RVXF motif not only has a PP1 anchoring function but also promotes the interaction of secondary, lower affinity binding sites, often resulting in an altered activity and/or substrate specificity of PP1 (1).
Mammalian genomes contain three genes that together encode four isoforms of PP1, namely PP1␣, PP1␤/␦, and the splice variants PP1␥ 1 and PP1␥ 2 (1,5). These isoforms are ϳ90% identical at the protein level and differ mainly in their extremities. With the exception of PP1␥ 2 , which is only expressed in the testis and the brain, the mammalian PP1 isoforms are ubiquitously expressed. PP1␣, PP1␤/␦, and PP1␥ 1 show an overlapping but distinct subcellular localization (8). Within the nucleus PP1␣ is mainly associated with the nuclear matrix, PP1␤/␦ is enriched in the non-nucleolar chromatin fraction, and PP1␥ 1 is predominantly targeted to the nucleoli. However, the overexpression of the nuclear interactor NIPP1, which is normally associated mainly with PP1␤/␦ (9), also results in a re-targeting of other PP1 isoforms to nuclear sites that contain NIPP1 (10), indicating that the steady-state subnuclear localization of PP1 isoforms is at least partially determined by their relative affinities for various targeting proteins. In accordance with this view, it was recently demonstrated that at least some interactors of PP1 contain isoform-specific binding sites (11).
Although PP1 is very abundant in the nucleus, the mechanism underlying its nuclear translocation is unknown. PP1 (36 -38 kDa) is sufficiently small to enter the nucleus passively. Yet, this is an unlikely mechanism for its nuclear translocation, because PP1 does not appear to exist in the cell as a free monomer (12). PP1 isoforms contain classical polybasic or bipartite nuclear localization signals (NLSs) that mediate transport by importins (13), but the functionality of these NLSs has never been explored. Another possible mechanism for nuclear translocation is piggyback transport involving the co-transport with an interactor that contains an NLS (14 -16). At least four nuclear interactors of PP1 have putative (PNUTS and Sds22) or established (NIPP1 and SIPP1) NLSs, making them candidate co-transporters of PP1 (17)(18)(19)(20). Also, strains of Saccharomyces cerevisiae that carried temperature-sensitive sds22 alleles showed a rapid loss of nuclear PP1 under restrictive conditions, providing additional evidence for a role of Sds22 in maintaining the normal nuclear localization of PP1 (21).
We have explored the mechanism of the nuclear transport and nuclear retention of EGFP-tagged PP1␥ 1 by site-directed mutagenesis of a putative NLS and interactor binding sites. The results revealed that the nuclear targeting as well as the nuclear retention of PP1␥ depends on proteins such as NIPP1 and PNUTS that contain both an NLS and an RVXF-binding motif. Sds22, which lacks an RVXF motif, does not appear to be involved in the nuclear targeting of PP1␥. Consistent with a role for NIPP1 in the nuclear targeting and function of PP1, we also show that nuclear proteins in NIPP1 Ϫ/Ϫ mouse embryos are hyperphosphorylated.

MATERIALS AND METHODS
Plasmids and Recombinant Proteins-The full-length sequences of rabbit PP1␣ and PP1␤/␦ and rat PP1␥1 were introduced between the XhoI and BamHI sites of pEGFP-C1 (Clontech), yielding expression vectors for EGFP-PP1 isoforms. NIPP1 and NIPP1-(143-224) were subcloned in the pGMEX-T1 vector (Amersham Biosciences) downstream of the glutathione S-transferase (GST) encoding cassette. In this vector the EcoRV sites were used for the cloning of NIPP1 and the XmaI/NotI sites for NIPP1-(143-224). Sds22 was subcloned in the BamHI site of the pSG5 vector (Stratagene) with a triple FLAG tag cassette inserted in its EcoRI site. The pcDNA1Neo plasmid (Invitrogen) encoding FLAG-tagged PNUTS was a kind gift of Dr. Y.-G. Kwon (22), whereas the pTactac vector encoding human PP1␥ 1 was kindly provided by Dr. D. Barford (23). The indicated mutants of PP1␥1, EGFP-PP1, and GST-NIPP1-(143-224) were made according to the QuikChange protocol (Stratagene) with the appropriate primers and templates. All constructs and mutants were verified by DNA sequencing.
