Determinants That Control the Distinct Subcellular Localization of p38α-PRAK and p38β-PRAK Complexes*

p38α and p38β MAPKs (mitogen-activated protein kinases) share about 80% of their protein sequence identity, but have quite different biological functions. One such difference is in regulating the subcellular localization of their downstream kinases, such as PRAK (p38-regulated/activated protein kinase or MK5). The p38α-PRAK complex is found in the nucleus, whereas the p38β-PRAK complex is exclusively localized to the cytosol. By generating a series of chimeric and point mutants of p38α and p38β, we found two amino acid residues (Asp145 and Leu156 in p38α, Gly145 and Val156 in p38β) that determine the distinct subcellular locations of p38α-PRAK and p38β-PRAK. The subcellular localization of MK2 (MAPK-activated protein kinase 2), another downstream kinase of p38, was regulated in the same manner as that of PRAK. We found that nuclear import, but not export, determines the subcellular localization of p38α-PRAK and p38β-PRAK. The published structure of the p38α-MK2 complex suggests Leu156 of p38α is involved in the interaction with the nuclear localization signal in PRAK. The difference at this residue between p38α and p38β may affect the nuclear localization signal in PRAK differently, and thereby influence the import of the complexes. Asp145 in p38α (or Gly145 in p38β) is located on a different surface patch, and further random mutagenesis revealed that mutation of Asp145, Thr123, and Gln325, the residues that can directly interact with importin α as predicted by modeling, but not mutation of the other 7 amino acid residues that cannot reach importin α, re-locate p38α-PRAK to the cytosol, suggesting that interaction with import machinery is involved in determining the subcellular localization of the p38α-PRAK and p38β-PRAK complexes. Last, we show that nuclear localization of PRAK is required for its role in inhibiting the proliferation of NIH3T3 cells. In conclusion, multiple determinants control the distinct subcellular localization of p38α-PRAK and p38β-PRAK complexes, and the location of PRAK plays a role in its function.

The physical location of a protein is essential for its biological function in eukaryotic cells, in which many proteins traffic continuously, whereas others selectively localize to different subcellular compartments. The best understood system for the transport of macromolecules between the cytoplasm and the nucleus is the nuclear import and export pathway (1)(2)(3)(4). In this pathway, nuclear import and export of proteins is mediated through the nuclear pore complex by a superfamily of transport receptors known collectively as karyopherins (1). These proteins recognize specific signal sequences, the nuclear localization signal (NLS) 3 and nuclear export signal (NES), in the proteins being transported. The typical NLSs contain a cluster of basic amino acids as is exemplified by the SV40 large T-antigen with the amino acid sequence PKKKRKV and is known to be bound with importin ␣ (5-7). NESs identified up to now are rich in hydrophobic amino acids, such as leucine or isoleucine. The traffic of some proteins between the cytoplasm and nucleus is mediated by a chaperon (8 -10). In these systems, the NLS or NES is not in the proteins being transported, but in the chaperon. The transport of p38␣ MAP kinase (mitogen-activated protein kinase) is chaperoned by either MK2 (MAP kinase-activated protein kinase-2) or PRAK (p38-regulated/activated kinase or MK5) that carries conventional NLS and NES sequences (11,12). However, other methods of p38␣ translocation may exist, since the homologous protein of p38 in yeast, Hog1, can be translocated into the nucleus in an importin ␣-independent mechanism without interacting with an NLS-containing chaperon (13).
The p38 group of protein kinases is one of the four subfamilies of the MAP kinase family that plays a role in a variety of biological processes that include cell proliferation, differentiation, senescence, and cell death (14 -17). None of the p38 group kinases contains NLS or NES sequences, but their subcellular localization can be regulated, at least in some cases, by their interacting proteins. It needs to be noted that they are not passively regulated by their chaperons. An interesting pair in this group is p38␣ and p38␤, which can in turn regulate the nuclear and cytoplasmic localization of their interacting proteins, such as PRAK, via an unknown mechanism (12). PRAK is not only a chaperon of p38␣ and p38␤, but also a substrate of the p38 group kinases that can inhibit Ras-mediated cell proliferation (18,19). Recently a study showed that PRAK plays a crucial role in Ras-induced senescence and tumor suppression (20). p38␣ and p38␤ have ϳ80% identical amino acid sequences but exhibit vastly different functions (16). The interaction of p38␣ with PRAK leads to both of them being co-localized to the nucleus. In contrast, the complex of p38␤ and PRAK results in cytosol localization. It is unclear how these two highly homologous proteins can so dramatically localize PRAK to the different subcellular compartments. To shed light on the mechanism by which p38␣ and p38␤ differently regulate localization of PRAK, we produced a series of chimeric proteins of p38␣ and p38␤, as well as their point mutants. We demonstrate that two amino acids (amino acids 145 and 156) in p38␣ and p38␤ are critical for them to differentially regulate the localizations of their binding partner PRAK, as well as MK2. We show that the subcellular location of p38␣-PRAK and p38␤-PRAK is determined by whether they can be imported into the nucleus, and we found that both the Asp 145 and Leu 156 in p38␣ are required for p38␣-PRAK importation into nucleus. Our modeling of p38␣-PRAK and p38␤-PRAK complexes based on the published crystal structure of the p38␣-MK2 complex suggests that residue 156 of p38␣/␤ interacts with PRAK docking peptide, which overlaps with its NLS. Data from the mutagenesis experiments and modeling implicate the involvement of Asp 145 in the interaction of the p38␣-PRAK complex with importin ␣. In addition, we found that PRAK-mediated inhibition of NIH3T3 cell growth is dependent on its nuclear localization, highlighting the importance of the location of PRAK on its biological function.
Immunofluorescent Staining-NIH3T3 cells were grown on glass coverslips in the cell culture medium, as described above, for 16 h. Expression vectors of FLAG-p38␣, FLAG-p38␤, and GFP-PRAK or various mutations of them were co-transfected into NIH3T3 cells in different combinations as indicated where necessary. 24 h post-transfection, cells were fixed with 4% polyformaldehyde in PBS for 10 min, followed by treatment with 0.2% Triton/PBS for 5 min. After rinsing three times with PBS, the coverslips were incubated with 10% normal rabbit blocking serum (Santa Cruz Biotechnology) in PBS for 30 min to suppress nonspecific binding of fluorochrome-conjugated IgG, followed by washing in PBS three times. Coverslips were then incubated with anti-FLAG M2 antibodies (Sigma) for 60 min at 37°C. After washing three times with PBS for 5 min each wash, cells were stained with rhodamine-conjugated rabbit IgG (diluted in 1.5-3% normal rabbit blocking serum) for 45 min at room temperature, followed by extensive washing with PBS. Cells were then stained with the DNA-binding dye 4Ј,6-diamidino-2-phenylindole (0.1 g/ml in PBS) for 10 min. Then the coverslips were subjected to rinsing with PBS, and finally mounted with GelMount (Biomeda Corp., Foster City, CA) and kept at 4°C. Slides were examined under an inverted fluorescence microscope (Olympus Iϫ51) for GFP fluorescence, rhodamine, and 4Ј,6-diamidino-2-phenylindole staining at wavelengths of 488/507, 540/625, and 360/460 nm, respectively. Images were captured through a ϫ100 objective lens by a DP50 microscope digital camera system (Olympus).
Cell Proliferation Assay-NIH3T3 cells were grown on coverslips in media as described above. When cells reached ϳ60% confluence, they were transfected with 1 g of expression plasmids of GFP, GFP-PRAK, GFP-PRAK(QTTG), or GFP-PRAK(R361Q) alone, or in different combinations with 1 g of p38␣, p38␣(D145G), p38␤, or p38␤(G145D), as indicated in Fig. 7, by using Lipofectamine 2000 transfection reagent. Cell proliferation was determined by BrdUrd intake assay, employing a 5-Bromo-2Ј-deoxyuridine Labeling and Detection Kit I (Roche) according to the manufacturer's instructions. In brief, 24 h post-transfection, the cell culture medium was changed with fresh medium supplemented with BrdUrd labeling solution (1 ϫ concentration). After incubating the cells at 37°C for 30 min, the BrdUrd labeling medium was aspirated and coverslips were washed in washing buffer three times. Cells were then fixed with the ethanol fixative (add 50 mM glycine solution to 70 ml of absolute ethanol to get 100 ml of fixative, pH 2.0) for 20 min. at Ϫ25°C, followed by rinsing three times with washing buffer. Fixed cells were incubated with mouse anti-BrdUrd working solution for 1 h at 37°C and washed three times. Hereafter, all coverslips were stained with rhodamine-conjugated anti-mouse secondary antibodies for 1 h at 37°C. After washing three times in wash buffer, coverslips were mounted with Gel-Mount (Biomeda Corp., Foster City, CA) and kept at 4°C. Slides were examined under a fluorescence microscope (Olympus Iϫ51) for GFP fluorescence and BrdUrd staining. To obtain the rate of nuclear BrdUrd intake, 600 cells positive for GFP were counted for each coverslip and the percentage of BrdUrdpositive nuclei in GFP-positive cells was calculated.
p53 Reporter Assays-NIH3T3 cells were transiently transfected with different combinations of plasmids expressing HA-PRAK, Ha-RasV12, p38␣, p38␤, p38␣(D145G), and p38␤(G145D), as indicated in Fig. 7, D and E, together with 0.5 g of the p53-dependent reporter plasmid PG-Luc. 24 h after transfection, luciferase activities were determined and normalized to the values of ␤-galactosidase activity derived from cotransfection with 0.5 g of the pCMV5-LacZ vector. The results are presented as mean Ϯ S.D. from at least five separate experiments.

Sequences in p38␣ and p38␤ That Regulate the Subcellular
Location of PRAK-Our previous study showed that interaction with p38␣ or p38␤ determines the nuclear or cytosolic location of PRAK in HeLa and HEK293 cells. It was also demonstrated that GFP-PRAK, a PRAK fusion with the reporter protein GFP, adequately mimicked the subcellular localization of PRAK (12). As controls in this study, ectopically expressed GFP-PRAK alone in NIH3T3 cells was found in the nucleus under a fluorescent microscope (Fig. 1A); FLAG-tagged p38␣ and p38␤ are generally diffused in the cells with some preference for FLAG-p38␣ in the nucleus and FLAG-p38␤ in the cytosol (Fig. 1A). The interactions of GFP-PRAK with FLAG-p38␣ and FLAG-p38␤ can be demonstrated by co-immunoprecipitation (Fig.  1B). The interaction of FLAG-p38␤ and GFP-PRAK leads to cytosolic localization of these two proteins, whereas the FLAG-p38␣ and GFP-PRAK interaction localize them to the nucleus (Fig. 1C).