Following GST pull-down, the precipitates were analyzed by immunoblotting with antibodies against GST (Santa Cruz Biotechnology) or EGFP (Fig. 4). The EGFP immunoprecipitates were used for immunoblotting with affinity-purified polyclonal antibodies against PNUTS, NIPP1 (25), and Sds22 (26) (Fig. 2). For the production of anti-PNUTS antibodies, a synthetic peptide (CGDPNQLTRKGRKRKTVTWPEE-GKL) comprising residues 384 -407 of PNUTS and an additional Nterminal cysteine was coupled to keyhole limpet hemocyanin and used to raise antibodies in rabbits. The antibodies were affinity-purified on albumin-coupled peptide linked to CNBr-activated-Sepharose-4B (Amersham Biosciences).
Confocal Microscopy and Immunocytochemistry-24 h after transfection, the COS1 or HeLa cells were washed twice with PBS and fixed for 10 min with 2% formaldehyde. Cell permeabilization was performed by a 10-min incubation in PBS supplemented with 0.5% Triton X-100. The permeabilized cells were washed three times for 10 min with PBS, pre-incubated for 20 min with PBS containing 3% bovine serum albumin, and then incubated for 90 min with either polyclonal antibodies against GST (Santa Cruz Biotechnology) or monoclonal antibodies against the FLAG tag (Stratagene). After three washes of 10 min each with PBS, the cells were incubated for 1 h with secondary anti-rabbit (for GST detection) or anti-mouse antibodies (for FLAG detection) that were labeled with tetramethylrhodamine isothiocyanate (TRITC; Sigma). Finally, the cells were washed three times for 10 min in PBS. Confocal images were obtained with a Zeiss LSM-510 laser-scanning confocal microscope (Jena, Germany) equipped with the Zeiss Axiovert  Nuclear Retention Assay-Cells transfected with EGFP-PP1␥ 1 were washed twice with "transport" buffer containing 20 mM HEPES, 110 mM potassium acetate, 5 mM sodium acetate, 2 mM magnesium acetate, 1 mM EGTA, and 2 mM dithiothreitol (27). Subsequently, the cells were permeabilized for 5 min with digitonin (40 g/ml) in transport buffer. After a wash with transport buffer, the permeabilized cells were incubated with transport buffer containing 200 M synthetic peptide NIPP1-(191-200) (KNSRVTFSED) or 200 M variant peptide KN-SRATASED. After 15 min, the green fluorescence was visualized by confocal microscopy.

RESULTS
The Nuclear Targeting of PP1␥ Requires a Functional RVXFbinding Channel-To study the subcellular localization of PP1␥ 1 , we transfected COS1 cells with a construct encoding a fusion of EGFP and rat PP1␥ 1 . In accordance with published data (10), the EGFP-PP1␥ 1 fusion was largely nuclear and enriched in the nucleoli (Fig. 1A). Two established inactive point mutants of PP1␥ 1 , i.e. EGFP-PP1␥ 1 -D64N and EGFP-PP1␥ 1 -H125A (28), showed the same subcellular distribution (not shown), demonstrating that the protein phosphatase activity of PP1␥ 1 is not required for its correct localization. The PSORT II program (29) predicts that the C-terminal KKKP sequence (residues 301-304) of PP1␥ 1 is a candidate NLS. However, the replacement of each of the three lysines in this sequence by an alanine did not affect the nuclear accumulation of EGFP-PP1␥ 1 (Fig. 1B), indicating that they are not part of a functional NLS. Because PP1␥ 1 did not appear to contain a functional polybasic or bipartite NLS, we reasoned that PP1␥ might be piggyback transported to the nucleus by its association with protein interactors that themselves contain one or more functional NLSs. We therefore generated two PP1␥ 1 variants that are mutated in the RVXF-binding channel, which is involved in the binding of most interactors of PP1 (see Introduction). EGFP-PP1␥ 1 -⌬286 -323, which is truncated just before the last ␤-strand (␤14) that lines the RVXF-binding channel (26), was largely cytoplasmic and sometimes showed a speckled distribution in the cytoplasm (Fig. 1C). A more subtle mutation of the RVXF-binding channel, as in EGFP-PP1␥ 1 -F257A, also largely abolished the nuclear accumulation (Fig.  1D). The corresponding mutations in other PP1 isoforms, as in EGFP-PP1␣-F257A and EGFP-PP1␤/␦-F256A, also resulted in a nuclear exclusion of these fusions (not shown). These data suggest that the RVXF-binding channel is required for the nuclear targeting of all PP1 isoforms. The cytoplasmic localization of the EGFP-PP1 fusions with a mutated RVXF channel also confirms that PP1 has no NLS that enables importinmediated transport to the nucleus. Identical data as shown in Fig. 1 were also obtained in HeLa cells (not shown).