To better understand the structure basis of the different functions of p38␣ and p38␤, a series of plasmids were constructed to express chimeric proteins (M1-M10) of p38␣ and p38␤ ( Fig. 2A). Each of these chimeric proteins was coexpressed with GFP-PRAK in NIH3T3 cells to determine their interaction with PRAK and their effects on PRAK subcellular localization. All of these chimeric proteins retained the ability to interact with GFP-PRAK (Fig. 2B). Analysis of the subcellular localizations of GFP-PRAK and FLAG-chimeric proteins revealed that the chimeric proteins contain-ing the p38␣ sequence from amino acids 132 to 170 (M2, M3, M4, M8, and M10) exhibited the same effect as wild type p38␣ in regulating the distribution of PRAK (Fig. 2C), whereas the chimeric mutants that contain the same sequence as p38␤ in the region from amino acid 132 to 170 (M1, M5, M6, M7, and M9) were co-localized with GFP-PRAK in the cytoplasm (Fig. 2C). Therefore, it is the sequence between amino acid residues 132 and 170 that determines the different functions of p38␣ and p38␤ in regulating the subcellular distribution of p38␣-PRAK and p38␤-PRAK.
There are four amino acids differences in the region from residues 132 to 170 between the p38␣ and p38␤ sequences (Fig.  3A). Eight mutants of p38␣ and p38␤ were generated by swapping each of these residues between p38␣ and p38␤. None of these mutations affected the interaction between p38␣ (or p38␤) and PRAK in the co-immunoprecipitation experiments (Fig. 3, B and C). However, Asp 145 to Gly (D145G) and Leu 156 to Val (L156V) mutations of p38␣ altered the subcellular localization of its complex with PRAK to the cytosol, much like the behavior of p38␤ in regulating PRAK subcellular localization, FIGURE 1. Distinct subcellular distributions of p38␣-PRAK and p38␤-PRAK complexes. A, subcellular localization of overexpressed GFP, GFP-PRAK, p38␣, and p38␤. 3 g of each expression plasmid, including GFPC1, GFPC1-PRAK, pcDNA3-FLAG-p38␣, and pcDNA3-FLAG-p38␤, was transfected into NIH3T3 cells. 24 h post-transfection, the cells transfected with p38␣ and p38␤ were stained with a mouse anti-FLAG M2 antibody and a rhodamine-conjugated rabbit anti-mouse secondary antibody (red). Localization of GFP-PRAK was viewed by a fluorescent microscope. Nuclei were displayed by staining of DNA with 4Ј,6-diamidino-2-phenylindole (DAPI) (blue). B, PRAK shows the same affinity for interacting with p38␣ and p38␤. 2 g of GFP-PRAK expression vector was cotransfected with 2 g of pcDNA3-FLAG-p38␣ or pcDNA3-FLAG-p38␤ into HEK 293T cells. 36 h after transfection, immunoprecipitation was performed to determine the interaction between PRAK and p38␣/␤. Abbreviations used herein are: IP, immunoprecipitation; IB, immunoblotting; and TCL, total cell lysate. C, subcellular locations of PRAK, p38␣, and p38␤ when they were coexpressed in different combinations. NIH3T3 cells were co-transfected with 1 g of GFPC1-PRAK and 3 g of pcDNA3-FLAG-p38␣ or 3 g of pcDNA3-FLAG-p38␤. After 24 h of transfection, the cells were fixed and analyzed by immunofluorescent staining, employing the same antibodies used in A. Overlapping localizations between PRAK and p38␣ or p38␤ are shown in the merged images (yellow).
whereas the other two mutations (I134L and K165R) did not cause any changes (Fig. 3D). In the complementary experiments, mutations of residues 145 and 156 in p38␤ to their counterparts in p38␣ changed the role of p38␤ in regulating PRAK subcellular localization to that of p38␣ (Fig. 3E). The amino acids at 145 and 156 also determine whether the mutants behave similar to p38␣ or p38␤ in localization when expressed alone (supplemental Fig. 1). Collectively, our data demonstrate that both of these two residues are crucial for p38␣ and p38␤ to regulate PRAK subcellular localization. A switch of either residue between p38␣ and p38␤ alters the molecular conduct of p38␣ and p38␤ in regulating PRAK subcellular localization.