To examine whether EGFP-PP1␥ 1 -⌬286 -323 and EGFP-PP1␥ 1 -F257A indeed display a reduced affinity for RVXF-containing interactors, we performed immunoprecipitations of the EGFP fusions from COS1 cell lysates with anti-EGFP antibodies ( Fig. 2A). The immunoprecipitation of EGFP-PP1␥ 1 -⌬286 -323 and EGFP-PP1␥ 1 -F257A did not result in the co-immunoprecipitation of two RVXF-containing interactors (NIPP1 and PNUTS) but showed an increased co-immunoprecipitation of Sds22, which does not contain a functional RVXF motif. The increased interaction of EGFP-PP1␥ 1 -⌬286 -323 with Sds22 has been reported previously and has been explained by a lack of competition with RVXF-containing interactors of PP1 (26). Our observation that the cytoplasmically retained mutants of EGFP-PP1␥ 1 show an increased interaction with Sds22 is initial evidence that the nuclear targeting of EGFP-PP1␥ 1 is not mediated by Sds22.
Nuclear Re-targeting of EGFP-PP1␥ 1 -F257A by the Overexpression of Nuclear Interactors with an RVXF Motif-The above data are consistent with the view that PP1␥ 1 is piggyback transported to the nucleus by association with proteins that contain both a PP1-binding RVXF motif and an NLS. In further agreement with this hypothesis we found that the cytoplasmic EGFP-PP1␥ 1 -F257A, which only showed a moder- ately decreased affinity for RVXF motifs (Fig. 2B), could be re-targeted to the nucleus by the co-expression of a GST fusion of NIPP1-(143-224) (Fig. 3C), a protein with an established NLS as well as a PP1-binding RVXF motif (19). Importantly, the co-expression of GST-NIPP1-(143-224)-V201A/F203A, which has a mutated RVXF motif but still contains a functional NLS, did not affect the cytoplasmic accumulation of EGFP-PP1␥ 1 -F257A (Fig. 3D). Intriguingly, the expression of GST-NIPP1-(143-224) re-targeted EGFP-PP1␥ 1 -F257A to the nucleus but not to the nucleoli (Fig. 3C). However, wild-type PP1␥ 1 (EGFP-PP1␥) also was no longer enriched in the nucleoli following the expression of GST-NIPP1-(143-224) (Fig. 3A). A similar subnuclear redistribution of PP1␥ 1 has been described following the expression of full-length NIPP1 (10) and can be accounted for by the accumulation of NIPP1 outside the nucleoli. As expected, the expression of GST-NIPP1-(143-224) with a mutated RVXF motif did not affect the subcellular distribution of EGFP-PP1␥ (Fig. 3B).
We have subsequently compared the ability of the major nuclear interactors of PP1 to re-target EGFP-PP1␥ 1 -F257A to the nucleus. EGFP-PP1␥ 1 -F257A was nearly exclusively nuclear following the co-expression of either full-length NIPP1 (Fig. 5A) or PNUTS (Fig. 5B). However, EGFP-PP1␥ 1 -F257A remained cytoplasmic when FLAG-tagged Sds22 was co-expressed despite the fact that this fusion was partially nuclear (Fig. 5C) and bound very tightly to EGFP-PP1␥ 1 -F257A (not shown). These data show that the nuclear targeting of PP1␥ is likely to be mediated by various nuclear interactors with an RVXF motif. In contrast, PP1 interactors such as Sds22 that are nuclear but lack an RVXF motif do not seem to be capable of co-transporting PP1 to the nucleus.
The Nuclear Retention of PP1␥ 1 Is Dependent on Interactors with an RVXF Motif-To examine whether the nuclear retention of EGFP-PP1␥ 1 is also dependent on its association with RVXF-containing interactors, we have added the RVXF-containing peptide KNSRVTFSED, corresponding to NIPP1-(191- 200), to digitonin-permeabilized cells. Digitonin was added at concentrations (40 g/ml) known to perforate the plasma membrane without affecting the nuclear envelope or other major intracellular membranes (27). In Fig. 6B it is shown that the addition of KNSRVTFSED, which is expected to compete with endogenous RVXF-containing proteins for binding to PP1, resulted in a complete loss of EGFP-PP1␥ 1 from the nucleus. By contrast, a similar peptide with a mutated RVXF motif (KN-SRATASED) did not have any effect on the nuclear localization of EGFP-PP1␥ 1 (Fig. 6C). Thus, the nuclear retention of PP1␥ 1 also depends on its interaction with RVXF-containing proteins.
A Deficiency of NIPP1 Is Associated with a Hyperphosphorylation of Nuclear Proteins-To examine whether the nuclear targeting and/or retention of endogenous PP1 also depends on proteins with an RVXF motif, we have initially tried to make use of commercially available PP1 antibodies to visualize endogenous PP1 after the overexpression of NIPP1 or PNUTS. However, PP1 antibodies from three different commercial sources could not be used for the immunodetection of endogenous PP1 in intact cells (not shown), in accordance with observations by Trinkle-Mulcahy et al. (30). As an alternative approach, we have therefore examined the consequence of the targeted disruption of the NIPP1 alleles in mice on the phosphorylation level of nuclear proteins, using anti-phosphothreonine antibodies. Indeed, if NIPP1 would be required for the nuclear targeting and/or retention of PP1, one would expect that the substrates of NIPP1-associated PP1 would be hyperphosphorylated in NIPP1 Ϫ/Ϫ cells. The targeted disruption of the NIPP1 alleles is associated with growth retardation and is embryonic lethal at ϳ7 dpc (31). In Fig. 7 it is shown that the NIPP1 Ϫ/Ϫ embryos at 6.5 dpc displayed a considerably increased nuclear protein phosphorylation level as compared with that of the wild-type embryos at 6.5 dpc. Wild-type embryos at 5.5 dpc, which correspond in size to the NIPP1 Ϫ/Ϫ embryos at 6.5 dpc, did not show a nuclear hyperphosphorylation on threonine (not illustrated). These data are consistent with a role for NIPP1 in the nuclear targeting and/or retention of PP1 and also show that NIPP1-associated PP1 has nuclear substrates. The latter conclusion also fits in nicely with previous observations that the NIPP1-PP1 complex, albeit inactive when reconstituted in vitro, can become an active protein phosphatase by the phosphorylation of NIPP1 (24). DISCUSSION By far, most proteins enter the nucleus either passively or by importin-mediated transport. However, neither of these mechanisms can account for the nuclear accumulation of PP1. Indeed, there is no evidence for the existence of free PP1, and none of the PP1 holoenzymes are small enough to diffuse through the nuclear pores passively. Also, PP1␥ 1 contains a consensus NLS, but this motif is not conserved in other PP1 isoforms (13) and its mutation did not affect the nuclear accumulation of PP1␥ 1 (Fig. 1B). We have obtained data strongly indicating that PP1 is translocated to the nucleus by the association with proteins that contain both an NLS and a PP1binding RVXF motif. First, a deletion or point mutation of the RVXF-binding channel of PP1 isoforms abolished the nuclear targeting of the phosphatase. Second, a PP1␥ 1 mutant with a decreased affinity for the RVXF motif (PP1␥ 1 -F257A) and a cytoplasmic accumulation could be re-targeted to the nucleus by the overexpression of nuclear interactors with a functional RVXF motif (Fig. 3). However, PP1␥ 1 -(⌬286 -323), which did not bind at all to RVXF-containing interactors (Fig. 4), could not be re-targeted to the nucleus. Both NIPP1 and PNUTS/ R111, which represent the major nuclear interactors of PP1 in mammalian cells (9), were able to translocate PP1␥ 1 -F257A to the nucleus, suggesting that they both represent true "nuclear targeting subunits" of PP1. Consistent with this view, we found that the disruption of the NIPP1 genes was associated with a nuclear hyperphosphorylation of proteins on threonine in mouse embryos. Interestingly, the established NLS of NIPP1-(143-224) is just N-terminal to the RVXF motif (19) and has also been implicated in the potent inhibition of PP1 by NIPP1 (32). Our finding that NIPP1-(143-224) and PP1␥ are co-transported (Fig. 3) strongly suggests that the binding of importin-␣ to this NLS and the binding of PP1 to the RVXF motif are not mutually exclusive.