The Subcellular Location of MK2 Can Be Regulated by p38␣ and p38␤ in the Same Way as That of PRAK-Several protein kinases, including MK2 and MNK1 (MAPK-interacting kinase 1), also interact with p38␣ (11,21,22), and it is of interest to determine whether they can be regulated by p38␣ and p38␤ to their respective subcellular locations in a manner similar to that of PRAK. Both MK2 and MNK1 can interact with p38␣ and p38␤ (Fig. 4A). When expressed individually in NIH3T3 cells, GFP-MK2 and GFP-MNK1 were localized to the nucleus and cytosol, respectively (Fig. 4B). This is consistent with results obtained by other groups (11,23,24). Coexpression of either p38␣ or p38␤ did not affect the cytosolic localization of MNK1 (Fig. 4C). This is not surprising, as MNK1 and PRAK are quite different in the structural location of their putative NLS. NLS in MNK1 is located at the N terminus but in PRAK it is located at the C terminus. As expected, the subcellular distribution of MK2, with its NLS being aligned well with that in PRAK, can be regulated by p38␣ and p38␤ in the same way as that of PRAK (Fig. 4D). Additionally, the mutants of p38␣ and p38␤ regulate the subcellular localization of MK2 in the same fashion as that of PRAK (Fig. 4, E and F).
Nuclear Import but Not Export Is Important in Determining the Subcellular Location of p38␣-PRAK and p38␤-PRAK-As we mentioned earlier, p38␣ and p38␤ do not contain NES and NLS sequences, whereas PRAK has both (Fig. 5A). To determine whether NES-mediated nuclear export plays a role of the cytosolic localization of p38␤-PRAK, we used a PRAK mutant, GFP-PRAK(SSS), in which the NES motif was mutated (12). Mutation of NES in PRAK did not affect the interaction of PRAK with p38␣/␤ (data not shown). As expected, mutation of NES did not affect nuclear localization of PRAK when it was expressed alone in NIH3T3 cells (Fig. 5B) or coexpressed with p38␣ (data not shown). Yet, when coexpressed with p38␤, GFP-PRAK(SSS) was mainly localized to the cytoplasm (Fig. 5C, bottom), reminiscent of the distribution pattern of wild type PRAK, indicating that NES in PRAK is not required for cytosolic locali- The blank part represents the amino acid sequence derived from p38␣, and the black part represents the sequence from p38␤. The amino acid numbers of the swapping sites were labeled on the top of the diagram. B, all mutants interact with PRAK at the same affinity as that of wide type p38␣/␤. 293T cells were cotransfected with GFPC1-PRAK alone or in combination with each of the FLAG-tagged p38 mutants (M1-M10). After 36 h of transfection, cells were harvested, followed by immunoprecipitation (IP) and Western blotting (IB) as described in the legend to Fig. 1B. C, determination of the domains in p38␣ and p38␤ that regulate the subcellular locations of PRAK. 1 g of each chimeric construct (M1-M10) was cotransfected with GFPC1-PRAK into NIH3T3 cells. After 24 h of transfection, cells were sequentially stained with mouse anti-FLAG M2 antibodies and rhodamine-conjugated rabbit anti-mouse secondary antibodies for FLAG-p38 (red). Localization of PRAK was displayed by GFP, and nuclei were visualized by 4Ј,6-diamidino-2-phenylindole (DAPI) staining (blue). The merged images (yellow) show co-localization between PRAK and p38␣/␤. TCL, total cell lysate.
zation of the p38␤-PRAK complex. Treatment with leptomycin B, an inhibitor of nuclear export that binds to the chromosomal region maintenance 1, had no effect on the cytosolic localization of p38␤-PRAK (Fig. 5C, middle), consistent with the notion that nuclear export has no role in the cytosolic localization of the p38␤-PRAK complex.
Because NLS overlaps with the docking motif in PRAK (Fig.  5A), mutation of the four amino acids RKRK to QTTQ in PRAK not only abolishes its nuclear localization but also eliminates its interaction with p38␣/␤ (12). To separate the docking function from NLS, a mutation of R361Q in PRAK was generated, which did not alter the PRAK interaction with p38␣ and p38␤ (Fig.  5D). The R361Q mutant is localized to the cytosol when expressed alone in NIH3T3 cells, similar to the QTTQ mutant (supplemental Fig. 2), indicating that the NLS function was interrupted. As shown in Fig. 5E, wild type PRAK was localized to the nucleus when coexpressed with p38␣. In contrast, the R361Q mutant was localized to the cytosol. This data suggests that NLS in PRAK is essential for the nuclear localization of p38␣-PRAK.
Asp 145 and Leu 156 in p38␣ Are Located on Two Different Surface Patches-The complex structures of the MK2 peptide and its complex with p38␣ were reported recently (25,26). Leu 156 in p38␣ is located in the vicinity that interacts with MK2. Asp 145 on the other hand is located on the other side of p38␣ close to the P-loop surface. Due to the high level of homology to p38␣, we believe Gly 145 and Val 156 in p38␤ are located at the same places as their corresponding Asp 145 and Leu 156 in p38␣. Based on the published structures of the p38␣-MK2 complex, models of p38␣-PRAK and p38␤-PRAK complexes were made (Fig. 6A). Leu 156 or Val 156 (purple spheres) are located in a ␤ strand that interacts with the docking peptides from PRAK (or other substrates such as MEF2A (red ribbon)). It is known that NLS in PRAK overlaps with its docking site for p38␣ and p38␤. Perhaps Leu 156 in p38␣ and Val 156 in p38␤ affects the conformation of the NLS differently for PRAK, therefore the binding with p38␤, but not with p38␣, may interfere with the interaction of NLS with importin ␣ for PRAK, preventing the nuclear import.