S. cerevisiae only expresses a single PP1 isoform termed Glc7. In accordance with our data, Glc7-F256A, which corresponds to PP1␥ 1 -F257A, also showed a decreased affinity for interactors with an RVXF motif but still interacted with Sds22 (33). Glc7-F256A also showed some subcellular localization deficiencies but, at variance with our results, Glc7-F256A was still associated with the nucleoli. We suggest that the cytoplasmic retention of PP1␥ 1 -F257A in mammalian cells is accounted for by the competition of other PP1 isoforms for binding to RVXF-containing interactors. In contrast, for the lack of competing PP1 isoforms Glc7-F256A can still accumulate in the nucleus in yeast. PP1␥ 1 -F257A could not be re-targeted to the nucleus by the overexpression of Sds22 (Fig. 5), which interacts with a fragment of PP1 that is remote from the RVXF-binding channel (26). This was an unexpected finding, because FLAG-Sds22 is  Paraffin-embedded transverse sections of NIPP1 wild-type (NIPP1 ϩ/ϩ ) and NIPP1 knock-out (NIPP1 Ϫ/Ϫ ) embryos of 6.5 dpc were immunostained with antibodies against phosphothreonine (Zymed Laboratories Inc.) and counterstained with Harris hematoxylin. The NIPP1 Ϫ/Ϫ embryos show a stronger staining with anti-phosphothreonine antibodies. No staining with the phosphothreonine antibodies was detected after preincubation of the antibodies with phosphothreonine (not shown). PrC, proamniotic cavity partially nuclear (Fig. 5C), is associated with PP1 in the nucleus (34), and interacts even better with PP1␥ 1 -F257A and PP1␥ 1 -(⌬286 -323) than with wild-type PP1 ( Fig. 2A). Our data therefore suggest that Sds22 migrates to the nucleus either independently of PP1 or as part of a complex that also contains a protein with a PP1-binding RVXF motif.
The (sub)nuclear targeting and retention of EGFP-PP1␥ 1 also appears to be mediated by RVXF-containing interactors, because virtually no EGFP-PP1␥ 1 was left in the nucleus after a short incubation of permeabilized cells with an RVXF peptide (Fig. 6B). Whereas some EGFP-PP1␥ 1 may have leaked out of the permeabilized cells, a considerable fraction remained cytoplasmic. These data indicate that there exist one or more cytoplasmic proteins that can bind PP1 independently of an RVXF motif and that are associated with structures (the cytoskeleton?) that do not leak out of permeabilized cells. Preliminary biochemical assays indeed provide evidence for the existence of a cytoplasmic protein or proteins that bind and inhibit exogenous purified PP1 in the presence of an RVXF peptide (not illustrated). We speculate that such inhibitory proteins keep newly synthesized PP1 from accidentally dephosphorylating proteins until it can form a holoenzyme by association with an RVXF-containing interactor.