Asp 145 in p38␣ or Gly 145 in p38␤ (red spheres) are located on the other side of the surface that interacts with PRAK (Fig. 6A). They may have no role in influencing the function of NLS in PRAK, but could alter the interactions with the protein(s) of import/export machinery to affect the traffic of p38␣/␤-PRAK complexes. Interaction with nucleoporin was shown recently to be important in ERK2 nuclear import (27). Sequence comparison between p38␣ and p38␤ revealed that the sequence corresponding to the nucleoporin binding region in ERK2 is conserved between p38␣ and p38␤, suggesting that the different functions of p38␣ and p38␤ in regulating PRAK localization is not related to their nucleoporin-binding ability.
To understand the role of Asp 145 (or Gly 145 ) in regulating PRAK subcellular localization, a series of point mutations were generated in p38␣ for comparative analysis of their roles in the regulation of p38␣ for PRAK localization. Amino acid 123 in p38␣ is Thr and in p38␤ is Ser. Converting Thr 123 in p38␣ to (Ile 134 , Asp 145 , Leu 156 , and Lys 165 of p38␣ were converted to the corresponding amino acids, Leu, Gly, Val, and Arg of p38␤, respectively) show the same affinity as wild type p38␣ in association with PRAK. GFPC1-PRAK was co-transfected with pcDNA3-FLAG-p38␣, or each of four point mutants of p38␣ into 293T cells. Immunoprecipitation (IP) and Western blot (IB) assay were performed as described in the legend to Fig. 1B. C, four point mutants of p38␤, L134I, G145D, V156L, and R165K (amino acids in these sites were replaced with the corresponding amino acids in p38␣), interact with PRAK in the same affinity as that of wide type p38␤. 293T cells were transfected with GFPC1-PRAK alone, or together with each of the four point mutations of p38␤. Coimmunoprecipitation and Western blot assays were performed as in B. D, two mutants of p38␣, D145G and L156V, show the same property as p38␤ in regulating the subcellular localization of PRAK. NIH3T3 cells were cotransfected with GFPC1-PRAK and each expression plasmid of the four point mutations of p38␣. The transfected cells were then fixed, stained, and analyzed as described in the legend to Fig. 2C. E, two mutants of p38␤, G145D and V156L, behave similarly to p38␣ in regulating the subcellular localization of PRAK. NIH3T3 cells were cotransfected with GFPC1-PRAK and each of the point mutants of p38␣. The transfected cells were then fixed and stained as in D for 24 h post-transfection. TCL, total cell lysate.
Ala relocated coexpressed PRAK to the cytosol (Fig. 6B). Amino acid 325 in p38␣ is Gln and in p38␤ is Glu. A Q325G mutation in p38␣ resulted in cytosolic localization of coexpressed PRAK (Fig. 6B). In contrast, PRAK coexpressed with p38␣ mutants containing single changes of H107T, L122G, L130G, I134G, R149G, V158A, or V319E were localized to the nucleus, just as with wild type p38␣ (Fig. 6C). A summary of point mutations on the three-dimensional structure of p38␣ complexed with PRAK is shown in Fig. 6D. All of these mutants retained the ability to interact with GFP-PRAK (Fig. 6E). Because the structures of importin ␣ in complex with NLS peptides are available (5-7), we included importin ␣ into our modeling (Fig. 6D). Based on this modeled structure, Asp 145 , Thr 123 , and Gly 325 in p38␣ are in the interacting distance with importin ␣ (Ͻ4 Å), whereas His 107 , Leu 122 , Leu 130 , Ile 134 , Arg 149 , and Val 158 are not able to interact importin ␣. Although Val 319 is in the distance for interaction with importin ␣, it may not be important for interacting with importin ␣ because its mutation did not affect the localization of p38␣-PRAK. Collectively, these data suggest that the interaction with importin ␣ is also required for nuclear import of the p38␣-PRAK complex, in which Asp 145 in p38␣ plays a vital role.

PRAK-induced Growth Inhibition of NIH3T3 Cells Depends on Its
Nuclear Location-It is of great interest to determine whether different subcellular localization of PRAK affects its function. We have shown previously that PRAK is essential in Ras-induced senescence, and overexpression of PRAK can inhibit the proliferation of NIH3T3 cells (18,20). We employed the BrdUrd intake assay to examine whether the inhibitory effect of PRAK on proliferation of NIH3T3 cells is dependent on its specific subcellular location. The data showed that GFP-PRAK robustly inhibited the proliferation of NIH3T3 cells compared with the control GFP (p Ͻ 0.001) (Fig. 7A). GFP-PRAK(QTTG) and GFP-PRAK(R361Q), two mutants that failed to be distributed in the nucleus, showed impaired inhibitory effects on the proliferation of NIH3T3 cells (Fig. 7A) compared with GFP-PRAK (p Ͻ 0.001), indicating that nuclear localization of PRAK is essential for its inhibitory effect on cell proliferation. Expression of GFP did not influence NIH3T3 cell proliferation (Fig.  7B). Coexpression of p38␣ and GFP (Fig. 7B, second column) showed some inhibitory effects on cell proliferation in comparison with GFP alone (Fig. 7B, first column) (p Ͻ 0.001). In contrast, coexpression of p38␤ with GFP (Fig. 7B, third column) exhibited little or no inhibitory effects on NIH3T3 cell proliferation (p Ͼ 0.05 compared with GFP alone). p38␣ enhanced the PRAK-mediated cell growth inhibition (Fig. 7B, fifth column versus the fourth column, p Ͻ 0.01). In contrast, p38␤, which docks PRAK to the cytosol, strongly blocked the inhibitory effect of GFP-PRAK on cell proliferation (Fig. 7B, sixth column), consistent with the idea that nuclear localization of PRAK is essential for its ability to inhibit cell proliferation. . The subcellular distributions of MK2, but not MNK1, can be regulated by p38␣ and p38␤ in the same manner as that of PRAK. A, both MK2 and MNK1 interact with p38␣/␤. Expression plasmids of p38␣, p38␤, MK2, and MNK1 were transfected in different combinations, as indicated, into NIH3T3 cells. 36 h after transfection, co-immunoprecipitation (IP) and Western blotting (WB) were sequentially performed to determine the interaction of p38␣/␤ with MK2 and MNK1. B, MNK1 is mainly distributed in the cytoplasm, and MK2 is primarily located in the nucleus. GFP-MNK1 or GFP-MK2 were transfected into NIH3T3 cells. 24 h posttransfection, the distributions of MNK1 and MK2 were displayed by GFP. C, p38␣ and p38␤ cannot regulate the distribution of MNK1. NIH3T3 cells were cotransfected with GFP-MNK1 and p38␣/␤. Both p38␣ and p38␤ were colocalized with MNK1 in the cytoplasm. D, p38␣ and p38␤ regulate the subcellular distributions of MK2 in the same way as that of PRAK (Fig. 1C). E, two mutants of p38␣, D145G and L156V, show the same property as p38␤ in regulating the subcellular localization of MK2. This is the same case with PRAK (Fig. 3D). NIH3T3 cells were cotransfected with GFP-PRAK and each expression plasmid of the four point mutations of p38␣ followed by staining and analyzing as described earlier. F, two mutants of p38␤, G145D and V156L, behave as p38␣ in regulating the subcellular localization of MK2. TCL, total cell lysate.
To confirm that the effect of p38␤ on PRAK-mediated inhibition of NIH3T3 cell proliferation is due to its docking of PRAK to the cytosol, we used a p38␣ mutant (D145G) that redistributes PRAK to the cytosol and p38␤(G145D) that docks PRAK to the nucleus, to determine their effects on the PRAKmediated inhibition of cell proliferation. As predicted, p38␣(D145G) (Fig. 7C, second column) had no effect on cell proliferation in comparison with wild type p38␣ (Fig. 7B). On the contrary, p38␤(G145D) (Fig. 7C, third column) had an enhanced inhibitory effect on cell proliferation. Coexpression of p38␣(D145G), which docks PRAK to the cytosol, eliminated PRAK-mediated cell growth inhibition (Fig. 7C, fifth column), whereas cotransfection of p38␤(G145D), which causes PRAK to be localized to the nucleus, showed enhanced effects on PRAK-mediated suppression of cell proliferation (Fig. 7C, sixth  column).
Our previous work showed that PRAK-mediated transactivation of p53 plays a role in Ras-induced inhibition of cell growth (20). To determine whether different subcellular distributions of PRAK affect its function in regulating transcriptional activity of p53 upon Ras stimulation, we used a reporter gene assay. As shown in Fig. 7D, PRAK (sixth column) can enhance p53 transcriptional activity by comparison to the control group (first column), and this action can be inhibited by p38␤, but not p38␣ (seventh and eight columns). As anticipated, p38␣-(D145G) (Fig. 7D, ninth column) showed inhibitory effects on p53 transactivation, and p38␤(G145D) (Fig. 7D, tenth column) acted the same as p38␣ on p53 transactivation. Overexpression of Ha-RasV12 in NIH3T3 cells induced the expression of the p53 reporter gene (Fig.  7E, sixth column), and this effect was inhibited by p38␤, but not p38␣ (Fig. 7E, eighth and seventh columns, respectively). As expected, these opposing effects of p38␣ and p38␤ can be reversed by swapping their amino acid residue 145 (Fig. 7E,  ninth and tenth columns). Collectively, we conclude that different subcellular localizations of PRAK, regulated by p38␣ and p38␤, may affect its ability to regulate p53 transactivation.

DISCUSSION
Proteins containing typical NLS and NES mediate their nuclear import and export through interactions of their NLS with a carrier protein complex containing importin ␣, or their NES with chromosomal region maintenance 1 (1-4). However, protein-protein interactions can also influence the subcellular localization of a protein containing NLS and NES signals (4,28,29). PRAK contains both NLS and NES, and it alone is mainly localized to the nucleus (Fig. 1A) (12). Interestingly, interaction with p38␣ and p38␤, both lacking NLS and NES signals and being distributed in both the nucleus and cytoplasm, leads to a solely nuclear distribution of the p38␣-PRAK complex and to an exclusively cytosolic localization of the p38␤-PRAK complex (Fig. 1). Our mutagenesis analysis has revealed that the differences in the two surface patches on p38␣ and p38␤ can influence the subcellular localization of p38␣-PRAK and the p38␤-PRAK complex (Figs. 2, 3, and 6). Structural modeling analysis suggests that the sequence in p38␣ and p38␤ that FIGURE 5. Nuclear import, but not export of PRAK, is required for the distinct locations of p38␣-PRAK and p38␤-PRAK complexes. A, a schematic diagram of PRAK showing its NES and NLS. The mutant PRAK(SSS) generated by converting three Leu residues in the NES to Ser, mutant PRAK(QTTG) produced by replacing four basic residues, Arg, Lys, Arg, and Lys in NLS with Gln, Thr, Thr, and Gly, and mutant PRAK(R361Q) generated by converting Arg to Gln, are indicated at the bottom. B, GFP-PRAK(SSS) is localized to the nucleus similar to GFP-PRAK. NIH3T3 cells were transfected with GFP-PRAK or GFP-PRAK(SSS), and PRAK was displayed by GFP. C, mutation of NES in PRAK does not influence the cytoplasmic localization of the p38␤-PRAK complex. pcDNA3-FLAG-p38␤ was co-transfected with GFPC1-PRAK or GFPC1-PRAK(SSS) into NIH3T3 cells. 24 h after transfection, the cells transfected with GFPC1-PRAK were treated with (ϩ) or without (Ϫ) 5 ng/ml leptomycin B (LMB) for another 6 h, and the cells transfected with GFPC1-PRAK(SSS) were left untreated. All the cells were then subjected to analyze the location of PRAK and p38␤. D, the PRAK mutant R361Q shows the same affinity as PRAK in binding with p38␣ and p38␤. HEK 293T cells were transfected with the expression plasmids of p38␣, p38␤, PRAK, and PRAK(R361Q) in different combinations as indicated. 36 h after transfection, co-immunoprecipitation (IP) and Western blotting (WB) were performed to determine the interactions between PRAK(R361Q) and p38␣/␤. E, PRAK(R361Q), the mutant impaired for nuclear import, sequestrates p38␣ in the cytoplasm. GFPC1-PRAK or GFPC1-PRAK(R361Q) were cotransfected with p38␣ into NIH3T3 cells. The locations of PRAK and p38␣ were determined. TCL, total cell lysate.
directly interacts with the NLS of PRAK may affect the function of the NLS differently (Fig. 6A), and some other sequences in p38␣ may interact with importin ␣ and thus play a role in determining the nuclear localization of the p38␣-PRAK complex (Fig. 6D).
We show that overexpressed p38␣ is distributed more in the nucleus than in the cytoplasm, and coexpression of p38␣ with PRAK leads to both of them being exclusively localized to the nucleus, indicating that the p38␣-PRAK complex has a unique property in subcellular localization in comparison with PRAK and p38␣ alone (Fig. 1, A and C). This notion is supported by coexpression of PRAK and p38␤, which shows that p38␤-PRAK is exclusively localized to the cytosol, significantly different from the localization of PRAK and p38␤ when expressed alone (Fig. 1, A and C). It is possible that endogenous p38␣ and p38␤ are partially in free form and some are docked with their partners, and therefore both nuclear and cytosolic localization of p38␣ and p38␤ have been observed. It is also possible that mammalian p38␣/␤ can be transported, as its yeast counterpart Hog1, in a NLSindependent mechanism (13), and thus both nuclear and cytosolic distribution can be observed. Because endogenous PRAK is primarily cytosolically localized, it most likely docks with protein partners such as p38␤. Because nuclear-cytosolic shuttling of PRAK was observed (12), the regulation of PRAK docking interaction could be important for the function of PRAK. Indeed, we show that PRAK-induced cell growth inhibition requires its nuclear localization (Fig. 7).
Our results show that the distinct locations of p38␣-PRAK and p38␤-PRAK are related to the function of the NLS of PRAK, but not its NES (Fig. 5), indicating that the interaction with import machinery was affected by complexing PRAK with p38␣ or p38␤. This idea was supported by the data that the mutations affect p38␣-PRAK importing are on the amino acids that can directly interact with importin ␣ in our modeling analysis and the mutations that did not influence p38␣-PRAK nuclear localization cannot reach importin ␣. These data suggest the cytosolic localization of p38␣(D145G)-PRAK and p38␣(L156V)-PRAK is due to either D145G or L156V mutations impairing nuclear import of the complex. However, it is still unclear why a single mutation of p38␤ on either residues 145 or 156 is sufficient to lead to nuclear localization of the p38␤-PRAK complex (Fig. 3E). Perhaps some sequences of p38␤ have some affinity for proteins of the import machinery, which allows nuclear import of p38␤-PRAK after a single mutation on either residue 145 or 156. Although NES appears to have no role in determining the subcellular location of the p38␣/ ␤-PRAK complex, NES is required for PRAK shuttling between the nucleus and cytosol (12). It was recently shown that interaction with nucleoporin is important for the nuclear import of ERK2 (27). Sequence alignment between ERK2 and p38␣/␤ showed that the region in p38␣ and p38␤ that corresponds to the nucleoporin binding region in ERK2 is conserved, suggesting that the difference between p38␣ and p38␤ is unlikely to be due to nucleoporin binding. Although the difference between p38␣ and p38␤ may not be related to nucleoporin binding, interaction with nucleoporin is still likely to be involved in the p38␣-PRAK complex nuclear import. Based on our data and previous published data by us and others (11,12), we propose that localization of p38␣/␤ and their downstream kinases PRAK and MK2 is determined by multiple factors, including NLS and NES, docking interactions, and their affinity for importin ␣ and other proteins of the importing machinery.
It is known that p38␣ and p38␤ have many different functions (16). Although the structural basis for the differences between these two proteins is not fully understood, different subcellular localizations are likely to play a role in the different functions of these two closely related proteins. We have recently found that PRAK plays a role in transactivation of p53 (20). It appears that nuclear localization of PRAK is required for PRAK to activate p53, as overexpression of p38␤ inhibits PRAKmediated p53 reporter gene expression (Fig. 7D). p38␣ is apparently responsible for the activation of PRAK in Ras-induced p53 activation, as p38␤ overexpression had an inhibitory effect on Ras-induced p53-dependent gene expression, whereas p38␣ overexpression had a slight enhancing effect (Fig.  7E). The different subcellular localizations of PRAK, resulting from its interaction with p38␣ and p38␤, are also likely to be responsible for the distinct effects of p38␣ and p38␤ on PRAK-mediated inhibition of NIH3T3 cells (Fig. 7, A-C). In light of the data in this article and in previous publications, we propose that different subcellular localizations of PRAK, regulated by p38␣ and p38␤, affect its ability to regulate p53 transactivation and thus cell proliferation. . PRAK-induced growth inhibition depends on its nuclear localization. A, QTTG and R361Q, two mutants of PRAK impaired for nuclear import, show diminished inhibitory effects on cell growth compared with wild type PRAK. NIH3T3 cells were separately transfected with GFP, GFP-PRAK, GFP-PRAK(QTTG), and GFP-PRAK(R361Q). After 24 h of transfection, cell proliferation was determined by a BrdUrd intake assay as described under "Experimental Procedures." Results are represented as mean Ϯ S.D. of five independent experiments. The statistical analyses were performed by employing unpaired Student's t test, and the significance is indicated as a p value. *, p Ͻ 0.001 compared with control GFP (the first column); #, p Ͻ 0.001 compared with GFP-PRAK. B, p38␣ enhances and p38␤ inhibits PRAK-induced growth inhibition of NIH3T3 cells. NIH3T3 cells were transfected with GFP, GFP-PRAK, p38␣, and p38␤ in different combinations as indicated. BrdUrd intake assay was performed as above. The data presented are mean Ϯ S.D. of five independent experiments. Statistical analyses were performed as in A. *, p Ͻ 0.001 compared with cells transfected with GFP (first column); #, p Ͼ 0.05 compared with control GFP. OE, groups cotransfected with GFP-PRAK, fourth to sixth columns), p Ͻ 0.001 compared with the corresponding groups transfected with GFP (first to third columns). ƒ, p Ͻ 0.01 compared with the group transfected with GFP-PRAK alone (fourth column). ૾, p Ͻ 0.001 compared with the group transfected with GFP-PRAK alone. C, the function of p38␣ and p38␤ in regulating the growth inhibition mediated by PRAK can be swapped by interchanging amino acid residue 145. Expression vectors of GFP, GFP-PRAK, p38␣(D145G), and p38␤(G145D) were transfected into NIH3T3 cells in different combinations as indicated. The data are shown as the mean Ϯ S.D. of five independent experiments. *, p Ͻ 0.001 compared with control GFP (first column); #, p Ͼ 0.05 compared with control GFP. OE represents p Ͻ 0.001 when groups transfected with GFP-PRAK were compared with the corresponding groups transfected with GFP (first to third columns), individually. ૾, p Ͻ 0.001 compared with the group transfected with GFP-PRAK alone (fourth column). ƒ, p Ͻ 0.05 compared with the group transfected with GFP-PRAK alone. D, different effects of p38␣ and p38␤ on PRAK-stimulated p53 transactivation. NIH3T3 cells were transfected with different combinations of expression vectors: p38␣, p38␤, p38␣(D145G), and p38␤(G145D), together with p53 reporter PG-Luc and pCMV-lacZ. Luciferase activities were determined 24 h after transfection and normalized to the values of ␤-galactosidase activity. The data shown are the mean Ϯ S.D. of five independent experiments. *, p Ͻ 0.001 compared with the control (first column); #, p Ͻ 0.05 compared with the group transfected with PRAK alone (sixth column). OE, p Ͻ 0.001 compared with the group transfected with PRAK alone (sixth column). E, Ras-induced p53 activation is affected differently by p38␣ and p38␤. Expression vectors of Ha-RasV12, p38␣, p38␤, p38␣(D145G), and p38␤(G145D) were transfected into NIH3T3 cells in different combinations as indicated. Luciferase activity was determined as described in D. *, p Ͻ 0.001 compared with the control (first column); #, p Ͻ 0.05 compared with the group transfected with Ha-RasV12 alone (sixth column). OE, p Ͻ 0.001 compared with the group transfected with Ha-RasV12 alone (sixth column).